Development of Chest X-Ray Interpretation Skills in Trainee Radiologists: Changes in Detection Accuracy and Visual Search Behaviour Following a Two-Day Training Course

Detecting and interpreting abnormalities in chest X-rays (CXR) is a complex visual search task. Differences between novice and expert performance in medical image perception are well documented but less is known about how early training shapes trainees’ detection accuracy, error types, and gaze patterns. Novices are often taught systematic viewing strategies, yet these may be ineffective without sufficient knowledge of normal variants and pathological appearances. Educational interventions aim to improve knowledge and recognition of normal and abnormal features, but their impact on scanning behavior and error taxonomy has rarely been examined. The present study examined the early development of CXR interpretation skills by assessing changes in detection accuracy and visual search behaviour following a two-day training course. To this end, first-year radiology trainees completed a CXR detection task while their eye movements were recorded before and after training. Images varied in difficulty, including cases with single obvious or subtle abnormalities and cases with multiple findings. Trainees showed modest improvements in detecting abnormalities and near-ceiling performance for cases with obvious findings. However, false-positive rates persisted on over 60% of abnormal cases and over a third of normal cases, and error rates for subtle- and multiple-abnormality cases still exceeded 70% after training. Most errors arose when trainees fixated abnormalities but did not mark them as abnormal, suggestive of faulty decision-making rather than inadequate scanning or failure to recognize suspicious regions. These findings suggest that early CXR training should prioritize strengthening discrimination of subtle pathology through perceptual learning and calibration of decision thresholds.