A new detector design

The use of PET for small animal studies in the development of new pharmaceuticals and therapies is becoming more widespread. In order to study the physiological and functional processes on the smallest scale, possibly even on the molecular level, a very high spatial resolution is required whilst retaining a high sensitivity. Current systems achieve resolutions of between 1.5 and 2 mm full-width at half maximum (fwhm) based upon detectors composed of pixellated scintillators and position-sensitive PMTs. However, in order to achieve further improvements in terms of spatial resolution in such systems, the scintillation pixels need to be made ever smaller, creating a number of technical problems and elevated cost. Also, such systems do not easily provide "depth resolution", or information on where, along the length of the crystal, the gamma ray interacted. Such depth of interaction (DOI) information is required for a reduction of parallax errors from off-center sources and for a better rejection of random coincidences.
We propose a different type of detector design that is based upon the classic Anger camera principle and should provide high spatial resolution, moderate DOI resolution and the potential for high sensitivity. This would provide the basis for a next-generation, high-resolution, small-animal PET system.
The design presented and studied here is based upon the same Anger principle, but miniaturized and with improved performance from the use of modern detector technologies. We propose using multiple, independent layers of the miniature Anger camera, stacked together, in order to create a detector head that has intrinsic DOI, see Figure 1.

Figure 1: A single detector head consisting of three module layers on the left and an exploded view of a module layer on the right showing the pixellated Silicon Photomultiplier (SiPM).

Monte Carlo Simulation

To evaluate the performance of the detector, a simulation of the camera using the Monte Carlo simulation package GEANT4 is performed.
During the different simulations, single annihilation gamma rays with an energy of 511 keV are fired in different positions and with different incident angles into a single detector head, see Figure 2. The annihilation photon is tracked within the detector. Only those events where an annihilation photon deposits at least 50 keV are used for further evaluations, i.e. a low-energy detection threshold is used such that the module layers become sensitive to Compton scatters.
Moreover the generation of optical photons via scintillation in the crystal is simulated. For this purpose the optical properties of all materials and their surfaces are specified. In order to get intrinsic DOI information, our camera uses only the signal obtained from the innermost "active" layer, i.e. the layer closest to the center of the tomograph, where energy was deposited by an annihilation photon. Consequently, the generation of optical photons is restricted to this innermost layer. Hence, for one primary gamma ray event a single "active" layer is determined for a detector head, in which subsequently optical photons are generated.
The generated optical photons are then tracked until they reach the surface of the Silicon Photomultiplier array. Here, the wavelength dependent quantum efficiency of the photo-multiplier is taken into account. Finally, only those optical photons which hit a sensitive area of the pixellated SiPM, contribute to the hit estimation algorithm to computed the coordinates in the xy-plane of the detector.

Figure2: Evaluation of a single detector head by firing a parallel 511keV photon beam in the center. Optical photons are generated via scintillation and tracked to estimate the hit position.

Expected detector performance

Our simulation results are very promising. Figure 3 shows that the spatial resolution in terms of fwhm will be ~0.4mm in the center of the detector, if a very thin light pipe is used. Currently, we are working on hit estimation algorithms to improve the performance of the detector close to the edges.
Moreover, a complete tomograph with 4 heads will be simulated to study effects like: distance to the object, sensitivity, etc., which will influence the final design of the camera as well. The results will be compared to the performance of existing tomographs, like the YAP-(S)PET system, which uses cameras based on a matrix of finely pixellated crystals. For this kind of study, the simulation will be enhanced, including positron annihilations from ¹⁸F sources within a phantom. The finite positron range, as well as deviations from collinearity of the annihilation photons, will be taken into account.
Additionally, to take full advantage of the DOI information appropriate image reconstruction algorithms will be developed. In order to use the precisely determined line-of-response (LOR) endpoints, a well-modeled EM algorithm might be applied first, which takes into account that every LOR ends within one of the three LSO layers, resulting in a reduction of parallax error. Moreover, since there is no hardware-based discretization of the LORs' x and y coordinates, binning is introduced at a later stage allowing minimization of the convolution effects of voxel and sinogram/planogram binning on the final images. The final goal of this design is to achieve images where spatial resolution is dominated by the finite positron range and non-collinearity of the annihilation photons.

Figure 3: Spatial resolution in terms of fwhm along the diagonal axis of the detector for different light pipe thicknesses.

Literature

S.Moehrs, A.Del Guerra, D.J.Herbert, M.A.Mandelkern, "A detector head design for small-animal PET with silicon photomultipliers (SiPM)", Physics in Medicine and Biology, 51(2006), 1113-1127.

D.J.Herbert, S.Moehrs, N.D'Ascenzo, N.Belcari, A.Del Guerra, F.Morsani, V.Saveliev, "The Silicon Photomultiplier for Application to high-resolution gamma cameras for PET", accepted for publication in Nuclear Instruments and Methods in Physics Research Section A (2006).