7
Original Research Auto-SENSE Perfusion Imaging of the Whole Human Heart Herbert Ko ¨ stler, PhD, * Jo ¨ rn J.W. Sandstede, MD, Claudia Lipke, MD, Wilfried Landschu ¨ tz, PhD, Meinrad Beer, MD, and Dietbert Hahn, MD Purpose: To show the application of auto-sensitivity en- coding (SENSE)—a self-calibrating parallel imaging tech- nique—to first pass perfusion imaging of the whole human heart. Materials and Methods: The self-calibrating parallel imag- ing method auto-SENSE was implemented for a saturation recovery turbo-fast low-angle shot (FLASH) sequence on a 1.5-T scanner using a standard four-element body phased array coil. By reducing the acquisition time per slice by a factor of two compared to conventional turbo FLASH imaging, the number of imaged slices could be doubled to six to ten with an unchanged temporal resolution of one image per heartbeat. This technique has been tested in eight healthy volunteers for contrast-enhanced heart perfusion imaging. Results: Auto-SENSE heart perfusion imaging with im- proved coverage of the human heart could be performed successfully in all volunteers. A first quantitative compar- ison of perfusion values between the auto-SENSE and the non-SENSE techniques shows good agreement. Conclusion: Auto-SENSE allows perfusion imaging of the whole human heart without gaps. Key Words: magnetic resonance imaging; myocardial per- fusion; parallel MRI; SENSE; ultrafast imaging J. Magn. Reson. Imaging 2003;18:702–708. © 2003 Wiley-Liss, Inc. DIFFERENT PARTIAL PARALLEL imaging methods have been demonstrated that increase the speed of magnetic resonance (MR) acquisition by decreasing the number of phase encoding steps, e.g., simultaneous acquisition of spatial harmonics ( SMASH ), sensitivity encoding (SENSE), and sensitivity profiles from an ar- ray of coils for encoding and reconstruction in parallel (SPACE RIP) (1–3). These methods use the sensitivity of the elements of phased array coils as prior knowledge to reconstruct missing information due to skipped phase encoding steps. In most techniques, coil sensitivity maps are acquired in extra scans before the acquisition of the parallel images. For k-space based methods, a principle of self calibration has been proposed (4). In the extensions of this technique (5,6) the center of k- space is densely filled, while for high spatial frequencies the sampling density is reduced. Recently, the same sampling scheme has been used in combination with a generalized encoding matrix (GEM) reconstruction (7). Self-calibration avoids errors that occur if the coil sen- sitivities change between the moment when the sensi- tivity maps are acquired and the actual scan. However, compared to the methods with an extra calibration scan, the self-calibrating techniques need to acquire extra lines in k-space, i.e., the lines to fill the center of k-space densely. Therefore, these self-calibrating tech- niques in general do not reach the same maximum acceleration factor as the techniques where the coil sensitivities are already known before the scan. A sampling strategy in which distinct sets of under- sampled data, e.g., even and odd lines in k-space, are acquired in an alternating fashion was proposed by Ma- dore et al (8). In the UNFOLD (unaliasing by Fourier- encoding the overlaps using the temporal dimension) re- construction (8), full field of view images can be obtained by temporal filtering. The alternating reduced field of view data sets may also be used for reconstruction by parallel imaging methods (9,10) or by a combination of UNFOLD and parallel techniques, i.e., TSENSE (11). First-pass contrast-enhanced myocardial perfusion MR imaging was first proposed in 1990 (12). Since then, a number of technical improvements of the acquisition sequences have improved the clinical practicability (13). However, for clinical usage, there still exists the need for a robust sequence covering the whole heart. In this work, first-pass perfusion imaging of the whole human heart without gaps is shown. This is achieved using the parallel imaging technique auto-SENSE, a technique that reconstructs full field of view images from alternating reduced field of view data sets. Re- duced scan time per image is used for improved cover- age of the heart, maintaining a temporal resolution of one image per heartbeat for all slices. MATERIALS AND METHODS All experiments were carried out on a clinical 1.5-T whole-body scanner (Magnetom Vision, Siemens Medi- Institut fu ¨ r Ro ¨ntgendiagnostik, Universita ¨t Wu ¨ rzburg, Josef-Schneider Str. 2 – Bau 24, 97080 Wu ¨ rzburg. *Address reprint requests to: H.K., Institut fu ¨ r Ro ¨ntgendiagnostik, Uni- versita ¨t Wu ¨ rzburg, Josef-Schneider Str. 2 – Bau 24, 97080 Wu ¨ rzburg. E-mail: [email protected] Received March 3, 2003; Accepted August 21, 2003. DOI 10.1002/jmri.10419 Published online in Wiley InterScience (www.interscience.wiley.com). JOURNAL OF MAGNETIC RESONANCE IMAGING 18:702–708 (2003) © 2003 Wiley-Liss, Inc. 702

