Effects of Practice on Detection of Abnormalities in Chest X-Rays

Abstract

The performance of radiology residents and staff radiologists in detecting and identifying abnormalities in chest x-rays was examined. 120 chest films were each read five times on five different sessions by each radiologist. Half of the films contained up to eight possible types of abnormality. Practice improved the detectability of abnormalities, as indexed by ROC-based measures, in both groups of readers. However, greater benefits of practice were noted for the residents than for the staff. Staff radiologists, but not the residents, had reporting criteria that were close to the optimal criterion for the diagnostic setting. Accuracy of disease identification also improved with practice in both groups. The results indicate that training and practice can lead to the stabilization of reporting criteria and improvement of diagnostic accuracy in chest x-ray diagnosis.