Cookies on this website

We use cookies to ensure that we give you the best experience on our website. If you click 'Accept all cookies' we'll assume that you are happy to receive all cookies and you won't see this message again. If you click 'Reject all non-essential cookies' only necessary cookies providing core functionality such as security, network management, and accessibility will be enabled. Click 'Find out more' for information on how to change your cookie settings.

OBJECTIVE. The objective of our study was to describe the characteristics of polyps viewed but then dismissed incorrectly by radiologists at endoluminal CT colonography (CTC), eye movements during these errors, and features provoking false-positive diagnoses. MATERIALS AND METHODS. Forty-two radiologists viewed 30 endoluminal CTC videos, each depicting a polyp, while their eye movements were tracked. Half of the videos had computer-assisted detection (CAD), and half did not. Classification errors were defined when proven polyps were seen but dismissed. Eye movements during these errors and during correct polyp identifications were compared with multilevel modeling. Polyps were divided subsequently into "difficult to classify" and "easy to classify" using a classification error threshold of more than 15%. Polyp diameter, height, and subjective conspicuity and the proportion of time viewed were compared between groups. RESULTS. Eye tracking revealed that 97% of false-negative polyp diagnoses were nonetheless preceded by the reader observing the polyp. The difficult polyps were significantly smaller than the easy polyps (mean diameter, 5.4 vs 8.2 mm, respectively p = 0.014) and were subjectively less conspicuous (median score, 4 vs 2; p = 0.0032). Readers spent proportionally less time viewing difficult polyps than viewing easy polyps (29.0% of the time they were on-screen vs 42.6%, respectively; p = 0.01) regardless of the presence of CAD. CONCLUSION. Even small and subjectively inconspicuous polyps attract reader gaze, but they are nonetheless ignored. These errors are made rapidly even with CAD. Efforts to improve reader performance at CTC should focus on decision making rather than detection alone.

Original publication




Journal article


American Journal of Roentgenology

Publication Date





W424 - W431