Abstract
Scatter in SPECT images may mask decreased uptake of radiopharmaceuticals in the left ventricle of the heart, which can alter the diagnostic outcome of the study. The newly developed dynamic SPECT (dSPECT) method, which reconstructs 4D images from a standard acquisition protocol, provides additional temporal information, which may be helpful to recognise such artefacts. Each voxel carries a time signature, which is different for different organs. In this paper, we investigate whether this signature can be used to detect and remove scatter. Time activity curves (TACs) from segmented data are tested for their potential to locally identify scatter according to a simple model. The investigation is carried out on artificial artefacts in real patient data as well as on existing scatter. Tests on the artificial artefacts showed that scatter can indeed be detected and removed while tests on real data revealed that the simplified model may suffice to remove the majority of local scatter.