Random correction using large coincidence window for the clear-PEM system

Abstract

The Clear-PEM system is a dedicated PET scanner optimized for breast imaging. Because of the limited field-of-view (FOV) and the moderate timing resolution, the imaging results of the system might be degraded by random noise. This noise is more pronounced at the region that is near to the torso, which could cause false-positive or inconclusive diagnostics. Because of the high number of lines of responses (LORs) comparing to the number of acquired coincidences, list-mode reconstruction is required to maintain efficiency and accuracy. A new acquisition strategy is presented in this abstract in order to largely increase the statistics of acquired random events, without the requirement of hardware to collect single counts. During data acquisition, a large coincidence window of 90 ns is set in the readout electronics. All data within this window are collected in list-mode and, afterwards, classified by the acquisition software into prompt counts (0-4 ns), ignored counts (4-20 ns) and random counts (20-90 ns). A smooth correction image is estimated using those collected random counts. Reconstruction is afterwards performed considering the correction image with multiplication of the ratio between coincidence window width and random window width. An experimental study was performed on a breast-torso phantom. Results show that this approach can increase the statistics of recorded random coincidences over 17 folds, leading to a pronounced improvement of random correction effect. Because of the adoption of a 90 ns coincidence trigger in the hardware that can filter out most of the counts, it yields a much less computational burden to the acquisition system in comparison to the correction method using single count rate.