Timing in silicon PET probesAndrej Studen, Jožef Stefan Institute, Slovenia

On behalf of the MADEIRA collaboration (University of Michigan, Ohio State University, IFIC Valencia, JSI & University of Ljubljana)

Introduction

Simulations show that PET image could be enhanced by placing a high resolution probe close to the imaged object. The probe operates in concert with the external ring. Events in external ring only supplement full tomographic data albeit with low resolution providing means of artifact suppression [1].

High resistivity silicon in form of spatially segmented cells is a competitive sensor material for PET imaging [2] since it provides:● Potentially excellent spatial resolution● Excellent energy resolution for event classification● Robustness & compactness for probe mechanics● Possibility of operation in magnetic fields● Mature development and processing

The only serious drawback is relatively low stopping power, which is normally compensated by increasing the thickness of the sensors. This increases charge collection time, compromising timing resolution required for synchronization with external ring.

The aim of the present study was to characterize performance of the assembly developed within the MADEIRA collaboration and seek solutions for reliable operation of silicon in PET assemblies.

Timing measurement setupThe timing performance was measured on a model prototype with:●1 mm thick silicon sensor with 256 pads with a pad size of 1.4 mm

●A pair of VATAGP7, 128 parallel channel application specific integrated circuits (ASICs) designed by GammaMedica-IDEAS, Inc.

●Custom made hybrid/PCB board for sensor & electronics

●Further downstream - standard issue electronics based on VME standard

Each ASIC channel is composed of a charge-sensitive preamplifier, whose output is split into a pair of CR-RC 2 shapers. The fast shaper with a shaping time of 200 ns is coupled to a leading edge discriminator for a common trigger. The slow shaper with a shaping time of 500 ns provides energy of interaction with resolution of 2 keV FWHM.

The model was exposed to radiation from 22-Na positron source. A LYSO/PMT assembly with PMT anode coupled to a constant fraction discriminator served as a timing reference (timing resolution approx. 1-2 ns FWHM). The CFD trigger and VATAGP7 trigger were fed to a time-to-analog converter coupled to peak sensing ADC. For each event we recorded the timing difference and interaction energy as provided by the ASIC.

Results & Comparison to simulation

The dominant feature is the time walk, which can be compensated by known charge amplitude. Cutting the left histogram at a fixed energy (right plot) a timing distribution much wider than jitter only (shaded histogram) is obtained. The additional broadening is due to variation in impact position of annihilation photon causing modification of pulse shape at discriminator input.

The broadening was checked by simulation using: ●GEANT 4 for realistic tracks of electrons excited by photon interaction

●Realistic electric field in the sensor for propagation of electron-hole pairs

●TCAD (part of Synopsis suite) generated weighting field for proper calculation of induced signal

●Virtual electronics reflecting VATAGP7 architecture

Excellent agreement was found.

Contribution of the impact position to the broadening is reduced at higher detector biases, where charge collection times are shortened due to higher electron & hole velocities.

Impact of poor timing resolutionThe timing window for event synchronization has to match that of the external PET scanner. Most scanner operate at timing windows between 6 and 12 ns [3], depending on scintillator material. The impact can be best seen looking at the efficiency, that is the portion of events in with delay of silicon trigger within the specified window. The events were classified according to interaction energy and for each energy bin (2 keV) a delay giving maximum efficiency was chosen. According to the plot (right) for efficiency exceeding 50 % the probe has to be operated above 400 V and timing windows should exceed 20 ns.

Novel detector geometry for improved timing

To improve the timing performance we propose a novel double- decker sensor arrangement (top). Rather than operating a single 1 mm thick sensor, a pair of sensors with ½ the thickness are used. The top detector is flipped and the corresponding channels are interconnected, giving equal sensor volume per electronic readout channel. This arrangement allows straightforward means of compact stacking of multiple sensor layers, overcoming the stopping power deficiency.

The performance of the arrangement was simulated using the same tools as for 1 mm thick sensor. The impact can be observed in the Figures: For 12 ns window (left) and a detector bias of 430 V, the efficiency of the sensors is practically 100 % throughout the Compton spectrum. The efficiency at 6 ns window (right) is maximized at 80 % for a bias of 430 V. Further observations:●Equivalent over-depletion (marked “same U/U

FD” in the right Figure) favors thinner detectors (due

to larger pad/thickness ratio)●Reducing the shaping time (150 75 ns) has little effect on overall efficiency→●Jitter has a negligible contribution for current setup. The capacitance is currently dominated by routing of channels, which is not expected to change dramatically in the double-decker arrangement. The actual values remain to be measured on an actual prototype.

Conclusions● There is a great agreement between measured data and simulation in determining the timing performance of 1 mm thick silicon sensors used as PET probes.

● Based on the agreement, the variation of photon interaction position significantly contributes to broadening of timing distribution.

● The effect can be reduced by over-biasing the sensor.● Current sensors allow operation at timing windows down to 20 ns.● According to a simulation, the double-decker arrangement with interconnected channels reduces this window to values used in present PET scanners (100 % efficiency @ 12 ns and 80 % efficiency @ 6 ns)

AcknowledgmentsThe work was carried out within the Collaborative Project "MADEIRA" (www.madeira-project.eu), cofunded by the European Commission through EURATOM Seventh Framework Programme (Grant Agreement FP7-212100). Authors would also like to acknowledge support from the NIH grant NIH R01 EB430.

Literature[1] Tai et al. Virtual-Pinhole PET. J Nucl. Med. 49 (3) 2008.[2] S. J. Park et al. A prototype of very high-resolution small animal PET scanner using silicon pad detectors. Nucl. Inst. Meth. A 570 (3) 2007.[3] G. B. Saha. Basics of PET imaging: Physics, Chemistry , and Regulations. Springer NY 2005.

The work was carried out within the Collaborative Project "MADEIRA" (www.madeira-project.eu), cofunded by the European Commission through EURATOM Seventh Framework Programme (Grant Agreement FP7-212100). Authors would also like to acknowledge support from the NIH grant NIH R01 EB430.

6 ns12 ns

8th thematic workshop: International Symposium On Advanced Intraoperative Imaging of Radioisotopes and Presymposium workshop TOF PET (04-06 September 2009)Baia delle Zagare, Mattinata, Italy