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

ObjectiveTo measure the effect of abnormality-prevalence expectation on naïve observer performance during lesion detection on chest radiographs.MethodsA multi-reader, fixed-case receiver-operating characteristic (ROC) and eye-position analysis study to assess the effect of prevalence expectation on observer performance was conducted. Sixteen diagnostic radiography students (naïve observers) were divided into four prevalence expectation groups (four in each group) and each was asked to interpret thirty (15 abnormal) postero-anterior (PA) identical chest image sets twice to decide if pulmonary nodular lesions were present. Prior to each viewing they were told that the images contained a specific number of abnormal images: group 1: 9 & 15; group 2: 15 & 22; group 3: 9 & 22; group 4:15 & 15.ResultsROC-analysis demonstrated that no significant effect could be measured as a function of prevalence (p > 0.05). However, sensitivity analysis showed a significant change in Group 3 (p = 0.0237). Eye-positional analysis showed one significant change, which was found in Group 1 for mean fixation duration on a lesion (p = 0.0458).ConclusionOverall, the findings of this study showed evidence that the rudimentary performance of naïve observers is altered due to changing prevalence expectation rates.