Can medical image perception exist as distinct discipline?
Medical image perception research measures the human observer’s ability to perform specific diagnostic tasks using real or simulated medical images and compares the observer performance with predictions from quantitative models. The following three characteristics of medical image interpretation limit the ability of traditional psychophysical models to predict human observers performance on these tasks: • The information-bearing variation of image brightness can be degraded by visible stochastic noise (e.g., film grain and quantum mottle). Stochastic noise is particularly important in the new “digital” modalities, such as CT, MR, PET and SPECT. Most investigations in traditional perception research use visual stimuli that are not degraded by stochastic noise. • Biological variability (e.g., variation in the sizes and locations of diagnostically important details) degrades a human observer’s diagnostic performance. • Most radiological tasks are complex interpretations for which the form