Auto-SENSE perfusion imaging of the whole human heart

Embed Size (px)

Citation preview

Original Research

Auto-SENSE Perfusion Imaging of the WholeHuman Heart

Herbert Kostler, PhD,* Jorn J.W. Sandstede, MD, Claudia Lipke, MD,Wilfried Landschutz, PhD, Meinrad Beer, MD, and Dietbert Hahn, MD

Purpose: To show the application of auto-sensitivity en-coding (SENSE)—a self-calibrating parallel imaging tech-nique—to first pass perfusion imaging of the whole humanheart.

Materials and Methods: The self-calibrating parallel imag-ing method auto-SENSE was implemented for a saturationrecovery turbo-fast low-angle shot (FLASH) sequence on a1.5-T scanner using a standard four-element body phasedarray coil. By reducing the acquisition time per slice by afactor of two compared to conventional turbo FLASH imaging,the number of imaged slices could be doubled to six to tenwith an unchanged temporal resolution of one image perheartbeat. This technique has been tested in eight healthyvolunteers for contrast-enhanced heart perfusion imaging.

Results: Auto-SENSE heart perfusion imaging with im-proved coverage of the human heart could be performedsuccessfully in all volunteers. A first quantitative compar-ison of perfusion values between the auto-SENSE and thenon-SENSE techniques shows good agreement.

Conclusion: Auto-SENSE allows perfusion imaging of thewhole human heart without gaps.

Key Words: magnetic resonance imaging; myocardial per-fusion; parallel MRI; SENSE; ultrafast imagingJ. Magn. Reson. Imaging 2003;18:702–708.© 2003 Wiley-Liss, Inc.

DIFFERENT PARTIAL PARALLEL imaging methodshave been demonstrated that increase the speed ofmagnetic resonance (MR) acquisition by decreasing thenumber of phase encoding steps, e.g., simultaneousacquisition of spatial harmonics ( SMASH ), sensitivityencoding (SENSE), and sensitivity profiles from an ar-ray of coils for encoding and reconstruction in parallel(SPACE RIP) (1–3). These methods use the sensitivity ofthe elements of phased array coils as prior knowledge toreconstruct missing information due to skipped phaseencoding steps. In most techniques, coil sensitivity

maps are acquired in extra scans before the acquisitionof the parallel images. For k-space based methods, aprinciple of self calibration has been proposed (4). Inthe extensions of this technique (5,6) the center of k-space is densely filled, while for high spatial frequenciesthe sampling density is reduced. Recently, the samesampling scheme has been used in combination with ageneralized encoding matrix (GEM) reconstruction (7).Self-calibration avoids errors that occur if the coil sen-sitivities change between the moment when the sensi-tivity maps are acquired and the actual scan. However,compared to the methods with an extra calibrationscan, the self-calibrating techniques need to acquireextra lines in k-space, i.e., the lines to fill the center ofk-space densely. Therefore, these self-calibrating tech-niques in general do not reach the same maximumacceleration factor as the techniques where the coilsensitivities are already known before the scan.

