Recent advances on X-ray imaging with a single photon counting system I. Introduction II. The system: microstrip detectors, RX64 ASICs, testing methods

Recent advances on X-ray imaging with a single photon counting system

I. Introduction II. The system: microstrip detectors,

RX64 ASICs, testing methodsIII. Energy resolution and efficiencyIV. Spatial resolutionV. Imaging results - mammographyVI. Imaging results - angiographyVII. Summary and outlook

Luciano Ramello – Univ. Piemonte Orientale and INFN, Alessandria

NURT 2003, October 27-31, 2003

G. Baldazzi1, D. Bollini1, A.E. Cabal Rodriguez2, C. Ceballos Sanchez2 ,W. Dabrowski3,

A. Diaz Garcia2, M. Gambaccini4, P. Giubellino5, M. Gombia1, P. Grybos3, M. Idzik3,5,

J. Lopez Gaitan10, A. Marzari-Chiesa6, L.M. Montano Zetina7, F. Prino8, L. Ramello8,

A. Sarnelli4, M. Sitta8, K. Swientek3, A. Taibi4, E. Tomassi6, A. Tuffanelli4,

P. Van Espen9, P. Wiacek3

1 University and INFN, Bologna, Italy; 2 CEADEN, Havana, Cuba;3 University of Mining and Metallurgy, Cracow, Poland; 4 University and INFN, Ferrara, Italy; 5 INFN, Torino, Italy; 6 University of Torino, Torino, Italy; 7 CINVESTAV, Mexico City, Mexico; 8 University of Eastern Piedmont and INFN, Alessandria, Italy; 9 University of Antwerp, Antwerp, Belgium; 10 Univ. de los Andes, Colombia

I. Introduction

Introduction (1) We are developing a single photon counting system for X-

ray imaging in the 15-50 keV range Spatial resolution is defined by the detector segmentation

(presently 100 m pitch strips) Energy resolution is determined mainly by the low-noise

front-end ASIC Rate capability (converted photons/mm2/s) is defined by the

timing characteristics of the ASIC and by the pixel size (presently 100 x 300 m2)

Medical applications of this system are those requiring high dynamic range of counts, good energy resolution; furthermore, they must be compatible with scanning mode

I. Introduction

Introduction (2) One-dimensional silicon array for scanning mode

imaging:• Good spatial resolution with reduced number of channels

• Spatial resolution in silicon limited by Compton scattering and parallax error, pitch smaller than about 50-100 micron not really useful

Advantages of digital single photon X-ray imaging:• Higher detection efficiency with respect to screen-film systems

• Edge-on orientation (parallel incidence) preferred for

energies above 18 keV

• Double energy threshold with simultaneous exposure possible

• Easy processing, transferring and archiving of digital images

I. Introduction

Introduction (3) Subtraction imaging: removes background structures Dual energy technique: isolates materials characterized

by different energy dependence of the linear attenuation coefficient Alvarez and Macovski 1976

Quasi-monochromatic beams: implement dual energy techniques in a small-scale installation, no synchrotron

[see NIM A 365 (1995) 248 and Proc. SPIE Vol. 4682, p. 311 (2002)]

First application: dual energy angiography at iodine K-edge (33 keV), possible extension to gadolinium K-edge (50 keV)

Second application: dual-energy mammography (18+36 keV)

Silicon efficiency vs. X-ray energyI. Introduction

Front configuration • 70 m Al shield

(could be reduced)• 300 m active Si

Edge configuration• 765 m insensitive

silicon • 10 mm (now) or 20

mm (later ?) active silicon

Photoelectric conversion in the active volume

simple calculation with cross-sections from XCOM data base of NIST

GaAs: a better alternative ?I. Introduction

Photoelectric conversion in the active volume

Front configuration for GaAs, Edge configuration for Si

GaAs is the best choice for 20 keV mammography

Si in edge mode (10 mm) is almost equivalent to GaAs for angiography

II. System

Silicon microstrip detectors AC coupling:

Bias Line with

FOXFET biasing Guard ring

essential to collect surface currents

Designed and fabricated by ITC-IRST, Trento, Italy

guardring bias line first strip (AC contact)

DC contact (to p+ implant)

Detector test: I-V measurements

400-strip detector from ITC-IRST, Trento, Italy:

Ibias(60 V) = 18.9 nA Istrip(60 V) 47.2 pA

Ibias(100 V) = 25.0 nA Istrip(100 V) 62.5 pA

10-10

2

4

6

10-9

2

4

6

10-8

2

Cor

rent

e di

fuga

(A)

100806040200Tensione di polarizzazione inversa (V)

Corrente di guard ring Corrente di bias line

Temperatura = 25.5 °C

Keithley 237 provides reverse bias,

HP 4145B measures currents, for bias line (serving 400 strips) and for guard ring.

