An evaluation of missing data compensation methods for a PET camera by comparing to no-gap data

We evaluated three methods for filling gaps in PET projection data that may results from gaps between detectors or from defective detectors. These methods include linear interpolation, constraint Fourier space, and forward projection of the images reconstructed from incomplete data iteratively. We compared reconstructed images from the gap-filled data to the images reconstructed from data without gaps acquired by rotating camera gantry during data acquisition. We collected data for a 20-cm flood phantom and the Hoffman brain phantom using the HOTPET camera that has a spatial resolution of about 3 mm. HOTPET Camera has the capability to be rotated during data acquisition and in standard operation it is rotated to fill the gaps in sinograms. Images reconstructed from data without gantry rotation and without missing data compensation were dominated by artifacts. Filling the gaps with even the simple and fast linear interpolation resulted in images with significantly less artifacts; however, it was less effective than the other two methods. Images from gap-filled data by the other two methods were artifact free but image quality was slightly degraded. Filling the gaps with forward projection of the iterative reconstructed images produced the best results; however, it was the slowest method.