A sampling strategy in which distinct sets of under-sampled data, e.g., even and odd lines in k-space, areacquired in an alternating fashion was proposed by Ma-dore et al (8). In the UNFOLD (unaliasing by Fourier-encoding the overlaps using the temporal dimension) re-construction (8), full field of view images can be obtainedby temporal filtering. The alternating reduced field of viewdata sets may also be used for reconstruction by parallelimaging methods (9,10) or by a combination of UNFOLDand parallel techniques, i.e., TSENSE (11).

First-pass contrast-enhanced myocardial perfusionMR imaging was first proposed in 1990 (12). Since then,a number of technical improvements of the acquisitionsequences have improved the clinical practicability(13). However, for clinical usage, there still exists theneed for a robust sequence covering the whole heart. Inthis work, first-pass perfusion imaging of the wholehuman heart without gaps is shown. This is achievedusing the parallel imaging technique auto-SENSE, atechnique that reconstructs full field of view imagesfrom alternating reduced field of view data sets. Re-duced scan time per image is used for improved cover-age of the heart, maintaining a temporal resolution ofone image per heartbeat for all slices.

MATERIALS AND METHODS

All experiments were carried out on a clinical 1.5-Twhole-body scanner (Magnetom Vision, Siemens Medi-

Institut fur Rontgendiagnostik, Universitat Wurzburg, Josef-SchneiderStr. 2 – Bau 24, 97080 Wurzburg.*Address reprint requests to: H.K., Institut fur Rontgendiagnostik, Uni-versitat Wurzburg, Josef-Schneider Str. 2 – Bau 24, 97080 Wurzburg.E-mail: [email protected] March 3, 2003; Accepted August 21, 2003.DOI 10.1002/jmri.10419Published online in Wiley InterScience (www.interscience.wiley.com).

JOURNAL OF MAGNETIC RESONANCE IMAGING 18:702–708 (2003)

© 2003 Wiley-Liss, Inc. 702

cal systems, Erlangen Germany) using the standardfour-element body phased array coil.

The study was approved by our institution’s ethicscommittee, and written informed consent was obtainedfrom all volunteers. Auto-SENSE perfusion imagingwas tested in eight healthy volunteers.

Imaging Parameters

First-pass perfusion images were acquired with a mod-ified arrhythmia insensitive multislice, saturation re-covery turbo-fast low-angle shot (FLASH) imaging tech-nique (14). A bolus of Gd-DTPA contrast agent (0.025mmol/kg bodyweight, flow 4 mL/second) was automat-ically injected into an antecubital vein followed by aflush of 20 mL saline using a power injector (Medrad,Volkach, Germany). The following sequence parameterswere chosen: repetition time 2.4 msec, echo time 1.2msec, delay between saturation and begin of FLASHsequence 10.0 msec, flip angle 8° or 18°, partial echoacquisition (last 80 points of a symmetric 128-pointecho), number of phase encoding steps 24–40 depend-ing on the heart rate, field of view (FOV) 300–400 mmwith reduction to 3/8 in phase encoding direction forthe aliased images, and slice thickness 8–10 mm. Thetime saved by acquisition of a reduced FOV was used tomeasure 6–10 slices per heartbeat. No gaps were leftbetween the slices. Forty to 62 consecutive images wereacquired for each location during 40–62 heartbeats.The image plane was chosen double oblique corre-sponding to a short axis view of the heart. Compleximages were calculated for every coil element and everyslice at every heartbeat. The phase encoding lines weresampled in reverse centric order, i.e., the outer lines ink-space, are sampled immediately after the saturationpulse and the center of k-space after about 80 msec(Fig. 1). Thus, the contrast of the originally proposedsequence (14) could be maintained.