Reverse bias voltage (V)

Lea

kag

e cu

rren

t (A

)

II. System

Detector test: C-V measurements

Keithley 237 provides reverse bias,HP 4284A injects sinusoidal signal to measure C:• V = 500 mVV = 500 mV• f = 100 kHzf = 100 kHz

28

26

24

22

20

18

V0(

Vol

t)

108642Posizione

0.35

0.30

0.25

0.20

0.15

0.10

0.05

0.00

1/C

2 (p

F-2

)

100806040200Tensione di polarizzazione inversa (V)

V0 = (23.16 ± 1.06) Volt

Full depletion voltage is constant across detector

Reverse bias voltage (V)

II. System

Strip-by-strip measurements

• VVBB = 60 V = 60 V• Contacts needed:Contacts needed: 0. Backplane0. Backplane

1.1. Strip Strip i2.2. Strip Strip (i+1)3.3. Bias lineBias line

Measuring strip current, IMeasuring strip current, Istripstrip

Measuring inter-strip resistance, RMeasuring inter-strip resistance, Rstripstrip

70

65

60

55

50

45

40

I strip

(pA

)

4003002001000Numero strip

500

400

300

200

100

Rst

rip(G

)

4003002001000Numero strip

)()(

2

dVIdVI

dVR

stripstripstrip

II. System

The RX64 ASIC (1)

RX64 - Cracow Univ. of Mining and Metallurgy design: single channel layout

- charge-sensitive preamplifier

- shaper

- discriminator (2 discriminators in the latest version)

- pseudo-random counter (20-bit) [not shown]

II. System

detector

test capacitor Ct

The RX64 ASIC (2)

RX64 - Cracow U.M.M. design - (28006500 m2)

- 64 front-end channels (preamplifier, shaper, 1 or 2 discriminators),

- 64 pseudo-random counters (20-bit),

- internal DACs: 1 or 2 for 8-bit threshold(s) setting and two 5-bit for bias settings

- internal calibration circuit (square wave 1mV-30 mV),

- control logic and I/O circuit (interface to external bus).

II. System

RX64 ASIC testing II. System

Probe card testing before assembly on PCB becomesconvenient when productionyield is low:

• Power consumption test• Test of the counter section• Full test of the analogue performance of the 64 channels, using both HIGH and LOW discriminator/counter sets

The test is performed using the same power supplies, cables, DAQ hardware and software as for the final assembled system

System assemblyManual wire bonding (detector - chip)

II. System

Automatic wire bonding (detector - pitch adapter - chip)

Noise and gain evaluation method200

150

100

50

0

Co

nte

gg

i

340320300280260240Soglia (mV)

200

150

100

50

0

Con

tegg

i

340320300280260240Soglia (mV)

15

10

5

0

340320300280260240Soglia (mV)

x0 = 291.4 ± 0.446sigma = 11.34 ± 0.51

1Obtain Counts vs.

Discriminator Threshold

(threshold scan)

2Smoothing of Counting

Curve

Error function Fit, or …

3Differential Spectrum

Gaussian Fit

extract mean and

III. Energy resolution and efficiency

Threshold uniformity (128 channels)

Calibration pulse of 5300 electrons (internal voltage step applied to C

test = 75 fF)

Mean threshold (from gaussian fit) for 128 channels:• Threshold spread

%• Small syst. difference

(4%) between chips


Linearity vs. injected charge (1)40

30

20

10

0

600500400300200100Soglia (mV)

calib DAC = 4 calib DAC = 10 calib DAC = 16 calib DAC = 6 calib DAC = 12 calib DAC = 18 calib DAC = 8 calib DAC = 14 calib DAC = 20

Differential spectra obtained with internal calibration:each value of the Calibration DAC produces on the test capacitor Ct (75 fF) a pulse of given charge


Linearity vs. injected charge (2)600

500

400

300

200

(m

V)

90008000700060005000400030002000Elettroni in ingresso

a = 5.1 ± 1.8b = 0.064966 ± 0.000486

• the RX64 chip is strictly linear up to the RX64 chip is strictly linear up to 55005500 electrons input charge electrons input charge (i.e. up to (i.e. up to 20 keV X-ray energy20 keV X-ray energy))• aa straight line fit straight line fit within linearity range gives offset (a) & gain (b)within linearity range gives offset (a) & gain (b)

Injected charge (electrons)


Gain uniformity (128 channels)

Scan with 10 different amplitudes (4-22 mV)

Circuit response reasonably linear up to 8000 electrons (29 keV) for T

peak= 0.5 s

<Gain> = 61.6 V/el.

