Embed Size (px)
Citation preview
fMRI of Human Primary Visual Cortex at Submillimeter
Resolution: Ûcuiar Dominance and îontrasi Perception in
Am blyopia
Bradley Gordon Goodyear
Graduate Program in Medical Biophysics I
Submitted in partial fùlfillment of the requirements for the degree of
Doctor of Philosophy
Faculty of Graduate Studies The University of Western Ontario
London, Ontario August, 1999
O Bradley G. Goodyear 1999
National Library 1+1 of Canada Bibliothéque nationale du Canada
Acquisitions and Acquisitions et Bibliographie Services services bibliographiques
395 Wellington Street 395. nie Wellington OttawaON K1AON4 Ottawa ON K1 A ON4 Canada Canada
The author has granted a non- exclusive licence ailowing the National Library of Canada to reproduce, loan, distribute or sel1 copies of this thesis in microforrn, paper or electronic formats.
The author retains ownership of the copyright in this thesis. Neither the thesis nor substantial extracts kom it may be printed or otherwise reproduced without the author's permission.
L'auteur a accordé une Licence non exclusive permettant a la Bibliothèque nationale du Canada de reproduire, prêter, distribuer ou vendre des copies de cette thèse sous la forme de microfiche/film, de reproduction sur papier ou sur format électronique.
L'auteur conserve la propriété du droit d'auteur qui protège cette thèse. Ni la thèse ni des extraits substantiels de celle-ci ne doivent être imprimes ou autrement reproduits sans son autorisation.
Abstract
Functicnûl rnrwetic resocriice imaging ft?.fRI) of the humm Snin exploits the
blood-oxygenation-level-dependent (BOLD) contrast of magnetic resonance images to
identify areas within the cortex that respond to a presented stimulus or task. At
submillimeter spatial resolution, fMRI bccomes quite difticult, cvrn at a magnetic field
strength of 4 Tesla, since the signal-to-noise ratio (SNR) within image voxels is
significantly reduced. This makes reliable detection of a BOLD response a forbidding
task.
The object of this thesis was to provide methods to (a) irnprove the signal-to-noise
ratio for high spatial resolution functional magnetic resonance images and to (b)
appropnately prescribe image orientations and locations to optimize the BOLD response
within selected regions of interest. These methods included the construction and
implementation of a radio frequency (RF) surface coil to improve SNR and RF
homogeneity throughout the image and a functional scout imaging technique that uses
receiver phase cycling to create functional maps as the scanner collects the image data.
Using the newly constructed coil and a modified echo plana imaging sequence
and image reconstruction scheme, tMRI studies at submillimeter spatial resolution were
performed to identify the ocular dominance colurnns within the human pnmary visual
cortex (VI) of healthy individuals with normal or conected-to-normal vision.
iii
Area V1 was identified using a fMRI experiment that demonstrated a
correlation between the magnitude of the BOLD response within V1 and the stimulus
î û ~ i ~ t This iûir&tim a-% not dctcctcd in highcr ûrder t is -d mis.
Functional MRI studies of VI of individuals who have unilateral amblyopia were
performcd to identifi the neuronal correlate of the reduction of contrast sensitivity of the
arnblyopic eye. The pooled fMRI response to rnonocular stimulation of the amblyopic
eye was significantly reduced, and reflected the decrease in contrast sensitivity. High
resolution tMRi revealed that the ocular dominance columns of the amblyopic eye were
significantly reduced in size for individuals who developed amblyopia during infancy.
The results of this work demonstrate a fMRI technique that can reliably resolve
functional units of the cortex on a submillimeter scale. In addition, this technique cm be
used to investigate brain plasticity at the cortical columnar level resulting from
developmental visual disorders or trauma.
Key words: functional magnetic resonance imaging, IMRI, BOLD, submillimeter
resolution, pnmary visual cortex, V 1, amblyopia, contrast, contrast sensitivity .
Co-Authorsbip
T l . iric iuiiowiiig thesis curitains niaieriai Crorn ibtx previousiy pubiished
manuscripts (Chapters 3 ,4 and 5 ) , a fourth and fifth manuscript submitted for publication
(Chapten 6 and 7), and a sixth manuscript in a submission format (Chapter 2). Additional
data to supplement these manuscripts are presented in Appendices A through D.
Al1 of the experimental work presented within this thesis was performed by
Bradley Goodyear.
The principal author of al1 original manuscripts. versions of which appear in
Chapters 3, 4, 5, 6, and 7 was Bradley Goodyear (Chapters 3. 4, 5, 6, and 7). The
remaining authors and their contributions are: Ravi Menon - senior author and thesis
supervisor (Chapters 3. 4. 5, 6 . and 7), Joseph Gati - technical assistance (Chapter 3),
David Nicolle - provided patient volunteers and helpful discussions (Chapter 6 and 7),
and Keith Humphrey - assistance in psychophysics experimental design and helpful
discussions (Chapter 6).
For copyright releases, see Appendix E.
Acknowledgements
1 wouid like to thank my supervisor Dr. Ravi Menon for his expertise, enthusiasm,
assistance, and advice duing the last four years. it was a pleasure working with him.
To my wife, Angela, thank you for your love and understanding. I couldn't have
accomplished this without your support, encouragement, and patience. Thank you for
always being a subject at the last minute, and helping me to keep my sanity when al1
those around me were losing thein.
To the rest of my family - my father, Gordon. rny rnother. Louise, and rny
brothers and sisters. Boyd, David. Ted, Glenda, and Hazel - thank you for your love,
support, and always being just a phone cal1 away.
A thank-you to loe Gati. Dr. Keith Hurnphrey. Dr. Melvyn Goodale, Dr. David
Nicolle, Dr. Tutis Vilis, Dr. Brian Rutt, Dr. Terry Thompson, and Dr. Aaron Fenster for
their input and helpful discussions.
As well, thank you to Enzo Barberi for his help and patience, to Chnstopher
Thomas (cubicle mate in crime), and to the rest of the guys in the lab.
Finally, to the rest of the students and staff at the Imaging Labs at the Robarts
Research tnstitute, thanks for making the past four years so enjoyable, and making RRI
one of the most pieasurable and stimuiating places to work.
vii
Table of Contents
Page . .
CERTIFICATE OF EXAMINATION ................................................ ii
ABSTRACT ................................................................................. iii
CO-AUTHORSHIP .~.............~........................................................... v ACKNO WLEDGEMENTS ................................................................. vi
TABLE OF CONTENTS .................................................................... viii
LIST OF TABLES ............................................................................ xii
LIST OF FIGURES .......................................................................... xiii
LIST OF APPENDICES .................................................................... xvii
........................................ LIST OF ABBEVIATIONS AND SYMBOLS xviii
CHAPTER 1 INTRODUCTION ......................................................... 1
.................... 1 . 1 Introduction to Functional Magnetic Resonance Irnaging 1
....................... 1.1. I Clrigins of Functional Magnetic Resonance Irnaging 2 1.1.2 Image Acquisition: T2 *-weighted Gradient- Recalled Echo Planar
.......................................................................... Imaging .3 1.1.3 Black Design of Etperimental Paradigm and Data Ana lysis ............. 7
1.2 Issues to Address for High Resolution tMRI .................................. 8
..................... 1.2.1 Increasing SNR Using Quadrature RF Surface Coils 9 ..... 1.2.2 Optimking SNR and BOLD Contrust within Submillimeter Voxels I I
...... I . 2.3 Mmimizing the Effective Resolution of T2 *-weighted MR Images 12 ............ 1.2.4 Increasing the Temporal Resolution of thefMRI Time Series 14
.................. 1.2.5 Maintuining Spatial Specif city of the BOL D Response l j .................................... 1.2.6 Reducing Artvacts Due jo Head Motion 16
... 1.3 Submillimeter Units of Cortical Activity : Ocular Dominance Colurnns 17
............................... 1.3.1 Architecture of Ocular Dominance Columns 17 .... 1.3.2 The Distribution o f Ocdur Dominance Columns within the Cortex II
1.4 The Neural Basis of Amblyopia ................................................. 22
1.4.1 Definit ion and Classifcaliion of Amblyopia ................................. 22 1 - 4 2 Behavioral Measurements in Arnblyopia .................................... 23 1.4 3 Physiologieal Measurements in Amblyopia ................................. 25
1.5 Hypotheses .......................................................................... 26
1.7 References ........................................................................... 30
viii
CHAPTER 2 A DEDICATED QUADRATURE TRANSMIT/RECEIVE RF SURFACE COIL FOR HGH-RESOLUTION fMRI ............
........................................................................ 2.1 Introduction
2 2 Methods ............................................................................
.............................................................................. 2.3 Results
......................................................................... 2.4 Discussion
.......................................................................... 2.5 References
................................. CHAPTER 3 THE FUNCTlONAL SCOUT IMAGE
3.1 Introduction ......................................................................... 3.2 Methods .............................................................................
3.2.2 Single Slice FLASH Mapping ................................................ ...................................................... 3 . 2 2 iMulii-slice EPl 1Mupping
............................................................ 3.3 fiesults and Discussion 52
................................................ 3.4 Conclusions 56
3.5 References ........................................................................... 57
CHAPTER 4 EFFECT OF LUMINANCE AND CONTRAST ON BOLD fMRI .......................... RESPONSE IN HUMAN VISUAL AREAS 58
4.1 Introduction ........................................................................ 58
............................................................................ 4.2 Methods 60
.............................................................................. 4.3 Resutts 62
.......................................................................... 4.4 Discussion 64
.......................................................................... 4.5 References 68
CHAPTER 5 SUBMILLIMETER FUNCTIONAL LOCALIZATION IN .......................................... HUMAN STRIATE CORTEX
5.1 Introduction ........................................................................ ............................................................................ 5.2 Methads
52.1 Erperimentd Paradigm ...................................................... 5.2.2 jMRI ...................................................................*......... 5.2.3 Anai'ysis .........................................................................
5.3 Results and Discussion ............................................................ 5.4 Conclusions ............................. .. ...................................... 5.5 References ..........................................................................
CHAPTER 6 A NEURONAL CORRELATE OF SUPRATHRESHOLD CONTRAST PERCEPTION IN HUMAN AMBLYOPIA ..........
6.1 Introduction ........................................................................
6.2 Methods ............................................................................ ......................................................................... 6.2.1 Subjects
............................................................ 6.2.2 Behavioral Testing ............................................................ 6.2.3 Functional Imnging
/ - . . O -. 4 ÜUÎU Anuiysis .................................................................. .............................................................................. 6.3 Results
......................................................................... 6.4 Discussion
......................................................................... 6.5 References
C W T E R 7 A NEURAL SUBSTRATE FOR THE DOhIINANT EYE IN HUMAN AMBLYOPIA AND NORMAL VISION ..................
........................................................................ 7.1 Introduction
............................................................................ 7.2 Methods
......................................................................... 7.2.1 Subjects 7.2.2 Stirnzrlus Presentation .........................................................
............................................................ 7.2.3 Functional Imaging 7.2.4 Data Analysis ..................................................................
7.3 Results and Discussion ........................................................... 7.4 References .........................................................................
CHAPTER 8 SUMMARY AND FUTURE DIRECTIONS .......................... 8.1 S m a r y ...........................................................................
8.1.1 A Quadrature RF Surface Coil for High Resoltition fMRI ............... 8.1.2 The Functional Scout Image ................................................. 8.1.3 Contrast Modulation of the BOLD Response in Human Visual Cortex . 8.1.4 Submillimeter Fwnctional Localizat ion in Human Striate Cortex ....... 8.1.5 Neuronal Correlates of Suprathreshold Contrast Perception in
Human dmblyopia ............................................................ 8.1.6 A Neural Substmte for the Dominant Eye .................................
8.2 Future Directions ......................... ,. ..................................... 6 2.1 Furure Directions fur Contrasr Moduktion of the BOLD Response ... 8.2.2 Future Directions for Submillimeter Functional Localization .......... 8.2.3 Future Directions offMRI Studies of Human Amblyopia ................
8.3 References .......................................................................... APPENDK A: TIME COURSES OF MR SIGNAL DURING BINOCULAR
PHOTIC STIMULATION USNG DIFFERENT L W A N C E LEVELS ........................ ... ................................... 152
APPENDIX B: OPTIMIZING MR AM> VISUAL STIMULUS PARAMETERS FOR HIGH RESOLUTION FMRI STUDIES OF OCULAR DOMINANCE ..................... .. ................................... 153
APPENDIX C: THE INDEPENDENCE OF THE OCULAR DOMINANCE OF THE BOLD RESPONSE ON SPATIAL FREQUENCY ......... 157
APPENDIX D: CONTRAST AND LUMINANCE CALIBRATION OF THE PROJECTION SCREEN .............................................. 159
APPENDIX E: ETHICS APPROVAL ................................................ 161
APPENDIX F: COPYRIGHT RELEASES ........................................... 162
CURRICULUM VITAE .................................................................... 164
List of Tables Page
TABLE 6-1 : Classification of amblyopia for each subject by orthoptic assessrnent .................................................................. 99
TABLE 7- 1 : Classification of amblyopia for six patients with amblyopia developed during infancy and two patients with amblyopia
........................................... developed afier 2 years of age 123
xii
List of Figures Page
FIGURE 1 - 1 : a. Diagram of a segmented gradient-recalled echo planar imaging puise sequence used in WH. b. The resuiting Bspace trajectory for the puise sequence in (a). c. A single gradient-recalled echo centered about k, = O, collected during the readout gradient, Gx ... 5
FIGURE 1-2: A T2 *-weighted image of the visual cortex ............................ 6
FIGURE 1-3 : Block design paradigm for a tMRI experiment ....................... 8
FIGURE 1 4 : The normalized voxel PSF for a T2*-weighted 256 by 256 image colltcted using an EPI imaging sequence as described in Figure 1-1 ................................................................... 14
FIGURE 1-5: Sagittal anatomical localizer image of the visual cortex showing the calcarine sulcus .......................................................... 1 6
FIGURE 1-6: a. Diagram of a horizontal slice through the brain showing the retinocortical pathway and how al1 visual input From one visual field projects directly to one hemisphere. b. Layers of the prirnary visual cortex showing the terminations of projections from the LGN. and the ongin of projections to other areas of the cortex and to deeper brain structures ..................... l.. .................................. 1 9
FIGURE 1-7: a. Surface diagram of a patch of layer 4c of the primary visual cortex showing the arrangement of ocular dominance columns. b. Diagram of a cross-section of layer 4c of the pnmary visual cortex showing the arrangement of orientation-specific neural cells relative to the ocular dominance columns ................................................ 20
FIGURE 1-8: The contrast sensitivity function for the normal and arnblyopic eye. 24
FIGURE 2-1 : Diagram of the quadrature surface coi1 ................................. 38
FIGURE 2-2: a The impedance of one of the elements of the quadrature RF coil. b. Isolation of the two coil elements of the quadrature surface coil given as the difference between the signal transrnitted to one coil and the signal detected by the other coi1 ................................ 41
FIGURE 2-3: Tz*-weighted MR images of two 4-mm thick slices of the hurnan visual cortex obtained using (a) a two-element 8-cm diameter square- loop quadrature RF coil and (b) a 8-cm diameter square-loop linear RF surface coi1 placed at the occipital pole ............................ 42
FIGURE 2-4: Contour plots of SNR for the Tz*-weighted MR images of Figure 2-3 for one slice acquired using the quadrature d a c e coil and the linear surface coi1 .................................................................. 43
FIGURE 2-5: Contour plots of the percentage increase in image SNR obtained using the quadrature surface coil for the same two slices in
................................................................... Figure 2-3 43
xiii
FIGURE 3- 1 : The functional scout imaging experiment .............................. 49
FIGURE 3-2: Functionai scout maps of a 10-mm-thick oblique axial slice produced using a receiver-phase-cycled FLASH sequence with (a) 2, (b) 4, and (c) 8 stimulation-control cycles. (d) FLASH-derived activation map produced using a Student's ?-test showing pixels exhibiting a positive
.............. response during photic stimulation at a p value of 0.0 1 53
FIGURE 3-3: Multi-slice EPI-derived hinctional scout maps of three 5-mm-thick sagittal slices through the visual cortex produced using receiver phase cycling with (a) 1, (b) 2, (c) 4, and (d) 8 stimulation-control cycles.
... The corresponding TI-wcightèd anatomical slice is shown in (e) 55
FIGURE 4-1 : BOLD fMRI activation maps of visual cortex showing areas exhibiting a positive response to a flickenng red LED stimulus with a luminance
............................ of(a)0.5,(b)2.2,(c)85,and(d)250cd/m2 63
FIGURE 4-2: The number of pixels within V 1 and extrastriate showing a positive response to the LED stimulus for one subject whose activation maps
................................................... are s h o w in Figure 4-1. 64
FIGURE 4-3: Mean MN response within the ROI encompassing VI and the ROI located within extrastriate for (a) a single subject and (b) six subjects ....................................................................... 65
FIGURE 4-4: BOLD MRI map showing pixels that are cornrnon to al1 activation maps .......................................................................... 66
FIGURE 5-1: TI-weighted anatomical image of a typical oblique slice in the human visual cortex ................................................................. 74
FIGURE 5-2: Schematic of the 'sliding window' approach to reconstructing the ....................................... multi-segment multislice EPI data 77
FIGURE 5-3: T2*-weighted images of three contiguous oblique/axial slices of the visual cortex of one subject prescribed parallel to the caicarine sulcus ........................................................................ 8 1
FIGURE 5-4: The distribution of the magnitude of image intensity over the time course of a baseline experiment within regions of interest within V 1 and within the background noise outside the head .................................. 82
FIGURE 5-5: a. Activation map of voxels, overlaid on the corresponding anatomical slice, demonsating a significant fMRI response to the three monocular right-eye stimulation penods. b. Activation map of voxels demonstrating a significant fMRI response to the three monocular left-eye stimulation periods. c. Average time course showing the tMRI response for the voxels shown in (a). d. Average time course showing the fMRI response for the voxels s h o w in (b). 83
FIGURE 5-6: a. Activation map of the voxels kom the maps of Figure 5-5 that show higher levels of activation during right-eye stimulation and left-eye stimulation. b. Activation rnap of the same slice showing
only those voxels of the activation maps in Figure 5-5 that show significantly higher activation during right-eye stimulation and significantly higher lefi-eye stimulation. c. The fMRI response of each voxel in (a) and (b) during right-eye stimulation minus the fi.49.I response within the s m c wxet dukg lefi-eye stimulûUon, divided into 5% bins and nonnalized to 1 .............................. 84
FIGURE 5-7: Maps of ocular dominance for 4 different subjects .................... 87
FIGURE 5-8: Activation map of activity during the baseline expenment with no visual stimuli ................................................................ 88
FIGURE 5-9: (a) Time-course during the visual stimulation paradigm for a single subject for pixels identified as right and left eye ODCs. (b) Average timecourses across al1 subjects. (c) Distributions of the number of
........................ pixels as a function of fiactional signal change 90
FIGURE 6-1: Method used to determine perceived contrast. a. The variable contrast standard fiequency is paired with each 22% contrast test fiequency. b. Response function obtained using a two-alternative temporal forced choice method ...................................................... 102
FIGURE 6-2: Behavioral and tMRi measurements of perceived contrast above threshold for participants with (a) normal vision and (b) unilateral amblyopia ................................................................... 106
FIGURE 6-3: Average MM response for activated image voxels exhibiting a significmt correlation with perceived contrast measured with the corresponding eye for participants with (a) normal vision and (b) unilateralarnblyopia ....................................................... 108
FIGURE 6-4: a. Functional maps of voxels correlating with the perceived contrast measured with the amblyopic eye and the non-amblyopie eye, overlaid on two matornical slices for one participant with unilateral strabismus and anisometropia. b. Average fMRI response of the activated voxels
......................................................................... in(a) 108
FIGURE 6-5: Average fMRi response of voxels whose response magnitude as a function of spatial fiequency correlated with perceived contrast measured wirh the non-amblyopic eye .................................. 109
FIGURE 6-6: The number of voxels activated in response to monocular stimulation as a function of spatial fiequency for participants with (a) normal vision and with (b) unilateral arnblyopia ........................................ 110
FIGURE 7-1 : Maps of ODCs overlaid on sagittal anatomical images of the medial bank of the visual cortex of one hemisphere for A. two subjects with right-eye dominant nomal vision and B. one subject with lefi-eye exotropia/anisometropia and one subject with right-eye exotropia/anisometropia ................................................. 1 26
FIGURE 7-2: Maps of ODCs overlaid on corresponding transverse anatomicd images for A. one subject with left-eye dominant normal vision and
one subject with right-eye dominant normal vision and B. one subject with right-eye exotropialanisometropia and one subject with left-eye exotropia/anisometropia ................................................... 127
FIGURE 7-3: The percent of cortical area of Vlc occupied by ODCs of the arnbiyopic eye as a function of visuai acuity of the eye for six aduits with arnblyopia since infancy ............................................. 130
FIGURE 7-4: The average fMRI response measured within ODCs of 1 1 subjects ................ with normal vision and al1 8 subjects with amblyopia 13 1
FIGURE 7-5: Distribution of the ocular dominance of the fMRI response of each voxel in al1 maps for al1 1 1 subjects with normal vision and for al1 8 subjects with unilateral amblyopia .................................... 133
FIGURE A- 1 : The average tMRl signal within the ROIS of Figure 4- 1 for (a) V 1 and (b) extrastriate cortex ................................................. 152
FIGURE B-1 : (a) Ratio of the area of V lc occupied by the dominant eye to the area occupied by the non-dominant eye, calculated as descnbed in Chapter 7 for two subjects with normal visiori. (b) Number of voxels within fùnctional maps of ocular dominance ................................... 154
FIGURE 8-2: The ratio of cortical area occupied by dominant eye colurnns to non- ............ dominant eye columns as a function of image voxel size 1 5 5
FIGURE B-3: (a) Percentage of cortical area of V l c occupied by the ocular dominance columns of the dominant eye using 4 different EPI sequences for two subjects and using a 4-second visual stimulus. (b) Nurnber of activated voxels in the maps of ocular dominance for the same EPI sequences in (a) ............................................ 156
FIGURE C-1: Ocular dominance of al1 voxels in the ODC maps for 8 subjects with amblyopia at three spatial frequencies .................................. 157
FIGURE D- 1 : (a) The contrast and luminance at the projection screen as a function of the contrast of the CRT of the stimulus-controlling cornputer measured using a Minolta CS- 100 Chroma Meter. (b) Same as in (a), except a DC offset was added or subtracted to the luminance at each contrast level to maintain a mean luminance that did not Vary as a h c t i o n of contrat by more than 2% ................................... 160
List of Appendices Page
APPENDLX A:
APPENDIX B:
APPENDIX C:
APPENDIX D:
APPENDIX E:
APPENDIX F:
TIME COURSES OF MR SIGNAL DURING BINOCULAR PHO'L'lCl S'l'lMULA'I'lON USING DIFFERENT LUMINANCE
.................................................................. LEVELS 152
OPTIMIZING MR AND VISUAL STIMULUS PARAMETERS FOR HLGH RESOLUTION FMRI STUDIES OF OCULAR
.......................................................... DOMINANCE 153
THE INDEPENDENCE OF THE OCULAR DOMINANCE OF THE BOLD RESPONSE ON SPATIAL FREQUENCY ......... 157
CONTRAST AND LUMINANCE CALIBRATION OF THE .............................................. PROJECTION SCREEN 159
ETHICS APPROVAL ................................................. 16 1
........................................... COPYRJGHT RELEASES 162
xvii
List of Abbreviations and Syrnbols
O
C A
00
2-D
BI balun
B o
BOLD
C
CNR
C P ~
CRT
CSF
DC
EPI
FID
FLASH
fMRI
FOV
FT
FWHM
GE
GRE
Gx
G-v
G: Hb
- fiequency
- induced emf
- pnsm diopters
- magnetization precessional frequency
- two-dimensional
- altemating transverse magnetic field
- balanced-to-unbalanced transformation network
- main static magnetic field
- blood-oxygenation-level-dependent
- capacitance
- contrast-to-noise ratio
- cycles per degree of visual angle
- cathode ray tube
- contrast sensitivity fùnction
- direct current
- echo planar imaging
- free induction decay
- fast low-angle shot
- functional magnetic resonance imaging
- field of view
- Fourier transform
- full-width at half-maximum
- GeneraI Electric
- gradient-recalled echo
- readout gradient
- phase encoding gradient
- slice selection gradient
- hemoglobin
xviii
Hb02
i kx
L
LED
LGN
MP
MR
MW
NIH
NMR
OD
ODC
OIS
OS
PCF
PET
PSF
C B F
rCBV
RF
rf
ROI
SF
SNR
STEAM
T
Tl
Tt
T2 *
- oxyhemoglobin
- x-direction in k-space
- y-direction in k-space
- inductance
- light emitting diode
- lateral geniculate nucleus
- magnetization prepared
- magnetic resonance
- magnet resonance imaging
- National Institutes of Health
- nuclear magnetic resonance
- right eye
- ocular dominance column
- optical imaging of iritrinsic signals
- lefi eye
- perceived contrast function
- positron emission tomography
- point-spread function
- regional cerebral blood flow
- regional cerebral blood volume
- radio frequency
- radio Frequency
- region of interest
- standard spatial fiequency
- signai-to-noise ratio
- stimulated echo acquisition mode
- Tesla
- longitudinal relaxation time constant
- transverse relaxation time constant
- transverse relaxation tirne constant including al1 magnetic field inhomogeneities
xix
T2 ' - transverse relaxation time constant specific to perturbations induced by blood deoxygenation
TE - echo time
TF - test standard frequency
TI - inversion time
TR - repetition time
VI - primary visual cortex
Vlc - central visual field representation of primary visual cortex
V2, V3 - higher order extrastriate visual cortical areas
VEP - visual evoked potential
xc - capacitative reactance
XL - inductive reactance
Chapter 1
Introduction
1.1 Introduction to Functional Magnetic Resonance Imaging
1.1.1 Origins of Functional Magnetic Resonance lmaging
Magnetic resonance imaging (MRI). originally developed in 1973 ( 1.2). has
become a powerful imaging modality in diagnostic neuroradiology and neuroscience,
since it c m noninvasively provide anatomical images of the brain with high resolution
and contrast. However, as demonstrated in 1990. MRI c m also produce images with
blood-oxygenation-level-dependent (BOLD) contrast to provide functional information
about the brain (3.4). In 1992. this technique was applied in human studies to map
cortical activity in response to sensory stimulation (5,6) and task performance (7). Now
more commonly known as BOLD contrast functional MN. or BOLD NEU, this
technique is applied in neuroimaging centers throughout the world to study both normal
and pathological brain function.
BOLD contrast relies on the brain's microvasculature response to local
metabolisrn associated with increases in neural activity. This response is an increase in
the flow of oxygenated blood to the site of neural activity (8). BOLD fMRI exploits the
magnetic properties of the main transport mechanisms within blood vessels for oxygen
delivery to tissue, which are the hemoglobin (Hb) macromolecules contained within the
red blood cells. Oxyhemoglobin; that is; hemoglohin with 4 b o n d oxygen moleder ,
and deoxyhemoglobin, hemoglobin lacking bound oxygen molecules, exhibit different
magnetic properties when placed in an extemal magnetic field (9). Deoxyhemoglobin is
paramagnetic with a relatively strong positive susceptibility, producing a magnetic field
that is additive to the main extemal field. That is, the susceptibility difference between
the blood vesse1 and the surrounding tissue olters the magnitude of the surrounding
rnagnetic field, creating an inhomogeneous magnetic field environment. While in this
environment, hydrogen nuclei (protons) within water molecules will experience a change
in the resonant frequency of precession of their magnetic moments (which are a result of
the intrinsic proton spin) about the mis of the main magnet field. This causes a loss of
phase coherence between proton spins, and a reduction in MR signal.
However. the hemodynamic response to stimulation or a cognitive task is an
increase in the flow of oxygenated blood to cortical sites of' neural activity,
overcompensating the increase in oxygen extraction (1 0). This hyperoxygenation phase
takes several seconds to occur. and decreases the local deoxyhernoglobin concentration.