The point spread functions in phase encoding direc-tion of the original saturation recovery turbo-FLASHsequence and the auto-SENSE sequence were calcu-lated assuming linear signal increase starting at zerofrom the first to the last sampled line in k-space. Thisassumption is valid for saturation recovery turbo-FLASH sequences with small flip angles as long as theacquisition time for an image is much shorter than theT1-relaxation time. The magnitude of the calculatedpoint spread functions of the saturation recovery se-quences with linear order (used for the non-SENSEsequence) and reverse centric order (used for auto-SENSE) of the phase encoding steps are shown in Fig.2. From this simulation, it can be seen that the reversecentric order in a saturation recovery sequence acts like

Figure 1. Graphic representation of the auto-SENSE perfusion imaging sequence: after electrocardiogram (ECG) triggering, allspins are saturated by a 90°-pulse. Then a FLASH module with reverse centric ordering of the phase encoding steps acquires animage of the first slice. After new saturations, additional slices are imaged during the same heart beat. Depending on the heartrate, between six and 10 slices can be chosen. From one heartbeat to the next, the acquired k-space lines are shifted accordingto the dynamic auto-SENSE sampling scheme.

Figure 2. Point spread functions (PSF) for saturation recoverysequences with different order of the phase encodings. Themagnitude of the PSF is displayed, assuming a linear signalincrease and either linear order (used for non-SENSE perfu-sion imaging) or reverse centric order (used for auto-SENSE) ofthe phase encodings.

Auto-SENSE Myocardial Perfusion Imaging 703

a low-pass filter. Consequently, it dramatically reducesthe side-lobes of the point spread function at the ex-pense of a moderate broadening of the main lobe.

At every second heart beat, all k-space lines wereshifted by half of the distance between two adjacentk-space lines (Fig. 3). This data acquisition scheme fordynamic auto-SENSE corresponds to that proposed forthe UNFOLD technique (8).

Determination of Coil Sensitivity Maps

Auto-SENSE differs from SENSE (2) and TSENSE (11)in the way the necessary coil sensitivities are deter-mined. For auto-SENSE images have to be chosen inthe time domain where the object does not change ormove. With the information of these images, the k-space was filled densely and an unaliased image withfull FOV was reconstructed for every coil element.These images of the different coil elements were com-bined to one image by using the “sum of squares.”Relative coil sensitivities were determined by dividingthe images of the coil elements by the combined image.These relative coil sensitivities were sufficient for aSENSE reconstruction, because the spatial scaling ofthe local coil sensitivities is transferred to the finalimage, but does not lead to any reconstruction artifactsas long as this scaling is the same for all coil elements(1,15). These time-independent coil sensitivities al-lowed Cartesian SENSE reconstruction of unaliasedimages from each set of small-FOV-images from thedifferent coil elements.

To determine maps of the relative coil sensitivities,two threshold values were used in our study. Thehigher threshold value was chosen lower than the min-imum pixel value of the myocardium, and the lowerthreshold lower than the minimum of the lung in thecombined full FOV image. A first sensitivity map wascalculated on a pixel-by-pixel basis by dividing the im-ages of the coil elements by the combined image. Toobtain a second, “smooth” map, images were filtered,i.e., the k-space was multiplied by a radial exponentialdecaying function with a decay to 1/e in 1/16 kmax. Ifthe pixel value in the combined image of all coil ele-ments was above the higher threshold value (i.e., in theheart, liver, and chest wall), the value of the “unfiltered”map was chosen for the final coil sensitivity map. Be-tween the two threshold values (i.e., in the lung) thevalue of the “filtered” map was taken. The coil sensitiv-ity values were set to zero if the signal intensity was

below the lower threshold value (i.e., noise) in the com-bined image. No other techniques such as local polyno-mial fitting or extrapolation were necessary. The result-ing coil maps of all four coil elements of one slice aredisplayed in Fig. 4.

Comparison to Standard Perfusion Imaging

In one volunteer, the perfusion exam was repeated fourtimes with a delay of 15 minutes between contrastagent injections. The first and third exams were per-formed using the auto-SENSE sequence (six sliceswithout gaps, reverse centric order of phase encodingsteps) while the second and fourth exams used theoriginal non-SENSE sequence (three slices, gap 10 mm,linear order of phase encoding steps) (14). Low dose ofcontrast agent (0.025 mmol/kg bodyweight) was usedto allow for repeated contrast medium applications andfor quantification.

