An investigation of the challenges in reconstructing PET images of a freely moving animal

Imaging the brain of a freely moving small animal using positron emission tomography (PET) while simultaneously observing its behaviour is an important goal for neuroscience. While we have successfully demonstrated the use of line-of-response (LOR) rebinning to correct the head motion of confined animals, a large proportion of events may need to be discarded because they either 'miss' the detector array after transformation or fall out of the acceptance range of a sinogram. The proportion of events that would have been measured had motion not occurred, so-called 'lost events', is expected to be even larger for freely moving animals. Moreover, the data acquisition in the case of a freely moving animal is further complicated by a complex attenuation field. The aims of this study were (a) to characterise the severity of the 'lost events' problem for the freely moving animal scenario, and (b) to investigate the relative impact of attenuation correction errors on quantitative accuracy of reconstructed images. A phantom study was performed to simulate the uncorrelated motion of a target and non-target source volume. A small animal PET scanner was used to acquire list-mode data for different sets of phantom positions. The list-mode data were processed using the standard LOR rebinning approach, and multiple frame variants of this designed to reduce discarded events. We found that LOR rebinning caused up to 86 % 'lost events', and artifacts that we attribute to incomplete projections, when applied to a freely moving target. This fraction was reduced by up to 18 % using the variant approaches, resulting in slightly reduced image artifacts. The effect of the non-target compartment on attenuation correction of the target volume was surprisingly small. However, for certain poses where the target and non-target volumes are aligned transaxially in the field-of-view, the attenuation problem becomes more complex and sophisticated correction methods will be required. We conclude that there are limitations with the LOR rebinning approach and simplified attenuation correction for freely moving animals requiring the development and validation of more sophisticated approaches.