High resolution gamma detector for small-animal positron emission tomography

In this study, the performance of continuous miniature crystal element (cMiCE) detectors with LYSO crystals of different thickness were investigated. Potential designs of a cMiCE small animal positron emission tomography scanner were also evaluated by an analytical simulation approach.;The cMiCE detector was proposed as a high sensitivity, low cost alternative to the prevailing discrete crystal detectors. A statistics based positioning (SBP) algorithm was developed to solve the scintillation position estimation problem and proved to be successful on a cMiCE detector with a 4 mm thick crystal. By assuming a Gaussian distribution, the distributions of the photomultiplier signals could be characterized by mean and variance, which are functions of scintillation position. After calibrating the detector on a grid of locations, a 2D table of the mean and variance can be built. The SBP algorithm searches the tables to find the location that maximizes the likelihood between the mean and variance of known positions and the incoming scintillation event.;In this work, the performance of the SBP algorithm on cMiCE detectors with thicker crystals (6 and 8 mm) was studied. The stopping power of a cMiCE detector is 40% and 49% for 6 and 8 mm thick crystals respectively. The intrinsic spatial resolution is 1.2 mm and 1.4 mm FWHM for the center and corner sections of a 6 mm thick crystal detector, and 1.3 mm and 1.6 mm for center and corner of an 8 mm thick crystal detector. These results demonstrate that the cMiCE detector is a promising candidate for high resolution, high sensitivity PET applications.;A maximum-likelihood (ML) clustering method was developed to empirically separate the experimental data set into two to four subgroups according to the depth-of-interaction of the detected photons. This method enabled us to build 2-DOI lookup tables (LUT) (mean and variance lookup tables for front group and back group). Using the 2-DOI SBP LUTs, the scintillation position and DOI could be estimated at the same time. The experimental measured misclassification rate for the 8 mm thick crystal detector is approximately 25%. The ML clustering method also provided a better fit to the distributions of the experimental signals, especially for the skewed ones. It therefore led to a significant improvement in the intrinsic spatial resolution in the corner region of the detector.;In order to eliminate the effort in calibrating a cMiCE detector, a parametric positioning method was studied. Gaussian, Cauchy, and parametric models for the light distribution inside the crystal were tested. From the diagnosis of the sum of squared residues and the goodness of fit to the experimental data, the parametric model was found to be the best fit to the light distribution. It was also the best performer in terms of intrinsic spatial resolution and DOI resolution. Using the parametric model, the intrinsic spatial resolution is 1.1 mm and 1.3 mm FWHM for the center and corner regions of the 8 mm thick crystal detector respectively. The DOI resolution is 3.2 mm FWHM.;Another variation of the SBP algorithm was tried to reduce the number of readouts need to be digitized. Several themes of different trade-offs between the readout number and spatial resolution were tested. The results show that excluding the PMT channels with less 1% of the total signal or digitizing only the nearest 21 channels around the channel with the maximum signal are the best choices, while the intrinsic spatial resolution is not compromised.;An analytical simulation approach was developed to investigate how the choice of cMiCE detectors affect image figures of merit for mouse-imaging cMICE PET scanners. For a high resolution imaging system, important physical effects that impact image quality are positron range, detector point-spread function and coincident photon count levels (i.e., statistical noise). Modeling of these effects was included in an analytical simulation that generated multiple realizations of sinograms with varying levels of each effect. To evaluate image quality with respect to quantitation and detection task performance, four different figures of merit were measured: (1) root mean square error; (2) a region of interest SNR (SNRROI); (3) non-prewhitening matched filter SNR (SNRNPW); and (4) recovery coefficient. The results indicate that positron range and non-stationary detector point-spread response effects cause significant reductions of quantitation (SNRROI) and detection (SNRNPW) accuracy for small regions, e.g., a 0.01 cc sphere. A cMiCE detector with 6 mm thick crystal is better for quantitation, while the one with 8 mm thick crystal is better for detection. DOI capability makes a major impact on the FOMs. cMiCE detector with 8 mm thick crystal + 2-DOI capability proved to be the best candidate for both quantitation and detection.