A perceptually-based algorithm provides effective visual feedback to radiologists searching for lung nodules

The gaze durations of fixation clusters generated by radiologists searching for nodules in chest images varies according to the degree of interest in the local image structure. The authors present a computer algorithm that filters out short-duration clusters, feeding back only the image locations fixated for >1000 ms, measured during the initial search of the image. The feedback algorithm highlighted 72% of initially fixated nodules, and 34% of initially unreported nodules were correctly identified in a second view with feedback, compared with only 21% without feedback. This computer algorithm, based on the radiologist's gaze duration data, effectively localizes potential nodule sites for feedback and enhances diagnostic performance.<
>