Analysis and Visualization of Medical Images using Eye-tracking

Clinical procedures rely heavily on visual data, particularly in diagnostic radiology. Eye-tracking technology was employed in this study to examine the relationship between visual search features and diagnostic performance in capsule endoscopy. Conventional visual search models, developed for static 2D images, may not be suitable for cross-sectional stack imaging used in computed tomography (CT) and magnetic resonance imaging (MRI) due to their dynamic nature. The study aimed to address this gap by analyzing the visual search patterns of an expert examining capsule endoscopy images. A framework integrating an eye tracker was developed, and a dataset of 536 images was collected. The eye-tracking system showed high accuracy in capturing gaze data. Expert visual search patterns were found to follow predominantly circular trajectories, with distinct optical flow characteristics. Variations in fixation du ration and location were noted, with certain image types exhibiting drilling and scanning patterns. These findings highlight the potential for using eye-tracking data to improve diagnostic accuracy and contribute to radiology education.