Drawing the Line Between Perception and Interpretation in Computer-Aided Mammography

Screening mammography calls for a combination of perceptual skills to find what may be faint and small features in a complex visual environment, and interpretive skills to rate their diagnostic significance. Evidence suggests radiologists' performance of this task can be improved by computer-aided prompting of target features. The introduction of computer-aided mammography provides an interesting case study of `allocation of function' issues. One is where to `draw the line' between perception and interpretation when determining the system's functional role. Our investigations indicate that radiologists find a system which is `perceptually acute', but `interpretatively naive', more acceptable than predicted by earlier work. We present evidence that this is because drawing the line in this way helps radiologists to understand, and to monitor, the system's behaviour. A second issue concerns the impact of computer-aided mammography on existing practices