This will, in tuni, decrease the magnitude of the locally induced suceptibility. Hence, the
hemodynamic response will actually be detected as an increase in MR signal. The goal of
BOLD fMRI is to identiQ localized increases in image voxel intensity within MR images
that are sensitive to the increase in blood oxygenation caused by the hemodynarnic
response of the local capillary bed of gray matter responding to a stimulus or task.
However, large vessels adjacent to
downstream from the actual site of
gray matter tissue and large draining veins M e r
neural activity also contribute to the BOLD signal.
Nonetheless, the local 1 y indiiced magnetic field i~honlogeri.eities mendorod & w e d~ EQ?
extend far beyond the vesse1 wall. Hence, the BOLD effect in large vessels should remain
localized. In any case, strategies exist to elirninate these contributions within MR images,
and some of these strategies will be discussed in Section 1.1.3 and in Section 1.2.1 when
discussing high spatial resolution tMRI.
1.1.2 Image Acquisition: Tr *-weighted Gradient-Recalled Echo Planar imaging
The accrual of phase incoherence between water proton spins resulting From local
magnetic field gradients surrounding the deoxygenated red blood cells is characterized by
a time constant T? '. This time is referred to as a transverse relaxation time, since phase
incoherence leads to a relaxation of the transverse magnetization (i.e.. MR signal) to
equilibrium as the phase of proton spins becomes randomly distributed over the
transverse plane. In the absence of this effect, there already exists spin dephasing due to
other mechanisms such as differing chemical environments, and these are charactenzed
by a relaxation tirne constant denoted by T2. In tandem. T2 and T?' contribute to an
a~parent relaxation time constant denoted by Tr*, which is shorter then either T2 or T2 ',
and is given by
A decrease in local magnetic field inhomogeneity causes an increase in Tz*, and hence
the BOLD response will be manifested as a small but detectable increase in image voxel
intensity if MR images me ro!!ec!ed vdh T2* &$!hg.
The accrual of phase incoherence is reversible if the inhomogeneities are static
over the time required to collect the MR signal. This reversibility c m be exploited using
spin-echo imaging techniques which refocus the phase incoherence. However, to achieve
T2* weighting, imaging techniques utilize the initial decay (fiee induction decay, or FID)
of the MR signal that is created with a radio frequency (RF) excitation pulse. One such
technique is Gradient-Recalled Echo (GRE) imaging which uses magnetic field gradients
to refocus MR signal to f om what is termed a gradient-recalled echo. To characterize the
evolution of the BOLD response. it is necessary to collect a time series of images with
adequate temporal resolution. To do this, GRE imaging can be used within an echo plana
imaging (EPI) sequence. as shown in Figure 1-la with the data collection gnd (k-space)
shown in Figure 1-1 b. Images are formed by a two-dimensional Fourier reconstruction of
the k-space data that consist of one gradient-recalled echo for every k, line, centered
along k, = O. as s h o w for exarnple, in Figure 1-1 c. The data array for every second echo
must be reversed since it was collected in the opposite direction. This could lead to
modulation along the phase-encode direction within the data set, however, this can be
corrected during post-processing.
Echoes collected nearest k, = O contribute to most of the signal intensity in the
resulting image. It has been demonstrated using fast low-angle shot (FLASH) imaging
sequences that collecting
fiom the inflow of blood
these echoes first reduces large vesse1 contributions resulting
into the imaging plane during image acquisition (1 1). Once a
Figure 1 - 1 : a. Diagram of a segmented gradient-recalled echo planar imaging pulse
sequence used in MM. The excitation pulse (RF) creates transverse magnetization, and
the G: gradients select the location of the x-y imaging plane. G, is the readout imaging
gradient during which MR signal is collected (Le., during the flat portions of the gradient
waveform: 1 4 2 , 3-4. etc.). GdV is the phase-encode gradient, which is increased
incrementally between readout gradients to move the trajectory in the k, direction. The
time fiom the RF pulse to when the data point nearest the center of k-space is collected is
denoted as the echo tirne, TE. The gradients along al1 three axes at the end of the
sequence spoil the remaining magnetization to prevent sfimulated echoes after subsequent
RF pulses (in the case of a short repetition time, TR). b. The resulting k-space trajectory
for the pulse sequence in (a). Numbers along the trajectory correspond to the labeled time
points in (a). c. A single gradient-recalled echo centered about k, = O, collected during
the readout gradient, Gx.
In multi-slice irnaging, either al1 segments of k-space are collected for each slice
before imaging the next slice, or the same segment is collected for dl slices before
cr!!ectir?g the zext ccpei. t . Tke !atter tetb~?Iqx is refmed ?û ûs ~! i~~- inter!eû~izg , ûiid
can improve the image signal-to-noise ratio (SNR) although it does increase the image
collection time per slice. This is explained in more detail in Chapter 5.
To enhance T2* weighting and BOLD contrasr, imaging parameters are usually
chosen such that the center of the first echo, Le., the point (k, = O. k, = O), is collected at
TE -- T2 * after the rf excitation pulse (1 2). However, this value of TE is usually decreased
somewhat to increase the SNR of the image and reduce susceptibility-induced motion
sensitivity. An example of a T2*-weighted MR image collected at a magnetic field
strength of 4 Tesla (T) is s h o w in Figure 1-2.
Figure 1-2: A T2*-weighted image of the visual cortex. This image was collected using
an 8-cm quadrature surface coi1 (described in Chapter 2) and a segmented gradient-
recalled echo planar imaging sequence with a k-space trajectory as show in Figure 1-1 b
(8 segments, 7 interleaved slices, 128x1 28 matrix, 14 cm square field of view, echo time
(TE) = 15 ms, repetition time (TR) = 250 ms, RF flip angle = 25').
1.1.3 Block Design of Experimentd Paradigm and Data Analysis
!nten~ity differencer betwpen ~*-v..eighted Imges c~!!ec!ed d~riirg r cûgzitive
task and during a resting state are not apparent upon inspection, since the BOLD response
in the capillaries of gray matter tissue corresponds to a change in image intensity that is
typically less than 6%. Xlthough subtracting an image collected during rest from an
image collected during the task would reveal image voxels that show an intensity change.
such a result is not reliable due to the subtle intensity differences involved. This is one
reason why most tMRI experiments investigating differences between two conditions
employ a block design paradigm (13,14), as s h o w in Figure 1-3. The paradigm consists
of repeated task periods altemating with resting control States. and continuous image
collection during each task and control period. Image voxels that exhibit an intensity
modulation that parallels the block design paradigm are then identified by statistical
analysis of the time series of images on a voxel-by-voxel basis using a Student's 1-test, or
by cross-correlating the time series with an expected hemodynamic response, at a stated
confidence level.
However. these statisticai analyses require that ail data measurements are
independent of one another. This is, of course, violated in fMRi since the intrinsic delay
of the hemodynamic response acts as a low-pass filter that convolves with the time series.
As a result, points in the series will be independent only if their sepration is on the order
of about 6 seconds (1 4). One way to account for this during data analysis is to express the
actual number of independent measurements in the time senes as the total number of
images divided by the number of images within a 6 second period (14). A more accurate
way is to correlate an average time series for image voxels within gray matter with itself
This gives an estimate of the temporal separation of independent images for an individual
subject (15), which is typically 5 to 8 seconds. Once the nurnber of independent images
has been determined, the confidence lwzl of the statistic cm be corrected.
task task task task
image number
Figure 1-3: Block design paradigm for a tMFU experiment. Images are collected
continuously throughout the expenment as task and control penods are altemated.
control
1.2 Issues to Address for High Resolution fMRI
L
control
A few investigations of brain function using EPI have been performed at high-
spatial resolution (16,1?), however, most stcdies remsin st re!ative!y low in-plane
resolution (> 2x2 mm2) due, in part, to limitations of conventional MR scanners. These
limitations include available MR signal at the operating magnetic field strength, and
h
either the slow nse time of the imaging gradients to maximum strength or the inability of
fast gradient systems to operate continuously. To image cortical function on a
control
submillirneter scale within these limitations, experimental techniques are needed that can
control control
provide T2*-weighted MR images with a suficient SNR, and a fMRI time series with a
temporal resolution that is adequate to fully characterize the BOLD response within
imiividlw! Imrge voxe!s. !fi rc.,i!io~~, tlie EOLE respnse bas to rem~in spti3!ly spcific
to the functional units of interest within imaging planes whose locations and orientations
have already been optirnized.
1.2.1 Increasing SNR Using Quadrature RF Surface Coils
Image SNR cm be increased using specialized RF coils. These coils provide a
relatively small rnagnetic field, Bi, that is perpendicular to the direction of the main static
field, B,. and is at a frequency that is as close to o, as possible to maximize the
interaction between BI and the magnetization (Le., resonance). This interaction is in the
fom of an applied torque that ' tips' the axes of precession away fiom Bo and towards the
transverse plane. RF coils are sirnply conducting loops whose natural inductance. L,
creates an inductive reactance (X, = j d . where j = G) which is nulled by the
distribution of lumped-element capacitance, C, around the loop to create capacitative
reactance, X, = o-' , equal in magnitude to Xr. This requires the relation
1
Thus, at a chosen operating frequency, namely u = a, the impedance of the coil is
purely resistive with no associated phase, ensuring that no power delivered to and fiom
the coil is refiected when the resistance of the coil if matched to the resistance of the
comecting coaxial cable.
In addition to transmitting RF energy to a smple, the RF coil can be used for
receiving MR signal. The decay of the transverse rnagnetization (or Free induction decay,
FID) created hy B! is detected as an nrrillating vnltsge rince each mtathg source ~f
magnetization induces its own voltage (or emf, 8 in the coil. Just as BI is strongest
nearest the coil. so is 5 for sources of magnetization nearest the coil. This correspondence
bctween 5 and BI is known as the principir of reciprocity. Hznce, the signai measured in
an MR spectroscopy or irnaging experiment will be proportional to the induced emf (1 8).
The most embedded regions of the cortex under investigation in this thesis, the
primary visual cortex, are at most 5 to 7 cm beneath the skull at the posterior of the brain.
This permits the use of surface coils for functional imaging. These types of coils provide
considerable SNR gains over head volume coils for cortical regions near the skull.
however, RF homogeneity is not presewed, since MR signal decays rapidly with distance
fiom the coil.
The Bi field used to create transverse magnetization can be thought of as two
fields each with an amplitude of +B, rotating in opposite senses. Only one of these two
halves rotates in the correct sense, Le., in the same sense as the precessing magnetiatition.
The other, rotating in the wrong sense, has a negligible effect. Thus, half of the power
associated with BI is wasted. This power loss can be avoided by using two quadrature
coils which are fed altemating currents 90" out of phase, such that when the curent is
maximum in one coil, it is minimum in the other. This makes full use of the counter-
rotating half of BI, and the increased efficiency upon transmission is an improvement of
a o f the SNR upon reception (19). The 90' phase lag of the transrnitted current and the
recombination of signal upon reception are both achieved using a quadrature hybnd
1 1 0 3n\ \ L U,LU)r
1.2.2 Optirniahg SNR and BOL D Contrasr within Submillirneter Voxels
It has been demonstrated that BOLD contrast increases with magnetic field
strength when TE has been optimized (1 2), suggesting that a higher magnetic field is best
for BOLD WRI studies. However, T2 * for gray matter tissue decreases with increasing
magnetic field strength, leading to a faster signal decay. The MR scanner used in this
thesis was a Unity INOVA VaridSiemens 4 Tesla whole-body system. This magnetic
field strength provides adequate MR signal for submiilimeter studies. The T2* of gray
matter at 4 Tesla is 30-35 milliseconds (12), which is sufficiently long to provide
adequate T2* weighting within these submillimeter voxels.
The prescribed volume of image voxels is govemed essentially by the available
gradient strength. The imaging system used in this thesis was equipped with 25 mT/m
whole-body, actively-shielded gradients with a rise time of 500 p to maximum snength.
This gradient strength is sufficient to achieve submillimter image resolution. Gradients
that provide 40 mT/m with whole-body systems, and 100 mT/m with head-insert
gradients, will theoretically allow studies at even greater spatial resolution. However, at
present, these systems have other hardware limitations involving component overheating.
Therefore, continuous operation of the gradients is not achievable. For the system used in
this study, the actual limitations of imaging at subrnillirneter resolution are the available
signal within image voxels, the time required to obtain the desired gradient strength, and
L!C tirne rpqrikc! m c~! !ea the imge d m As 2 conseqcence îf these tidzg restrictions,
an image voxel will actually contain additional signal arising from outside the prescribed
voxel, thereby decreasing the effective image resolution. This is cornmonly referred to as
T2* bluning, and will be discussed M e r in Section 1.2.3.
1.23 bfàxirnizing the Effective Resolution of T2 *-weighted MR Images
One segment of k-space data takes tens of milliseconds to collect using an EPI
imaging sequence as described above. During this time. the magnetization is
continuously dephasing due to Tl* effects as previously described, and the magnitude of
echoes along the collected train in the phase-encoding gradient, Gv, direction decreases
exponentially. A Fourier transform of ihis exponentially decaying envelope of echo
magnitude is a Lorentzian lineshape that describes an image voxel whose width extends
beyond its prescribed dimensions in the y-, or phase-encode, direction. There is no
appreciable increase in voxel size in the readout direction. The profile of the effective
voxel width is commonly referred to as the voxel point-spread function (PSFI. The longer
it takes to collect a train of echoes, or the longer the echo train itself, the more T2* decay
cm occur. and the wider the resulting PSF. One way to combat this effect is to minimize
the time between echoes. However, this leads to a significant reduction in image SNR as
the bandwidth (in Hz) per pixel is increased. Another way is to decrease the number of
echoes collected after an excitation pulse by increasing the number of segments in k-
space. This technique can introduce image artifacts, however, these cm be removed
during post-processing.
Figure 1-4 shows how the voxel PSF decreases as the number of collected echoes
per segment is decreased for a 256 by 256 image (14 cm by 14 cm) wiîh 2.8 milliseconds
between echoes in the phase-encode direction, and using a value of T2* = 33 ms for gray
matter. An estimate of the effective voxel width is the full-width at half-maximum of the
corresponding PSF. In this thesis, either 128 by 128 images or 256 by 256 images were
collected with 16 echoes per k-space segment, corresponding to approximately a 32%
increase in voxel size in the phase-encode direction. Thus, when prescribing a voxel site
of 0.545 mm. as in some of the studies in this thesis. the effective voxel width will be
0.7 19 mm in the phase-encode direction.
Collecting 8 echoes per segment can funher reduce the effective image voxel
width. However, the resulting imaging sequence becomes significantly less like EPI in
nature. In addition, if imaging time is to be kept constant, there will be a significant
reduction in image SNR due to the reduced time between image segments which imposes
restrictions on the magnitude of the flip angle of the RF excitation piilse.
124 1 26 130 130 132
image voxel
Figure 1-4: The normalized voxel PSF for a T2*-weighted 256 by 256 image collecte(
using an EPI irnaging sequence as descnbed in Figure 1 - 1. The labels on the horizontal
a i s correspond to the center of the nth image voxel. The values on the left portion of each
curve indicate the nurnber of echoes per k-space segment for the corresponding PSF. The
values on the right portion of each curve are the percentage increases in image voxel
width in the phase-encode direction, calculated frorn the full-width at half-maximum of
the corresponding PSF. The vertical lines about the 128'~ voxel show the prescribed voxel
width.
1.2.4 Increasing the Temporal Resolution of the fMR1 Time Series
The use of EPI enables fast irnaging dong with rnultislice coverage of anatomical
ueas of intrrrsi. N%an imaging at hi& spatial resoiution, however, it iakes severai
seconds to image the entire primary visual cortex, since this requires a number of slices,
especially when trying to minimize slice thickness. The resulting temporal resolution will
be inadequate to locate and characterize the peak of the BOLD response within the fMRI
time senes. Using a slice-interleaved image acquisition as outlined in Section 1.1.2 and as
described in Chapter 5, the time between points in the fMRI t h e series can be reduced to
the tirne required to collect one segment for al1 slices, if images are reconstnicted using
sets of temporally continuous k-space segments for the corresponding slice. This doesn't
identi@ the peak of the BOLD response with more accuracy.
1-25 Maintaining Spatial Specijkity of the BOLD Response
Using stimuli of prolonged duration cm lead to spatial 'bleeding' of the vascular
response to surrounding tissue not necessarily involved in the task (21), thereby
destroying the spatial specificity of the BOLD response (22). When investigating
hct ional differences between submillimeter anatomical units that are adjacent to each
other, the vascular response of one can totally mask the response of the other if the spatial
specificity of the vascular response is lost. [t has been demonstrated that to maintain
spatial specificity of the vascular response, stimuli duration should be less than the time it
takes for the hyperoxygenation phase of the hemodpamic response to saturate (-5-6
seconds) (2 1). At the same time, however, stimulus duration has to be sufficient to ensure
that the difference in magnitude of the BOLD response under different conditions can be
measured reliably and reproducibly.
To ensure that slices selected for imaging encompass the anatomical areas of
interest, it is important that slice locations and orientations are optimized pnor to the
Iplreso!uticr. tqxrirnent. This wi!! r!rc reduce tke !ike!ihccd rf missivg v e r s sf
neural activity specific to the stimulus or task. The investigations in this thesis were
concemed solely with the primary visual cortex, which lies primarily dong the calcarine
sulçus, and is easily discemable within a sagittal anatornicai image of the visual conex, as
show in Figure 1-5. This image can easily be obtained pnor to the functional imaging
experiment to accurately position slices for investigation using high-resolution MRI.
However, this is not necessarily the case for studies of cortical areas that are not as easily
identified in quick anatomical localizer images, such as higher order visual areas. Hence,
a method of grossly localizing the anatomical areas of interest is required. Chapter 3
addresses this issue. and describes a quick functional imaging alternative to an
anatomical scan.
Figure 1-5: Sagittai anatomical localizer image of the visual cortex showing the calcarine
sulcus. Slices for investigations of the primary visual cortex are usually prescnbed either
parallel or perpendicular to this sulcus.
1.2.6 Reducing Artifacfs Due to Head Motion
Of primzry c m c e m Ir! high-resr!ctior? 'h.!PJ ir tire immbi l i~9? icn cf the
participant's head during imaging, while ensuring that the participant remains
cornfortable. The use of a bite-bar, which involves having the participant bite down on a
piece of dental wax secured to an immovable platform, is one method to minimize
movements of the skull. Although effective, this method of head imrnobilization cm be
uncornfortable and disconcerting to participants recruited from the general population or
fiom a patient population.
Another method involves bean-filled vacuum bags that, when evacuated, ngidly
conform to the shape of the head to restrict head movement. In this thesis, however. head
movement was restricted using a well-padded head vise that fits snuggly to the sides of
the head of the participant with an additional padded bar that fits across the participant's
forehead. This device was constructed in-house. Any remaining artifacts due to head
motion, or motion of the brain within the cerebral spinal fluid due to physiological
pulsation, c m be corrected during post-processing of the image data using the motion-
correction algorithms of SPM96 (23).
1 3 Submillimeter Units of Cortical Activity: Ocular Dominance Columns
A demonstration of the capability of high spatial resolution IMR[, incorporating
the proposed solutions to the problems as discussed above, would be to resolve
submillimeter functional units of activity within the brain. Nature has provided such units
within the human primary visual cortex (area V 1), namely the ocular dominance colurnns
(ODCs). ODCs are groups of neural cells that receive input from one eye only, and are
exclusive to the primary visual cortex.
First. however. as a review of how visual information amves at the primary visual
cortex, consider Figure 1-6a, which is a diagram of a horizontal slice through the brain
showing the major components of the retinocortical pathway. Visual information that
falls on the retina enters the brain via the optic nerve and along the optic tract to a laminar
thalamic structure called the lateral geniculate nucleus (LGN). Each hemisphere of the
brain directly receives input from one visual field only. That is, visuai information that
falls on the nasal retina (Le.. the nose side) of one eye will cross over at the optic chiam
to the contralateral hemisphere and join with visual information that falls on the temporal
retina (i.e.. the temple side) of the ipsilateral eye. From the LGN, visual input travels
along the optic radiations and terminates within the primary visual cortex at the posterior
of the brain for early stage processing before being projected to higher order visual areas
of the cortex or to deeper brain structures.
In humans, the primary visual cortex runs mainly dong the calcarine sulcus, is
approximately 2 mm thick, and is also a larninar structure. YS s h o w in Figure 1-6b. As
s~,cfiï, iii riwc i-6b, dJ5e Iirvjections Som gïe Lûii ielTknate wmn Iayzr $. is -witiih
this layer (layer 4c) that visual input from each eye is segregated into altemating groups
of neural cells called oczrl~r dominance columns. Above and below layer 4c, neurons also
exhibit a bias towards one rye, howcver, they can also be influenced by the other eye.
most especially above and below the borders of the ocular dominance columns. These
neurons are referred to as binocular neurons.
Figure 1-6: a. Diagram of a horizontal slice through the brain showing the retinocortical
pathway and how al1 visual input from one visual field projects directly to one
hemisphere. b. Layers of the primary visual cortex showing the terminations of
projections from the LGN, and the origin of projections to other areas of the cortex and to
deeper brain structures. Layers are drawn to reflect relative size.
Studies of ocular dominance column architecture were pioneered by Hubei and
Wiesel, where they demonstrated that ocular dominance columns in macaque monkey
srriatc cûrien zïe atmpd as a k ~ ~ â t i ï î g siâbs that rm imgcïïtidly dong th3 îuiiicol
surface (24,25). An exarnple of this arrangement is s h o w in Figure 1-7a. In histological
studies, this anangement has also been demonstrated in human primary visual cortex,
where it was found that the width of ocular dominance colurnns ranges t'rom 0.5 mm to
1 .O mm (26). In addition, running perpendicular to the ocular dominance slab are groups
of neural cells that are specific to orientation (Figure 1-7b). The full 180 degrees of
orientation are repeatedly represented across the slab. and these orientation columns are
nearly an order of magnitude smaller than the ocular dominance columns (26). However,
the separation of orientation-specific cells representing the same orientation is on the
same order as the size of an ocular dominance coiumn.
Investigations of the arrangement of ocular and orientation dominance within the
primary visual cortex of awake animals have been performed using optical imaging of
intrinsic signals (OIS) (27-29). Using the reflectance properties of oxygenated blood,
intrinsic signals can be shown to be spatially specific to the sites of these ocular
dominance cnlumns (for a review of the OIS technique, see ref 30).
Figure 1-7: a. Surface diagram of a patch of layer 4c of the primary visual cortex
showing the arrangement of ocular dominance columns (e.g. black = lefl eye column,
white = right eye column). b. Diagram of a cross-section of layer 4c of the pnmary visual
cortex showing the arrangement of orientation-specific neural cells relative to the ocular
dominance columns (L = left eye column. R = right eye column).
1.3.2 The Distribution of Ocular Dominance Columns within the Cortex
On average, in normal pnmary visual cortex, the amount of cortical area devoted
to the central visual field (-20 degrees of visual angle) is thought to be equal for the left
and right eye. However, 97% of the population have a dominant eye (65% right eye; 32%
left eye) (3 1,32). That is. when viewing a target or scene within the central visual field,
one eye dominates the visual input. To date, there have been no investigations of how eye
dominance in the central visual field correlates with the size, distribution, or neural
activity of ODCs in normal developed visual cortex in an atternpt to identi& the neural
basis of this behaviorai phenomenon. One reason for this is that there is no non-invasive
way of performing studies of the distribution of ODCs throughout the entire central
visual field representation, and it is difficult to assess the eye dominance of nonnally
developed animals. The only investigation of eye dominance at the cortical column level
is an animal model, where a dominant eye was created by suturing the eyelid of the
fellow eye during the animai's infancy. The result was a marked increase in ODC width
at the expense of the visually deprived eye (33-35).
For the peripheral visml field (>20 degrees), is has been demonstrated that the
corresponding ODCs for the eye closest to that visual field are larger (35). In the far
periphery (>60 degrees, say) this is not dificult to imagine, since only one eye, and hence
dominant eye. can be used due to the obstruction of the bridge of the nose. This visual
angle, and beyond, corresponds to the monocular crescent in the primary visual cortex
where there are no ODCs, just a large group of neurons devoted to the eye used for the
far periphery .
Both the animal mode1 studies of the central visual field and the studies of the
cortical representation of the peripheral visual field in normally developed cortex suggest
that in normally visual cortex, there may be a bias in cortical are% albeit subtle, towards
the ODCs of the dominant eye even for the central visual field representation. The ability
of MRI to interrogate the entire cortex in a non-invasive marner rnay be able to suppon
or refute this hypothesis.
1.4 The Neural Basis of Amblyopia
Amblyopia is a developmental disorder of vision associated with a loss of visual
acuity in one of the eyes uith no detectable ocular pathology. and occurs in nearly 4% of
the North American population (36.37). Although defined by a loss of visual acuity,
amblyopia is more accurately described by a nurnber of abnormalities in spatial vision
including spatial acuity and spatial contrast sensitivity (38,39). Two visual disorders
commonly associated with amblyopia are anisometropia, a refiactive error of the eye that
causes a blur of the retinal image, and strabismus, a misalignment of the optical axes
where one eye is tumed inward (esotropia) or turned outward (exotropia). In strabismus.
the image that hlls on the fovea of the strabismic eye is different from the image on the
retina of the non-deviated eye. This eventually leads to a suppression of the image that
falls on the fovea of the strabismic eye. The alignrnent of the strabismic eye is commonly
corrected for cosmetic reasons; however, visual acuity is rarely restored to normal.
Depending on the reversibility or severity of the strabismus or anisometropia,
patients may not necessarily develop amblyopia. In any case, the affected eye is
commonly treated by training the eye in a nurnber of monocular tasks while occluding the
vision of the prefened eye. Although this can lead to a marked improvement in acuity.
normal vision is rarely restored. A suggested reason for this is that amblyopia commonly
develops during infancy or early childhood while the visual cortex is still developing. If
treatment is initiated after this critical period of development, then the neural
consequences of amblyopia may be irreversible (36,39).
1.12 Behavioral Measuremenis in Amblyopia
In the clinic. amblyopia is assessed by (a) the reduction of visual acuity as
measured using an eye chart, (b) the refractive properties of the lens, and (c) the angle
through which the eye has deviated bom normal while binocularly viewing objects
placed in different regions of the visual field. Although these measurements abide by the
definition of amblyopia, they fail to accurately describe the deficits of the amblyopic eye
during everyday visual experience, which is filled with objects that differ in shape. size.
position, luminance. and contrast, as well as other properties. One method of measuring
how the arnblyopic eye responds to these properties is by comparing the contrast
sensitivity function (CSF) of each eye. Contrast sensitivity is the reciprocal of the
contrast required to detect a target, and is usually measwed as a function of the spatial
frequency of sinusoidal gratings in cycles per degree (cpd) of visual angle. The CSF can
also be measured at different luminance, target size, and position in the visual field,
making the CSF an excellent mearure of the visual expenenre of the mblyopic eye.
As shown in Figure 1-8. the CSF for normal vision is not constant across spatial
fiequencies, but rather peaks between 2 and 5 cpd (40,41), and drops significantly at high
spatial Frequencies. In the amblyopic eye, the CSF is reduced (see also Figure 1-8), and
the difference between the preferred eye and the arnblyopic eye increases with spatial
frequency (42-44). Strabismic and anisometropic amblyopia demonstrate similar contrast
sensitivity functions, except at low spatial fiequencies, where the anisometropic eye can
t \
-nomaîeye + - - amblyopic eye
spatid f requency t cydes per degree)
Figure 1-8: The contrast sensitivity hc t ion (CSF) for the normal and amblyopic eye.
Contrast sensitivity can also be probed at contrasts above threshold (i .e.,
suprathreshold). In that case, the resulting measurement is not one of sensitivity, but
rather one of perceived contrast. It has been demonstrated that with increasing contrast,
the perceived contrast function (PCF) approaches a constant across al1 spatial fiequencies
(45). However, at relatively low contrasts (~30%) with moderate-to-low luminance. the
?CF cm stil! dernonstrate the percepal deficits of the mblyopic cye. In addition, when
presented with low-contrast grating patterns, amblyopes demonstrate impairments in
reaction tirne (46,47).
1.4.3 Physiological Measurernents in Am blyopia
The i,-urs! hsis fer ~ ~ h ! ; l ~ p k i~ Iriimms hm net been st~!ied ex?ei,sire!y.
