Task Performance On Constrained Reconstructions: Human Observer Performance Compared Wit Sub-Optimal Bayesian Performance

We have previously described how imaging systems and image reconstruction algorithms can be evaluated on the basis of how well bina~-~sc~nation tasks can be performed by a machine algorithm that “views” the reconstructions.1-3 Algorithms used in these investigations have been based on approximations to the ideal observer of Bayesian statistical decision theory. The present work exa ‘nes the performance of an extended family of such algorithmic viewing tomographic images reconstructed from a small number of views using ridge Maximum Entropy software, MEMSYS 3. We investigate the effects on the performance of these observers due to varying the parameter a; this parameter controls the stopping point of the iterative reconstruction technique .and effectively determines the smoothness of the reconstruction. Fo detection task considered here, performance is maximum at the lowest values of a stu these values are encountered as one moves toward the li t of maximum likelihood estimation while maintaining the positivity constraint intri A breakdown in the validity of a Gaussian approximation used r probability) was observed in this region. Measurements on e same task show that they perform comparably to the best machine observers in the region of highest machine scores, i.e s of a. For increasing values of a, both human and machine observer Pe de. The falloff in human performance is more rapid than that of the machine observer at the largest values of a (lowest performance) studied. This behavior is common to all such studies of the so-called psychometric function.