Compton Identification Within Single ‘Pixels’ D. Scraggs, A. Boston, H. Boston, R. Cooper, J. Cresswell, A. Grint, A. Mather, P. Nolan University of Liverpool C. Hall, I. Lazarus Daresbury Laboratory A. Berry, T. Beveridge, J Gillam, R. Lewis University of Monash Introduction References Pulse Processing and Interaction Classification Possible Interaction Signals SmartPET Specifics Scintillation detector arrays are the current medium of choice for nuclear medicine, driven by availability, affordability and a proven ability to produce results. However, semiconductor detectors offer superior energy resolution, allowing very precise photonic measurements. This becomes a very interesting property when exploited by advanced digital electronics allowing the use of Pulse Shape Analysis [1] (PSA) and Gamma Ray Tracking (GRT). The SmartPET project couples planar germanium detectors, with superb energy resolution, to digital electronics for the production of a combined PET/SPECT scanner. The scanner is currently at a prototype and further development stage. Results thus far include: PET imaging of a point source ~ 1mm spatial resolution Successful uncollimated SPECT imaging in Compton camera mode. [1] K. Vetter et al, NIM A452 (2000) 223, [2] I Lazarus, private communication, [3] I. Y. Lee NIM A463 (2001) 250, [4] WaveLab, Donoho et al Funding Many thanks to the Medical Research Council for their support of this project and many others projects; the results of which have been used by the author. SmartPET utilises double sided orthogonal strip planar germanium detectors with a 5mm pitch. Total crystal dimensions are 74mm by 74mm by 20mm through depth. The actual active area presented to radiation is 60mm by 60mm, allowing a guard ring of 7mm width. The guard ring prevents electric field lines warping at the edge of the active area. The raw granularity of the detector is thus 5mm by 5mm by 20mm through depth. Strip signals are processed by 80MHz, 14bit GRT4 VME modules. Analysis of strip signal waveforms gives an increased resolution of interaction position. The detectors are housed in a rotating gantry which allows both PET and SPECT configurations to be achieved. PET configuration requires the source to be placed between both detectors, simultaneous interactions then define a line of response (LOR). Whilst in SPECT configuration the source is required to be placed on one side of both detectors in composite. One detector then acts to scatter a photon into the second detector via Compton scattering. A cone surface is then defined on which the source must lie. Single detector crystal and electrode configuration. The strips create 5x5mm effective pixels Detectors 1 and 2 sitting in rotating gantry. Detectors are shown in PET mode Schematic of PET mode, in which the source lies on LOR. Right: Intersecting cones from Compton reconstruction define possible source locations The most probable interaction of a photon above 200keV in Germanium is Compton scattering. Fortunately, Compton scattered events can be easily rejected; unless the scatter occurs within a single ‘pixel’, in which case no Compton suppression is currently possible. SmartPET relies upon PSA to improve upon the raw granularity of the detectors. This is a valid and very useful technique which has been proven to work. Unfortunately, PSA applied to a signal which is the result of a Compton interaction will invariably result in a mean position of interaction and not an initial position of interaction, thus resulting in image blurring. Signals resulting from Compton scattering through depth show discontinuities on their leading edges. This is due to the cessation of collection of one particular assortment of charges and the continued collection of other charges. Photons scattering parallel to electrode planes should not show discontinuities and are therefore expected not to be distinguished from their photoelectric alternatives. It has been possible to extract Compton scattered events and study their discontinuities for the purpose of developing a method to initially remove and then include scattered photons in images. Above: Compton scattered signal showing considerable discontinuity. Compton scattered signal not showing any perceptible discontinuity The analysis of signals as filtered by the trigger pattern of the electrodes is insufficient to reject all Compton scattered events. PSA analyses points along the signal but cannot differentiate between the signals shown previously, where one is clearly very different from the other. A possible solution to the raw analysis of signals is prior pulse processing. By transforming the signals into the frequency domain discontinuities become apparent and manageable. An appropriate form of signal processing for SmartPET signals is the wavelet transform, in which a small oscillatory wave ( ) is used to probe the original signal ( ) over a range of frequencies ( ) at specific locations ( ). Frequency is represented by wavelet scale. The wavelet transform is defined as: and contains wavelet coefficients for the various convolutions between the mother wavelet over various scales or frequencies and locations. A large coefficient measures a good match between the wavelet and signal at that particular location and frequency. Scale Signal Location Coefficient The two peaks at low scale or high frequency show a signal discontinuity, and can be used to indicate that the signal does indeed originate from a Compton scatter. Unless some other non-photoelectric phenomenon occurred. Applying this method to data has proved very encouraging. Of definite Compton event signals the technique correctly identified Indeed, scatters parallel to electrode planes will not show any discontinuities. It is also intuitive that scattering through depth has a significantly more adverse effect on image quality, given the larger variance of mean interaction positions. A detector coincidence scan is currently being performed to create a test group of photoelectric signals. If the technique works an adaptive grid will be constructed to recover useful information from Compton signals. A wavelet transform of the first signal shown previously was performed [4]