Hence, the actual mechanisms and cortical areas underlying arnblyopia are not accurately
known. However, it is thought that although the activity in striate cortex may explain
somz aspects of the visual deficits witnessrd in behavioral measurements, a full
description can only be achieved by incorporating the abnormalities in the neural activity
within other areas of the visual cortex (48,49), in addition to V 1.
Nonetheless, a large body of research has demonstrated a modification in the
structural and functional properties of the primary visual cortex in animal models of
amblyopia (for a review. see ref. 50). In non-human primates with experimental
strabismus or anisometropia, it has been demonstrated that there is a severe decrease in
binocularly activated neurons (5 1,52). This is explainable for strabismus since visual
input at the corresponding locations on the retina are uncorrelated, and hence can lead to
a decrease in binocular interaction. In anisometropia, blurring of the retinal image might
cause an incoherence in the activity of high spatial fiequency cortical neurons innervated
by the affected eye. Support for this explanation is evident in macaque studies that
demonstrate a loss of binocular neurons (52). This may aiso help to explain the decrease
in contrast sensitivity of the eye at high spatial fiequencies. However, this explanation
does not seem plausible for the reduction in contrast sensitivity witnessed in strabismus,
since presumably the strabismic eye has been exposed to clear images, albeit not on the
correct axis.
Single unit work has demonstrated a shift in the response of primary visual cortex
,, .,,,, ,,,. ,,. C--- ~ I G L L U ~ ~ S a w a y t i v i i i ambiyüpis tye, a ï d a sigïiifi~aiit iiiiîei~ii~e iri Uie spaiiai
properties of neurons in the prirnary visual cortex dnven by each eye (5 1). In support of
these findings, it has been show in some cases that strabismus increases the spacing
between columns of the sarne eye (53) . However. the degree of ODC shrinkage depends
on the experimental manipulation of the eyes. Moreover, there is a large intrinsic
variablility in the size of ODCs, making changes in the cortex resulting from amblyopia
difficult to discern (54).
1 .S Hypotheses
The main hypothesis to be tested in this thesis is that tMRI is capable of
producing reliable and reproducible maps of brain function on a submillimeter scale, and
that tMRI can also demonstrate cortical plasticity on a submillimeter scale resulting From
a visual disorder such as amblyopia. To test this hypothesis, a number of goals (or
specific aims) need to be achieved. These include:
a ) the construction of a specialized RF coi1 c m provide adequate SNR and
sensitivity to the entire primary visual cortex, ailowing functional studies on a
submillimeter scale,
b ) the development of rapid functional localizer experiment to help optimize
prescribed slice locations and orientations prior to a hi&-resolution experiment,
C) the development of a short functional experiment to facilitate the demarcation of
primary visual cortex, and
4) the Oerelopmcr.t cf ?.$Pd =A St.i~:.ic:ü! ex.;eri=.,er,:s meas- cûn:rast
perception in normals and in amblyopes and that demonstrate the possible
neuronal correlates of the deficits in contrast perception witnessed in amblyopia.
1.6 Tbesis OutIine
Chapter 2 describes the constru ction of a sp ecialized quadrature RF surface coil
for shidies of the pnmary visual cortex. The objective was to produce a coil that provided
signifiant SNR gains over similar linear and head volume coils.
Chapter 3 describes a fMRI technique to grossly localize anatomical regions of
interest. The method is essentially a subtraction of images collected during stimulation
and rest by means of inverting the phase of the MR receiver during one of these
conditions. More elaborate and spatially specific experiments can then be planned when
the proper orientation and location of slices are planned usine the results of this
hinctional scout experirnent.
Chapter 4 demonstrates that the fMRI response withinh primary visual cortex is
sensitive to changes in the contrast of visual stimuli. A quick functional experiment
involving the modulation of image contrast provides a method of isolating primary visual
cortex within fimctional images that obviously contain other cortical areas in addition to
primary visual cortex. Additional data showing sample fMRI t h e courses in V1 and in
~ X ~ Ü S ~ S ~ C m a s ûf th2 ~ i j ü a ! cûitcir xc shû;ili iïi Appeiidi~ A.
Chapter 5 demonstrates how fMRJ is capable of imaging ocular dominance
column distributions, and ttius can localize brain function on a submillimeter scale.
Chapters 7 shows that a bias in ocular dominance column size towards the preferred eye
is a possible neural substrate within the cortex for this dominant eye for humans who
have developed amblyopia during infancy. Appendix C demonstrates that the ocular
dominance of the fMRL response is independent of spatial frequency. An accurate
demonstration of this finding is possible only if imaging and stimulus parameters are
optimized to maintain the spatial specificity of the BOLD response. These parameters are
discussed and demonstrated in Appendix B.
Before the ocular dominance column distribution in the primary visual cortex of
amblyopes is discussed in Chapter 7, Chapter 6 demonstrates a neuronal correlate of the
deficits witnessed in the perception of contrast rneasured with the amblyopie eye. This
finding lends credence to the shift in ocular dominance to the preferred eye, as
demonstrated in Chapter 7. The luminance and contrast calibrations of the projection
screen are shown in Appendix D.
Finally, Chapter 8 provides a sumrnary of the thesis, and outlines possible fùture
directions of the present studies and outlooks for hi&-resolution W.
1.7 References
1. Lautebur PC. Image formation by induced local interaction: Examples employing
2. Mansfield P, Grannell PK. NMR 'diffraction' in solids? J. Phys. C6, L422 (1973).
3. Ogawa S, Lee TM, Nayak AS, GIynn P. Oxygenation-sensitive contrast in magnetic
resonance imaging of rodent brain at high fields. iMagn. Reson. Med. 14, 68-78
( IWO).
4. Ogawa S, Lee TM, Kay AR, Tank DW. Brain magnetic resonance imaging contrast
dependent on blood oxygenation. Proc. Nari. Acad Sci. USA 87,986899872.
5. Kwong K. Belliveau JW, Chesler DA, Goldberg IE, Weisskoff RM, Poncelet BP,
Kennedy DN, Hoppei BE, Cohen MS, Turner R, Cheng HM, Brady TJ, Rosen BR.
Dynamic magnetic resonance imaging of human brain activity during primary
sensory stimulation. Proc. Natl. Acad. USA 89, 5675-5679 (1 992).
6 . Ogawa S, Tank DW, Menon R, Ellermann JM. Kim S-G, Merkle H, Ugurbil K.
Intrinsic signal changes accompanying sensory stimulation: Functional brain mapping
with magnetic resonance imaging. Proc. N d . Acad. Sci. USA 89,595 1-5955 (1992).
7. Bandettini PA, Wong EC. Hinks RS, Tikofsky RS. Hyde JS. Tirne course EPI of
human brain function during task activation. Magn. Reson. Med. 25,390-397 ( 1 992).
8. Roy CS. Sherrington CS. On the regulation of the blood-supply of the brain J.
Physiol. (Lond) 1 1,8501 O8 ( 1890).
9. Pauiing L, Coryeii C. n i e magneric propenies and structure of hemogiobin. Proc.
Nad. Acad. Sci LISA 22,210-2 16 (1 936).
1 O.Fox PT, Raichle ME. Focal physiological uncoupling of cerebral blood fiow and
oxidative metabolism during sornatosensory stimulation in human subject. Proc. Noil.
Acad. Sci. USA 83, 1140-1 144 (1986).
1 1 .Menon RS, Ogawa S, Tank DW, Ugurbil K. 4 Tesla gradient recalled echo
charactenstics of photic stimulation induced signal changes in human primary visual
cortex. Mugn. Reson. Med. 30,3 80-3 86 (1 993).
12. Gati JS, Menon RS, Ugurbil K, Rutt BK. Experimental determination of the BOLD
field strength dependence in vessels and tissues. Magn. Reson. Med. 38, 296-302
(1997).
13. Rucher RLy Bandenini PA, O'Craven KM, Savoy EU, Peterson SE, Raichle ME,
Rosen BR. Detection of cortical activation during averaged single trials of a cognitive
taîk using functional magnetic resonance imaging. Proc. N d . Acad. Sci USA 93,
14878-14883 (1 996).
1 4 Menon RS, Goodyear BG. Subrnillimeter functional localization of in human striate
cortex using BOLD contrast at 4 Tesla: Implications for the vascular point-spread
function. iMugn. Reson. Md. 4 1.230-235 ( 1999).
15.Boynton GM, Demb JB. Glover GH, Heeger DJ. Neuronal basis of contrast
discrimination. Vision Res. 39,257-269 (1 999).
16.Thompson MR. Venkatesan R, Kuppusarny K, Celik A, Lin W. Kido DK, Haacke
EM. Increased-contrast. high-spatial-resolution, diffusion-weighted, spin-echo, echo-
plana imaging. Radiology 2 10,253-259 ( 1999).
1 7 .Chen W. Zhu XH, Thulbom KR, Ugurbil K. Retinotopic Mapping of lateral
geniculate nucleus in humans using functional magnetic resonance irnaging. Proc.
Natl. Acad Sci USA 96,2430-2434 (1999).
1 8. Chen CN, Hoult DI. Biomedical Magnetic Resonance Technologv, Mould RF. ed.
(Adam Hilger, Bristol, 1989).
19. Hoult DI, Chen CN, Sank VJ. Quadrature detection in the laboratory Frame. Mugn.
Reson. Med. 1 , 3 5 9 4 5 (1984).
20. Malherbe JAG. Microwave Transmission Line Couplers, Norwood MA, ed. (Artech
House, 1988).
2 1 .Malonek D, Grinvald A. Interactions between electrical activity and cortical
microcirculation revealed by imaging spectroscopy: Implications for functionai brain
mapping. Science 272,55 1-554 ( 1 996).
Menon RS, Ogawa S. Hu X, Strupp JP, Andersen P, Ugurbil K. BOLD based
functional M N at 4 Tesla includes a capillary bed contribution: echo-plana imaging
correlates with previous optical imaging using intrinsic signals. Magn. Reson. Med.
33,453-459 (1995).
23. Friston KJ, Jevard P. Turner R. Statistical parametric maps in functional imaging: a
general linear approac h. Human Brain Mapping 2, 189-2 10 ( 1995).
24. Hubel DH, Wiesel TN. Receptive fields and functional architecture of monkey striate
cortex. 1 Physiol. 195,2 1 5-243 ( 1968).
25. Hubel DH, Wiesel TN. Ferrier lecture: Functional architecture of macaque monkey
visual cortex. Proc. R. Soc. Lond. B. 198, 1-59 (1 977).
26.Horton JC, Dagi LR. McCrane EP, de Monasteno FM. Arrangement of ocular
dominance columns in human visual cortex. Arch. Ophthalmol. 108, 1025- 1 O3 1
( 1990).
27. Gnnvald A, Lieke EE, Frostig RD, Gilbert C. Wiesel RI. Functional architecture of
cortex revealed by optical imaging of intrinsic signals. Nature 324.36 1-364 (1986).
28. Ts'o DY, Frostig R. Lieke EE, Grinvald A. Functional organization of primate visual
cortex revealed by high resolution optical imaging. Science 249,J 17-420 ( 1990).
29. Frostig RD, Lieke EE. Ts'o DY, Grinvald A. Cortical functional architecture and
local coupling between neuronal activity and the microcirculation revealed by in-vivo
high-resolution optical imaging of intrinsic signals. Proc. Natl. Acud. Sci USA 87,
6082-6086 (1 990).
30. Bonhoeffer T, Grinvald A in Brain Mapping: The Methods, ch. 3: Optical irnaging
based on intnnsic signals: the methodoiogy. Toga AW, Mazziotta JC, eds., Academic
Press, Inc. (1996).
3 1. Porac C, Coren S. The dominant eye. Prychol. Bull. 83,880-897 ( 1 976).
32. Purves D, White LE. Monocular preferences in binocular viewing. Proc. Natl. Acad.
Sci. USA 91,833908342 (1994).
33. Wiesel TN, Hubel DH. Single-ce11 responses in striate cortex depnved of vision in
one eye. J. Neurophysiol. 26, 1003-1017 (1963).
34. Wiesel TN, Hubel DH. Extent of recovery from the effects of visual deprivation in
kittens. J. Neurophysiol. 28, 1060- 1072 (1965).
35. LeVay S. Wiesel TN, Hubel DH. The development of ocular dominance columns in
normal and visually deprived monkeys. J. Comp. Neurol. 19 1, 1-5 1 (1980).
3 6 .Ciuffreda KH. Levi DM, Selenow A. Amblyopia: basic and clinical aspects.
Butterworth-Heinemann, Stoneharn (1 99 1 ).
37. Daw NW. Visual development. Plenum Press, New York (1 995).
38. von Noorden GK. Binocular vision and ocular moiility, 4" ed., C.V. Mosby, St. Louis
( 1 990).
39. Levi DM, Carkeet A. Amblyopia: a consequence of abnormal development. In Eurly
visual development: normal and abnormal, Simmons K, ed.. pp. 3 9 1 -408, Oxford
University Press, New York (1 993).
40 .Campbell FW, Robson JG. Application of Fourier analysis to the visibility of
gratings. J. Physio.1 (Lond.) 197,55 1-556 ( 1 968).
41 .Blakemore C. Campbell FW. On the existence of neurones in the human visual
system selectively sensitive to the orientation and size of retinal images. J. Physiol.
(Lond.) 2O3,23 7-260 ( 1 969).
42. Bradley A, Freeman RD. Contrast sensitivity in anisometropic amblyopia Invest.
Ophthamol. Vis. Sci. 2 1,467-476 ( 198 1 ).
43. Hess RF, Campbell FW, Greenlaugh T. On the nature of the neural abnormality in
human ambly opia: Neural aberrations and neural sensitivity loss. PfIuegers Arch.
Gesamte. Physiol. 337,20 1-207 (1 978).
44.Hess RF, Pointer JS. Differences in the neural basis of human amblyopia: the
distribution of the anomaly across the visual field. Vision Res. 25, 1577-1594 (1985).
4S.Georgeson MA, Sullivan GD. Contrast constancy: deblurring in human vision by
spatial fiequency channels. J. Physiol. 252,627-655 (1975).
46. Levi DM, Harwerth RS, Smith EL. Humans deprived of normal binocular vision have
binocular interactions tuned to size and spatial frequency. Science 206, 852-854
(1 979).
47. Loshin DS, Levi DM. Suprathreshold contrast perception in functional amblyopia
Doc. Ophthamol. 55,213-236 (1983).
48. Kiorpes L, Kiper DC, O'Keefe LP, Cavanaugh JR, Movshon JA. Neuronal correlates
of amblyopia in the visual cortex of macaque monkeys with experimental strabismus
and anisometropia. J Neurosci. 1 8,64 1 1-6424 ( 1998).
49. Thiele A, Bremrner F. Ilg UJ, Hofhann KP. Visual response of neurons from areas
V 1 and MT in a monkey with late onset strabismus: a case study. Vision Res. 3 7 , 8 5 3 -
863 ( 1 997).
5O.Movshon JA, Kiorpes L. The role of experience in visual development. In
Development of sensory systems in mammals, Coleman JR ed., New York, Wiley
Press, pp. 1 5 5-202 ( 1 990).
5 1 .Movshon JA, Eggers HM, G i v i MS, Hendrickson AE, Kiorpes L, Boothe RG.
Effects of early unilateral blur on the macaque's visual system. III. Physiologcial
observations. J. Neurosci. 7, 1340- 135 1 ( 1 987).
52.Horton JC, Hocking DR, Kiorpes L. Pattern of ocular dominance columns and
cytochrome oxidase activity in a macaque monkey with naturally occumng
anisometropic amblyopia. Vis. Neurosci. 14,68 1 -689 ( 1 997).
53. Lowe1 S. Ocular dominance column development: strabismus changes the spacing of
adjacent columns in cat visuai cortex. j. Aieurosci. 14.743 1-7468 (1994).
54. Horton JC, Hocking DR. Intnnsic variability of ocular dominance column periodicity
in normal macaque monkeys. J. Neurosci. 16.7228-7339 ( 1996).
Chapter 2
A Dedicated Quadrature TransmitPReceive RF Surface Coi1
for High-Resolution fMRI
by Bradley G. Gooàyear
2.1 Introduction
The optimization of image signal-to-noise ratio (SNR) is critical in high
resolution fùnctional magnetic resonance imaging (MN) studies of the huma. cortex
because of the small MR signal within subrniilimeter image voxels. Although SNR can
be improved somewhat by clever pulse sequence design. RF coil optimization is an
obvious first step. A single-loop circular surface coil of diarneter d gives the highest SNR
for a volume at depth d (1,2), however, RF homogeneity drops rapidly with distance from
the coil, and regions of interest are !imited to areas comparable with the dimensions of
the coil (3), making high resolution functional MRI studies of the cortex with this type of
coil design a difficult task.
Theoretically, a two-element quadrature coil should provide a fi improvement
in image SNR (4), and if the coil could conform to the shape of the head, RF
homogeneity over the volume of interest would be greatly improved. However, such an
arrangement is not a simple one since the two coils must also be well isolated fiom each
other to eliminate the mutual inductance between them that drastically reduces the coil's
ûierr!! scnsiti~it;. 3:d efficie~cj (3). This r n i t ~ a l induç:muicc cx be t!imim:ed Sy
overlapping the two coils, with the amount of overlap determined empirically. When
isolation has been achieved, each element of the quadrature coil behaves as an
independent coil, where each has been matched in impedance (50 R) to the preamplifier
and coaxial cable connection.
In this paper, we demonstrate the construction, evaluation, and implementaion of
a quadrature RF coil consisting of two identical 8-cm diameter square-loop elements.
This coil has been used in both low resolution (5) and high resolution (6,7) fMRI studies
of the human primary visual cortex. We have compared the newly constructed coil to a
single element RF coil of a configuration identical to one of the two elements of the
quadrature coil to demonstrate the improvement in SNR.
2.2 Methods
The quadrature RF coil was mounted to the outer side of a 18-cm diameter
cylindrical section of acrylic. Al1 measurements of impedance, inductance and
capacitance were made using a Hewlett-Packard Network/Spectn.un Analyzer Mode1
41 95A (Hewlett-Packard, Pa10 Alto, CA). Figure 2-1 shows a planar view diagram of the
RF coil. Each element of the coil had an effective diameter of 8 cm, and was constnicted
from 1 cm wide copper strips formed in the shape of a square with each corner trirnmed.
The inductance, L, of each square loop was measured to be approximately 155
nH. At the precessional frequency of water protons at 4 Tesla (CO, = 170.3 MHz), the
capacitance. C . required for a resonant circuit (from o, = l / d z ) with purely resistive
impedance was 5.63 pF. This capacitance was distributed over 8 insertion points,
requiring 45.1 pF at each point. However, the RF coil connects to a 50-Ohm coaxial cable
and a 50-Ohm preamplifier. To rnauimize the power t r a d e r efficiency of the coil, the
coil impedance while loaded (i.e., placed in the proximity of the sample) must be equal to
50 Ohms. This was achieved by altering the values of each capacitor around the loop
which changes the impedance charactenstics of the coil. For fine tuning of the magnitude
of the resistance, one capacitor was replaced with a variable capacitor, and another
variable capacitor was insened to tune the resonant frequency to 170.3 MHz. The final
values of capacitance were obtained by adjusting the values of the variable capacitors
while monitoring the loaded impedance on the display of the spectrum analyzer.
In t h i s configuration, the voltage measured at the input to the coi1 was balaneed:
that is, positive on one side and negative on the other. To keep the shield of the
connecting coaxial cable at ground, the voltage m u t be unbalanced such that al1 curent
to and fiom the coil will be passed along the center conductor of the coaxiai cable. Thus.
a balanced-to-unbalanced transformation network (or 'balun') was included between the
coil and the coaxial connection (8,9). The isolation between the coils wtiile loaded was
assessed by exciting one coil with a narrow band of RF Frequencies about 170.3 MHz and
measuring the detected signal in the other coil.
Figure 2-1: Diagrarn of the quadrature surface coil. Each coil element has an effective
diameter of 8 cm. Capacitance is given in pico-Farads (pF). and inductance is given in
nano-Henrys (nH). The tuning and matching capacitors are adjusted to provide a loaded
coil impedance of 50 Ohms that is purely resistive at 170.3 MHz. The 'balun' network at
the base of each coil transforms the measured voltage to an unbalanced signal. The center
conductor of the coaxial cable is connected at '+', and shield of the coaxial cable is
connected at ground.
Using the coil with a VaridSiemens (Paio Alto, CA; Erlangen, Germany) Unity
INOVA 4 Tesla whole-body MR system, the power required for a 90' RF pulse within the
region of interest (i.e., the primary visual cortex) was determined by obtaining a one-
dimensional sagittal profile (2 ms sinc RF waveform; 256 points; 20 cm field of view
(FOV); echo time (TE) = 25 ms; volume repetition tixne (TR) = 5 s, slice thickness = 10
n~), 3 1 identif;.ing the tc=smissiw p c x : !ex! nccessq :û üchicvc thc zxxirn;;?;
signal intensity along the profile. T2*-weighted MR images parallel to the calcarine
sulcus within the visual cortex of one healthy volunteer subject were collected using a 3-
slice, 16-segment interleaved EPI gradient-recalled echo pulse sequence (0.55 mm x 0.55
mm prescribed in-plane resolution; TE = 15 ms; volume TR = 4 s; RF flip angle = 25';
4 mm slice thickness). with centric ordering of k-space and a navigator echo for every
segment (6). The subject's head was immobilized using a well-padded. plexiglass head
vice.
To compare the quadrature coil with a single-element linear RF coil, a linear coi1
was created by opening one of the coil loops (Le.. removing the capaciton) and re-tuning,
in situ. the remaining coil to 170.3 MHz. This coil was then positioned under the occipital
pole of the sarne subject. Al1 calibrations and imaging experirnents were then repeated,
ensuring that the position of the head relative to the coil was the sarne as the previous
experiments. Images were collected along the sarne orientation as with the quadrature
coil.
The SNR within images was rneasured by averaging the image intensity within a
selected region of interest within p h a r y visual cortex and dividing by the average image
intensity within a region of interest outside of the head. The value of image intensity at
each pixel in each image was then scaled to reflect the SNR at that pixel. Contour plots of
image SNR were then calculated for both the quadrature coil and linear coil images.
Since the images for the linear coil were collected in a separate imaging session, the
,,,,,, ,-,-,, ,-- ,-, - ---:&t as- - nns # A # ( 9 CI\ ;iiiagca w c i e ied&cd wiui iiiuSF C U ~ C ~ C ~ wiih die quadralure ç d ushg ar:wvo ( lu).
To show the percentage increase in image SNR when using a quadrature coil, new
images were created by subtracting the value of SNR at each pixel within images taken
using the linear coil from the conesponding value within images created using the
quadrature coil. The result was then divided by the SNR of each pixel within the images
taken using the linear coil. Contour plots of percentage increase in image SNR were then
calculated.
2.3 Results
Figure 2-2a shows the impedance (resistance and phase) of the quadrature coil
when placed near the occipital pole at the back of the head. At the operating frequency,
170.3 MHz, the magnitude of the impedance is 50 Ohms, and has no associated phase.
Figure 2-Zb shows that the signal detected by one coil when the other coil was excited by
a narrow band of frequencies is minimum at .te qerating frequency. The mount of
isolation (48 dB) while loaded renders one coil as essentially 'invisible' to the other coil
at the operating frequency of the coil.
Figure 2-2: a. The impedance (magnitude and phase) of one of the elements of the
quadrature RF coil. The peaks in the magnitude plot show the resonance frequencies of
the coaxial cable connected to the coil. b. Isolation of the two coil elements of the
quadrature surface coil given as the difference (in dB) between the signal transmitted to
one coi1 and the signal detected by the other coil. The differences reaches a maximum of
48 dB at f 70.3 MHz.
For the quadrature coil, the power level required for a 90' pulse (as measured
with the VarianfSiemens system) was between 30 and 31 dB. For the linear coil, the
required power was 32 dB. Hence, the increase in sensitivity of the quadrature coi1 results
in a 1-2 dB decrease in the power required to achieve a 90' flip of the rnagnetization.
This should be reflected in images obtained with the quadrature coil since MR signal
measured in volts is proportional to the image intensity. Examples of T2*-weighted
images as obtained in an fMRI experiment are shown in Figure 2-3. The measured
average signal intensity within a region of interest (ROI) in the pnmary visual cortex of
images obtained with the quadrature coil was 1.46 dB greater than the signal measured in
the same ROI in the images collected using the linear coil. This identically reflects the
increase in coi1 sensitivity apparent fiom the decrease in power required for the 90" pulse.
By inspection of Figure 2-3, it is difficult to identiS> improvements within the image in
terms of signal intensity. However, it is apparent that image intensity is more unifonn
acrnrr the i m a p (lefi-tn-right) for t!x qiizdysi_ti~-te coi! imzges r a h r firn a r d i d fd!-cff
of image intensity with distance from the coi1 as within the linear coil images. This is
more clearly demonstrated in coutour plots of image SNR as s h o w in Figure 2-4. Along
midline, there is litîle gain in image SNR. However, the uniformity of image intensity
from left-to-right is obvious.
Figure 2-3: T2*-weighted MR images of two 4-mm thick slices (14 cm FOV) of the
human visual cortex obtained using (a) a two-element 8-cm diameter square-loop
quadrature RF coi1 and (b) a 8-cm diameter square-loop linear RF surface coil placed at
the occipital pole (bottom of image, indicated by the white dots). Al1 MR parameters
were the same for (a) and (b).
Figure 2-4: Contour plots of SNR for the T?*-weighted MR images of Figure 2-3 for one
slice acquired using the quadrature surface coil (left) and the linear surface coi1 (right).
Figure 2-5 shows the relative increase in image SNR. The quadrature coil
provides 10-25% improvement in image SNR on mis, and 50-100% improvement off
Figure 2-5: Contour plots of the percentage increase in image SNR obtained using the
quadrature sudace coi1 for the same two siices in Figure 2-3.
2.4 Discussion
We have demonstrated that a quadrature surface coil that conforms to the
curvature of the back of the head provides a substantial increase in image SNR compared
to a linear coil whose dimensions are identical to one of the elements of the quadrature
coil. Of course a linear coil tvhose dimensions are on the order of the total size of the
quadrature coil would provide more coverage of the visual cortex, however, image SNR
would be greatly sacnficed. Thus, we tèel that the cornparison of the quadrature coi1 with
a linear coil with dimensions the same as one of the elements of the quadrature coil is the
rnost appropriate, and in no way provides any biases towards the quadrature coil
performance.
The increase in image uniformity allows direct comparisons of anatomical
structures that lie on the same SNR contour since any confounds due to differences in
image SNR with changes in depth have been reduced. Since many areas of the visual
cortex are retinotopically organized, a more uniform functional image should also make
inferences regarding eccentricity or polar angle much easier.
The curved nature of the quadrature coil provides greater coverage and RF
penetration, however a improvement in image SNR is not realized. Although the
quadrature coil provide only a 10025% increase in image SNR on axis, this is still a
substantial improvement. At high-resolution, the MR signal within a voxel is greatly
reduced, and a 10% increase in signal may be the difference in whether or not a BOLD
response c m be detected. An increase in image SNR also leads to a decrease in the
variance of measurements of signai intensity fiom image to image provided the MR
~&n?lor rtalii!ity i~ deqliate. is ~f k p r t m ~ e te .!EiKi Lbk ! e ~ h ~ ~ b l i ~ Y--
relies on measurement variance to identifj areas of brain activity.
However, RF coi1 design is just the first step in improving the spatial and
temporal resolution of functional MM, and it is clever pulse sequence design and the
spatial specificity of the cerebral microvasculature that will be the ultimate determinants
of the limits of the spatial resolution of MM.
46
2.5 References
Edelstein WA, Foster TH, Schenck JH. In Proceedings of the SMRM 41h Annual
Meetino. p. 964 (1987).
Roemer PB, Edelstein WA. In Proceedings of the SMRM 6lh Annual Meeting, p. 410
(1987).
Roemer PB, Edelstein WA, Hayes CE, Souza SP, Mueller OM. The NMR phased
array. Magn. Reson. Med. 16, 192-225 (1989).
Hoult DI, Chen CN, S a n k VJ. Quadrature detection in the laboratory Frame. Magn.
