Randoms Mean Value Estimation with Exact Method for Ring PET Scanner

Techniques for randoms mean value estimation from noisy random data were examined for the case of radioactivity concentration unevenly distributed in the trans-axial field of view (FOV) for ring PET scanners. The randoms mean values are required data in order to apply ordinary Poisson reconstruction algorithm. Then noise contribution from random data must be eliminated. Current randoms mean value estimation technique we used is based on fan-sum approximation approach. This methods produces biased random mean value estimation toward the edge of the trans-axial FOV in the case of non-uniformly distributed radioactivity concentration. An exact Defrise's method was examined in this paper. This method accurately estimated singles count rate per detector group. Random sinogram was constructed from these estimations. It was found that application of exact method reduced trans-axial bias in random mean value estimation. Parameter selection was investigated.