Henning Schubert

Eye movements during reading of CT scans: How do radiologists make their diagnosis?

H. Schubert¹, S. Elsner¹, R.W. Günther¹, W. Huber², J. Neuser³

¹Department of Diagnostic Radiology, Universitätsklinikum der RWTH Aachen, D-52057 Aachen, Germany (e-mail:schubert@rwth-aachen.de);
²Neurolinguistics, Department of Neurology, Universitätsklinikum der RWTH Aachen, D-52057 Aachen, Germany;
³Institute of Medical Psychology, Universitätsklinikum der RWTH Aachen, D-52057 Aachen, Germany

Introduction. Since many years eye movements in cognitive psychology are recorded to analyse structure and use of knowledge. Analysing eye movements during reporting in radiology has been an issue since the end of the 1960's (Kundel & Wright, 1969, Radiology, 93: 315-320). But almost all studies since then use conventional radiographs for their experiments. The analyses were focused on visual search strategies during the viewing of chest radiographs, satisfaction of search looking for lung nodules and detecting tumours in Mammograms (Beard, 1997, Journal of Digital Imaging, 10: 14-20). There are only a few studies analysing eye movements reading computed tomography (CT) scans. (Beard, 1990, Journal of Digital Imaging, 3: 230-237).

Hypothesis. Since eye movements are conducted by cognitive processes in reading (Just & Carpenter, 1980, Psychological Review, 87: 329-354), we expect that also eye movements during reading CT scans reflect reading-strategies of the radiologist (as they do in reading conventional radiographs and text) and that radiologists with different expertise use different strategies. To know and understand these reading-strategies will help:

to increase teaching efficiency for students and residents
to avoid missing important pathological findings
to develop computer aided scanning models
to design workstations for CT interpretation under ergonomic issues

Method. The study consists of reading 3 different sets of CT's: cranium, chest and abdomen. Each set consists of 5 normal scans and 15 scans with pathological findings: 5 lesions that are easy, 5 that are moderate and 5 that are difficult to detect. Eye movements were registered with the pupil-corneal-reflection method (Debic 84). We will examine the following subjects:

10 radiology residents with little CT experience
10 board-certified radiologists with great CT experience

Reading time for each item is 1 min. After reading, the subject has to report the findings and differential diagnosis looking on the blank screen. Eye movements are recorded continuously. The data analysis comprises number and duration of fixations and refixations, the fixation path and correctness of findings and diagnosis.

Results. Acquisition and data analysis is not completed yet but first results show differences of eye movements between the two subject groups.