Analyzing the eye movement of dentists during their reading of CT images

In order to evaluate the eye movements of dentists when they were interpreting radiographs, ten normal computed tomography (CT) images and ten images with pathologic lesions were shown to eight dentists, and the pattern of their eye movement was qualitatively analyzed. Six fixation point parameters were calculated, including the time required to discriminate between normal and pathologic images (X1), the total fixation point count (X2), the total travel distance between fixation points (X3), the average time spent on each fixation point (X4), the total gaze fixation time (X5), and the maximum gaze fixation time spent on each image (X6). When the subjects were interpreting pathologic images, X2 and X4 were shorter; however, when they were viewing normal images, time was spent on observing multiple fixation points before the completion of the interpretive process. While pathologic images were recognized through top-down processing, there was a tendency for normal images to be recognized through bottom-up processing. The results of discriminant analysis, using a linear discriminant function, indicated that the independent variables X2 and X4 and the dependent variable X5 were the only variables that contributed significantly to differentiating between normal and pathologic images. The linear discriminant function was Z = 9.0 × 10−2 × X2 + 3.0 × X4 − 2.1 (discriminant score: Z < 0, pathologic image; Z ≧ 0, normal image). When the mean value of each individual's gaze fixation data was substituted into the discriminant formula, the hit rate for normal and pathologic images was discriminated at 94% (15 of 16).