A pilot study of eye movement during mammography interpretation: eyetracker results and workstation design implications.

Digital mammography can potentially improve mammography image and interpretation quality. On-line interpretation from a workstation may improve interpretation logistics and increase availability of comparison images. Interpretation of eight 4k- x 5k-pixel mammograms on two to four 2k- x 2.5k-pixel monitors is problematic because of the time spent in choosing which images display on which monitors, and zooming and roaming on individual images that are too large to display completely at full resolution. The authors used an eyetracker to measure radiologists viewing behavior during mammography interpretation with film on a viewbox. It was observed that a significant portion of the mammographers' time is spent viewing "comparison pairs" (typically two or more comparisons per case), such as the left mediolateral and craniocaudal images or old and new images. From the eyetracker measurements, we estimated that the number of image display, roam, and zoom operations decreases from an average of 64 for one monitor to 31 for four monitors, with the largest change going from one to two monitors. We also show that fewer monitors with a faster response time is superior to more monitors with a slower response time. Finally, the authors demonstrate the applicability of time-motion analysis to mammographic workstation design.

Effects of delayed processing on mammographic phantom object detection.

OBJECTIVES: Delayed processing of films is a common occurrence in mobile mammography screening programs. The effects of such delayed processing on radiologists' detection of phantom test objects are investigated. METHODS: Twelve screen-film combinations were exposed using a phantom and developed after delays of 0, 1, 2, 3, 5, and 7 days. Films were scored by three board-certified radiologists specializing in mammography, based on visibility of calcifications, masses, fibrils, and line pairs. RESULTS: Variance analyses of mean scores indicate that, across screen-film combinations, there are no statistically significant decreases in object detection associated with delayed processing. The analyses do indicate, however, a statistically significant difference in scores unrelated to delay, attributable to the screen-film combination used. CONCLUSIONS: Delayed processing of mammography film does not lead to a decrease in the visibility of phantom test objects, despite large decreases in overall image optical density.