Small (3.5%) systematic difference between chips


Rate capability of the RX64

1k 10k 100k0

20

40

60

80

100(a)

Counting rate [1/s]

Effi

cien

cy [%

]

Tp=1.0s Tp=0.7s Tp=0.5s

1k 10k 100k0

5

10

15

20

25

30(b)

Tp=1.0s Tp=0.7s Tp=0.5s

Counting rate [1/s]

Gai

n [m

V/k

eV]

Efficiency Gain

Counting rate [1/s]

Test with random signals, 8 keV

Three different shaping times T(peak): 1.0, 0.7, 0.5 s

Sufficient performance for imaging applications up to 100 kHz / strip

Counting rate [1/s]

100

0100 k 100 k10 k10 k


Gain and Noise summary (I)

Detector with 128 equipped channels (2 x RX64):• RMS value of noise = 8.1 mV ENC = 131 electrons

• RMS of comparator offset distribution = 3.2 mV: 2 times smaller than noise (common threshold setting for all channels)

Module T(peak) Gain ENC (el.)

Det. + 2 x RX64 Short 61.6 131

6 x RX64 Short 63.7 176

6 x RX64 Long 82.8 131

Fanout + 6 x RX64 Short 63.7 184

Fanout + 6 x RX64 Long 82.8 148


Calibration setups for X-ray detectorCu-anode X-ray tube with Cu-anode X-ray tube with

fluorescence targetsfluorescence targets

241241Am source with rotary Am source with rotary target holdertarget holder


Pb collimator

Fluorescencetarget X-ray tube

Board with detector

Calibration results (single strip)

150

100

50

0

Co

un

ts

500400300200100

Threshold (mV)

Source Am+Rb target Source Am+Mo target Source Am+Ag target Tube+Cu target Tube+Ge target Tube+Mo target Tube+Ag target Tube+Sn target

CuE (K) = 8.0 KeV

GeE (K) = 9.9 keV

RbE (K) = 13.4 keV

MoE (K) = 17.4 keVE (K) = 19.6 keV

AgE (K) = 22.1 keVE (K) = 24.9 keV

SnE (K) = 25.3 keVE (K) = 28.5 keV


Gain and Noise summary (II)450

400

350

300

250

200

150

(m

V)

24222018161412108Energia (keV)

CuGe

Mo

Ag

Sn

Rb

Mo

Ag

Retta calibrazione con la sorgente Retta calibrazione con il tubo

6 x RX64 + fanout + detector, T(peak) Long

GAIN ENC30

improved amplif. setting

ENC50

241Am source 62.8 V/el. 154 el. 179 el.

X-ray tube 63.7 V/el. 151 el. 182 el.

internal calib. 64.6 V/el. 141 el. 164 el.


Matching between channels

RX64 chip: 64 channels measured simultaneously with common threshold(absolutely essential for practical applications)


The Double Threshold chip

First RX64-DT chip measured: spectra obtained with moving hardware window of 14 mV (5 LSB threshold DAC) by 1 LSB steps.


ENC = 196 electrons

The conversion efficiencyIII. Energy resolution and efficiency

Detector was exposed to same beam flux in FRONT and EDGE modeThe (not well kown) absolute beam flux cancels in the ratio: Counts(EDGE) / Counts(FRONT)Experimental results compare well with GEANT 3.21 simulations

Quasi-monochromatic beam at 6 energies (18-36 keV)

Fluorescence setup with 4 targets (15.7-25.0 keV)Preliminary analysis

IV. Position resolution

The micro X-ray beam X-ray tube (Mo anode) with

capillary output at MiTAC, Antwerp University

Si(Li) detector to measure fluorescence at 90 degrees

CCD camera with same focal plane as X-ray beam

optional Mo/Zr filters to reduce intensity and change energy spectrum

X, Y, Z movements with 1 m precision


Measuring the position resolution X-ray tube (Mo anode)

operated at 15 kV and 40 kV Silicon detector in front

configuration (Al protection removed)

Mo or Zr filter Horizontal scan (in/out of

beam focus) by 1 mm steps to check focus Vertical scan (across strips)

by 10 m steps to measure position resolution


The MicroBeam Vertical scan of a 25 m

diameter Ni-Cr wire, tube at 15 kV

Si(Li) detector counts vs. wire position for Ni K

peak: observed RMS of 28.5 m

Deduced beam RMS after deconvolution of wire is not much smaller

Beam RMS decreases with increasing tube kV (while beam halo becomes more important)


