Investigation of quantitative errors due to LOR rebinning motion correction for freely moving small animals with microPET

As part of development of a microPET based system to enable the brain imaging of a freely moving animal while simultaneously observing its behaviour, we previously investigated the feasibility of reconstructing a motion-tracked volume of interest (VOI) in the presence of an extraneous activity compartment with unknown motion. We observed that the fraction of lost events during line of response (LOR) rebinning-based motion correction can be very high during certain poses but the errors are complicated by attenuation losses in the extraneous compartment. In this study, we investigated this issue further. The aim was to (a) design an experiment that isolates the effect of lost events from attenuation in the extraneous compartment, and (b) determine the relative contribution of lost events to quantitative errors during LOR rebinning. A dual phantom study was performed using a hot-rod phantom and uniform phantom moving independently and imaged with a microPET Focus 220 scanner. The hot rod phantom was continuously tracked during acquisition and motion corrected using LOR rebinning and multiple acquisition frame (MAF) methods, while the uniform phantom executed motions that were unaccounted for. We found that LOR rebinning results in up to 87% loss of events in some poses. This not only causes severe artifacts due to pose-dependent loss of projection data but also causes poor quantitative accuracy since the loss of counts can only be approximately corrected by a global scale factor. We conclude that alternative approaches to LOR rebinning, such as direct list mode reconstruction, are required for motion correction of freely moving animals.