Visual Search Patterns in Roentgen Diagnosis

Abstract

The surprising frequency of reader error in roentgen diagnosis constitutes a challenging problem, the magnitude of which has been ably documented by Garland (6, 7). Some avoidable reader error results from our failure to record critical shadows. More commonly, however, reader error appears to be due to failure to perceive shadows which have been adequately recorded. More precise control of “film quality” than is already possible with existing equipment and materials is not likely, therefore, to increase greatly the accuracy of roentgen diagnosis but greater understanding of the nature of perceptual error may result in significantly improved performance (9). The factors responsible for errors of perception are, of course, most elusive, but it seems reasonable to assume that they derive, first, from the reader's failure to scan or search the film adequately, and second, from his failure to organize the recorded images effectively. Despite the fundamental importance of visual search in roentgen diagnosis, there have been, to our knowledge, no objective studies of film-scanning patterns; our concepts of optimum visual search technics are based almost wholly on introspective analysis. Objective studies (3, 4) of visual search patterns in photographic interpretation suggest that these concepts may be misleading. It appeared useful, therefore, to undertake an objective study of the visual search technics of radiologists in an effort to elucidate the physiologic processes involved in roentgen diagnosis and to test the validity of present concepts of optimal film-scanning patterns. Method: Various methods have been described for analyzing search patterns by recording the eye movements of observers (1, 2, 5), but because of its simplicity, the following modification of a technic described by Brandt was ultimately selected for this pilot study. A series of paper roentgenograms of normal and abnormal chests were prepared and presented for interpretation to each of the trained radiologists selected as subjects. The presentations were made with a fixed reading distance in a dimly lighted room. The observers were provided with a spotlight source of variable diameter with which to illuminate the roentgenograms. During practice runs, they were told to move the light as they wished over the surface of the film, but to adjust the source to illuminate, at any instant, an area no larger than they thought was necessary for “comfort and accurate interpretation.” No subsequent adjustment of the light source was permitted. The readers were finally told that they would be rated on the accuracy of their interpretations; speed and search pattern were not mentioned.