Skip to Main Content
We have previously developed a windowed image reconstruction method that allows for region-of-interest (ROI) reconstruction from a subset of TOF PET data. As a result, the method may improve the reconstruction for a ROI by discarding data that are not particularly information-bearing to it but containing high statistical noise or unwanted events (scatter or random). In our previous paper, we have evaluated the performance of this method with numerical experiments and showed that it can improve the image signal-to-noise ratio (SNR) when considering only the true events. We also showed that it can recover the area near a high activity area better in comparison with the conventional filtered back-projection (FBP) algorithm for non-TOF PET and the convolved ramp and Gaussian with confidence weighting (CRG-CW) method for TOF-PET. In this work, we more extensively evaluate the performance of this windowed image reconstruction method by including also scatter and randoms in simulation. We evaluate the SNR of the reconstructed images in defined ROIs as well as the variations of the SNRs with the size of the window function. Our results indicate that the windowed image reconstruction method can generate images with higher SNRs relative to FBP images. In addition, when the ROI is small and nearby a high-activity structure, the windowed reconstruction method can generate better SNRs than do the CRG-CW method.