Detection or decision errors? Missed lung cancer from the posteroanterior chest radiograph

A test bank of verified chest radiographs was compiled for visual search experiments to investigate radiology performance in the detection of early lung cancer. A measure of the physical characteristics of the lesions was derived to determine the conspicuity (chi) of the nodules and to investigate possible causes of failed detection. Observer performance was measured by alternate free response operating characteristic (AFROC) methodology and was supplemented with visual search recording. Correlation of AFROC scores and the chi values was poor but inspection of the visual search recordings showed that most nodules were fixated. Fixations on missed lesions produced average dwell times greater than three times the minimum duration thought to be associated with detection. We conclude that the majority of errors were failures of decision rather than detection and comment on the implications of this for strategies to improve diagnostic effectiveness.