A Sensitivity Model for Multi-Pinhole SPECT

Abstract

Multi-pinhole single photon emission computed tomography (SPECT) has seen increasing use in small animal imaging, often with overlapping projections in order to improve the count sensitivity. Accurate image reconstruction requires knowledge of the pinhole sensitivity, which depends on source-to-pinhole distance, effective aperture diameter, keel length, and angle of incidence relative to the pinhole normal direction. A model is presented that accounts for these factors as a function of a set of pinhole-specific parameters (effective aperture diameter, normal direction vector, etc.). The model parameters are estimated by least squares analysis of point source data from a single tomographic acquisition. This method was applied to data acquired with a fourteen-pinhole small animal SPECT scanner with overlapping projections. The parameter estimation algorithm converged to a stable solution, and the count values predicted by the pinhole sensitivity model were generally within 10% of the measured count values. The estimated pinhole-specific parameters were of sufficient precision to detect subtle and otherwise unknown manufacturing differences between two batches of pinhole apertures. In reconstructed images, use of the measured parameters led to a noticeable reduction of artifacts and improvement of image quality. Overall, this method of characterizing pinhole sensitivity is robust and practical for multi-pinhole SPECT scanners.