Reson. Med. 1,339-353 ( 1 984).
Goodyear BG. Nicolle DA, Humphrey GK, Menon RS. BOLD tMRI response of
primary visuai areas to suprathreshold contrast in hurnan amblyopia (submitted).
Menon RS, Goodyear BG. Submillimeter functional localization in human striate
cortex using BOLD contrast at 4 Tesla: Implications for the vascular point-spread
fiction. Magn. Reson. Med. 4 1.230-235 ( 1999).
Goodyear BG, Nicolle DA, Menon RS. A Neural Substrate for the Dominant Eye in
Human Amblyopia and Normal Vision (submitted).
Frankel S. Reactance networks for coupling between unbalanced and balanced
circuits. Proceedings of the I. R. E. 32,486-493 (1 94 1).
Holcomb WG. Gore JC. An improved network for impedance matching and
simultaneous unbalanced-to-balanced transformation. Magn. Reson. Imag. 3,295-296
(1 985).
10. Fnston W. Jezzard P,Tumer R. Statistical pararnetric maps in functional imaging: A
generd linear approach. Human Bruin Mapping 2, 1 89-2 1 0 (1 995).
Chapter 3
The Functional Scout image*
by Bradley G. Goodyear, Jmeph S. Gati and Ravi S. Menon
3.1 Introduction
In MR studies aimed at elucidating the basic mechanisms of the brain's response
to sensory inputs or cognitive tasks, it is often necessary to use a single slice (as in high
temporal resolution BOLD based tMRI (l)), ultra-high resolution tMRï (2) or a single
voxel (as in S E A M spectroscopy studies of lactate (3,4)). To date, the slice or voxel has
been prescribed using anatomic features (e.g. the calcarine sulcus) rather than functional
maps. which often means that the areas of largest activation for the actual stimulus used
are missed in single slice studies, or partial-volumed with unactivated cortex in
spectroscopy studies. In our studies where a clinical scanner (GE Signa 1.5 T) was used
( 5 ) , functional map generation required the transfer of image data "off-line" for post-
processing to localize areas of brain function using statistical methods. This can take 10
minutes or more, while the subject must continue to lie still in the magnet. In our research
scanner environment (Varian Unity INOVA 4 T), the analysis software "Stimulate" (6)
resides on the workstation that runs the scanner and the analysis is much faster.
'A version of this chapter has been published. Goodyear BG, Gati JS, Menon RS. The functional scout image: immediate mapping of cortical function at 4 Tesla using receiver phase cycling. Magn. Reson Med. 38,183-186 (1997). O i997 John Wiiey & Sons, Inc.
in either situation, once a functional map of the volume is generated, one has to
gauge "by eye" the angulation, slice thickness or voxel size and set these on the scanner
since the planning triols o f the scanner are not incorprî!id in =y of the zv&!srl?!e
analysis software packages. As a more imrnediate alternative, functional mapping using
"on-line" reconstruction and a dedicated cornputer has recently been demonstrated (7).
However, this requues extensive hardware interfacing and computational power.
In this chapter. we demonstrate a new imaging method which provides scout
maps of visual cortex activity directly by virtue of the data acquisition scheme with no
image post-processing and no computational demands. To demonstrate this method, raw
data collected during photic stimulation and control states were subtracted through phase
altemation of the receiver between states, while the phase of the transmitted RF was kept
constant, as s h o w in Figure 3-1. The resulting data was averaged over the desired
number of cycles, and a BOLD signal difference map was produced upon 2-D Fourier
transformation. This can be performed in a multi-slice mode and the activation maps
compared with conventional processing techniques. The method amounts to doing a
magnitude reconstruction of the cornplex-difference k-space data corresponding to
stimulation and control. The robustness of this technique allows us to directly plan
double-oblique slices or voxels exactly matched to the desired activation regions using
the scanner's prescription tools. One can use functional scouts of this type to optimize the
coverage of a multislice data set, to control the angulation and slice thickness of a single
slice or optirnally match the angulation and voxel size of a spectroscopy study to the
activated region.
3.2 Methods
control ~ " ~ ~ a ~ n n
stim<lus s t 7 h h y stimulus off on I I Yf
receiver= 8 phase
.
receiver, phase
delay
Figure 3-1: The functional scout imaging expenment. Image accumulation for the
control period begins after an inter-volume delay in the console's acquisition memory.
During the control scans, the receiver phase is set to O*. The visual stimulus is gated on
after acquisition of the control images. Acquisition of the second k-space data set
summed in acquisition memory begins after another inter-volume delay with the receiver
phase now set to 180°. The transmitter phase remains at O0 during al1 acquisitions. The
stimulus is gated off immediately after acquisition. The experiment repeats for
subsequent cycles as the data continues to be summed to memory with a sign detennined
by the receiver phase.
Al1 experiments were performed on a Varian Unity INOVA 4 Tesla whole-body
acquire
imaging system (Varian, Pa10 Alto, CA; Siemens. Erlangen, Germany) equipped with 25
mTfm actively shielded whole body pdients. A distnbuted-capcitance, circulzr K.!
delay
surface coi1 with a 13 cm diarneter placed on the occipital pole was used for transmission
acquire
and reception of the RF signal. Stimulus-invoked signal changes were produced within
primary visual cortex of volunteers using LED goggles (Grass Quincy, MA) flashing at a
rate of 10 Hz.
Figure 3-1 demonstrates the concept used for al1 functional scout experiments.
The k-space data for al1 slices was collected in acquisition memory during the control
state (receiver phase = 0") after m inter-volume delsry. Fo!!owi% th.ir mpki?ic?n, the
visual stimulus was initiated at the beginning of the next inter-volume delay at which
time the receiver phase was set to 180°, while maintaining constant phase of the
transmitted RF (O0). The second k-space data set fiom the slices was summed (but ~5th a
negative sign due to the receiver phase) in acquisition memory after the delay, and the
stimulus was tumed off irnmediately after the scan. The inter-volume delay must be
sufficiently long to allow the hemodynarnic responsr to the presented stimulus to
stabilize, typically 5-8 seconds (1). The experiment was repeated for subsequent
stimulation-control cycles with continual summing of the data in acquisition memory.
The resulting intensity map generated by the 2-D Fourier transformation of the raw
difference k-space data demonstrates cortical and vascular regions responding to the
presented stimulus. Phase-cycling the receiver instead of the bansrnitter eliminated any
potential DC offset inherent in the receiver system. We also used a single steady state
dummy scan to condition the subject to the noise and hence achieve physiological
equilibrium. On our 4 T scanner, the stimulus is gated on and off by the pulse-program.
for optimum control and accuracy.
3.2.1 Single Slice FLASH Mapping
Activation maps were acquired using our functional scout technique with a
FLASH gradient-recalled echo sequence. Primary visual areas were located within high
resolution (256 x 256, 20 cm FOV) anatomical TI-weighted images acquired using a
magnetization-prepared (MP) FLASH sequence (TI = 1.2 s, TE = 6 ms, TR = 12 ms, flip
angle - 3û0, iuid i O mm siice uiickness j (8 j. From Uiis saginai siice, an oblique axiai
plane lying in the calcarine sulcus was prescribed. The k-space data sets for the FLASH
images (128 x 128, 16 cm FOV, TE = 25 ms, TR = 50 ms, flip angle = 22' and 10 mm
slice thickness) were acquired as s h o w in Figure 1. Each k-space data set was collected
in 6.4 seconds. and the inter-volume delay was 5 seconds. This protocol was repeated 2.3
and 8 times, providing three fünctional scout images upon reconstruction. One scan was
executed before image acquisition to acclimatize the subject. An additional 16 T,*-
weighted images were collected with the same imaging parameters, but without receiver
phase cycling between stimulation and control States. A functional map was produced
using Stimulate (6) with a Snident's t-test. and was overlaid on a T I -weighted anatomical
image (256 x 256) of the same oblique axial siice.
3.2.2 Mult i-..lice EP I Mapping
Activation maps were produced, as described above, for 1 1 sagittal slices through
the visual cortex (128 x 128,20 cm FOV, effective TE = 10 ms, TR = 60 rns and 5 mm
slice thickness) using a multi-slice EPI acquisition. Stimulation and control periods were
initiated 15 seconds before imaging and continued during data acquisition using the
scheme described in Figure 3-1. A series of maps were constructed on-line by 2D-FT
using 1, 2, 4 and 8 stimulation-control cycles. TI-weighted MP-FLASH images (256 x
256) of the same 11 slices were also acquired to serve as anatomical reference images for
the functional scout maps.
3.3 Results and Discussion
Figure 3-2 shows functional scout images produced using a receiver phase-cycled
FLASH gradient-recalled echo imaging sequence. These maps were obtained from a
single trial on a single subject. The activation maps show al1 pixels that change in
intensity in response to the applied stimulus since they were effectively denved from a
subtraction of images made during the two states. Figure 3-2(a) is a functional rnap
acquired using two stimulation-control cycles. This corresponds to a total scanning time
of less than one minute. Even with these few stimulus and resting states, mapping using
our functional scout technique can show areas of brain activation with reasonable clarity.
Figures 3-Z(b) and 3-2(c) were produced using 4 and 8 stimulation-control cycles.
respectively. These images show that these maps improve in contrat with increasing
nurnber of cycles. Particularly prominent are the high intensity areas due to BOLD signal
changes in major blood vessels, which do not necessarily represent the exact location of
neural activity (8). However, this gross spatial localization is sufficient for the purposes
of experimental planning. Figure 3-2(d) shows an activation rnap for the same
anatomical slice produced using a Student's t-test. This map was produced off-line using
Stimulate m i n g on a Sun SPARCstation 4 using data acquired with 8 stimulation-
control cycles without receiver phase cycling. Colored pixels represent a percentage
change in signal in pixels that have been determined statisticdy (p < 0.01) to respond to
the visual stimulus.
Figure 3-2: Functional scout maps (128 x 128) of a 10-mm-thick oblique axial slice
produced using a receiver-phase-cycled FLASH sequence with (a) 2, (b) 4, and (c) 8
stimulation-contml cycles. Pixel intensity indicates an image intcnsity difference (BOLD
signai difference) between stimulation and control state. (d) FLASHderived activation
map produced using a Studeat's t-test showhg pixels exhibithg a positive response
during photic stimulation at ap value of 0.01. Colors within vascular regions correspond
to >6% signal changes, wMe signal changes in cortical gray matter are in the 1-5 %
range. The occipital pole is located near the bottom of each image.
The functional scouts and the traditionally calculated functional map show
remarkable correspondence of activated areas. This similarity is certainly adequate for
planning additional experiments on the suhject in the same session.
Figures 3-3(a)-3-3(d) show functional scout images for the 3 rniddle slices
derived from an EPI acquisition in the same manner as above. Thcse maps were produced
from sagittal slices through the occipital pole including the calcarine fissure. Figure 3-
3(e) shows TI Oweighted FLASH anatomical images corresponding to the activation maps
along the same row. Contrast-to-noise clearly increases with additional stimulus cycles.
These results suggest that acquinng activation maps in this manner is also a fast and
efficient method of localizing brain function in a 3-D volume and with excellent
functional integrity .
Although our technique required no special hardware. the quality of the functional
map was likely due in part to the stability of the scanner and the robust BOLD effect at 4
T. Using simple features of the console. we cm automatically update the reconstmcted
functional map after every stimulation-control cycle. The experiment cm be halted when
the map is of the desired quality. An added benetit is that wbject conperation and metic?n
c m be assessed instantaneously as opposed to waiting until the session is over to analyze
the data. This makes our functional scout imaging technique a simple, practical method
of localizing b r i n function.
Figure 3-3: Multi-slice EPI-derived fùnctional scout maps (1 28 x 128) of three 5-mm-
thick sagittal slices through the visual cortex produced using receiver phase cycling with
(a) 1, (b) 2, (c) 4, and (d) 8 stimulation-control cycles. The occipital pole is located to the
left of each image. Pixel intensity indicates an image intensity difference @OLD signal
difference) between stimulation and control state. The corresponding TI-weighted
anatocnical slice (256 x 256) is shown in (e).
3.4 Conclusions
F ~ ~ q c t i ~ p d manc nf cortic- &&y !vere p r ~ d < ' ~ - d ~ ~ m ~ & ~ ~ & y yn th- S C = ~ ~ & S ----r- -A
console with the subject in the magnet. Maps were acquired, upon magnitude
reconstruction of complex-difference k-space data, using conventional single-slice
FLASH and multi-slice EPI sequences with receiver phase cycling between stimulation
and control States and constant transmitted RF. The functional scout image displays al1
pixels showing a BOLD signal difference, and allows the prescription of slices and
voxels for additional experiments using the planning tools provided by the scanner
manufacturer. Mapping using our new method delineates activated areas that are
qualitatively very similar to maps produced using conventional off-line analysis
techniques. This makes the functional scout image a simple, yet extremely powerful tool
in the localization of brain hct ion.
3.5 References
Menon RS, Ogawa S, Hu X, Strupp JP, Anderson P, Ugurbil K. BOLD-based
functional MRI at 4 Tesla includes a capillary heci contnh~itinn: Echn-planar imaging
correlates with previous optical imaging using intrinsic signals. Magn. Reson. Med.
33,453-459 (1995).
Menon RS, Ogawa S, Strupp JP, Ugurbil K. Ocular dominance coIumns in human V1
demonstrated by functional magnetic resonance imaging. J. Neurophysiol. 77, 2780-
2787 (1997).
Frahm J, Kruger G, Merboldt K-D, Kleinschrnidt A. Dynamic uncoupling and
recoupling of perfusion and oxidative metabolisrn during focal brain activation in
man, Magn. Reson Med. 3 5. 143- 1 48 ( 1 995).
Pritchard J, Rothman D, Novotny E, Petroff O. Kuwabara T, Avison M, Howseman
A, Hanstock C, Shulman RG. Lactate rise detected by 1H NMR in human visual
cortex during physiologic stimulation. Proc. Natl. Acad. Sci. (USA) 88, 5829-583 1
( 1 9%).
Gati JS, Menon RS, Ugurbil K, Rutt BK, Experimental determination of the BOLD
field strength dependence in vessels and tissue. Magn. Reson Med. 38, 296-302
( 1 997).
6. Strupp IP, Stimulate: A GUI based fMRI analysis software package, Neurolmuge
3(3), S607, June 1996.
7. Cox RW, Jesmanowicz A, Hyde JS, Real-time functional magnetic resonance
imaging, Magn. Reson. Med. 33,230-236 ( 1995).
8 . Menon RS, Ogawa S, Tank DW, Ugurbil K, 4 Tesla Gradient recalled echo
characteristics of photic stimulation-induced signal changes in the human pnmary
visual cortex, Mugn. Reson. Med. 30,380-386 ( 1 993).
Chapter 4
Effect of Luminance and Contrast on BOLD fMRI Response in
Human Visual ~ r e a s ~
by Bradley G. Goodyear and Ravi S. Menon
4.1 Introduction
The flow of oxygenated blood increases to areas of the brain that respond to a task
or sensory input (1). In a functional magnetic resonance imaging (MRI) experiment,
areas of functional activity are located by examining intensity differences between blood-
oxy genation-sensitive (or T ~ * -weighted) magnetic resonance images collected during and
in the absence of a presented stimulus. This mechanism has been termed blood-
oxygenation-level-dependent, or BOLD, contrast (2), and has been widely used, for
example, in tMRI studies of photic stimulation of visual cortex (3,4). Studies have
investigated the effect of the stimulus presentation frequency and flicker frequency on
NRI response (5,6), but little attention has been given to the effect of luminance and
contrast of the presented stimulus. Experiments using BOLD fMRI have also s h o w that
visual stimuli such as contrast-reversing grating s or c hec kerboard patterns are better than
'A version of this chapter has been published. Goodyear BG, Menon RS. Effect of luminance contrast on BOLD fMRt response in human primary visual areas. J. Neurophysiol. 79,2204-2207 (1998). O 1998 American Physiological Society.
difisely illuminated stimuli at eliciting a BOLD fMRI response in primary visual areas,
but have not addressed the underlying mechanisms responsible for these findings.
In more sophisticated fMRi experiments, it is common practice to examine
activation differences between multiple tasks. A difference in stimulus luminance or
contrast in presentations of n visual stimulus may lead to areas of activation unrelated to
the tasks if the area of the brain under investigation plays a role in the coding of
luminance or contrast. In studies aimed at elucidating the cortical response to monocular
input (e.g., ocular dominance column studies), an imbalance in stimulus luminance or
contrast may lead to the erroneous labelling of some areas since it may have been the
change in luminance or contrast that modulated the activity. To avoid this problem, it is
important to know how and where contrast and luminance are coded in the visual cortex.
Miyaoka et al. have investigated the uptake of labeled deoxyglucose in VI of
albino rats in response to a diffisely-illuminated visual stimulus (7). In their study, there
was no appreciable change in deoxyglucose uptake with increasing luminance. Albrecht
and Hamilton have made measurements of the electrical activity of V 1 neurons within cat
and monkey cortex in response to changes in local contrast of a visual stimulus (8). Their
results showed that there was an increase in neuronal finng rate (in spikedsec) with
increasing stimulus contrast, which was dependent on the spatial fiequency of that
contrast. In addition, this response saturated when the contrast approached two to three
orders of magnitude. To date, studies of contrast sensitivity in humans have been limited
to optical measurements of retina adaptation to stimulus contrast. In these studies, human
contrast sensitivity has been shown to increase with mean field luminance, but saturates
at low spatial frequencies (9). De Lange measured temporal contrast sensitivity using
contrast-reversing gratines. and found that at high mean field luminance the contrast
sensitivity maximized near 8- 10 HZ (10). This has also been shown at the cortical ievel in
V1 snidies using PET (1 1 ) and fMFü (5). PET studies have demonstrated that there are no
luminance effects on the regional cerebral blood flow for a visual stimuli subtending a
large field of view (12). Their results, however, were restricted by the resolvable image
pixel size within their matornical region of interest, and confounded by the difference in
spatial frequency of their two stimuli. Hence, there has been no direct measure of local
neuronal activity as si function of luminance or local stimulus contrast in humans. Given
the robust BOLD effect at a magnetic field strength of 4 Tesla, the current study
investigates the effect of stimulus luminance and contrast on fMRI response in primary
visual areas.
4.2 Methods
Six subjects, with no known visual deficits, participated in this study. Stimulus-
invoked signal changes were produced within pnmary visual cortex of volunteers using a
8 mm diameter red LED flickering at 8 Hz. The luminance of the LED was controlled
using a GRASS Instments (Quincy, MA) visual stimulator control box equipped with a
potentiometer. The LED was located approximately 26 cm fiom the subject's eyes,
subtending qproximately 2 degrees of visual field. The luminance of the flashing LED
was calibrated [in candelas per square meter (cd/m2)] at a distance of 26 cm using a
Minolta CS-100 Chroma Meter (Minolta Camera Co., Ltd., Japan). The background
Iuminance war !w!d a! î constmt vdile q~l! !n thP z.d$efit dmkness ir?side the bore cf
the MR scanner with the room lights extinguished. Four trials were performed, each
using a diffèrent, randomly selected, LED luminance (0.5,2.2, 85, 250 cdfrn'). Subjects
were asked to fixate on the fïickering LED stimulus throughout the experiment. This
provided a luminance-dependent stimulus at the fovea, and a contrast-dependent stimulus
with a change in mean field luminance near the edge of the LED (at -2-4" of
eccentricity ).
Al1 imaging experiments were perfonned on a Varian Unity INOVA 4 Tesla
whole-body imaging system (Varian, Pa10 Alto, CA; Siemens, Erlangen, Germany)
equipped with 25 mT/m actively shielded whole body gradients. A distributed-
capacitance, circula radio frequency (RF) surface coi1 with a 13 cm diameter was placed
under the occipital pole of the subject's head to transmit and receive the RF signal. An
oblique axial plane through the calcarine sulcus was prescribed within a high resolution
(256 x 256) FLASH gradient-recalled echo anatomical (Tl-weighted) image (13).
Experiments were performed using a single-slice e c h ~ planar Imaging (EH) sequence
(1 28 x 128 resolution, 20 cm field of view, TE = 10 ms, TR = 125 ms, and 10 mm slice
thickness). Thirty images were collected during both the photic stimulation and dark
control periods. This was repeated four times within each trial, giving a total of 240
images per trial.
Two anatomical regions of interest (ROIs) were selected for anaiysis. One ROI
encompassed primary visual area V l and the other included only extrastriate areas. For
eech fnd , 2 pixe! crgss-ccne!rticn (r = 9.40) :vrs perkmed v,.tv;.th LI inxpected pixel
tirnecourse using Stimulate (14) running on a Sun SPARCstation 4. The resulting map
displayed pixels showing a BOLD response to the presented stimulus. From this map, we
detcrmined the number and the mean percentage change in the intensity of these activated
pixels within each ROI. Activated pixels common to al1 trials were also selected to
rnonitor their mean percentage change with increasing LED luminance.
4.3 Results
Figure 4-1 shows activation maps overlaid on the corresponding anatomical
image for a LED luminance of (a) 0.5, (b) 2.2, (c) 85, and (d) 250 cdlm2. These results
are for a single subject. Activated areas (Le., pixels passing the correlation threshold) lie
within visual cortex. For sample time courses, see Appendix A. Figure 4-2 shows the
nurnber of activated pixels within the selected ROIs as a function of LED luminance. The
number of activated pixels (Le., the spatial extent of activation) wilhin the ioosely defined
V1 increases with increasing LED luminance. However, this trend is not seen within
extrastnate regions.
Figure 4-3(a) shows the mean percentage change in activated pixel intensity as a
function of LED luminance for the activation maps in Figure 4-1. The averaged data for
al1 subjects is shown in Figure 4-3(b). For the pixels within each of the defmed ROIS,
there was no sipnificant change in the mean fMRI response of activated pixels with
increasing LED luminance. However, for those pixels common to al14 maps. the tMRl
response incnased as the LED luminance was increased. Moreover, there was an increase
in the fMRI response of pixels common to any one LED luminance and dl higher
luminaace levels (not shown). There was no such trend in exîmstriate rrgions.
Figure 4-1: BOLD fMRi activation maps (overlaid on the comsponding matornical
image) of visual cortex showing areas exhibiting a positive response (correlation
tineshold r = 0.40) to a fickering red LED stimulus with a luminance of (a) 0.5, (b) 2.2,
(c) 85, and (d) 250 c d h 2 . The background luminance was maintained at the arnbient
room darkness. The occipital pole is located at the bottom of each image. One ROI
encompassing VI and one ROI within extrastriare are both outiineà in white.
a .c-;;e-~ . . . .- . , . 1 IO i oa Yi
normalked LED luminance
Figure 4-2: The number of pixels within V1 and extrastriate showing a positive response
to the LED stimulus for one subject whose activation maps are shown in Figure 4-1. The
LED luminance has been nomalized to the lowest level.
Figure 4 4 is a map of image pixels that exhibited an increasing trend in their
tMRi response as the LED luminance was increased, for the subject s h o w in Figure 4-1.
Al1 of these pixels lie within the defined ROI encornpassing VI. while no pixels are
present in higher visual areas outside the ROI.
4.4 Discussion
The maps in Figure 4-1 demonstrate that the flickering LED stimulus produces a
substantial amount of activity in the visual cortex. Although the stimulus subtended only
2" of visual field, the spatial extent of activity in the cortex, especially in VI, shows
remarkable spread. This was probably due to the cortical magnification factor, and couid
65
also be anributed to subject eye movement, which was not monitored for fixation,
causing changes in the position of the LED in retinotopic space. As mentioned above, one
ROI war selecrec! 10 mcnmpirr V!. 'fier virw! ZEE i.n dditinn !o VI (cg., W md
V3) may be included within this ROI. The borders of each of these areas cm be identified
using retinotopic phase mapping techniques (15), however these methods were not
awilablc at the time of this study.
1 14 1QO 1400 norrnalked LED luminance
1 1 Q 1 QU 1 flQQ normalhed LED luminance
Figure 4-3: Mean fMRi response (expressed as a percentage change in activated pixel
intensity) within the ROI encompassing VI and the ROI located within extrastriate for
(a) a single subject and (b) six subjects. The LED luminance has been normalized to the
lowest luminance level, and the percent change in (b) has been shown relative to the
percent change in the lowest LED luminance trial. Filled symbols represent al1 the pixels
activated in the defined ROIS for each trial. Open symbols represent only pixels activated
at ail LED luminance levels.
The results of F igw 4-3 that demonstrate no increase in the overall average fMRl
response with increasing luminance are in agreement with PET and VEP studies that
demonstrate no luminance dependence in the measured activity in the visual cortex
(1 1,12).
Figure 4-4: BOLD fMRI map (for the subject in Figure 4-1) showing pixels that are
common to al1 activation maps. These pixels exhibit an increase in their fMRI response
with each increase in LED luminance luminance. The outiined ROI (in white) is identical
to that in Figure 4-1.
Figure 4-3 also shows that there are individuai pixels that show an increase in
their fMRI response with increasing LED luminance, which is not supported by the
literature. Upon inspection of Figure 4-4, these pixels seem to be in areas more anterior
and ventral to where one would expect to get a response ta a stimulus subtending only 2"
of visual field based on the known retinotopic organization of the visual cortex. It is
possible, then, that these pixels may be coding the increasing contrast at the edge of the
LED when the LED is increased in luminance, and that those pixels whose response did
not increase were insensitive to luminance modulation within the 2" visual field or
exbibiteci 100% contrast gain.
Coding for local contrast has been detemined to take place in the visual pathway
as early as in the retina (16). The ambient light intensity during the day can Vary up to six
orders of magnitude. However, the range of contrasts in a typical visual scene spsn only
about two orders of magnitude. By coding contrast, neurons projecting from the retina
cm convey essential information about the retinal image despite enormous variation in
the ahsolute b e l of light (16). Assuming that the ROLD f M R l respnnse i s a direct
correlate of neural activity, our results for VI (and possibly V2 or V3) may suggest that
either more neurons are being recmited within the same imaged voxel as contrast is
increased. or that neurons activated in VI during low contrast levels are more highly
activated during high contrast levels. The latter interpretation would agree with previous
studies of cat (1 7) and macaque (1 8) primary visual cortex. The results of Tolhurst also
show that the contrast sensitivity function for the cat can predict simple ce11 responses in
visual cortex (17). This has not been discussed extensively in the monkey literature.
However, the lineshape of Figure 4-3 showing the mean percentage change of activated
pixels in V1 common to al1 trials supports human contrast sensitivity experiments for low
spatial frequency contrast-reversing gratings (9).
Shapley has shown using single ce11 recordings that cells within the magnocellular
layers of the macaque LGN increase their activity with contrast more readily than do
parvocellular cells (19). As MRI field strengths increase, tMRI studies of LGN and other
midbrain structures known to be sensitive to contrast or luminance [e.g.. the superior
colliculus (7)] will become more feasible. A visual paradigm designed to probe the
response of primary visual areas and these midbrain structures must therefore take
luminance and contrast into consideration.
68
4.5 References
Roy CS, Sherrington CS. On the regulation of the blood-supply of the brain. J.
Physiol. 1 1.85- 1 08 [ 1 890).
Ogawa S, Lee T-M. Kay AR, Tank DW Brain magnetic resonance imaging with
contrast dependent on blood oxygenation. Proc. Natl. Acad. Sci USA 87, 9868-9872
(1 990).
Tootell RBH, Reppas JB, Kwong KK, Malach R, Born RT, Brady TJ, Rosen BR,
Belliveau JW. Functional analysis of human MT and related visual cortical areas
using magnetic resonance irnaging. J. Neurosci. 15,32 15-3230 ( 1995).
Tootell RBH. Dale AM, Sereno MI, Malach R. New images from hurnan visual
cortex. TINS 19,48 1-489 (1 996).
Kwong KK. Belliveau JW, Chesler DA. Goldberg IE, Weisskoff RM. Poncelet BP.
Kennedy DN. Hoppel BE, Cohen MS. Turner R. Cheng H-M, Brady TJ. Rosen BR.
Dynarnic magnetic resonance imaging of human brain activity during primary
sensory stimulation. Proc. Nat/. Acad. Sci USA 89,5675-5679 (1992).
