The effect of abnormality-prevalence expectation on expert observer performance and visual search.

Abstract

To measure the effect of abnormality-prevalence expectation on experienced radiologists' performance during pulmonary nodular lesion detection on a chest radiograph.

A multiobserver receiver operating characteristic (ROC) and eye-position analysis study was performed to assess the effect of prevalence expectation on observer performance. Twenty-two experienced radiologists were divided into three groups and each was asked to interpret 30 (15 abnormal) identical posteroanterior chest images twice and decide if pulmonary lesions were present. Before each viewing, the radiologists were told that the images contained a specific number of abnormal images: group 1: 9 and 15; group 2: 15 and 22; and group 3: 15 and not told.

ROC analysis demonstrated that no significant effect could be measured as a function of prevalence expectation (P > .05). However, eye-position analysis showed significant increases in eye movements at higher prevalence expectation rates in terms of the number of fixations per image (group 1: P = .0001; group 2: P = .0001; group 3: P = .001) and the total scrutiny time of each image (group 1: P = .0001; group 2: P = .0283; group 3: P = .028).

Overall, findings of this study showed no evidence that the accuracy of expert radiologists is altered due to changing prevalence expectation rates. However, the time spent interpreting each image and the number of fixations increased at higher prevalence rates. Maintenance of diagnostic efficacy has been shown even when circumstances challenge normal observer behavior.