Beam profile in microstrip detector The minimum size of the beam is maintained for a

depth of focus of 3-4 mm


Position resolution results (1)102.5

102.0

101.5

101.0

100.5

100.0

Hit

cen

tro

id (

stri

p)

300250200150100500

Beam Position (m)

y=99.711 + 0.0098132x Si microstrip beam profile:Centroid (strip units) vs.Beam Position (m)

Simulation of the Centroidvs. Beam Position


Position resolution results (2)0.10

0.05

0.00

-0.05

-0.10

Ce

ntr

oid

- f

it (s

trip

un

its)

300250200150100500

Beam position (m)

Maximum deviation from straightline is ± 0.12 strips (12 m)

Later, beam halo has been reducedthanks to a 100 m pinhole

Preliminary analysis of latest data shows considerable reductionof maximum deviationfrom straight line

Dual Energy Mammography Dual energy mammography allows to

remove the contrast between the two normal tissues (glandular and adipose), enhancing the contrast of the pathology

Single exposure dual-energy mammography reduces radiation dose and motion artifacts

to implement this we need:• a dichromatic beam• a position- and energy-sensitive detector

V. Mammographic imaging

The dichromatic beam (1) W-anode X-ray tube operated at 50 kV Highly oriented pyrolithic graphite (HOPG)

mosaic crystal (Optigraph Ltd., Moscow) higher flux than monocrystals (also higher E/E)


-2 goniometer Bragg diffraction,first and second harmonics energies E and 2E are obtained

The dichromatic beam (2) A. Tuffanelli et al., Dichromatic source for the application of dual-energy tissue cancellation in

mammography, SPIE Medical Imaging 2002 (MI 4682-21)


incidentspectraat 3 energysettings …

… spectra after 3 cm plexiglass

(measured with HPGe detector)

Use of dichromatic beam it’s possible to tune dichromatic beam energies to

breast thickness, to obtain equal statistics at both energies better signal-to-noise ratio


The mammographic test (1) A three-component phantom made of polyethylene,

PMMA and water [S. Fabbri et al., Phys. Med. Biol. 47 (2002) 1-13] was used to simulate the attenuation coeff. (cm-1) of the adipose, glandular and cancerous tissues in the breast


E (keV) _fat _gland _canc PE PMMA water

20 .456 .802 .844 .410 .680 .810

40 .215 .273 .281 .225 .280 .270

By measuring the logarithmic transmission of the incident beam at two energies, with a projection algorithm [Lehmann et al., Med. Phys. 8 (1981) 659] the contrast between two chosen materials vanishes

The mammographic test (2) Low energy and high energy images were acquired

separately (no double threshold ASIC yet) with the 384-channel Si detector, covering a 38.4 mm wide slice of the phantom

After correction for flat-field and bad channels, the dual-energy algorithm was applied to the logarithmic images at the two energies, changing the projection angle to find the contrast cancellation angles for pairs of materials


For more details, see poster by C. Ceballos on Tuesday 28/10

Mammography test results (1)V. Mammographic imaging

The contrast cancellation angles for each pair of materials wereobtained, both from experiment and from MCNP simulation

Mammography test results (2)V. Mammographic imaging

1=detector

2=PMMA3=water4=PE

The PE pattern alone is visible in measured data at projection angle 36.5 ° (PMMA-water cancellation)

Simulations are in fair agreement with data for PMMA-water cancellation angles at 2 out of the 3 energy pairs; we are investigating problems due to low statistics at high energies and to uncertainty on PE sample composition

The angiographic test setup

Phantom with 4 iodine-filled Phantom with 4 iodine-filled cavities of diameter cavities of diameter 1 1 or or 2 mm2 mm

1.1. X-ray tube with dual-energy outputX-ray tube with dual-energy output

- each measurement each measurement 1.4 • 10 6 photons photons / mm/ mm2 2 (in 2+2 seconds)

2.2. Phantom made of PMMA + AlPhantom made of PMMA + Al

3.3. Detector box with two collimatorsDetector box with two collimators

X-ray tube with dual energy output

Detector box with 2 collimators

Phantom

VI. Angiographic imaging

Procedure for image analysis (I)

1. 1. MeasureMeasure Flat Flat fieldfield at both at both energiesenergies

1.1

1.0

0.9

0.8

0.7

0.6F

latfi

eld

norm

al.

3002001000canali

E = 31.5 keV

1.1

1.0

0.9

0.8

0.7

0.6

Fla

tfiel

d no

rmal

.