For quantitative comparison, normalized signal timecourses were obtained; to remove contamination,curves proportional to the signal time courses in theventricles were subtracted to minimize the variation ofthe baseline signal, i.e., the signal before the arrival ofthe contrast agent in the myocardium. Then, the timecourses were converted to curves of relative signalchange by dividing the measured signal time course by

Figure 3. Dynamic auto-SENSE k-space sampling for an acceleration factor of 2: the sparsely sampled k-space is shifted by halfthe distance between two adjacent lines from odd to even acquisitions. By combining odd and even acquisitions, a densely filledk-space can be obtained.

Figure 4. Relative B1-maps of the four coil elements for oneslice.

704 Kostler et al.

the baseline signal. The perfusion was calculated byconstrained deconvolution (Fermi function as resid-uum) (14) of the relative signal change time courseswith the relative signal change time course of the leftventricle. Eight sectors of the mid-ventricular slice wereanalyzed for the auto-SENSE and the non-SENSE ex-ams.

In these sectors, the SD of the curves of relative signalchange before the arrival of the contrast agent wascalculated as an experimental determination of thenoise in the myocardium.

RESULTS

Auto-SENSE imaging of the human heart could beperformed successfully in all eight volunteers. As allvolunteers could hold their breath for 40 heartbeats,dense k-spaces and resulting full FOV images werefilled from all 40 acquisitions at the same slice. UsingCartesian SENSE reconstruction with an increase ofthe FOV by a factor of two in phase encoding direc-tion, aliasing could be removed in all perfusion im-ages. Figure 5 shows an example of a perfusion examwith 10 slices, which allows a good coverage of thewhole human heart. All 10 images were acquired dur-ing one heartbeat. The perfusion images of one sliceover all 40 acquired heartbeats are shown in Fig. 6.There, the passage of the contrast agent through theright and left ventricle and the myocardium can beseen. The signal time courses in eight sectors of themyocardium of an auto-SENSE examination (thirdperfusion exam) corrected for contamination from theventricles and converted to relative signal changecurves are shown in Fig. 7 (open squares). For com-parison, the corresponding curves of a separate exam(second perfusion exam) using a standard perfusionimaging sequence (non-SENSE, solid circles) for thesame slice in the same volunteer are also displayed.As expected, both methods gave similar time coursesand up-slopes. The quantitative results of the deter-mination of the perfusion are shown in Table 1 andFig. 8. Both auto-SENSE and non-SENSE revealed

similar results regarding means and SDs of the per-fusion value: 1.24 � 0.10 and 1.27 � 0.10 for auto-SENSE and 1.19 � 0.05 and 1.16 � 0.14 for non-SENSE.

Noise in baseline images determined as SD of relativesignal change in the baseline images was 5.99% and4.80% for auto-SENSE and 3.08% and 2.16% for thenon SENSE examinations.

DISCUSSION

Auto-SENSE was implemented for heart perfusion im-aging on a clinical scanner using a standard bodyphased array coil. In all exams, the scan planes wereindividually adjusted double oblique, parallel to theshort axis of the heart. Although the coil was not spe-cially built for these orientations, successful auto-SENSE reconstruction was possible in all exams.

Auto-SENSE uses coil sensitivities acquired during aseries of dynamic scans for a SENSE reconstruction. Byinterleaving the acquisition of different sparse parts ofthe k-space, a dense k-space can be filled. Coil sensi-tivities can be extracted from a subset of the dynamicseries where no changes occur. In this study, all volun-teers could hold their breath for 40 heartbeats. There-fore, the whole series could be used to determine thecoil sensitivities. In auto-SENSE, no extra lines in k-space are necessary for the individual scans and thefull maximum acceleration factor of SENSE can bereached. In contrast to SENSE imaging, auto-SENSEdoes not need a separate prescan to determine the coilsensitivities. Therefore, auto-SENSE combines the ad-vantages of the self-calibrating and traditional calibra-tion methods: the elimination of miscalibration due tomovements between the two scans and achievement ofthe full maximum acceleration factor.