Thomas CG. Gati JS. Menon RS. Amplitude response and stimulus presentation
frequency response of hurnan primary visual cortex using BOLD EPI at 4 T. Magn.
Reson. Med. 40.203-209 (1998).
Miyaoka M, Shinohara M, Batipps M, Pettigrew KD, Kennedy C, Sokoloff L. The
relationship between the intensity of the stimulus and the metabolic response in the
visual sy stem of the rat. Acti. Neurol. Scand Suppl. S72, 16- 1 7 (1 979).
Albrecht DG, and Hamilton DB. Striate cortex of monkey and cat: contrast response
function. J Physiol. 48,217-237 (1982).
van Nes FL, Bouman MA. Spatial modulation transfer in the human eye. J. Opt. Soc.
Am. 57,40 1-406 (1967).
10. De Lange D m H. Research into the dynamic nature of the human fovea: cortex
systems with intermittent and modulated light. 1. Attenuation characteristics with
white and colored light. J. Opr. Suc. Am. 48,777-784 (1959).
1 1. Fox PT, Raichle ME. Stimulus rate dependence of regional cerebral blood flow in
human striate cortex, demonstrated by positron emission tomography. 1
Nemphysiol. 51, 1 109-1 120 (1984).
12. Fox PT, and Raichle ME. Stimulus rate determines regional blood flow in striate
cortex. Ann. Neurol. 17,303-305 (1 985).
13 .Menon RS, Ogawa S, Tank DW, Ugurbil K. 4 Tesla gradient recalled echo
characteristics of photic stimulation-induced signal changes in the human primary
visual cortex. Magn. Reson. Med. 30,380-386 (1993).
1 4. Strupp J.P. Stimulate: A GUI based tMRI analysis software package. Neuro Image
3(3), S607 (1 996).
15. Engel SA, Glover GH, Wandell BA. Retinotopic organization in human visual cortex
and the spatial precision of functional MN. Cereb. Cortex 7, 1 8 1 - 192 ( 1997).
1 6. Wandell BA. Foundations of Vision. (Sunderland, MA: Sinauer, 1 984), pp.476.
17. Tolhurst DJ, Dean AF. Spatial summation by simple cells in the striate cortex of the
cat. Exp. Brain Res. 66,607-620 ( 1 987).
18. Devalois EU, Albrecht DG, Thorell LG. Spatial frequency selectivity of cells in the
macaque visual cortex. J. Opt. Soc. ..lm. 67.779-784 (1982).
1 9. Shapley RM. Visual sensitivity and parallel retinocortical channels. Annu. Rev. Psy.
41,635-658 (1990).
Submillimeter Functional Localbation in Human Striate
by Bradley G. Goodyear and Ravi S. Menon
5.1 Introduction
Optical imaging using intrinsic signals (OIS) has dernonstrated a biphasic
response in the local hemodynamic response to neural activity. Using wavelength
resolved OIS. Malonek and Grinvald have demonstrated that subsequent to the onset of
neural activity, there is a transient increase in the tissue concentration of
deoxyhemoglobin [Hb], caused by an increase in local oxygen consumption with no
cornmensurate change in blood fiow or volume (1 ). Several seconds later, there is a large
increase ifi the regienal Slood volume (rCBV) md cerebrd bblood flow (CBF). whkh
vastly overcornpensates for the local oxygen consumption increase. This results in a
hyperoxygenation of the capillaries and venous vasculature due to an increase in tissue
oxyhemoglobin concentration [Hb02], relative to the resting condition. This temporal
:A version of this chapter has been published. Menon RS. Goodyear BG. Submillimeter functional Iocalization in human striate cortex using BOLD contrast at 4 Tesla: implications for the vascular point- spread function. Magn. Reson Med 41,230-235 (1999). O 1999 John Wiley & Sons, Inc.
signature is also thought to correspond to that seen in fMRI studies of the visual system at
very high rnagnetic fields (2,3). It has been argued to varying degrees that the early phase
of the response [the 'initial di$ in MRI parlance (1-?)! ir Mer bcdized te ti?e source of
neuronal activity than the hyperoxic phase because while the metabolic changes giving
rise to the increased oxygen consumption are well CO-localized with the neural activity,
the resulting vascular flow response is not as tightly coupled to neural activity. Since the
initial dip is very small and extremely difficult to map even at 4 Tesla ( 4 0 % of the
positive going Blood Oxygen Level Dependence (BOLD) signal), for practical
experiments it is the latter phase that is used to make maps of human brain function in
tMRi studies. On this basis, it has been argued that MRi cannot produce submillimeter
hnctional resolution in humans because the later hyperoxic vascular response spreads out
over many millimeters ( 1 ).
However, Das and Gilbert (4) have made careful measurements of the spread of
neural activity spatially beyond a point source of activity [the cortical point-spread
function (PSF using extracellular electrodes and the cortical PSF of the vasculature using
OIS at 6 10 nrn (which measures a combination of [Hb], [Hb02] and ce11 swelling). Using
visual stimulation in a feline mode1 in which a 1 mm x 1 mm patch of cortex
demonstrated spiking neurons, the mean diameter of the vascular response was found to
be 3.8 mm in extent (i.e. the cortical vascular PSF). They relate this to the fact that the
metabolically demanding subthreshold depolarization of neurons spreads fa. beyond the
spiking area due to extensive horizontal arborization of cortical neurons in the upper
layers of the cortex. Furthemore, it is believed that OIS is sensitive to increased
metabolic activity due to the transmembrane potentials that give rise to subthreshold
synaptic potentials (5). In this interpretation, the vascular response is coupled to neural
metaholic nctivity, and it is the cortical PSF nf the r'ihthshnlrl depc?!airation and hmre
metabolism that is mirrored by the vascular PSF.
Examination of Figure 3 of Malonek and Grinvaldis paper suggest that there may
be ment to both explanations. Using stimuli of 2 or 4 seconds in duration, it appears that
the early part of the hyperoxic response can yield almost as well localized functional
maps as those made during the initial deoxygenation phase. At later times ( N O seconds
since post stimulus onset) it certainly appears that the vascular response becomes
nonselective. Therefore, we have explored the capability of BOLD for measuring a
cortical point-spread function (PSF) using shon stimuli and the early part of the
hyperoxygenation phase of the BOLD response. The cortical PSF can be defined by the
area of cortex 'activated' by a small visual stimulus (4). The meaning of the term
'activated' depends on the technique being used to measure the cortical response.
Extracellular recordings measure spiking activity, while OIS and fMRl indirectiy
measure metabolic activity that can be associated with inhibitory or excitatory post-
synaptic potentials which are expected to extend some distance from the spiking neurons
due to horizontal connections (5). In order to detennine the PSF of the cortical vascular
response, a point source of spiking activity in the cortex is needed. Traditionally, one
might use a point source of light IO accomplish this. One could then examine how far the
activity-linked blood flow changes rxtended, but with WRI, this depends on the
somewhat arbitrary thres holds used in analy sis.
Another approach is to attempt to resolve point source pairs of known separation
as is done with a typical quality assurance resolution phantom. In humans, nature has
prnGded ri convenient resolotinn grid in the fonn of the ocular dominme column (CiDC)
corresponding to each eye in primary visual cortex, or visual area VI. Strictly speaking,
this binary nature of the innervation is found only in layer IV of V 1. In the layers above
and below that, horizontal connections blur the distinction somewhat; however, eye
dominance rernains consistent throughout the depth of the cortex (layers 141). These
columns have the advantage that they can be selectively and noninvasively activated
using photic stimulation of one eye (independent of stimulus size in degrees). Columns
corresponding to one eye altemate with those from the other as one follows the cortical
nbbon as shown SC hematically in Figure 5- 1.
In human striate cortex, the ocular dominance columns are sornewhat less than I
mm x 1 mm in cross-section and traverse parallel to the cortical surface (6); they can be
visualized as interdigitated patches along perpendicular sections of VI (< 1 mm center-
to-center). Functional MM of ODCs is a challenge because it requires submillimeter
spatial resolution while preserving sensitivity to the microvascular BOLD signal, in
addition to suppressing minute head motions. The factors goveming resolution at this
level are contrat-to-noise ratio ( C M ) and the spatial specificity of the physiological and
vascular response. In fMRI expenments using BOLD contrast, CNR can be enhanced
through the use of surface coils (7), physiological noise suppression schemes (8), pulse-
sequence design (9, 1 O), parameter optimization (1 l), and magnetic field strength (1 2).
However, as described above, physiology sets the ultimate limits on how tightly the
vasculature responds to neural activity in an ODC, and it is this response we wish to
characterize.
Figure 5-1: Ti-weighted anatomical image of a typical oblique slice in the human visual
cortex. The inset on the top left shows an expanded view of the cortical ribbon of gray
matter, in which the cortical surface abuts the cerebral spinal fluid (dark) and the white
matter lies below the gray matter. Any given segment of the cortical ribbon has 6
cytoarchitectonically defined layers as s h o w schematicaily in the inset on the top right.
This illustrates that the left and right eye inputs are segregated only in layer 4C, and that
horizontal connections between the ocular dominance columns occur in the superficial
layen of cortex. Interactions between these lateral neurons blur the sharp edges of the
columns seen in layer 4C, but preserve the general concept of ocuiar dominance in al1
Iayers.
Previous studies of ocular dominance using MRI (13) have suffered from
insufficient spatial resolution fiom two perspectives. First, the image resolution was
somewhat coarse relative to the actual nixe of the cnl!imni, giving r k e !O partial volume
effects. Also, by saturating the vascular response with a prolonged visual stimulus, the
nonspecificity of the tMRi response in the draining venules and veins yielded an
effectively lower spatial resolution as has been pointed out in optical irnaging and fi1R.I
studies (1. 14). In this work, we have exarnined the fiactional signal changes that occur in
an ODC when the corresponding eye is stimulated using a short duration contrast-
reversing checkerboard venus when both eyes or the opposite eye is stimulated, and have
shown that with the appropriate paradigm, these point sources that are approximately 700
p m apart can be resolved using fMRI.
5.2 Methods
Six healthy subjects (3 male, 3 female, 26 * 3 years (mean I SD)) with no known
visual deficits participated in the study. Written informed consent was obtained as per
NIH and local institutional guidelines. The study was composed of nine 30 second epochs
(i.e., 9 single triais), each consisting of 4 seconds of visual stimulation using a black and
white polar checkerboard 15' in diameter, reversing contrast at 8 Hz. During each trial, a
monocular [lefi eye (L) or right eye (R)] or binocdar (B) view of the checkerboard was
presented to the subject through a pair of liquid crystal shutter glasses (Translucent
Technologies Inc., Toronto, Ontario). Between presentations (26 sec), both shutters were
closed and translucent. The checkerboard remained on at a11 times tn ensure that the
subject did not become dark-adapted. The nine trials were presented in the following
order: B-L-R-B-L-R-B-L-R. During the experiment, each subject was instructed to keep
both eyes open and to fixate on the cross-hairs at the center of the checkerboard
whenever visible.
Functional imaging experiments were performed using a Varian Unity I N O VA 4
Tesla whole-body system (Varian NMR Instruments, Pa10 Alto, CA; Siemens, Erlangen.
Germany) equipped with 25 mTlm actively-shielded whole-body gradients. The subject
lay on the patient bed with the back of the head resting on a well-padded support. The
underside of this support housed a distributed capacitance. 8-cm diameter. quadrature
radio frequency (RF) surface coil to transmit and receive the MR signal. The subject's
head was imrnobilized using a well-padded head vise secured to the sarne platform that
held the coil and head support. This vise also housed a mirror above the subject's eyes
that was tilted at an angle to allow the subject to view the stimulus on a projector screen
placed near the subject's waist.
At the beginning of the imaging experiment, Ti-weighted sagittal MR images
were collected to prescribe oblique axial imaging planes parallel to the calcarine fissure.
During the experiment, Tr*-weighted MR images were collected using a EPI gradient-
recalled echo pulse sequence (256 x 256 rnatrix, 14 cm field of view, TE = 20 ms, TR =
400 ms' flip angle = 35": 4 mm slice thicknew) with hnth the segments (eight segments)
and slices (three slices) interleaved, as well as with centric ordering of k-space and a
navigator echo for every segment. The in-plane resolution was 547 x 547 p.
.4ccounting for Tl* bluning, the effective image resolution, as determined From the Ml-
width at half-maximum of the voxel point spread h c t i o n as discussed in Section 1.2.3.
was approximately 630 Pm. One segment of raw k-space data was collected for each
image plane before collection of the next segment of k-space, as s h o w in Figure 5-2.
1- Image 2-1 kepace segments [-Image 1 7 / I i \
8~~ce1'I I II I I I I'I I I I II I I
I I I I I I I I I m a o Figure 5-2: Schematic of the 'sliding window' approach to reconstnicting the multi-
segment multislice EPI data. The slices are acquired in segmented interleaved mode,
allowing considerable relaxation (in this case, 400 ms) between subsequent excitations,
and consequently a higher tip angle.
To decrease the time between points in the image time series, each image was
reconstructed using the 'sliding window' technique s h o w in Figure 5-2. Using this
o-rktbr? of MI? fluoroscopy ( ! 5 ) , LI image rvas reconstnicted frcri. f! cmtigucw
segments (one '%dowW) for that image plane. To reconstruct the next image in the time
series, the image '~ indow' ' was shified in time by one segment for that image plane.
Using this technique, the peak of the BOLD response is more discemable.
To locate V 1, an additional experiment was perfomed within the same imaging
session using checkerboards of different contrasts (IO%, 40%, 80%). We have previously
found that only V1 responds to contrast changes, making this experiment a robust method
to dernarcate the boundaries of V 1 (1 6) when retinotopic phase mapping techniques are
unavailable.
One experiment was perfonned in the absence of any visual stimuli to monitor
fluctuations in the baseline signal within V1 over the time course of the experiment
compared to the background thermal noise. As well, this data set was analyzed. as
described in the next section, to investigate the resulting map in the absence of any
stimulus.
Although the 'sliding window' technique decreases the tirne between points in the
fMRi time series of images, it is at the expense of temporally smearing the data. For this
reason, an average time series for selected regions within VI was correlated with itself to
determine the separation of independent images, which was taken as the full-width at
half-maximum (FWHM) nf the aritncmelation funrtion. The total number of
independent measurements for the experiment was then calculated by dividing the total
number of images after the sliding window interpolation by the number of images
contnined within the width of the autocorrelation function. For example, if the total
number of images in the time series was 400 and the width of the autocorrelation function
was 10 images. then the number of independent measures in the time series would be 40.
In this case, when analyzing the data with a t-test, a test period would have to contain at
least 1 1 images to contain 2 independent measurements (Le., 10 + 1 ) and 2 1 images to
contain 3 independent measurements (i.e., 10 + 10 + 1).
The collected data were analyzed using Stimulate (17) running on a Sun
Ultrasparc 5 (Sun Microsystems, Mountainview, CA) after applying a low frequency
filter to remove any baseline drift. Two activation maps for each subject were calculated
using a Student's t-test to determine image voxels that were significantly activated
[relative to the baseline state @ < 0.01)] during the right-eye and the left-eye monocular
viewing conditions. The final map of ocular dominance consisted of voxels fiom these
two monocular maps that demonstrated significant differences @ < 0.05) between the
activation levels under the two monocular conditions, measured as peak amplitude of the
hyperoxygenation phase of the BOLD tirne senes was also made. That is, if the activation
level of a voxel was significantly higher during right-eye stimulation over the three trials
than during the three left-eye stimulation trials, it was considered as right eye dominant;
similarly to obtain left-eye dominant voxels. We determined the mean activation within
this map during the rnonocular stimulation of the corresponding eye, dunng the
monocular stimulation of the other eye. and during binocular stimulation. We also
investigated the mean cluster size of the activated voxels in the ocular dominance column
maps by recording the number of voxels within each individual cluster (along the cross-
section of a colurnn band). .4 weighted average of the number of voxels per cluster was
then calculated, and converted to micrometen using the intrinsic image resolution (Le., 1
voxel wide = 630 pm).
5.3 Results and Discussion
Figure 5-3 shows T2*-weighted images of three slices from the same subject
acquired using the functional imaging sequence as described above. These high-
resolution images show remarkable detail, with linle blurring, and exhibit little or no
ghosting artifacts. The signal-to-noise within V 1, defined as the mean image intensity
within an ROI in VI divided by the standard deviation of the background noise in an ROI
well outside the head, was in excess of 60: 1. Using a 4 Tesla MR scanner, a head-
restraining vise, a navigator echo correction for every image segment, and a sliding
window data series interpolation for every image segment provides a time series of
images that demonstrate less than a 10% increase in the standard deviation of the image
intensity within V1 during baseline activity when compared to background noise, as
show in Figure 5-4.
Figure 5-3: Tz*-weighted images of three contiguous oblique/axial slices of the visual
cortex of one subject prescribed parallel to the calcarine sulcus. The prescribed image
resolution is 547 pm in-plane.
Figure 5-Sa is an image from one subject showing the location of voxels within
V1 that exhibited a significant increase in MR signal above baseline in response to the
three nght-eye monocular stimulation periods. An image of the same slice is s h o w in
Figure 5-5b, showing voxels exhibiting a significant response to the three left-eye
monocular stimulation periods. Both maps contain a significant number of common
voxels. The average time course for the activated voxels in Figure 5-5a is shown in
Figure Mc, and the average time course for the activated voxels in Figure 5-jb is shown
in Figure M d . Fm= these maps, the average time courses demonstrate no significînt
difference in the magnitude of the £'MN response during monocular left-eye or
monocular right-eye stimulation.
-4 -2 O 2 4 zeaero mean voxel Intenslty (%)
Figure 5-4: The distribution of the magnitude of image intensity over the time course of
a baseline experiment within regions of interest within VI and within the background
noise outside the head (nomalized to zero mean).
The map of Figure 5-6a shows the difference between the maps of Figure 5-Sa
and Figuer 5-Sb. If a voxel was more highly activated on average over the three nght-eye
monocular stimulation periods than during the three left-eye monocular stimulation
periods, it was color-coded red: otherwise, if a voxel was more highly activated by le%
eye stimulation, it was color-coded blue.
O 100 200 300 O lm m 300
-w image number
Figure 5-5: a. Activation map of voxels, overlaid on the comsponding anatomical slice,
dernonstrating a signifiant fMRJ response @ < 0.01) to the three monocular right-eye
stimulation periods. b. Activation map of voxels demonstrating a significant hlRI
response @ < 0.01) to the three monocular lefi-eye stimulation periods. c. Average time
course showing the fMRI nsponse (in percent relative to baseline) for the voxels shown
in (a). d. Average t h e course showing the fMRI response for the voxels show in (b).
'B' indicates the trials corresponding to binocular stimulation, 'L' indicates the trials
corresponding to left-eye monocular stimulation, and 'R' indicates the trials
comsponding to right-eye monocular stimulation.
rlght eye - lefi e e responre (normal 1: ed)
Figure 5-6: a. Activation map of the voxels fiom the maps of Figure 5-5 that show
higher levels of activation during right-eye stimulation (red) and lef't-eye stimulation
(blue). b. Activation map of the same sîice showhg only those voxels of the activation
maps in F i p 5-5 that show significantly higher activation dining right-eye stimulation
(red) and significantiy higher Ieft-eye stimulation (blue). c. The fMRI response of each
voxel in (a) and (b) during right-eye stimulation minus the fMRI response within the
same voxel during left-eye stimulation, divided into 5% bins and norrnalized to 1. The y-
axis indicates the percentage of pixels (or voxels) within each 5% bin.
As Figure 56a demoaseates, red and blue altemate as mal1 voxel clusters.
However, large solid areas are evident as well, possibly due to large vessels. The
distribution of the difference between the average fMRI response during right-eye
stimulation and the average fMRI response during lefi-eye stimulation is show in Figure
5-6c (i.e.' R minus LI. The majority of this distribution shows little or nn ciifference
between right-eye and left-eye stimulation (i.e., R - L = 0).
Of course, the value of each voxel in the map shown in Figure 5-6a is an average
of three responses. For each pixel in this map, the three response levels (in percent above
baseiine) during right-eye stimulation were compared, using a Student's t-test, to the
three response levels within the same pixel during left-eye stimulation. Figure 5-6b
shows the voxels of Figure 5-6a that show a significant difference @ < 0.05) between
right-eye and left-eye stimulation. As Figure 5-6b demonstrates, the large blob-like areas
have now disappeared, since these areas exhibit no difference in their activation levels
during right-eye and left-eye stimulation. The remaining voxels in Figure 5-6b altemate
in color as tiny clusten, consistent with the organization of ocular dominance columns
within a perpendicular section of V 1. The distribution of the difference in the activation
levels during right-eye and left-eye stimulation is also shown in Figure 5-6c. The
population of voxels showing no significant difference in right-eye and left-eye
stimulation were removed by performing the t-test, resulting in two distributions - one
showing significant right-eye ocular dominance (R - L > 0) and one showing significant
lefi-eye ocular dominance (R - L < O). This distribution of the ocular dominance of the
fMRI response and the organization of the pixels in Figure 5-6b lead us to believe that
these voxels reflect activity within the ocular dominance columns of the p n m q visual
cortex.
Figure 5-7 shows ODC maps for each of four subjects superimposed on the
corresponding anatomic slices. A consistent pattern of generally interdigitated activation
corresponding to each eye i s nated. This pattern i s s hrllrnnrk of ocular dominance
columns viewed perpendicular to their long axis (6, 18) (see Figure 5-1). We also note
that this pattem is consistent across al1 six subjects and three slices in Our study,
demonstrating that the methodology is quite robust. One might expect some partial
volume effect for columns whose axes might be at slight angles relative to the slice
normal, resulting in an overestimate of the column sizes. However, we found no
significant deviation tiom published data from post-mortem studies (6) although these
were done with cytochrome oxidase staining and may not reflect the sarne properties as
our metabolic rneasurement. Furthemore. even though the calcarine fissure usually has a
slightly different angulation between hemispheres, our results on column size. fractional
signal change and histograms of ocular dominance are independent across the
hemispheres, suggesting no systematic biases in the data. Our data averaged across al1
subjects shows that the mean cluster size (see Section 5.2.3) was 1.29 voxels,
corresponing to an ODC width of 0.71 on a side, consistent with post-rnortem studies (6,
1 9).
Figure 5-8 shows a map resulting from the analysis of the baseline data set. Only
a few spurious pixels are present after the statistical analyses, and their arrangement is
not consistent with any forrn of ocular dominance column arrangement. Hence, our
analysis techniques alone do not bias the formation of any columnar-like maps of
activity .
Figure 5-7: Maps of ocular dominance for 4 different subjects. Red (blue) voxels
indicate areas that are more highly activated during monocular stimulation of the right
(lefi) eye. Activated m a s lie within gray matter in V 1, and alternate in color on a scale
consistent with the size of ocular dominance columns in humaas (0.5 - 1 .O mm).
In a very eiegant series of experiments characterizing the modulation m f e r
function in the visual system, Engel et al. showed that the MRI signai could be localllcd
to within 1.1 mm at 1.5 T (19). As they also point out, this Limit depends on the signai-to-
noise rario of the measurement and that there is no theoretical limit why it c a m t be
exicnded, as we have. Ultirnately, they suspect, and we agree, that the lateral connections
in the horizontal layers of the cortex as well as the vascular demity (20), will set the
limiting resolution in hlRI, and these cm vary in different arcas of the braia.
Figure 5-8: Activation map of activity during the baseiine experiment with no visual
stimuli. The analysis of the data set was identical to that used for the maps of Figure 5-7.
Only a few spurious pixels are present, demonstrating no statistical biases.
In Figure 5-9(a), a t h e course for a single subject is shown for right eye ODCs
and left eye ODCs during the entire paradigm. The nine single trial epochs are clearly
visible. We have averaged these curves for both eyes, for each stimulation condition and
for ai l subjects in Figure 5-9(b). In what we identiQ as ODCs king driven directly by an
open eye Oeft ODCs driven by left eye averaged with right ODCs dnven by right eye),
the fractional signal change has a mean of 1.74 %, while in the adjacent columns of the
nonstimulated eye (right ODCs whiie Ieft eye stimulated averaged with Icft ODCs while
right eye stimulated) the mean fiactional BOLD change is 1.01 %. In other words, the
part of the fMRI response that allows functional mapping of ODCs (termed the mapping
signal in OIS (1)) is 72 % higher than the response in the inactive columns. We r e m to
this point later. Thus at lem at short pst-stimulus times, even the hyperoxygenation
phase gives considerable conaast beween the active and inactive columns, contrary to
the suggestion fiom optical imaging (1). To furtiier investigate the fMRI changes, we
examined the nature of the underlying distribution of the fiactional signal changes rather
than just their means. This was done by creating a histogram of d l the fractional signal
changes occurring in the stimulated ODCs and comparing it against the histogram of the
in.rtix.p G D 0 torrespedifig te the tye. 7Xis is s b . m in Figxe 5-9(c). Exh set cf
columns has the same underlying shape (slightly kurtotic) and width, but the stimulated
column pixels consistently exhibit a higher fractional signal change than the
conesponding points on the distribution for the inactive çolurnns. Thus the observation
that the active columns of the stimulated eye have 1.74 times the signal change compared
to the inactive columns of the other eye is likely a consequence of a point-by-point shift
in the entire distribution and is not just a shifi in the mean of a few pixels or a change in
the histograrn shape.
Typically. we have found draining veins usually produce greater than 5% signal
changes (2. 12. 15) and can be removed by thresholding the activation maps. However.
no post-hoc thresholding was needed to eliminate large fractional changes in this case.
probably because draining venous vessels might not be expected to meet the two criteria
used to make the maps. In fact the histograms show that no pixels exceeding a 4% signal
change met our cntena of significant change fiom baseline AND significant difference
between monocular States. This Fractional value is consistent with the combined changes
obsewed in the OIS signals for [Hb] and [Hb02]. We were also able to reliably
discriminate changes of under 0.4% in our data, due to the head imrnobilization, the
navigator correction and the intrinsic stability of the scanner. With these short duration
stimuli, the fMRl response never reaches saturation, and therefore the peak value of the
hyperoxygenation phase is presumably proportional to the local metabolic activity.
Figure 5-9: (a) Time-course dunng the visual stimulation paradigm [B = binocular
stimulation, L = left eye stimulation, R = right eye stimulation] for a single subject for
pixels identified as right and left eye ODCs. (b) Average timecourses across al1 subjects
[(i) = ODCs of lefi eye under left eye stimulation and ODCs of right eye under right eye
stimulation, (ii) = ODCs of both eyes under binocular stimulation, (iii) ODCs of lefl eye
under right eye stimulation and ODCs of right eye under left eye stimulation]. The gray
bar denotes the time the stimulus was on for. (c) Distributions of the number of pixels as
a function of fractional signal change for the conditions (i) and (iii) above. The entire
distribution shifts down in mean activation level between a condition in which the correct
columns are being stimulated and the condition in which the other eye's columns are
being stimulated. The means of the distributions correspond to the peak of the averaged
timecourses (i) and (iii) in (b).
An interesting observation was that with these short duration stimuli, the post
stimulus undershoot appears absent as does any initial dip (Figure 5-9(b)). While the
fnnner ohservatinn is c~nsistent witir the OIS !I?er&m (1) md the be!!oon n-!cde! of
Buxton and colleagues in the short stimulus regime (21), the latter is dificult to explain,
particularly in light of the fact that we must be observing capillary bed changes to make
the lefi-ri@ rye ODC discrimination. We postdate that ihis is stimulus intensity reiated.
In previous observations of the initial dip at 4 Tesla with fMRI, very bright stimuli have
been used, combined with extended periods of total darkness. In our expenment, the
luminance was rnoderate so as to avoid squinting and the translucency of the liquid
crystal shutter glasses ensured that no dark adaptation took place between trials.
Another curious feature is that of the binocular stimulation condition (Figure 5-
9(b)). During binocular stimulation, in pixels identified as ODCs the fractionai signal
change is less than when the ODC is being stimulated by the appropriate eye. There are
two possible explmations for this. The first is that the vascular reserve is not capable of
providing the full rCBF change to both sets of columns. We find this possibility rernote,
in pari because it destroys any expectation of linearity of the BOLD signal with metabolic
activity (22). If one spatially blurs the data by a factor of twol it would be expected that
with twice the input into the voxel (binocular condition), the BOLD effect should be
twice as large. In fact Our data suggest it is lower than the monocular state. Thus, the
more likely option is that inhibitory activity between columns reduces either the spiking
activity within the columns or reduces subthreshold membrane activity in the horizontal
layers, leading to reduced metabolic activity. Resolution of these alternatives would
require OIS measurements to be done in conjunction with single unit recordings at
multiple sites.