3002001000canali

E = 35.5 keV

2. 2. Normalize counts between the two energiesNormalize counts between the two energies

3. 3. Compute transmission in PMMA + AlCompute transmission in PMMA + Al1.0

0.8

0.6

0.4

0.2

0.0Tra

smis

sio

ne

te

oric

a

5004003002001000pixels

E = 31.5 keV

1.0

0.8

0.6

0.4

0.2

0.0Tra

smis

sio

ne

te

oric

a

5004003002001000pixels

E = 35.5 keV

<N(31.5 keV)> / <N(35.5 keV)> = 2.432


Procedure for image analysis (II)15

10

5

0

pixe

ls

3002001000pixels

161412108642

Con

tegg

i ( x

103 )

15000

12000

9000

6000

3000

Con

tegg

i

3002001000pixels

15

10

5

0

pixe

ls

3002001000pixels

6

5

4

3

2

1

Con

tegg

i (x1

03 )

6000

4000

2000

Co

nte

gg

i

3002001000pixels

15

10

5

0

pixe

ls

3002001000pixels

-0.8

-0.6

-0.4

-0.2

0.0

log

con

tegg

i

-0.8

-0.6

-0.4

-0.2

0.0

log

co

nte

gg

i

3002001000pixels

E = 31.5 keV E = 35.5 keV

logarithmic subtraction

5.315.35 lnln NN


Images vs. iodine concentration

-0.8

-0.6

-0.4

-0.2

0.0

log

co

nte

gg

i

3002001000pixels

-0.3

-0.2

-0.1

0.0

0.1

log

co

nte

gg

i

3002001000pixels

-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

log

cont

eggi

3002001000pixels

Cavity diameter = 1mm

15

10

5

0

pix

els

3002001000pixels

-0.8

-0.6

-0.4

-0.2

0.0

log

co

nte

gg

i

15

10

5

0

pixe

ls3002001000

pixels

-0.3

-0.2

-0.1

0.0

0.1

0.2

log

cont

eggi 15

10

5

0

pixe

ls

3002001000pixels

-0.15

-0.10

-0.05

0.00

0.05

0.100.15

log

cont

eggi

370 mg / ml 92.5 mg / ml 23.1 mg / ml


MCNP simulations: see C. Ceballos et al., AIP Conf. Proc. 682, 2003, pp. 185-191

Signal-to-Noise ratioSNR SNR defined as ratio betweendefined as ratio between CONTRAST (C CONTRAST (Css) and fluctuations in ) and fluctuations in

a given area (here 1x1 pixel) of the image (a given area (here 1x1 pixel) of the image (CCnn): ): SNR = Cs/Cn

50

40

30

20

10

0

SN

R

4003002001000Concentrazione (mg/ml)

cavità 4 teor. cavità 4 cavità 3 teor. cavità 3 cavità 2 teor. cavità 2 cavità 1 teor. cavità 1

100

80

60

40

20

0

SN

R

4003002001000Concentrazione (mg/ml)

cavità 4 teor. cavità 4 cavità 3 teor. cavità 3 cavità 2 teor. cavità 2 cavità 1 teor. cavità 1

d = 1 mm

d = 2 mm

SNR

SNR

Concentration (mg/ml)


VII. Conclusion

Summary

A relatively simple linear X-ray detector for scanning mode radiography was developed

Energy resolution (1.3 keV FWHM at 22 keV) is well suited for the available quasi-monochromatic beams

Efficiency in edge mode (10 mm Si) is sufficient for D.E. mammography and angiography at iodine K-edge

Imaging results with phantoms show interesting SNR values, detailed simulations using MCNP and GEANT 3 were developed

VII. Conclusion

Outlook

Exploit double threshold ASIC for D.E. Mammography (ASIC mass tests ongoing)

Build larger detectors for full-size imaging Measure DQE and MTF with microbeam Angiography: implement synchronization

with ECG Angiography: explore the Gadolinium

option at 50 KeV

VII. Conclusion

Thanks to ... The organizers of NURT 2003 for this nice

opportunity to present our results The Italian Ministry for Education, University and

Research (MIUR) The Polish State Committee for Scientific

Research INFN Torino for allowing access to technical staff

and bonding facilities ICTP Trieste for travel and subsistence support to

Cuban researchers The European Community for travel and

subsistence support for students under the ALFA II programme (contract AML/B7-311/97/0666/II-0042)

Documents

Recent advances on X-ray imaging with a single photon counting system I. Introduction II. The system: microstrip detectors, RX64 ASICs, testing methods