The signal-to-noise ratio (SNR) for the calculation ofcoil sensitivities is high in this auto-SENSE perfusionimaging study because the information of all 40 imagescontributes to the time independent full FOV images.Artifacts from Gibbs ringing may occur when the sen-sitivity maps are generated only from the inner part of

Figure 5. Auto-SENSE reconstructed images of 10 slices without gaps acquired during one heart beat.

Auto-SENSE Myocardial Perfusion Imaging 705

the k-space, which does not represent the full resolu-tion of the magnetization distribution. Dynamic auto-SENSE allows the generation of the sensitivity map atthe same resolution as the reconstructed images andthus eliminates these artifacts, assuming no motionoccurs at the frequency of the acquisition of full k-spaces.

Auto-SENSE perfusion imaging additionally im-proves the SNR of the coil sensitivity determinationbecause of the enhancement of the MR signal duringthe passage of the contrast agent. This allows the directdetermination of the coil sensitivities even in regionswhere the achieved MR signal is usually insufficient, forexample in the lungs.

The UNFOLD reconstruction (8) relies on temporalfiltering of aliased pixels and has been successfullytested in situations where a large part of the image doesnot change intensity. As artifacts, e.g., caused bybreathing or patient movement, and the passage of acompact contrast agent bolus are not confined to a

Figure 6. Auto-SENSE perfusion exam: 40 auto-SENSE reconstructed images of one slice acquired during 40 consecutive heartbeats showing the passage of the contrast agent through the right and left ventricle and the myocardium.

Figure 7. Normalized signal time courses of eight sectors ofthe myocardium acquired using the auto-SENSE sequence(open squares) or the standard saturation recovery turbo-FLASH sequence (non-SENSE, solid circles). Both sequencesshow similar results, especially a similar up-slope.

706 Kostler et al.

certain range of temporal frequencies, but more often toa limited period of time, the choice of filters may beextremely difficult in an UNFOLD perfusion exam ofsick patients. The same problem in choosing the rightfilter in the frequency domain arises for TSENSE (11). Infirst pass perfusion imaging, it may be easier to selectimages without artifacts in the time domain to deter-mine the coil sensitivities.

Temporal filtering after auto-SENSE reconstructiondoes not seem to be necessary for heart perfusion im-aging. Image artifacts such as those presented by Kell-man et al (11) or Madore (16) for SENSE reconstructionwere not noticed in our study. This might be due to theimproved determination of the coil sensitivities espe-cially in the lungs.

Although MR heart perfusion imaging was alreadyproposed in 1990 (12), its clinical impact is still ratherlimited. One reason is the incomplete coverage of theheart. Up to now, only one to four slices could be ac-quired per heart beat with the commonly used turbo-FLASH sequences (13,14), meaning that a completecoverage of the heart without gaps was not possiblewith a single contrast agent administration. To improveanatomic coverage, echo planar imaging (EPI)-hybridsequences have been proposed (17–21). Auto-SENSEheart perfusion imaging also allows a complete cover-age of the human heart but avoids EPI readouts andEPI related artifacts.

The parameters used for auto-SENSE perfusion im-aging were chosen to give the same contrast as thetechnique proposed by Jerosch-Herold for quantitativeheart perfusion imaging (14): low dose of contrast agentwas used to be able to determine the arterial inputfunction in the ventricle. The center of k-space wassampled at the same time after the saturation as in the

standard sequence. This was achieved by using reversecentric ordering of lines in k-space for auto-SENSE andlinear ordering in the standard sequence. As a conse-quence, the point spread function in phase encodingdirection of the auto-SENSE perfusion study showsdrastically reduced side lobes, which leads to reducedcontamination from distant pixels compared to thestandard sequence. This means that contaminationfrom the ventricles to the myocardial signal timecourses should be less in auto-SENSE perfusion imag-ing compared to linear ordered saturation recovery im-aging. The improved point spread function of centricordering might also be used for conventional saturationrecovery imaging. But this would lead to a delay insampling of the center of k-space and consequently to achange in image contrast and a decrease of the linearrange between contrast agent concentration and signalincrease.

