Previous studies investigating eye movements of radiologist during their image analysis have recorded gaze data on single images containing one or more objects, e.g., nodules in a CT scan. Typically the images are displayed for a fixed duration and the results are illustrated through aggregated visualizations e.g., fixation plotting, scan paths or heat maps. However, when analyzing search patterns while performing standard interpretation of volumetric chest CT data, paging through a stack of transverse sections warrants three-dimensional analysis of eye movements. To our knowledge this has not been studied.
Materials and Methods
Our solution for recording volumetric gaze data, using a remote corneal reflection system, relies on two-way communication between the eye tracker and our custom DICOM viewer. This enables real-time mapping between on-screen fixations and the 3D DICOM coordinates of the viewed images at a rate of 50 Hz, thereby linking our eye tracker data and the physical space in which the 3D image data were obtained.
Our approach to tracking and visualizing gaze data recordings within the acquired volumetric space utilizes the DICOM coordinate system for rendering graphics in 3D space. It produces several interesting visualizations, such as time-stamped 3D gaze paths, and 3D heat maps revealing direct and foveal dwell time per voxel, all fused with the acquired volume data for simultaneous viewing.
Our approach offers insight into the relationship between recorded gaze data and the volumetric space of the viewed medical images. We are employing it in an investigation of how paging through volumetric lung CT data affects nodule detection though pop-out feature detection in the peripheral visual field. More generally it will facilitate studies across a wide variety of 3D modalities and illuminate specific aspects of volumetric inspection that can be used for training and diagnostic support.