Optimal detection theory and image reconstruction-Comparison of performance in the presence of noise
The authors study the reconstruction problem as a detection problem and cast the problem as a binary hypotheses decision-making problem to decide the presence of a high-contrast object. Two formulations of the detection problem are presented: one involves direct optimal processing of the projection data; the second optimally processes the data after reconstruction. Results for a simple imaging scenario are presented in the form of receiver-operating-characteristic curves to show how direct optimal processing yields a considerable gain in the decision-making performance over that obtained when first using image reconstruction. Results for the second case are interpreted as providing an upper bound on the decision-performance of all postprocessing algorithms that use convolution backprojection.
