Iterative reconstruction algorithm which incorporates a high resolution anatomical image: simultaneous correction for 3-D detector response, attenuation, and scatter in SPECT neuroreceptor imaging

Computer simulation studies have shown that twofold to threefold cortical variations in neuroreceptor single photon emission computed tomography (SPECT) images may result from imaging artifacts. To differentiate these variations from physiological variations resulting from a disease state, an iterative reconstruction algorithm (IRA) which corrects for the artifacts was developed. The IRA utilizes both a realistic 3-D model of the imaging process and a realistic Hoffman 3-D brain model. For both the normal and the patient the radioactivity was set to unity in each gray matter voxel of a basal ganglia slice, but for the patient the radioactivity in the right parietal cortex was decreased by 30%. Filtered backprojection of both the normal and the patient resulted in circumferential cortical profiles which exhibited twofold to threefold cortical variations, although (except for the deficit in the right parietal cortex) these should theoretically be constant. For IRA, however, circumferential cortical profiles of both the normal and the patient were constant, except for the 30% deficit in the right parietal cortex. These results show that IRA can compensate for variations due to imaging artifacts to permit the quantitation of physiological variations.<>