Lesion conspicuity, structured noise, and film reader error

Abstract

The concept of conspicuity is introduced in an attempt to objectively quantitate radiographic observational error. Defined as a ratio between lesion contrast and surround complexity, the measure correlates well with the probability of detecting faint nodular lesions in chest radiographs. The concept helps in understanding why abnormalities are missed by radiologists. It is also used to explain why image-processing techniques advocated in the past did not yield improvements in diagnostic accuracy and to outline directions for the future. Preliminary results are presented which show that photographic subtraction can increase the conspicuity of simulated early lung lesions and improve their detection.