5.4 Conclusions
Our data show that is possible to reliably and robustly separate two different
neural populations that are approximately 700 p n apart using fMRi in the visual cortex
when using the hyperoxygenation phase of the BOLD response. Key features in
performing such segregation are differential mapping techniques, short duration stimuli
and appropriate stimuli with carefully characterized luminance and contrast. The use of
short duration stimuli is particularly important because it prevents the BOLD response
from saturathg and allows us to use the peak activation as a proportional measure of the
amount of neural activity in the voxel (22). A long interotrial interval allowed the BOLD
response to return to baseline (2). The use of mechanical head immobilization and a very
high field scanner are also very important tools since the head must be stationary and the
expenment as short as possible.
By demonstrating the hinctional resolvability of fMRi at 700 Pm' this distance is
an upper limit to the cortical vasculature point-spread function in the visual cortex. Such
fine resolution may not be achievable in al1 parts of the cortex, and will depend on the
availability of clever paradigms that allow differential mapping. Ultirnately, it is the
extent of the lateral connections and their inherent activity that will detemine the cortical
vascuiature PSF in different areas of the brain.
93
5.5 References
Malonek D, Grinvald A. Interactions between electrical activity and cortical
micr~nircdqt&n reupded $I i?n.@fi~ ~ ~ ~ ~ ~ ~ c ~ w : ~ ~ ~ ~ ~ c ; & ~ s fc: fi^&^^^
mapping. Science 272,55 1-554 (1996).
Menon RS, Ogawa S, Hu X, Strupp JP, Andersen P, Ugurbil K. BOLD based
functionai MRI at 4 Tesla includes a capillary bed contribution: Echo-planar imaging
correlates with previous optical imaging using intrinsic signals. Mup. Reson Med.
33 ,45349 (1995).
Hu X, Le TH, Ugurbil K. Evaluation of the early response in fMRI in individual
subjects using shon stimulus duration. Magn. Reson Med. 37,877-884 (1 997).
Das A. Gilbert CD. Long-range horizontal connections and their role in cortical
reorganization revealed by optical recording of cat visual cortex. Science 375, 780-
784 ( 1 995).
Gnnvald A, Lieke EE, Frostig RD, Hildesheim R. Cortical point-spread fùnction and
long-range lateral interactions revealed by real-time optical imaging of macaque
rnonkey primary visual cortex. J. Neurosci. 1 4,2545-2568 ( 1 994).
Honon JC. Dagi LR, McCrane EP, de Monasteno FM. Arrangement of ocular
dominance columns in human visual cortex. Arch. Ophthalmol. 108, 1025- 1 O3 1
(1991).
Lin CS, Rajan SS, Gold J. A novel muiti-segment surface coi1 for neuro-functional
magnetic resonance imaging. MW. Reson. Med. 39, 164-168 (1 998)
Le TH, Hu X. Retrospective estimation and correction of physiological artifacts in
tMRI by direct extraction of physiological activity fiom MR data. MW. Reson Med.
35,290-298 (1 996).
Yang Y, Glover GH, van Gelderen P, Patel AC, Mattay VS, Frank JA, Duyn JH. A
cornparison of fast MR scan techniques for cerebral activation studies at 1.5 tesla.
Magn. Reson. Med. 39,61-67 (1998).
10.Menon RS, Thomas CG, Gati JS. Investigation of BOLD contrast in fMRi using
multi-shot EPI. NMR Biomed. 10, 179- 182 (1 997).
11. Kim S-G, Hendrich K, Hu X, Merkle H, Ugurbil K. Potential pitfalls of bct ional
M N using conventional gradient-recalled echo techniques. NMR Biomed. 7, 69-74
(1 994).
12. Gati JS, Menon RS, Ugurbil K, Rutt BK. Expenmental determination of the BOLD
field strength dependence in vessels and tissue. Magn. Rrson. Med. 58, 296-302
( 1 997).
13. Menon RS, Ogawa S. Stmpp JP, Ugurbil K. Ocular dominance columns in hurnan V1
as demonstrated by functional magnetic resonance imaging. J. Neurophysiol. 7 . 2780-
2787 (1997).
14.Menon RS, Ogawa S, Tank DW, Ugurbil K. 4 Tesla gradient recalled echo
characteristics of photic stimulation-induced signal changes in the human primary
visual cortex. Magn. Reson. Med. 30,380-386 (1993).
15.Riederer SJ, Tasciyan T, Farzaneh F, Lee JN, Wright RC, H e a e n s RJ. MR
fluroscopy : Technical feasibility. Magn. Reson. Med. 8, 1 - 1 S (1 988).
16. Goodyear BG, Menon RS. Effect of luminance contrast on BOLD tMRi response in
human pnmary visual areas. J. Neurophysiol. 79, 2204-2207 (1 998).
17. Stmpp JP. Stimulate: A GUI based fMRI analysis software package. Neurolmage. 3,
S607 (1996).
18.Wiesel TN, Hubel DH, Lam DM. Autoradiographic demonstration of ocular
dominance columns in the monkey striate cortex by means of transneuronal transport.
Brain Res. 79,273-279 (1974).
19. Engel SA, Glover GH, Wandell BA. Retinotopic organization in hurnan visual cortex
and the spatial precision of functional MN. Cereb. Cortex 7 , 18 1 - 192 (1 977).
20. Zheng D, LaMantia AS, Purves D. Specialized vascularization of the primate visual
cortex. .l Neurosci. 1 1,2622-2629 (1 99 1).
21. Bwcton RB, Frank LR. A mode1 for the coupling between cerebral blood flow and
oxygen metabolism during neural stimulation. J. Cereb. Blood. Flow Metab. 17, 64-
72 ( 1 997).
22.Boynton GM, Engei SA, Glover GH, Heeger DJ. Linear systems analysis of
functional magnetic resonance imaging in human V 1 . J. Neurosci. 16, 4207-422 1
(1 996).
Chapter 6
A Neuronal Correlate of Suprathreshold Contrast Perception
in Human ~ r n b l ~ o ~ i a ~
by Bradley G. Goodjwr, David A. Nicolle, G. Keith Humphrey, and
Ravi S. Menon
6.1 Introduction
Strabismus, a misalignrnent of the optical axes. and anisometropia, a difference in
the refractive properties of the eyes, are two disorders commonly associated with
unilateral amblyopia (1 ,2). A visual deficit often witnessed in these forms of amblyopia is
a reduction in contrast sensitivity, that is, the reciprocal of the contrast required to detect
a visual target. In normal vision, contrast sensitivity is highest between 2 and 5 cycles per
degree (cpd) (for example, see ref. 3,4). For the arnblyopic eye, the reduction in contrast
sensitivity is more pronounced at higher spatial frequencies, and in some cases,
sensitivity is maximum at a spatial frequency that is lower than the maximum for the
preferred eye (5-7).
'A version of this chapter has bem submitted for publication.
At suprathreshold contrasts nearing -20-30%, the perceived contrast becomes
equal to the actual physical contrast, and is thus constant across al1 spatial fiequencies.
This known as contrast constancy. and suggests that spatial fiequency channels in the
visual cortex may be organized to compensate for attenuation at low contrast and high
spatial frequency (8). However, when luminance levels are low and when visual targets
extend into the periphery, there is a drop in contrast sensitiMty as well as a shift of the
contrast sensitivity function to lower spatial frequency (9-1 1). Under these viewing
conditions, the contrast at which contrast constancy occurs rnay be elevated compared to
that of foveal targets at relatively high luminance.
It has been dernonstrated in contrast-matching studies that the amblyopic and
preferred eye perceive targets to have the same contrast (12-14). However, it has also
been reported that contrast increment thresholds at suprathreshold contrast are elevated
for the arnblyopic eye (15,16). Together these studies suggest that although the
perception of suprathreshold contrast is not impaired for the amblyopic eye, there are
impairments related to level of contrast above detection threshold. These impairments are
also reflected in measurements of reaction time to grating patterns (14,17), which have
also been shown to increase with spatial Frequency, even when contrasts are equated
perceptually (1 8,19).
Animal models have suggested that neural substrates of amblyopia are evident in
visual areas as early as VI in the form of a massive loss of binocular neurons and a
marked shift in ocular dominance (20-26). However, in histological studies of human
visual cortex, no shift in ocuiar dominance has been found (27,28), most likely due to the
onset of arnblyopia being after the critical period of ocular dominance column
developmnt (20). E!e~h~~kysi~!~gi~~! shdies in m h l s hwe dei,vnstrteI t h t the
spatial frequency tuning of the remaining binocular neurons exhibits a reduced neuronal
firing rate at higher spatial frequencies (24,26), suggesting a change in the degree of
binoçular interaction (30). In support of this observation, histological studies have
demonstrated a reduction in rnetabolic activity within neurons specializing in binocular
processing (3 1).
Up to now, only positron emission tornogrnphy has been used in hurnan
neuroimaging studies to investigate amblyopia, and has demonstrated a decrease in global
signal within V 1 in response to monocular stimulation of the amblyopic eye (32). Spatial
frequency nuiing of neural activity has not been investigated. nor has imaging been used
to determine if visual cortical activity correlates with behavioral measures of contrast
perception. Using functional magnetic resonance imaging (fMRI) to measure the
response to different spatial frequencies at relatively low contrast rnay reveal neuronal
correlates of contrast perception with respect to the level of contrast above threshold for
the amblyopic eye.
6.2 Methods
Six participants with no known visual deficits were recruited fiom the academic
environment of the IJniversity of Western Ontario (UWO). Four participants with
unilateral arnblyopia developed during early childhood or infancy were recruited through
the Department of Ophthalmology at the London Health Sciences Centre. London,
Ontario, Canada. The classification of amblyopia for each participant was deterrnined by
orthoptic assessment, and is summarized in Table 6-1. During the experiment, each
participant's vision was optically corrected using their existing prescription lenses, if
necessary. Al1 participants had no previous experience with MR imaging. The
expenmental protocol was approved by the UWO Human Subjects Review Board.
Subject, Age, Sex Eye Refractive Correction Snellen Strabismic (sphere, cylinder, axis) V i s d Acuity deviation
(prisrn diopters)
Strab ismic DA, 42, M R +O30 20/20 L.XT., 50
L +O30 + 1 .O0 x 90 20/200+ 1 MA, 56, F R +6.00 + 1 .O0 x 97 20/400 R.XT., 40
L +5.75 + 1 .O0 x 86 20/3 O Strasbhmic and Anisornetrupic
LL, 39, F R +6.00 201250 RXT., 30 L +3.75 20/20
CB, 51, F R +2SO + 1 .O0 x 30 201 125 R.ET., 4 L + I .SO t2.00 x 180 20/25
Table 6-1: Classification of amblyopia for each subject by orthoptic assessment. L.XT. =
left exotropia, R.XT = right exotropia, R.ET = right esotropia.
To ensure that stimulus conditions were the sarne as those used during the fMRI
experiments. behavioral rneasurements of suprathreshold contrast perception were made
while the participant lay in the magnet. Al1 visual stimuli were presented on a projection
screen that was located 1.25 metrrs from the participant's eyes and mounted on the
patient bed around the participant's legs. An angled minor, positioned above the
participant's eyes, provided a hill view of the screen. The participant wore a pair of liquid
crystal shutter glasses (33), and each eye was tested separately by closing the eyepiece of
the fellow eye. The visuai stimuli consisted of vertical sinusoidal gratings at 6 spatial
frequencies (0.5, 1. 2 . 4 8 , 12 cpd) of known physical contrast (22%) and luminance (20
cd/m2). The 1 cpd grating was not included for participants with normal vision. The
physical contrast and luminance of each grating were measured using a Minolta CS-100
Chroma Meter (Minolta Camera Co., Ltd.. Japan; for sarnple calibration curves of
projection screen contrast, see Appendix D). Contrast was defined in the usual way as the
luminance of the bright regions of the grating minus the luminance of the dark regions,
divided by twice the mean. Each target was confined to a circle that subtended 12' of
visual angle, and was reduced in contrast near the edge of the circle to eliminate sharp
edges. The surrounding area of the display was a luminance-matched grey.
The perceived contrast at each spatial fiequency was obtained using a temporal
two-alternative forced choice procedure, as outlined in Figure 6-1. The subject was asked
to match the 22% physical contrast of the test fiequency (TF) to the variable contrast of a
standard spatial frequency (SF), which was cornrnon to d l spatial frequencies tested.
During each trial, there was a 500 ms presentation of one of the test spatial fiequency
eratings. as shown in Figure 6 - l a at 22% physical conmt. This was followed hy a Sn0 - ms presentation of an isoluminant grey screen matched in mean luminance to the
gratings. Finally, there was a 500 rns presentation of the standard spatial fkequency (4 cpd
for participants with normal vision; 1 cpd for participants with ambiyopia) that differed in
contrast, or not, fiom that of the test frequency. At al1 other times, the screen was an
isoluminant grey. The participant was then asked to choose the grating that contained the
greater contrast. This was repeated for a number of physical contrasts of the standard
frequency (12%. 17%, 22%, 27%, 32%) which were paired with the test frequency in
random order. Five responses were recorded at each contrast. This procedure was then
repeated for the remaining test frequencies. The contrast levels of the standard frequency
were based on results of pilot studies, and provided a symmetrical range about 22%
contrast to reduce biases in participants' responses. For each test frequency, the
percentage of participant responses indicating that the standard frequency grating
contained greater contrast was plotted as a function of the standard fiequency contrast to
produce a response function. as shown in Figure 6-lb. The appropriate selection of
standard frequency contrasts and randomization of presentations provide a response
function that is well characterized. Thus, the contrast of the standard frequency matching
each test frequency was taken as the 50% point on a fitted line to the corresponding
response function (35). Plotting the resulting contrast values as a hinction of spatial
fiequency provided a rneasurement of the suprathreshold perceived contrast function.
The threshold contrast of the standard frequency was then determined using a two
alternative forced-choice method while narrowing the range of contrast of the grating
about detection threshold. The level of perceived contrast above threshold was then
calculated in multiple units of contrast threshold for each eye. Upon completion of the
behavioral tasks, the participant was removed From the magnet for a 10- 15 minute rest
be fore irnaging .
Figure 6-1 : Method used to detemine perceived contrast (simulated data, for illustrative
purposes only). a. The variable contrast standard frequency (SF) is paired with each 22%
contrast test frequency (TF). b. Response function obtained using a two-alternative
temporal forced choice method. Smooth line represents a curve fitted to the data points
6.2.3 Functional Imaging
Al1 imaging expenments were performed on a Un* iNOVA Varîan/Siemens 4
Tesla whole-body MR scanner, equipped with 25 mT/m whole-body gradients. An 8-cm
diameter, quadrature radio fiequency (RF) surface coil was placed at the back of the
participant's head to transmit and receive the MR signal. The participant's head was
im-m&i!ized usifig a wc!!-pcd&d h e ~ d whi& ws 1ll0~7ted CE~Q ~be ypme nlltfnnn r ----- A-A
that housed the RF coil.
Seven imaging planes were prescribed parailel to the calcarine fissure, which w s
identified in each participant within one TI-weighted sagittal localizer image. The
functional image data acquisition was a T2*-weighted, 8-segment, slice-interleaved, echo-
planar imaging sequence (32) (128x128 matrix, 14x14 cm field of view, 3 mm slice
thickness, echo time (TE) = 15 ms. repetition time (TR) = 500 ms, flip angle = 35').
During image acquisition, each 22% physical contrast grating was presented
separately, in random order, a total of 3 times at each spatial frequency, with each
presentation lasting 3 seconds every 20 seconds. The remaining 17 seconds of each epoch
was a luminance-matched grey screen. The total imaging time was 6 minutes for each
eye.
To measure reaction times, the participant was instnicted to press a button when
each grating was detected. At the end of the experiment, anatornical reference images
were collected using a three-dimensional, gradient-recalled echo FLASH imaging
sequence (256x256~32 rnatrix, 14~14~4 .8 cm field of view, inversion tirne (TI) = 0.5 s,
TE = 6.3 ms, TR = 11.7 ms, flip angle = Il0). Before data analysis, al1 images were
motion-corrected using SPM96 (36), leaving 5 slices for andysis.
6.2.4 Data Analysis
T TA-- z St~dcnt's ;-test incoïpûrzting iï rime itclay ta d i ~ w - fur inirhsic
delay of the hemodynamic response, image voxels showing a significant (p < 0.01)
increase in signal above baseline in response to monocular stimulation (which we term as
'activated' voxels) were used to produce a functional map of activation at cach spatial
fiequency. Regions of interest were selected using the anatomical reference images to
include primary visual cortex (VI) and possibly other early visual areas (V2). For each of
the resulting functional maps, the average fiactional signal change (or fMN response)
and the pooled fMRI response within the selected regions of interest were recorded and
averaged over subjects. Measurements of perceived contrast above threshold, MN
response, and reaction tirne were correlated by linear regression. converting the resulting
correlation coefficient to an equivalent p value with the appropriate degrees of freedom
(= 28 for normals; = 22 for amblyopes), and conecting for small sarnple sizes. Each
measurement was also correlated across eyes to determine differences between the
normal and arnblyopic eye. In an additional analysis, individual activated voxels were
identified whose fMRI response as a function of spatial frequency correlated significantiy
(J> < 0.05) with the perceived contrast function.
The wer~ge perceiveci contwt &ove threshold frr pfiicipmts 4 t b mm-!
vision is s h o w in Figure 6-2a(i). The curves in the figure, and in al1 subsequent figures,
were obtained by fitting the data, S, to a bi-exponential of the form
h -cm S = a ~ e , (6- 1
where o is spatial fiequency, and a, b, and c are the fitted parameters. The perceived
contrast is the same for the left and right eye. Figures 6-2a(ii) and 6-2a(iii) show the
pooled tMRI response and reaction rate (i.e., the inverse of the reaction time) as a
function of spatial frequency, respectively. Over all subjects, both the pooled MRI
response and the reaction rate correlate with perceived contrast @ < 0.05), and the left
and right eye monocular viewing conditions produce identical responses.
Figure 6-2b shows the results for participants with amblyopia. Although the
perceived contrast was the same for both eyes, the perceived contrast above threshold
measured with the amblyopic eye was decreased in magnitude and also showed a shift to
lower spatial frequencies. For al1 participants, the difference between perceived contrast
above threshold measured with the amblyopic and non-amblyopie eye increased as
spatial Frequency increased. The pooled fMRI response (Figure 6-2b(ii)) and reaction
rates (Figure 6-2b(iii)) also showed a correlation with perceived con- above threshold
( p < 0.05), and these responses were reduced during monocular stimulation of the
amblyopic eye ( p < 0.0 1). The reduction in reaction rate for the amblyopic eye is in
agreement with past studies (17), as well as the reduction in the pooled activity within
early visual areas during monocular stimulation of the amblyopie eye (29).
Figure 6-2: Behavioral and fMFü measurements of "effective contrast (Le., perceived
contrast above threshold) for participants with (a) normal vision (n = 6) and (b) unilaterd
arnblyopia (n = 4). (i) The matching contrast of the standard spatial frequency (in units
above threshold contrast) for the given test fiequencies that presented at 22% physical
contrast. (ii) Pooled £MN response in early areas of the visual cortex, defined as the
product of the nurnber of activated voxels in the functional map and their average fMRJ
response. (iii) Reaction rate (reciprocal of reaction time). Error bars represent the
standard error of the mean.
To determine if the fMN response as a function of spatial frequency within
individual image voxels showed the same trend as conhast perception, we used perceived
contrast as an input correlatian fûnction to correiate with the fMRl rsspnnse tn monocrilar
stimulation of the corresponding eye. Figure 6-3 shows that for voxels that significantly
correlated with perceived contrast @ < 0.05), the average fMRI response for each eye
was the same, regardless if the eye ivas arnblyopic or not.
In an additional analysis, we used both the perceived contrast measured with the
arnblyopic and the preferred eye to correlate with the fMRi response to monocular
stimulation of the amblyopic eye. Figure 6-4a shows functional maps overlaid on two
anatomical siices for a representative subject, and Figure 6-4b shows the mean activation
for voxels that showed a significant correlation @ < 0.05). The yellow voxels of the
functional map correlate with perceived contrast of the amblyopic eye, and the red voxels
correlate with perceived contrast of the preferred eye. The yellow voxels show a response
that is reduced in magnitude and shifled to lower spatial frequencies. The difference
between the magnitude of the responses of the yellow and red voxels becomes greater as
spatial frequency increases.
spaîiai ft equency <cyt#eg>
Figure 6-3: Average tMRI response for activated image voxels exhibithg a significant
correlation @ = 0.05) with perceived coneast measured with the correspondhg eye for
participants with (a) nomial vision (n = 6) and @) unilaterai amblyopia (n = 4).
spaiil brquency (clF#eg)
Figure 6-4: a. Functional maps of voxels correlating with the perceived contrast
measwed with the amblyopie eye (yeiîow) and the non-amblyopie eye (red), overlaid on
two anatomid &ces for one participant with unilateral strabismus and anisometropia b.
Average fMRI response of the activateci voxels in (a).
When we isolate the red voxels and compare their magnitude to the fMRl
response of voxels activated during monocular stimulation of the prefened eye from
Figure 6-3b. we see, as shown in Figure 6-5, that the average tMRI response as a function
of spatial Frequency is identical to the fMRI response to monocular stimulation of the
pre ferred eye.
amblyopie eym opreferted eye
spatial aequency (cyaü egl
Figure 6-5: Average fMRI response of voxels whose response magnitude as a function
of spatial Frequency correlated with perceived contrast measured with the non-amblyopie
eye. Error bars represent the standard error of the mean.
Since we have s h o w that the pooled MRI response reflects the perceived
contrast above threshold measured with the arnblyopic eye (Figure 6-2), the results
shown in Figure 6-5 suggest there may be a decrease in the number of activated voxels in
response to stimulation of the amblyopic eye that could demonstrate a decrease in
neuronal recruitment. Figure 6-6a shows that for normal vision, the number of activated
voxels as a function of spatial frequency is the same for the left and right eye. The
nurnber of voxels activated in response to monocular stimulation of the amblyopic eye,
however, is less, but this difference is significant @ < 0.05) only at higher spatial
fTPqllyrli~': (Fizlre &m.
left eye o right e y
spaîial t equency (cl(c#eg 1
Figure 6-6: ïhe nurnber of voxels activated in response to monocular stimulation as a
function of spatial frequency for participants with (a) normal vision (n = 6) and with (b)
unilateral arnblyopia (n = 4) (*, p < 0.05). Error bars represent standard enor of the mean.
6.4 Discussion
An interesting feature of Figure 6-2a(i) is that the data point at 4 cpd lies below
the fitted curve. This spatial frequency was the standard frequency, and was presented to
each participant repeatedly for nearly 10 minutes before it became one of the test
Frequencies. It has been demonstrated that prolonged a d o r repeated exposure to a
spatial Frequency can result in adaptation, resulting in a reduced response to that
frequency during subsequent presentations (e.g., 4). In our case, we suspect that the
visual system had become adapted to the 4 cpd stimulus, resulting in a reduced response
in our behavioral measurement at this frequency. As well, there is a slightly reduced
did not dissipate over the participant's rest period. However, as seen in Figure 6-2a(iii), a
decrease in reaction rate is not apparent at 4 cpd. It has been demonstrated that adaptation
to a spatial fraquençy p r o h x s a broadly tuned reduction in reaction rate (37). We
probably did not see this effect due to inefficient sampling of the reaction rate curve.
With our existing data, however, we cannot verify that these observations are indeed due
to adaptational effects. In any case, these were merely post hoc observations and not
goals of this study. Similar features are apparent within Figure 6-2b at 1 cpd.
Measurements of contrast sensitivity or suprathreshold contrast perception in
amblyopia are usually presented case by case since accurate measurements for single
individuals can be obtained by averaging many responses, and there is considerable
variability among amblyopes (13). It was not the goal of this study to make
generalizations about contrast perception in amblyopia, nor to classifi the type and
degree of amblyopia based on Our measurements. Rather, we have demonstrated that
there are significant differences from normal vision, and that behavioral and
physiological measurements of contrast perception are correlated. In addition, to relieve
participant boredom and reduce participant motion resulting from lengthy imaging scans,
we used only 5 repetitions of the standard fiequency within trials of the behavioral tasks,
and only 3 presentations of each spatial frequency target during imaging. There were,
however, significant correlations on an individual basis as well.
Using short duration stimuli (< 4 seconds) is key in high spatial resolution
imaging studies to maintain the spatial specificity of the fMRI response by preventing
saturation of the local hemodynamic response (38). A non-saturated hemodynamic
response is aiso more likely to maintain proportionality between neural activity and
BOLD fMRI signal (35). Using our techniques, we have demonstrated that the neural
activity in early visual areas of the visual cortex is dependent on spatial fiequency, and is
correlated with perceived contrast. In addition, we have shown that psychophysical
measurements of behavior can actually be correlated with fMRI data to localize cortical
areas that dernonstrate the behavior.
Even though the pooled activity in early visual areas reflects the perceptual
contrast deficits in amblyopia, Our data suggest that the average localized neural activity
is the sarne for both eyes, even in the presence of amblyopia. This is consistent with
electrophysiological studies demonstrating that there is no reduction in the neuronal
firing rate for cells which preferentially fire in response to stimulation of the amblyopie
eye (26). In Our study, image voxels contained both preferentially firing and non-
preferentially firing cells, and the image resolution used in this study cannot isolate these
types of cells. However, our results demonstrate that fMRI of arnblyopia using Our
technique, even at moderate spatial resolution, is sensitive to the reduced localized neural
activity as a function of spatial fiequency, as well as to neuronal populations whose
spatial fiequency tuning seems to be spared. This can be explored M e r in a study at
higher spatial resolution to segregate these ce11 types by investigating their response to
monocuiar and binocular stimulation.
It has also been suggested that the ocular dominance columns for the amblyopie
eye may have shrunken as a result of certain types of infantile amblyopia (20.21). and
that this may be the neural basis of amblyopia. Since tMRI has been shown to be able to
image ocular dominance column distributions (35), a submillimeter spatial resolution
study of the visual cortex of amblyopes using fMRI may be able to support or contradict
this hypothesis. Nonetheless, the results of this study demonstrate that tMRi is a usehl
tool in the investigation of brain plasticity resulting from amblyopia throughout the entire
visual cortex, and can also be used in similar investigations of other visual or cognitive
disorders.
6.5 References
von Noorden, GK. Binocular Vision and Ocular Motiliv (Mosby, St. Louis, 1990).
Ciuffreda KJ, Levi DM, Selenow A. Amblyopia: Basic and Clinical Aspects
(Butierworth-Heinemann, Boston, 199 1).
Campbell FW, Robson JG. Application of Fourier analysis to the visibility of
gratings. J. Physiol. ILond. ) 197. 55 1-556 ( 1 968).
Blakemore C, Campbell FW. On the existence of neurones in the human visual
system selectively sensitive to the orientation and size of retinal images. J. Physiol.
(Lond) 2O3,23 7-260 ( 1 969).
Bradley A, Freeman RD. Convast sensitivity in anisometropic amblyopia. Invesr.
Ophthamol. Vis. Sci. 2 1,467-476 ( 198 1).
Hess RF, Campbell FW. Greenlaugh T. On the nature of the neural abnormality in
human amblyopia: Neural aberrations and neural sensitivity loss. Pfluegers Arch.
Gesamte. Physiol. 337.20 1-207 ( 1978).
Hess RF, Pointer IS. Differences in the neural basis of hurnan amblyopia: the
distribution of the anomaiy across the visual field. Vis. Res. 25, 1577-1 594 (1985).
Georgeson MA, Sullivan GD. Contrast constancy: deblurring in hurnan vision by
spatial frequency channels. J. Physiol. 252,627-655 ( 1975).
Campbell FW, Green DG. Optical and retinal factors affecting visual resolution. J.
Physiol. 186,558-578 (1965).
