Deconvolution of Compton scatter in SPECT

Abstract

A deconvolution algorithm has been developed which compensates for Compton scattering in SPECT images. Compton scatter is modeled as a convolution of the nonscattered projection data with an exponential function. Deconvolution of the total (scatter + nonscatter) projection data yields compensated true projection. Using Monte Carlo methods, the scattered and nonscattered components of a SPECT image are simulated thus allowing a comparison of scatter compensated results with direct nonscatter results. The quality of the compensation is evaluated by comparing the ratio of total to direct counts with the ratio of compensated to direct counts. This deconvolution technique has been developed and evaluated for experimentally acquired SPECT data as well as for simulated data.