Study on inter-crystal compton scatter recovery for a high-rate pre-clinical PET scanner

Abstract

however, the peak height was highest with the image where NN recovery was applied.

This thesis study is about scanner sensitivity enhancement for a high-rate pre-clinical PET to enable a truly real-time tracer dynamics studies. As the method, inter-crystal Compton scatter recovery was used to recycle triple coincidence events which were previously discarded to enhance the scanner sensitivity. A conventional Maximum Energy (ME) method and a novel multi-stage Neural Network (NN) method were employed to analyze the effectiveness of sensitivity enhancement and the accuracy of finding the correct LORs. An ETHZ SAFIR prototype scanner which has small scintillator size and the one-to-one coupling block detector design where an inter-crystal Compton scatter recovery may be more useful. For a point source, the sensitivity was enhanced by around 1.4 times with the ME method and around 2.2 or 1.5 times with two different NN methods. The accuracies were 58%, 66% and 71% respectively. For a NEMA NU4 IQ phantom source, the sensitivity enhancements were 1.1 and 1.2 times and the accuracies were 60% and 67%. For a reconstructed point source image, the FWHM did not change before and after the recovery