10. Daitch JM, Green DG. Contrast sensitivity of the human peripheral retina. Vision Res.
9,947-952 (1969).
1 1. Hilz R, Cavonius CR. Functional organization of the peripheral retina: sensitivity to
periodic stimuli. Vision Res. 14, 1333-1 337 (1974).
12. Hess RF, Bradley A. Contrast perception above threshold is only minimally impaired
in human amblyopia Nature 287,463-464 (1980).
13. Hess RF, Bradley A, Piotrowski L. Contrast coding in amblyopia. 1. Differences in
the neural basis of arnblyopia. Proc. R. Soc. Lond. Series B 217,309-330 (1983).
14. Loshin DS, Levi DM. Suprathreshold contrast perception in bct ional amblyopia.
Doc. Ophthamol. 55,213-236 (1983).
15. Ciufieda KJ, Fisher SK. Impairment of contrast discrimination in amblyopie eyes.
Invest. Ophthamol. Vis. Sci. (SuppZ.) 28, 1 O3 ( 1987).
1 6. Ki per DC, Riorpes L . S uprethreshold conirast sensitivity in experimentally strabismiç
monkeys. Vision Res. 34, 1 575- 1583 (1 994).
17. Levi DM, Harwerth RS, Smith EL. Humans deprived of normal binocular vision have
binocular interactions tuned to size and spatial frequency. Science 206, 852-854
( 1 979).
18.Breitmeyer BG. Simple reaction timr as a measure of the temporal response
properties of transient and sustained channels. Vision Res. 1 5, 14 1 1 -141 2 (1 975).
1 9. Harwerth RS. Levi DM. Reaction time as a measure of suprathreshold grating
detection. Vision Res. 18, 579-1 596 (1 978).
20. Hubel DH, Wiesel TN, LeVay S. Plasticity of ocular dominance columns in monkey
stiate cortex. Phil. Trans. R. Soc. Lond. (Biol.) 278, 377-409 (1977).
2 1. LeVay S, Wiesel TN, Hubel DH. The development of ocular dominance colurnns in
normal and visually deprived monkeys. J. Comp. Neurol. 19 1, 1-5 1 (1 980).
22. Crawford MLJ, von Noorden GK. The effects of short-term experimental strabismus
on the visual system in Macaca mulatta. Invest. Ophtharnol. Vis. Sci. 18, 496-505
(1 979).
23. Crawford MLJ, von Noorden GIS. Optically induced concornmitant strabismus in
monkeys. Invest. Ophthamol. Vis. Sci. 19, 1 105-1 109 (1 980).
24. Movshon JA, Eggers KM. Gizzi MS, Hendriclcson A, Kiorpes L, Boothe RG. Effects
of early unilateral blur on the macaque's visual system: HI. Physiological
observations. 1 Newosci. 7, 1340-1 35 1 (1987).
25. Lowe1 S. Ocular dominance column development: strabismus changes the spacing of
adjacent columns in cat visuai cortex J. Neurosci. 14,745 1-7468 (1994).
26. Kiorpes L, Kiper DC, OTKeefe LP, Cavanaugh SR, Movshon JA. Neuronal correlates
of amblyopia in the visual cortex of macaque monkeys with experirnental strabismus
and anisometropia. J. Neurosci. 18,64 1 1-6424 (1 998).
27. Horton JC, Stryker MP. Amblyopia induced by anisometropia withour shrinkage of
ocular dominance columns in human striate cortex. froc. Nat!. .4cad. Sci. 90, 5494-
5498 (1993).
28.Horton JC, Hocking DR. Pattern of ocular dominance columns in human striate
cortex in strabismic amblyopia. Vis. Neurosci. 1 3,787-795 ( 1996).
29. Daw, NW. Visual Development (New York, NY: Plenum Publishing Corp, 1995).
30.Holopigian K. Blake R, Greenwald MJ. Selective losses in binocular vision in
anisometropic amblyopes. Vision Res. 26,52 1-630 (1 986).
3 1 .Horton JC, Hocking DR. Kiorpes L. Pattern of ocular dominance columns and
cytochrome oxidase activity in a macaque monkey with naturally occumng
anisometropic arnblyopia. Vis. Neurosci. 14.68 1 -689 ( 1 997).
32. Demer JL, von Noorden GK, Volkow ND. Gould KL. Imaging of cerebral blood flow
and metabolism in arnblyopia by positron emission tomography. Am. J. Ophthamol.
105,337-347 (1988).
3 3. Milgrarn P. A spectacle-mounted liquid-crystal tac histoscope. Behmior Res. Methods
Inst. Comp. 19,449-456 (1 987).
34. Wethenll GB, Levitt H. Sequential estimation of points on a psychometic îunction.
Brit. J. Math. Stat. Psych. 18, 1-10 (1965).
35. Menon RS, Goodyear BG. Submillimeter functional localization in hurnan striate
cortex using BOLD contrast at 4 Tesla: Implications for the vascular point-spread
bc t ion . Magn. Reson. Med. 41,2300235 (1999).
36. Fnston KJ, Jezzard P. Turner R. Statistical parametric maps in functional imaging : A
general linear approac h. Human Brain Mapping 2, 1 89-2 1 0 ( 1 99 5).
3 7 .Menees SM. The effect of spatial frequency on the latency of spatial contrast
detection. Vision Res. 38,3933-3942 (1 998)
38 . Das A, Gilbert CD. Long-range horizontal co~ect ions and their role in cortical
reorganization revealed by optical recording of cat visual cortex. Science 375, 780-
784 (1995).
CBrpter 7
A Neural Substrate for the Dominant Eye in Human
Amblyopia and Normal vision**
by Bradley G. Goodyear, David A. Nicolle, and Ravi S. ikfenon
7.1 Introduction
Arnblyopia. a decrease in visud acuity in one or both eyes caused by form vision
deprivation or abnormal binocular interaction during infancy or early childhood, occurs
in approximately 2% of the Arnencan population (1,2). Strabismus, a misalignment of the
optical axes, and anisometropia, a difference in the refractive properties of the eyes. are
two disorders that commonly lead to amblyopia (12). Although the affected eye can be
surgically corrected, vision is not often restored to normal, suggesting that amblyopia has
an underlying neural basis. Non-human primate models of strabismus and anisornetropia
suggest that anomalies in neuronal b c t i o n are evident in visual areas as early as primary
visual cortex, or VI, in the form of a massive loss of binocular neurons, suggesting a
change in the degree of binocular interaction (3,4).
"A vnsion of this chapter has been submitted for publication.
There is also evidence of a rnarked shift in the ocular dominance of neural
activity, both in anisometropia (5-7) and in profound cases of strabismus (8,9),
suggesting a reduction of geniculocortical ir?pu?s in.!̂ !zyer 4c of V! h m the zmblyopic
eye. This could possibly lead to a reduction in the size or spacing of the ocular
dominance colurnns (ODCs) within layer 4c if amblyopia develops within the critical
penod of ODC development ( IO) , as demonstrated in animals visually depnved at or
shonly after birth (1 1,12). However, histological studies of naturally occurring
anisometropia in monkeys have not detected ODC shrinkage (1 3).
The only studies of V1 in human amblyopes conclude that there is no apparent
ODC shrinkage as a result of anisometropic (1 4) or strabismic arnblyopia (1 5). Although
these results are most likely legitimate since the age of onset of the amblyopia was
undoubtedly after the critical period of ODC development (IO), the conclusions were
based on an average over the entire reconsmcted cortex, including the peripheral visual
field representation of V1 where ihere is a known predominance of the contralateral eye
(1 6-1 9). This method could overlook any predominance within the central visual field (O0
to SO).
To date, no studies have investigated column size or distribution in relation to
behavioral measurements of the visual deficits commonly witnessed in strabismic and
anisometropic amblyopia, which are known to be different in the penphery (20,21).
Visual acuity loss in the central visual field, however, is well characterized clinically for
both strabismus and anisometropia, and demonstrates the dominance of the unaf5ected
eye. Even in the absence of amblyopia, the vast rnajority of humans preferentially use one
eye (right eye, 65%; lefi eye, 32%) to align a target in the central visual field (22,23). The
domin& PYP CSII he deterzhed hehaviord!y i~rii.g ses-f-r digin.ent tests (34,15),
however. the neural basis for this is not known.
Neuroimaging studies of human amblyopia using positron emission tomography
(PET) (26) and functional magnetic resonance imaging (WRI) (27) support previous
tindings, having demonstrated a decrease in the pooled neural activity within V1 in
response to monocula. stimulation of the arnblyopic eye. However, imaging has not been
able to identify the underlying neural mechanisms due to poor image resolution. We
wished to investigate the correlation between the preferred unaffected eye of human
adults with amblyopia or the dominant eye of adults with normal or corrected-to-normal
visual acuity with ODC size, distribution, or activity within the central visual field
representation of V 1, which we denote as V 1 c. This was achieved using a high-resolution
fMRI technique that has been demonstrated to reliably produce maps of ocular
dominance (see Section 7.2.3), and using visual stimulation that was constrained to the
central visual field.
7.2 Methods
Eleven subjects [five male, six female, 26 3 years or age (mean * SD)], with
nomai or corrected-to-normal visual acuity were recniited by ttntten informed consent.
Subjects were tested to determine the dominant eye in the central visual field using
variants of two near-far alignrnent tests. In the Porta test (24), subjects were instructed to
hold a pencil in one hand vertically at arm's length, and align the pencil with a point on a
distant wall with both eyes open. Subjects were then asked to view the pencil
monocularly and report the eye with which they saw the pencil as being aligned with the
distant point. In the Miles test (25), subjects were asked to extend their arrns in front of
them with their palms facing toward them, and focus on the same distant point. Subjects
were then instnicted to slowly bnng their hands together until they could just resolve the
distant point. Subjects were asked to report the eye with which they could still see the
distant point. The reported eye in each test was the same, and was considered the
subject's dominant eye. Four subjects were left-eye dominant, and seven were right-eye
dominant. The dominant eye for subjects with amblyopia (Table 7-1) was always the
unaffected eye.
7.2.2 Stimulus Presentation
Either a 66%-contrast checkerboard reversing contrast at 8 Hz (for normals) or
66%-contrast vertical sinusoidal gratings (3 experimental runs at 3 spatial frequencies:
0.5, 2, and 4 cycleddegree) drifting at 2 cycles per second and reversing direction every
500 ms (for amblyopes) was projected onto a screen positioned near the subject's waist,
1.25 rn from the subject's eyes. The subject viewed the stimulus either rnonocularly (R or
L) or binocularly (B) through a pair of liquid crystal shutter glasses (28) for 4 seconds at
32 second intervals. Nine trials were presented in the following order: B-L-R-B-L-R-B-L-
R. The 28 second latency consisted of both liquid crystal shutters being closed, which
were translucent to prevent dark-adaptation. The mean luminance of the stimuli was 60
cd/m2.
The visual stimuli were contained within a circle that subtended 15" of visual
angle (the central visual field). The remainder of the screen was grey, matched in mean
luminance to the stimuli. Subjects were asked to fixate a small grey dot (1° in diameter) at
the center of the circle to the best of their ability during the entire experiment. No
monitoring of subject eye movement was performed.
Subject, Age, Sex Eye Refiaction Correction Snellen Strabismic (sphere, cylinder, axis) Visual Acuity Deviation
(prism dioptm)
Stra bismic MA, 56, F R +6.00 + 1 .O0 x 97 201400 R.XT. 40
L +5.25 + 1.75 x 86 20130
BC, 28, M R +4.50 +OS0 x 90 20/20 L.XT. 30 L +3.75 +OS0 x 98 30160
Strasbismic and Anisometropic
LL, 39, F R +6.00 201250 R.XT. 30 L +3 .O75 20130
MO, 39, F R + 1-50 20/20 L.ET. 20 L +3.50 20170
LC, 30, F R +2.75 + 0.25 x 170 20120 L.XT. 35 L +5.75 + 1.25 x 160 20/200
MS, 64, F R -t 1 .O0 20/25 L.XT. 30 L + 1 $75 + 1 .O0 x 005 20/ 1 O0
DA, 42, M R +OS0 20120 L.XT. 50 L +O.OO + 1 .O0 x 90 20/200
CB, 51, F R +2.50 + 1 .O0 x 30 20/ 125 R.ET. 4 L +1.50 + 2.00 x 180 20125
Table 7-1. Classification of amblyopia for six patients with arnblyopia developed during
infancy (< 1 year of age) and two patients (DA and MO) with amblyopia developed after
2 years of age. Subjects were recruited through the Department of Ophthalmology at the
London Health Sciences Centre. London, Ontario, Canada, and gave written informed
consent. Cycloplegic refiaction was determined during orthoptic assessment. and vislia1
acuity was tested using best correction. Anisometropia was defined by a greater that I
diopter difference between the eyes in either the sphencal or cylindrical correction.
[LU. = left exotropia. R. XT. = nght exotropia. L. ET. = left esotropia. R. ET = right
esotropia].
7.2.3 Functional Imaging
411 irnaging war yerfonned on a VmianlSiemens Unity l.AJOV-4 4 Tesla whole-body
system (Palo Alto, CA; Erlangen, Gennany). A distributedtapacitance, 8-cm diameter,
quadrature RF surface coi1 was used to transmit and collect the MR signal. The subject's
head was immobilized using a well-padded, plexiglass head vice. Functional images were
collected using a 3-slice, 16-segment interleaved EPI gradient-recalled echo pulse
sequence (0.55 mm x 0.55 mm prescribed in-plane resolution; echo time = 15 ms:
volume repetition time = 4 s; RF flip angle = 25'; 3 mm slice thickness), with centric
ordering of k-space and a navigator echo for every segment (29). The optimization of
imaging and stimulus parameter is presented in Appendix B.
V l c was localized in a separate imaging experirnent within the same session by
isolating areas showing a significant increase in fMRi response with an increase in the
contrast of a visual stimulus presented in the same visual space as used in the remriining
experiments (?0,3 1).
7.2.4 Data Ana fysis
The details of the analysis are given in Chapter 5. Al1 time series of images were
corrected for low-fiequency drift and for motion using SPM96 (32). Individual functional
maps of image voxels showing a significant rise in MR signal above baseline were
created for each condition (i.e., B, R, and L) by regrouping corresponding trials and
correlating the t h e series of images with a boxcar function that incorporated a temporal
delay for the hemodynamic response. The statistical significance of the correlation was
or ât p = O.?!. The rr?qyhrrde of t,he MR. sigznA st t!x p&!k cf tlro fiilRI resycnse md m
average magnitude over a narrow range in between each trial were used to caiculate the
magnitude of the fMRI response above baseline (expressed as a percentage increase in
MR signal) uithin each voxel of each map. For each voxel in the R rnap, if the average
fMRI response over the three trials within an activated voxel during R was significantiy
greater @ < 0.05) than the response over the three L trials in the L map, that voxel was
considered part of the right-eye ODC mosaic. Similarly for the left-eye ODC mosaic.
Relative area was then calculated based on the relative area occupied by the voxel
clusters corresponding to dominadpreferred-eye ODCs and to non-dominant/amblyopic-
eye ODCs (for more details, see Chapter 5).
7.3 Results and Discussion
Figure 7-1 shows functional maps demonstrating the arrangement of ocular
dominance columns within human Vlc for subjects with normal vision (Figure 7-1 A) and
with unilateral arnblyopia (Figure 7- 1 B). Our maps of ocular dominance are identical in
appearance to those obtained by optical imaging of monkeys (33,34) and histology of
both monkeys (1 8,19,35) and hurnans (36). Where the cortex is interrogated tangentially
(as indicated by the lack of white matter/grey matter contrast beneath the map in Figure
7-l), ODCs altemate as thin stripes less than 1 mm in width, demonstrating how ODCs
run tangentid to the sUTf8ce of the prirnary visual cortex. The striped appearance is lost
where the cortex is interrogated dong other orientations, and maps do not show the entire
ODC rnosaic. AU maps do not show the entire ODC masaic, since voxels exhibiti. an
insignificant predominance in their response due to some partial-voluming were
exciuded. Maps of ocular dominance can also be obtained in tramverse sections of the
cortex (Figure 7-2). allowing more efficient coverage of VI to make quantitative
meamirements of columnar activity and spatial distribution.
Figure 7-1: Maps of ODCs overlaid on sagittal anaiornical images (posterior located to
the right) of the medial bank of the visual cortex of one hemisphere for A two subjects
with right-eye dominant normal vision and B. one subject with lefi-eye
exotropidanisometropia (subject LC, left) and one subject with Rght-eye
exotropidanisometmpia (subject U, right). The calcarine sulcus is indicated by the black
arrow in each image. Red (green) indicates lefi (right) eye ODCs. The bright white area
to the right of each image is due to rapid blood flow within the sagittal sinus. Scale bar =
1cm.
Figure 7-2: Maps of ODCs overlaid on comsponding transverse anatomical images for
A. one subject with lefbeye dominant (left) and one subject with right-eye dominant
normal vision (right) and B. one subject with right-eye exotropia/anisometropia (subject
LC, lefi) and one subject with lefi-eye exotropia/anisometropia (subject LL, right). Red
(green) indicates left (right) eye ODCs. Colors alternate within gray matter throughout
the map as tiny clusters of differiag widths as cotumnç are sampled dong diRering
orientations due to the folding nature of the cortex. A striped appearance is evident in
some portions of each map since slices were predbed parallel to the posterior superior
bank of the cdcarine sulcus. Scale bar = 1 cm.
The best maps for amblyopes were obtained when using a 2 cycleddegree visual
stimulus since it produced the most dense rnaps of ODCs. Although ODC maps were
more sparse when using higher spatial fiequency stimuli, there was no difference in the
measurement of relative area occupied by ODCs as a fùnction of spatial fIequency (see
Appendix Cl, and the average fMPl orponre within ~ x e l s e~hihiting z rigriifirmt a d
measurable response was also the same. This was also demonstrated in previous studies
(27), and reflects the deficits in the spatial frequency tuning of the pooled neural response
in arnblyopia (37).
The investigation of al1 maps for al1 subjects with arnblyopia since infancy
revealed that the ODCs of the amblyopie eye occupied 40.7% of Vlc . This was both
significantly less than 50% @ < 0.0001) and significantly less than that of ODCs of the
non-dominant eye of subjects with normal vision @ = 0.0002), which also occupied
significantly less than half of Vlc (46.0%, p < 0.00 1). For normal vision. there was an
insignificant bias in area towards the ODCs of the contralateral eye (50.6%, p = 0.27),
mirroring the subtle bias found in histological studies of the operculum of macaque VI
(18,19). These macaque studies have also shown areas within the operculum of some
specimens where ODCs of one the eyes were predorninant in both hemispheres (1 8,19).
However. this was not attributed to the dominant eye since the monkeys were not tested
behaviorally to detemine their dominant eye before processing of the brains.
The ODCs of the preferred eye of the two subjects with amblyopia since
childhood (subject DA: age 4, subject MO: age 2), however, occupied the sarne relative
area of V l c (subject DA: 53.0%; subject MO: 52.3%) as ODCs of the dominant eye of
normal vision. This could be attributed to the late onset of the amblyopia, or the fact that,
at least in one case (subject DA), the amblyopia is purely strabismic, which has been
demonstrated in one case study to not affect the relative size of ODCs (15). However, for
one siihject who developed only strabirmic m!n'-!ynpi~ dixkg h facy (c~bjec? BC, see
Table 7-l), we did observe a shift in ocular dominance, although it was the smallest
effect seen in our patient population. We believe the result for this subject to be valid,
however, due to the subject's visual acuity. In combination with all the othar subjects
who developed arnblyopia during infancy, there is a logarithmic relationship between the
area of Vlc occupied by ODCs of the arnblyopic eye and visual acuity of the eye (Figure
7-3). According to o u results, most cases exhibit a ratio of the relative size of ODCs that
is less than 1.5-to-1. This would be difficult to measure using any technique that doesn't
sample the entire area of V 1 c. In this way, fMRI has the advantage of being able to non-
invasively interrogate al1 of Vlc, several times in the same session, if necessary.
A reduction in ODC area inferred from fMRi could result from an imbalance in
the monocular stimulation of each eye, since we rely on differences in MM magnitude
to identify ODCs. However, the average magnitude of the MRI response within
prefened-eye ODCs and amblyopic-eye ODCs is the sarne when the corresponding eye is
monocularly stimulated (Figure 7-4), at least for areas showing a significant and
measurable response. As a check. we repeated our study for one subject with corrected-
to-normal visual acuity in both eyes, only this time the subject had the corrective contact
lens removed from the non-dominant eye, significantly reducing its visual acuity. The
cortical area of V l c occupied by the dominant eye (53.5%) was practicaily the same as
when the non-dominant eye was corrected to normal visual acuity (54.1%). Hence, we
believe ou - measurements indeed reflect a reduction in the cortical area that preferentially
responds to monocular stimulation of the amblyopic eye, at least for the central visual
field representation.
Non-dominant eye of norrnals
Figure 7-3: The percent of cortical area of V l c occupied by ODCs of the amblyopic eye
as a function of visuai acuity of the eye (expressed as a percent of normal visual acuity,
e.g., 20/20 = 100% and 20/200 = 10%) for six adults with amblyopia since infancy. Each
data point is the average of three measurements of relative ODC area calculated from
ODC maps created fiom three separate experiments during the same imaging session.
The seventh data point at the far right [i.e., (100,46.0)] is the value for the non-dominant
eye of 11 subjects with normal or corrected-to-normal visual acuity. The fitted curve is a
logarithmic function, with the given best-fit correlation value, corrected for small
sarnples. For 1 % visual acuity [i.e. ln(1) = O)], the fined curve predicts a relative column
area of 32.3 1% for ODCs of the amblyopic eye. Error bars represent the standard error of
the mean.
Figure 7-4: The average fMFü response measured within ODCs of I I subjects with
normal vision (N) and al1 8 subjects with amblyopia (A) (*expressed as a percentage
above baseline, and normalized to the response measured within ODCs of the dominant
eye of normal vision during monocular stimulation of the dominant eye). The first group
of 4 bars represents the fMRI response measured within ODCs dunng monocular
stimulation of the corresponding eye, the second group represents the response measured
within ODCs during binocular stimulation, and the last group of bars represents the
response measured within ODCs during monocular stimulation of the fellow eye. The
magnitude of tiir OIIRI response of edch group of bars are ail significantly different from
the response of any other group, however, there are no statistically significant differences
within each group of bars.
During binocular stimulation, the magnitude of the fMRI response within ODCs
was less than the response during monocular stimulation of the correspondkg eye @ =
0.01) (Figure 74). This c m possibly be attributed to inhibitory horizontal connections
within the upper layers of the cortex which, during excitatory activity within a column,
inhibit the activity in nearby columns of the fellow eye (38). This response does not
qpear to be reduced in amblyopia since the relative average fMRI responses within
ODCs of the preferred eye and the arnblyopic eye during binocular stimulation are the
same (Figure 7-4). suggesting that the same relative amount of inhibitory activity is
taking place within the upper horizontal layers. As well, the relative average MRI
response within ODCs during rnonocular stimulation of the non-corresponding eye is also
the same, even in amblyopia.
Other striking observations frorn our ocular dominance maps for amblyopia are
the reduction in the nurnber of voxels that non-preferentially respond to either eye and the
skewness of the distribution of ocular dominance of the tMRI response to higher values
(Figure 7-59, supporting previous fmdings of a reduced binocular interaction within V1 of
amblyopes (3,4). However, the results of Figure 7-4 suggest that the activity of the
remaining neurons contrihuting to a measurable fMRI response is quite normal. The
distribution of the ocular dominance of the fMRI response (Figure 7-5) also shows a
slight shift in ocular dominance toward the dominant eye in normal vision, and a more
marked shiR toward the preferred unafTected eye in amblyopia (5-9).
Figure 7-5: Distribution of the ocular dominance of the tMRI response of each voxel in
al1 maps for al1 11 subjects with normal vision and for al1 8 subjects with unilateral
amblyopia. The ocular dominance of each voxel in the ODC map was defined as the
fMRI response within a voxel during non-dominant/arnbiyopic-eye monocular
stimulation subtracted from the fMRI response within the same voxel during
dominantlpreferred-eye monocular stimulation. One (1) represents the maximum value of
ocular dominance measured within dominantlpreferred eye ODCs such that the horizontal
scale ranged from 100% non-dominantlamblyopic-eye dominant to 100%
dominant/preferred-eye dominant.. This range is divided into discrete bins of 5%. Zero (O)
represents <5% relative difference between the fMRI responses during
dominantlpreferred-eye and non-dominant/amblyopic-eye monocular stimulation.
Since it is thought that a local increase in regional cerebral blood flow and
metabolism, and hence fMRI signal, is correlated with an increase in synaptic activity
(39-41), our results are consistent with the interpretation that, even in the presence of
amblyopia, the synaptic activity per tissue volume within a cortical column when
monocularly stimulated by its correspondhg eye is the same regardless of whether the
column re~resents the dominant or non-dominant eye, at Ieast for activity attributhg to a
significant and measurable fMRI response. Our results dso suggest that the dominant eye
has a larger cortical representation within the central field representation of V 1. We have
also demonstrated that high-resolution fMRI, like optical imaging, is sensitive to the very
tightly regulated vascular response on the scale of the ocular dominance column.
Functional MM, however, offers the advantage of noninvasive studies, and now, as we
have show here, can provide useful information at high resolution regarding cortical
plasticity in humans resulting from amblyopia and possibly other visual, or even
cognitive, disorden.
References
von Noorden GK. Binocular Vision and Ocular Mofility (Mosby, St . Louis, 1990).
Ciuffreda N, Levi D&i, Sçicïiuw A. Arrr$iyupiu; ausic und Ginicai Aspects
(Butterworth-Heinemann, Boston, 199 1).
Crawford MLJ, Harwerth RS, Chino YM, Smith EL. Binocularity in prism-reared
monkeys. Eye 1 O, 16 1 - 166 (1996).
Smith EL Chino YM, Ni J, Cheng H, Crawford MLJ, Harwerth RS. Residual
binocular interactions in the striate cortex of monkeys reared with abnormal binocular
vision. J Neurophysiol. 78, 13534362 (1997).
Hendrickson AE, Movshon JA, Eggers HM, Giv i MS, Boothe RG, Kiorpes L.
Effects of early unilateral blur on the macaque's visual systern. 11. Anatomical
observations. J. Neurosci. 7. 1327-1 339 (1 987).
Movshon JA, Eggers HM, Gizzi MS, Hendrickson A, Kiorpes L. Boothe RG. Effects
of early unilateral blur on the macaque's visual system: III. Phy siological
observations. J. Neurosci. 7, 1 340- 1 3 5 1 (1 987).
Kiorpes L, Kiper DC, O'Keefe LP, Cavanaugh JR, Movshon JA. Neuronal correlates
of amblyopia in the visuai cortex of macaque monkeys with experimenfal strabismus
and anisometropia. J. Neurosci. 1 8,64 1 1 -6424 ( 1 998).
Eggers HM, G i a i MS, Movshon JA. Spatial properties of striate cortical neurons in
esotropic maceques. Vis. Sci. 25, S2?8 (1984).
Fenstemaker SB, George KH. Kiorpes L, Movshon JA. Soc. Neurosci Abstr. 323,
2363 (1997).
10. Daw NW. Visual Developmenf (Plenum Press, New York, 1995).
11. Hubel DH, Wiesel RI, LeVay S. Plasticity of ocular dominance columns in monkey
striate cortex. Phil. T , w u . R. Soc. Lond (BioZ.) 278,377409 ( 1 977).
12. LeVay S, Wiesel TN, Hubel DH. The development of ocular dominance columns in
normal and visually depnved monkeys. J. Comp. Neurol. 1 9 1, 1-5 1 (1 980).
13 .Horton JC, Hocking DH, Kiorpes L. Pattern of ocular dominance columns and
cytochrome oxidase activity in a macaque monkey with naturally occumng
anisometropic amblyopia. Vis. Neurosci. 14,68 1-689 (1 997).
14. Horton JC, Stryker MP. Amblyopia induced by anisometropia without shrinkage of
ocular dominance colurnns in human striate cortex. Proc. nlarl. -4cad Sci LtS.4. 90,
5494-5498 ( 1 993).