The results of the quantitative perfusion evaluationshow comparable perfusion values for the original andthe auto-SENSE technique. The image noise is higherin this auto-SENSE study by a factor of about two. Ifdedicated coils are used, and consequently, the g-factorfor the parallel imaging is reduced, the theoreticallypossible factor of �2 might be achieved. The determinedperfusion values are similar in Auto-SENSE and nonSENSE studies and the standard deviation of the per-fusion values is acceptable. This should be further as-sessed in future studies, but it fits to the result of atheoretical study which states that temporal and spa-tial resolution, but not image noise are the limitingfactors in MR heart perfusion imaging (22). On theother hand, reduced SNR in parallel perfusion imagingmay be compensated in future studies by using higherfield strength, pulse sequences with improved signal tonoise ratio (23) or techniques that allow the use ofhigher contrast agent doses (24,25).

CONCLUSION

Auto-SENSE allows contrast agent enhanced satura-tion recovery turbo FLASH MR perfusion imaging withcomplete coverage of the human heart without gapswith a temporal resolution of one image per heartbeat.This is achieved by reduced scan time per image whilespatial and temporal resolutions are preserved. Auto-SENSE is a parallel imaging technique that allowsSENSE reconstruction without the need for a separateprescan to determine the coil sensitivities. For dynamicstudies, the scan time per image is identical to that ofthe SENSE technique and no extra lines in k-spacehave to be acquired.

Table 1Perfusion Values in Eight Sectors of a Mid-ventricular Slice in Four Perfusion Examinations of a Volunteer

Sector 1 Sector 2 Sector 3 Sector 4 Sector 5 Sector 6 Sector 7 Sector 8 Mean SD

Exam 1, Auto-SENSE 1.27 1.27 1.41 1.08 1.28 1.18 1.23 1.15 1.24 0.10Exam 2, Non-SENSE 1.19 1.14 1.18 1.14 1.23 1.27 1.12 1.23 1.19 0.05Exam 3, Auto-SENSE 1.19 1.23 1.21 1.25 1.36 1.24 1.49 1.20 1.27 0.10Exam 4, Non-SENSE 1.30 1.14 1.09 0.92 1.06 1.12 1.29 1.35 1.16 0.14

Figure 8. Perfusion values in eight sectors of the myocardiumin four consecutive perfusion examinations of a volunteer.Auto-SENSE and the non-SENSE method shows comparablemean values and SDs.

Auto-SENSE Myocardial Perfusion Imaging 707

ACKNOWLEDGMENT

The authors acknowledge Mark Griswold for helpfulcomments and discussion.

REFERENCES

1. Sodickson DK, Manning WJ. Simultaneous acquisition of spatialharmonics (SMASH): fast imaging with radiofrequency arrays.Magn Reson Med 1997;38:591–603.

2. Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P. SENSE:sensitivity encoding for fast MRI. Magn Reson Med 1999;42:952–962.

3. Kyriakos WE, Panych LP, Kacher DF, et al. Sensitivity profiles froman array of coils for encoding and reconstruction in parallel (SPACERIP). Magn Reson Med 2000;44:301–308.

4. Jakob PM, Griswold MA, Edelman RR, Sodickson DK. AUTO-SMASH: a self-calibrating technique for SMASH imaging. Simulta-neous acquisition of spatial harmonics. MAGMA 1998;7:42–54.

5. Heidemann R, Griswold M, Haase A, Jakob PM. VD-AUTO-SMASHimaging. Magn Reson Med 2001;45:1066–1074.

6. Griswold MA, Jakob PM, Heidemann RM, et al. Generalized auto-calibrating partially parallel acquisitions (GRAPPA). Magn ResonMed 2002;47:1202–1210.

7. McKenzie CA, Yeh EN, Ohliger A, Price MD, Sodickson DK. Selfcalibrating parallel imaging with automatic coil sensitivity extrac-tion. Magn Reson Med 2002;47:529–538.

