How do radiologists do it? The influence of experience and training on searching for chest nodules

Abstract 

Four observer groups with different levels of expertise were tested to investigate the nature of expert performance. The task was the detection and localisation of significant pulmonary nodules in postero-anterior views of the chest. One hundred and twenty digitised chest images were used. The observer groups were 8 experienced radiologists, 5 experienced radiographers before and after six months training in chest image interpretation, and 8 undergraduate radiography students. Eye tracking was carried out to investigate differences in visual search strategies between observers. Detection performance was measured with an Alternate Free Response Operating Characteristic technique. Performance measures showed the experienced group of radiologists plus radiographers after training were better at the task than the remainder (t-test p=0.046). Differences were shown in the eye-tracking parameters between the groups: saccadic amplitude (ANOVA p=0.00047), number of fixations before and after training (t-test p=0.041), and scrutiny time per decision and per film for the experienced versus the inexperienced observers (t-test p=0.02). Visual coverage reduced with increasing level of experience but this result did not reach significance. Generally there were distinct differences in the search strategies between the experienced and inexperienced observers and we discuss the significance of these findings. We believe the results support some recent theoretical models of expert performance and that the findings may prove to be helpful in ‘fast-track’ educational programmes of image interpretation for non-radiology practitioners.

Keywords: Expert performance, Pulmonary nodules, AFROC, Eye-tracking, Fixation patterns