15.Horton JC. Hocking DR. Pattern of ocular dominance columns in human striate
cortex in strabismic amblyopia. Vis. Neurosci. 13,787-795 ( 1996).
16. Florence SL, Kaas JH. Ocular dominance columns in area 17 of old world macaque
and talapoin monkeys: complete reconstmctions and quantitative analysis. Vis.
Neurosci. 8,449-462 ( 1992).
17. Rosa MGP, Ganas R Fiorani M, Soares JGM. Laminar, columnar and topographie
aspects of ocular dominance in the prîmary visual cortex of Cebus monkeys. Exp.
Brain Res. 88,249-264 ( 1 992).
18. LeVay S, Connolly M, Houde J, Van Essen DC. The complete pattern of ocular
dominance stripes in the striate cortex and visual field of the macaque monkey. J.
Neurosci. 5,486-501 ( 1985).
19. Horton JC, Hocking DR. Intrinsic variability of ocular dominance column periodicity
in normal macaque monkeys. J. Neurosci. 16,7228-7239 (1996).
2 O .Sireteanu R, Fronius M. Naso-temporal asyrnmetries in human amblyopia:
consequence of long-term interocular suppression. Vision Res. 22, 10%- 1 O63 (1 98 1).
2 1 .Hess RF, Pointer JS. Differences in the neural basis of human amblyopia: the
distribution of the anomaly across the visual field. Vision Res. 25, 1577-1 594 (1 985).
22. Porac C. Coren S. The dominant eye. Psychol. Bull. 83,880-897 (1976).
23. Purves D, White LE. Monocular preferences in binocuiar viewing. Proc. Natl. Acad.
Sci. U.S.A. 91,8339-8342 (1994).
24. della Porta GB. De Refractione: Optics Parte: Libri Novem, Ex Oflcina Horatij'
Sulvania, Apud Io Iacobum Carlinum and Anotnium Pacem (Naples, 1593).
25. Miles WR. Ocular dominance in human adults. J. Gen. Psychol. 3,4 12-420 ( 1 930).
26. Demer L, von Noorden GK, Voikow ND, Gould KL. Imaging of cerebrai blood flow
and metaboiism in arnblyopia by positron emission tomography. Am. J. Ophthamol.
105,337-347 (1988).
27.Goodyear BG, Nicolle DA, Humphrey GK, Menon RS. A Neuronal Correlate of
Suprathreshold Contrast Perception in Hurnan Amblyopia Inferred fkom Functional
Magnetic Resonance Imaging (submitted).
28. Milgrarn P. A spectacle-mounted liquid-crystal tachistoscope. Behavior Res. Methods
Inst. & Comp. 19,449-456 (1987).
2 9. Menon RS. Goodyear BG. Submillimeter fùnctional localization in hurnan striate
cortex using BOLD contrast at 4 Tesla: Implications for the vascular point-spread
function. Magn. Reson. Merl. 4 1,230-235 ( 1999).
30. Goodyear BG. Menon RS. Effect of luminance contrast on BOLD fMRI response in
human primary visual areas. J. Neurophysiol. 79,2204-2207 (1 998).
3 1. Tooteil RBH, Hadjikhani NK, Vanduffel W, Liu AK, Mendola JD, Sereno MI, Dale
AM. Functional analysis of primary visual cortex (VI) in humans. Proc. Natl. Acad.
Sci. LI.S.A. 95, 81 1-817 (1998).
32. Friston KJ, Jepard P,Tumer R. Statistical parametric maps in fùnctional imaging: A
general linear approach. Human Brain Mapping 2, 189-21 0 (1 995).
33. Grinvald A, Frostig RD, Siegel RM, Bartfeld E. High-resolution optical imaging of
fùnctional brain architecture in the awake monkey. Proc. Natl. Acad. Sci. (I.S.A. 88,
ll559-ll563 (1991).
3 4 .Malonek D, Grinvald A. Interactions between electrical activity and cortical
microcirculation revealed by irnaging spectroscopy: implications for functional brain
mapping. Science 272,55 1-554 ( 1 996).
35. Horton JC, Hocking DR. An adult-like pattern of ocular dominance columns in striate
cortex of newborn monkeys prior to visual experience. J. Neurosci. 16, 1791 -1 807
(1 996).
36.Horton JC, Dagi LR, McCrane EP, Monasterio FM. -4rrangement of ocular
dominance columns in human visual cortex. Arch. Ophthalmol. 108, 1025-103 1
(1 990).
37.Levi DM. In Spatial Vision, Regan D, ed., vol. 10, ch. 8 (CRC Press Inc., Boca
Raton, 1 99 1 ).
38.Das A. Gilbert CD. Long-range horizontal connections and their role in cortical
reorganization revealed by optical recording of cat visual cortex. Science 375, 780-
784 ( 1 995).
39. Jueptner M, Weiller C. Review: does measurement of regional cerebral blood flow
reflect synaptic activity? -- implications for PET and fMRI. Neuroimage 2, 148-156
( 1 995).
40 .Akgoren N, Mathiesen C, Rubin 1, Lauritzen M. Laminar analysis of activity-
dependent increases of CBF in rat cerebellar cortex: dependence on synaptic strength.
Am. J. Physiol. 273, H 1 166-H 1 1 76 ( 1997).
4 1. Iadecola C, Yang G, Ebner TJ, Chen G. Local and propagated vascular responses
evoked by focai synaptic activity in cerebellar cortex. J .Veurophysiol. 78. 65 1 -659
(1 997).
Chapter 8
Summary and Future Directions
8.1 Summary
8.1.1 A Quadrature RF Surface Coi! for High Resolution fMRI
The reduced MR signal within submillimeter image voxels makes high resolution
tMRi a difficult task. RF coil design is an obvious first step in improving image SNR.
Since a two-element quadrature surface coi1 provides a fi improvement in image SNR
(1) over a linear coil whose dimensions are identical to one of the coil e1eme::ts of the
quadrature coil. we chose to construct a coil that was not only quadrature, but was also
mounted to fit the curvature of the head. We have demonstrated that the mutual
inductance between the coils can be eliminated by empirical detemination of the overlap
of the two coil elements, thus increasing the overdl sensitivity and efficiency of the coil
(2).
In Ti*-weighted oblique axial images of the human visual cortex, the quadrature
s d a c e coi1 provided a substantial increase in image SNR (10 - 25% medial, 25 - 100%
lateral) compared to a linear coil whose dimensions are identical to one of the elements of
the quadrature coil. In addition, RF homogeneity was greatly improved across the slice.
As demonstrations of the capabilities of the quadrature coil, it was used in both low
resolution (3) and high resolution (4,5) fMRl studies of the human primary visual cortex.
8. l .Z The Functional Scout Image
To date, comrnon practiçe is to use anatomical features (e.g., the caicarine suicus)
rather than functional maps to prescribe image planes for MRi studies of the brain.
Using this method, areas of largest activation may be missed or partial-volumed with
unactivated cortex. Functional mapping using "on-line" reconstruction and a dedicated
cornputer has recently been demonstrated. requinng extensive hardware interfacing and
computational power (6).
We have proposed a new imaging method which provides scout maps of activity
directiy by virtue of the data acquisition scheme with no image post-processing and no
cornputational demands. Raw data collected during a desired nurnber of task and control
states are subtracted through phase altemation of the receiver between states, while the
phase of the transmitted RF is kept constant. Upon 2-D Fourier transformation of the
data, a BOLD signal difference map is produced. Double-oblique slices or voxels can
then be exactly matched to the desired activated regions using the scanner's prescription
tools*
As a demonstration of this method, we produced functional maps of visual
activity in the visual cortex resulting from binocular photic stimulation using both
conventional single-slice FLASH and multi-slice EPI sequences. Mapping, using our new
method, delineated activated areas that were qualitatively very similar to maps produced
using conventional off-line analysis techniques. This illustrates that the functional scout
image is a simple, yet extrernely powerful tool in the localization of brain function.
8.1.3 Contrast iblodulation of the BOLD Response in Human Visual Cortex
In fMRi studies of the visual cortex, a difference in stimulus contrast across
conditions may lead to false conclusions regarding the BOLD response to other
properties of the stimuli if the response is modulated by changes in stimulus contrast. To
avoid this problem, it is important to know if and where contnist is coded in the visual
cortex when designing a visual paradigm for a tMRI study.
Animal studies have demonstrated an increase in neural activity in V1 with
increasing stimulus contrast (7,8). Although, in humans, coding for local contrast has
been determined to take place in the visual pathway as early as in the retina (9), there has
been no measure of local neuronal activity as a function of local stimulus contrast within
human visual cortex. We determined the effect of contrast on the spatial extent and level
of activation in V 1 and extrastriate areas using BOLD fMRI.
Our results demonstrated that the pooled BOLD response in VI increased with
increasing stimulus contrast, supporthg previous animal studies (1 0,11). In exwstnate
cortex, however, no contrast modulation of the BOLD response was detected. This result
suggests that contrast modulation may only occur in VI, and that a simple contrast
mc?dulation functiona! eqer-iment may he usefi~l i ~ . demmating V 1 ( I I ) instead of
relying on flat-mappinglretinotopy techniques (1 3).
8.1.4 Submiliimeter Fzmctional Locaiizarion in Huntan Striate Cortex
The results of optical imaging experiments has led some to suggest that fMRI is
not capable of resolving functional units on a submillimeter scale in hurnans because the
hyperoxic vascular response of each unit spreads out over many millimeters (14). Using
stimuli of 2 or 4 second duration, optical imaging has demonstrated that the early part of
the hyperoxic response can yield well-localized functional maps (14). By attempting to
resolve ocular dominance columns (ODCs) within human VI, we explored the possibility
of imaging brain fùnction on a submillimeter scale. We took advantage of the features of
the early part of the hyperoxygenation phase of the BOLD response to a Csecond (bief)
visual stimulus.
By examining the fractional signal changes that occur in an ODC when the
corresponding eye is stimulated versus when the opposite eye is stimulated, we have
shown that these point sources of neural activity that are -700 CL m apart (15) can be
resolved using fMRI. This is an upper limit to the cortical vasculature point-spread
function in the visual cortex.
Such fuie resolution in other areas of the cortex will depend on the availability of
clever paradigms that allow differentia! rnapping. U!?imzte!y, i? is ~ h e point-spreac!
function of the cortical vasculature that will determine the limits of functional
resolvability.
8.1.5 Neuronal Correlates of Suprathreshold Contrast Perception in Human Amblyopia
A reduction in contrast sensitivity is a visual deficit commonly witnessed in
amblyopia (1 6- 18). At low suprathreshold contrasts (< 30%). amblyopic visual deficits
are related to "effective" contrast above detection threshold (19-21). Although primate
models have suggested that neural substrates of amblyopia are evident in visual areas as
early as V 1 (2 1-26), the neural basis of human amblyopia is not well investigated, nor
understood.
Positron emission tomography studies have demonstrated a decrease in global
signal within human V 1 in response to monocular stimulation of the amblyopic eye (27),
however, spatial frequency tuning of neural activity in human V1 has not been
investigated, nor has imaging been used to determine if visual cortical activity correlates
with behavioral measures of contrast perception. We used fMRI to measure the neuronal
response to monocular presentations of differing spatial frequencies at relatively low
contrast (22%) to identify a neuronal correlate of contrast perception and "effective"
contrast above threshold for the amblyopic eye.
We have demonstrated that the neural activity in early visual areas of the visual
cortex is dependent on spztki! frPq~ency, m.d is cnrre!~ted vnth percei~ed c m t m ? . !r?
addition, we have s h o w that psychophysical measurernents of behavior can actually be
correlated with fMRI data to localize cortical areas that demonstrate the behavior. The
pooled activity in early visuai areas retlects the perceptual contrast deficits in amblyopia,
however, the average localized neural activity is the same for both eyes, even in the
presence of amblyopia. These results are consistent with electrophysiological studies
demonstrating that there is no reduction in the neuronal firing rate for cells which
preferentially fire in response to stimulation of the amblyopie eye (21), and with the
notion that the reduction of contrast sensitivity in amblyopia may be due to a decrease in
neuronal recruitrnent (28).
Our results demonstrate that WRI of amblyopia is sensitive to the reduced neural
activity as a function of spatial frequency, as well as to neuronal populations whose
spatial fiequency tuning seems to be spared, making fMRl a useful tool in the
investigation of brain plasticity resulting fiom amblyopia.
8.1.6 A Neural Substrate for the Dominant Eye
The vast majority of humans preferentially use one eye to align a target in the
central visual field (29,301, however, the neural basis for this is not known. In the case of
amblyopia, the predominance of the unaffected eye may be the result of a shifi in the
ocular dominance of neural activity in V 1 (21,24-26). This could possibly have resulted
the critical penod of ODC development (3 1).
+-
Nruroimaging has not been able to identify the undrrlying neural mechanisms of
amblyopia due to poor image resolution. We investigated the correlation between the
preferred unaffected eye of human adults with amblyopia or the dominant eye of adults
with normal or corrected-to-normal visual acuity with ODC size, distribution, or activity
within the centrai visual field representation of V 1 using a high-resolution MN.
Our results suggest that the synaptic activity per tissue volume within a cortical
column when monocularly stimulated by its corresponding eye is the sarne regardless of
whether the colurnn represents the dominant/preferred or non-dominant/amblyopic eye.
In addition, the ODCs of the dominant eye cover a larger cortical area within the central
field representation of V 1. The ODCs of the amblyopic eye are significantly reduced in
size only if amblyopia was developed during infancy.
8.2 Future Directions
8.2.1 Future Directions for Contrast Modulation of the BOLD Response
As mentioned above. retinotopic paradigms are commonly used to demarcate
different visual areas in the visual cortex. To accurately perform the demarcation, the
cortex is usually displayed in a flattened representation. Presently, the software needed to
perform 'flattening' is not in place in this laboratory. When the software becomes
available, an interesting experiment would be to compare the results of luminance
contrast modulation with retinotopy to more accurately define visual areas that are
contrast-modulated and to what degree.
8.2.2 Fziture Directions for Submillimeter Functional Localization
As the magnetic field strength of MR scanners becomes greater, the more
available signal there will be within submillimeter voxels. A head gradient insert can
provide the necessary gradient fields for submillimeter imaging in far less time than
whole-body gradients. These two factors should make functional imaging on a
submillimeter scale less dificult, and may also deterrnine the lirnits of spatial resolution
of fMR.1.
8.2.3 Fictwe Directions offMRl Studies of Hman Amblyopia
As a result of the development of strabismus dunng infancy, the retinotopic
organization of the primary visual cortex may have been altered. This may be a
contnbuting factor to a decrease in the size of the ocular dominance columns of the
amblyopie eye. An obvious experiment would be to retinotopically map the primary
visual cortex of human amblyopes using monocular stimulation. This may prove to be
difficult since strabismics have difficuity fixating a target. However, monitoring of eye
movernents within our magnet environment may be possible in the near tùture.
A number of experiments investigating contrast sensitivity and suprathreshold
contrast perception can be performed that involve retinotopic mapping. It would be
interesting to see how the BOLD response to supratheshold contrast targets changes with
eccentricity, which can then be repeated at different contrasts. Behavioral measurements
of incremental contrasts could determine the amount of contrast required to perceive a
change in stimulus contrast. The results of these behavioral expenments could then be
compared to similar NRI expenments to determine if a perceived change in contrast
translates to a measurable change in the BOLD response.
The number of possible experiments involving amblyopia are too numerous to
mention. However, many basic and well-established psychophysicai results
demonstrating visual deficits in amblyopia have only speculative neural substrates.
Functional MRI would be an invaluable tool to investigate the many questions that
investigators studying human amblyopia have lefi unanswered.
Refe rences
Hoult DI, Chen CN, Sank VJ. Quadrature detection in the laboratory M e . Magn.
Reson. Med. 1: 339-353 (1984).
Roemer PB, Edelstein WA, Hayes CE, Souza SP, Mueller OM. The NMR phased
array. iuagn. Reson. Med. 16, 192-225 ( 1 989).
Goodyear BG, Nicolle DA. Humphrey GK, Menon RS. A Neuronal Correlate of
Suprathreshold Contrast Perception in Human Amblyopia Inferred from Functional
Magnetic Resonance Irnaging (submitted).
Menon RS, Goodyear BG. Submillimeter functiond localization in human striate
cortex using BOLD contrast at 4 Tesla: Implications for the vascular point-spread
function. Magn. Reson. Med. 41.230-235 ( 1 999).
Goodyear BG, Nicolle DA. Menon RS. A Neural Substrate for the Dominant Eye in
Human Amblyopia and Normal Vision (submitted).
Cox RW. Jesmonowicz A. Hyde JS. Real-time functional magnetic resonance
imaging, Magn. Reson Med. 3 3,230-236 ( 1 995).
Miyaoka M, Shinohara M, Batipps M, Pettigrew KD, Kennedy C, Sokoloff L. The
relationship between the intensity of the stimulus and the metabolic response in the
visual system of the rat. Acti. Neurof. Scand Suppl. S72, 16-1 7 ( 1 979).
Albrecht DG, and Hamilton DB. Striate cortex of monkey and cat: contrast response
function. J Physiol. J8,2 1 7-23 7 (1 982).
Wmdell BA. Foundations of I'ision. (Sunderlaid, Shz~ici, 1954), pp.176.
10. Tolhurst DJ, Dean AF. Spatial summation by simple cells in the striate cortex of the
cat. Erp. Brain Res. 66,607-620 (1 987).
1 1 . Devalois RL, Albrecht DG, Thorell LG. Spatial frequency selectivity of cells in the
macaque visual cortex. J. Opt. Soc. Am. 67,779-784 (1982).
12. Tootell RBH, Hadjikhani NK, Vanduffel W, Liu AK, Mendola SD, Sereno MI, Dale
AM. Functional analysis of primary visual cortex (VI) in humans. Proc. N d Acud.
Sci US-A. 95, 8 1 1-8 17 (1 998).
1 3. Tootell RBH, Mendola SD, Hadjikhani NK, Ledden PJ, Liu AK, Reppas JB, Sereno
MI, Dale AM. Functional analysis of V3A and related areas in human visual cortex
J Neurosci. 17,7076-7078 (1997).
1 4 .Malon& D, Grinvald A. Interactions between electrical activity and cortical
microcirculation revealed by imaging spectroscopy: implications for functional brain
mapping. Science 272,55 1-554 ( 1996).
15.Horton JC, Dagi LR, McCrane EP. Monasterio FM. Arrangement of ocular
dominance columns in human visual cortex. Arch. Ophthdmol. 108, 1025- 1 O3 1
(1 990).
16. Bradley A. Freeman RD. Contrast sensitivity in anisometropic amblyopia. Invesf.
Ophrhamol. Vis. Sei. 2 1,467-476 (1 98 1).
17. Hess RF, Campbell FW, Greenlaugh T. On the nature of the neural abnormality in
human amblyopia: Neural aberrations and neural sensitivity loss. PfIuegers Arch.
Gesamte. Physiol. 337,20 1-207 ( 1978).
18.Hess RF, Pointer JS. Differences in the neural basis of human amblyopia: the
distribution of the anomaly across the visuai field. Vis. Res. 25, 1577-1 594 (1985).
19. Crawford MLJ, von Noorden GK. Optically induced concommitant strabismus in
monkeys. Invesf. ûphthamol. Vis. Sci. 19, 1105-1 109 (1980).
20. Movshon JA, Eggers HM, Gin5 MS, Hendrickson A, Kiorpes L, Boothe RG. Effects
of early unilateral blur on the macaque's visual system: III. Physiological
observations. .L Neurosci. 7, 1340-1 35 1 (1987).
21. Kiorpes L, Kiper DC, O'Keefe LP, Cavanaugh JR, Movshon JA. Neuronal correlates
of amblyopia in the visual cortex of macaque monkeys with experimental strabismus
and anisometropia. J. Neurosci. 18,64 1 1-6424 ( 1998).
22. Crawford MLJ, Harwerth RS, Chino YM, Smith EL. Binocularity in prism-reared
monkeys. Eye 10, 161-166 (1996).
23.Smirh EL Chino YM, Ni J, Cheng H, Crawford MLJ, Hanverth RS. Residual
binocular interactions in the striate cortex of monkeys reared with abnormal binocular
vision. J Neurophysiol. 78, 1353- 1362 (1 997).
24Hendrickson AE, Movshon JA, Eggirs HM, G i u i MS, Boothe RG, Kiorpes L.
Effects of early unilateral blur on the macaque's visual system. II. Anatornical
observations. J. Neurosci. 7, 1327-1 339 ( 1 987).
25. Movshon JA, Eggers HM, Gizzi MS, Hendrickson A, Kiorpes L, Boothe RG. Effects
of ear ly unilateral blur on the macaque's visual system: III. Phy siological
observations. J. Neurosci. 7, 1340- 135 1 ( 1 987).
26. Eggers HM, Gizzi MS, Movshon JA. Spatial properties of striate cortical neurons in
esotropic macaques. Vis. Sci. 25, S278 (1984).
27. Demer L. von Noorden GK, Volkow ND, Gould KL. Imaging of cerebral blood tlow
and metabolism in amblyopia by positron emission tomography. Am. J. Ophrharnol.
105,337-347 (1988).
28. Levi DM. In Spatial Vision, Regan D. ed., vol. 10, ch. 8 (CRC Press Inc., Boca
Raton, 1 99 1 ).
29. Porac C, Coren S. The dominant eye. Psychol. Bull. 83,880-897 (1976).
30. Pwes D, White LE. Monotulm preferences in binocul=r vievring. Proc. nlatl. Acad.
Sci. U.S.A. 91,833998342 (1994).
3 1. Daw NW. Visual Development (Plenum Press, New York, 1995).
Appendix A
Time Courses of MR Signal during Binocular Photic
Stimulation using Different Luminance Levels
Figure A-1 shows time courses of the average £MRI signal within the selected
ROIs shown in Figure 4- 1 , demonstmting that in V 1 the tMRi signal nses to higher levels
when photic stimulation is at a higher luminance, whereas in extrastriate cortex this is not
the case.
a Y -z 3 S= lncreasing u = C 3 II LED
intensity
-
image number Image number
Figure A-1: The average fMRI signal within the ROIs of Figure 4-1 for (a) V1 and (b)
extrastriate cortex. The tirne courses are stacked for clarity. In both (a) and (b), the
highest LED intensity is represented by the bottom-most trace, with LED intensity
increasing as one moves down the figure.
Appendix B
Optimizing MR and Visual Stimulus Parameters for High
Resolution fMRI Studies of Ocular Dominance
As demonstrated in Chapter 7, the ocular dominance columns of the dominant eye
occupy significantly more temtory of the central visual field representation of the
primary visual cortex (area V l c). Resolving ocular dominance columns and making
conclusions about their distribution using MRI would be impossible without careful
optimization of imaging and stimulus parameters.
Figure B-l(a) shows how the relative area of V lc occupied by each eye is
dependent on the visual stimulus duration, maximizing at approxirnately 4 seconds. Four
seconds is short enough to avoid saturation of the BOLD response, yet long enough to
reliably detect a difference between the BOLD response within a voxel during lefi-eye
and right-eye monocular stimulation. Figure B-I(b) shows the same dependence on the
number of activated voxels passing the ststistical threshold.
To demonstrate that the image resolution used in our study was sufficient, we
investigated the ratio of the cortical area occupied by dominant eye colurnns to the
cortical area occupied by non-dominant eye columns as a function of image voxel size.
With the existing data, we applied Gaussian filters of varying widths upon image
reconstruction to emulate a range of voxel sizes, and then reapplied our analysis for
producing maps of ocular dominance columns. The result is shown in Figure 8-2. As the
voxel size becomes larger. the ratio of occupied area tends to 1. and any bias towards the
dominant eye has been obliterated. Only when the voxel size falls below 1 mm does the
dominant eye bias begin to become apparent, consistent with reported sizes of human
ocuiar dominance columns. The fitted line extrapolates to 1.1 87 at infinite resolution.
This corresponds to 54.396 of the cortical area being occupied by the dominant eye. Our
reported value of 54.0% at an image resolution of 0.55 mm was within the standard error
of the mean of the projected value, demonstrating that the voxel size used in this study
was sufficient to make this determination.
2 4 6 nimulua dut adontr) stimulation du ration <SI
Figure B-1: (a) Ratio of the area of Vlc occupied by the dominant eye (D) to the area
occupied by the non-dominant eye (ND), for two subjects with nomal vision. (b)
Number of voxels within functional maps of ocular dominance (i.e., number of voxels
passing the statistical threshold), normalized to the number of voxels representing the
ODCs of the dominant eye in the map created for a stimulus duration of 4 seconds.. For
both (a) and (b), four separate irnaging experiments were performed, each with 4 lefi-eye
and 4 right-eye monocular stimulation periods. All maps were created using the same
technique as descnbed in Chapter 7. Enor bars represent the standard error of the mean.
Figure 8-2: The ratio of cortical area occupied by dominant eye colurnns to non-
dominant eye columns as a function of image voxel size. Error bars represent the
standard error of the mean for 11 subjects with nomal vision. The fitted cuve has been
extrapolated to an infinite number of voxels, projecting a ratio of 1.1 87 (i.e., 54.3%
dominant eye, 46.7% non-dominant eye).
Figure B-3(a) demonstrates that collecting images using a 14-cm FOV, 16-shot
(or segment) or an 8-cm FOV, 8-shot EPI sequence perform equally well at
demonstrating the bias of the dominant eye. Both of these techniques involve the
collection of 16 echoes d e r each RF pulse. Figure B-3(a) also shows that a 14-cm FOV,
8-shot or an 8-cm FOV, 4-shot EPI sequence (Le., 32 echoes per RF pulse) does not
demonstrate such a bias. In this case, the ocular dominance columns cannot be resolved
due to T2* bluning (see Chapter 2). Figure B-3(b) shows that the 8-cm FOV, 8-shot EPI
sequence further decreases the arnount of T2* blurring, resulting in more voxels passing
the statistical threshold of significance. Mthough the 8-cm FOV, &hot and 14-cm FOV,
16-shot sequence both involve 16 echoes per RF pulse, the 8-cm FOV, 8-shot sequence
need only collect 128 frequency-encode data points instead of 256.
14cmi16 shot
t4cmf t6 14cmt8 shot stwt
8 cm/8 thot
B cm/4 shot
Figure B-3: (a) Percentage of cortical area of Vlc occupied by the ocular dominance
columns of the dominant eye using 4 different EPI sequences for two subjects and using a
4-second visual stimulus. (b) Nurnber of activated voxels in the rnaps of ocular
dominance for the same EPI sequences in (a).
scaled version of any other, reflecting the difference in the density of the ocular
dominance maps. A 2 cyclesldegree visual stimulus provided the greatest number of
activated voxels, and hence the most dense maps of the ocular dominance column
distribution within V 1 c.
Appendix D
Contrast and Luminance Calibration of the Projection Screen
The visual stimuli used in this thesis were projected through a mesh-screen
window onto a screen placed at the edge of the magnet bore. The transmission properties
of the mesh and the projection screen lead to a loss or gain in luminance znd contrast.
This must be carefully evaluated if any inferences are to be made regarding contrast and
luminance modulation of the BOLD response in the visual cortex, since the contrast
provided by the computer controlling the visual stimuli will not be the same as the
contrast on the projection screen.
Figure D-l(a) shows how the contrast and luminance at the projection screen
differs from the contrast of the CRT screen of the stimulus-controlling computer. By
adding a DC offset to the output luminance to the projector at each contrast level, the
mean luminance as a function of contrast can be held constant. This was repeated at a
nurnber of luminance levels of the CRT screen (20.40,60 cd/m2).
0.0 0.2 0.4 0.6 0.8 1.0 3 CRT Contmt a 0.0 0.2 0.4 0.6 0.8 1.0
CRT Contrast .Y
Figure D-1: (a) The contrast and luminance at the projection screen as a function of the
contrast of the CRT of the stimulus-controlling computer measured using a Minolta CS-
100 Chroma Meter (Minolta Camera Co., Ltd., Japan). The luminance of the CRT was
kept constant at 60 cdm'. (b) Same as in (a), except a DC offset was added or subtracted
to the luminance at each contrast level to maintain a mean luminance that did not Vary as
a f ic t ion of contrast by more than 2%. Similar curves and calibrations were performed
at a luminance of 20 cdm2 and 40 cd/m2.