8. Madore B, Glover GH, Pelc NJ. Unaliasing by Fourier-encoding theoverlaps using the temporal dimension (UNFOLD), applied to car-diac imaging and fMRI. Magn Reson Med 1999;42:813–828.

9. Kostler H, Landschutz W, Sandstede J, Lipke C, Hahn D. SENSE-Imaging of the perfusion of the heart (Abstract). Fortschr Rontgen-str 2000;S172:143. (Ger)

10. Koestler H, Landschutz W, Sandstede J, Lipke C, Hahn D. Heartperfusion imaging using SENSE (Abstract). Radiology 2000;217(P):464.

11. Kellman P, Epstein FH, McVeigh E. Adaptive sensitivity encodingincorporating temporal filtering (TSENSE). Magn Reson Med 2001;45:846–852.

12. Atkinson DJ, Burstein D, Edelman RR. First-pass cardiac perfu-sion: evaluation with ultrafast MR imaging. Radiology 1990;174:757–762.

13. Wilke N, Jerosch-Herold M, Wang Y, et al. Myocardial perfusionreserve: assessment with multisection, quantitative, first-pass MRimaging. Radiology 1997;204:373–384.

14. Jerosch-Herold M, Wilke N, Stillman AE. Magnetic resonancequantification of myocardial perfusion reserve with a Fermi func-tion model for constrained deconvolution. Med Phys 1998;25:73–84.

15. Bydder M, Larkman DJ, Hajnal JV. Combination of signals fromarray coils using image based estimation of coil sensitivity profiles.Magn Reson Med 2002;47:539–548.

16. Madore B. Using UNFOLD to remove artifacts in parallel imagingand in partial-Fourier imaging. Magn Reson Med 2002;48:493–501.

17. Ding S, Wolff SD, Epstein FH. Improved coverage in dynamic con-trast-enhanced cardiac MRI using interleaved gradient-echo EPI.Magn Reson Med 1998;39:514–519.

18. Reeder SB, Atalar E, Faranesh AZ, McVeigh ER. Multi-echo seg-mented k-space imaging: an optimized hybrid sequence for ultra-fast cardiac imaging. Magn Reson Med 1999;41:375–385.

19. Panting JR, Gatehouse PD, Yang GZ, et al. Echo-planar magneticresonance myocardial perfusion imaging: parametric map analysisand comparison with thallium SPECT. J. Magn Reson Imaging2001;13:192–200.

20. Schwitter J, Nanz D, Kneifel S, et al. Assessment of myocardialperfusion in coronary artery disease by magnetic resonance: acomparison with positron emission tomography and coronary an-giography. Circulation 2001;103:2230–2235.

21. Epstein FH, London JF, Peters DC, et al. Multislice first-pass car-diac perfusion MRI: validation in a model of myocardial perfusion.Magn Reson Med 2002;47:482–491.

22. Kroll K, Wilke N, Jerosch-Herold M, et al. Modeling regional myo-cardial flows from residue functions of an intravascular indicator.Am J Physiol 1996;271:H1643–H1655.

23. Schreiber WG, Schmitt M, Kalden P, Mohrs OK, Kreitner K-F,Thelen M. Dynamic contrast-enhanced myocardial perfusion im-aging using saturation-prepared TrueFISP. J Magn Reson Imaging2002;16:641–652.

24. Christian TF, Aletras AH, Balaban RS, Arai AE. Absolute quantifi-cation of high contrast dose perfusion imaging can be repeated attwo hours despite shortened baseline T1 relaxivity. In: Proceedingsof the 10th Annual Meeting of ISMRM, Honolulu, 2002.

25. Gatehouse PD, Elkington A, Pennell DJ, Firmin DN. A method formeasuring the blood input bolus during first-pass myocardial per-fusion imaging with high contrast agent concentration. In: Pro-ceedings of the 10th Annual Meeting of ISMRM, Honolulu, 2002.

708 Kostler et al.