To the Editor.
—Drs Egglin and Feinstein1 provided us with a new and commendable insight with regard to the way the diagnostic judgments of radiologists are modified by the perceptual context in which test results are presented to them. The method and arguments they presented in the article convinced me that our thought process in front of an imaging study follows Bayesian rules. The radiologists in their study appeared to use the early cases in the series they had to read as a "pilot" group: their concrete perceptions and influences were used to calibrate their readings for the rest of the cases in the series. We would like to believe that in our daily practice we also use pretest probability estimates. However, we infer the pretest probability from risk factors and the results of tests (clinical or otherwise) that are presented to us by a narrative of the clinical
Vermont AI. Radiologists' Interpretations: The Problem of Context Bias. JAMA. 1997;277(8):628. doi:10.1001/jama.1997.03540320030025