Quantitative metrics to characterize radiologists’ search patterns during multiparametric MRI assessment for prostate cancer

Assessing prostate cancer (PCa) on multiparametric magnetic resonance imaging (mpMRI) has been associated with interobserver variability. Studying the search patterns of expert radiologists captured using eye tracking technology can improve lesion identification accuracy and minimize interobserver variability. However, no studies have investigated radiologist search patterns in PCa mpMRI. The objective of this pilot study was to develop a 3-dimensional (3D) mpMRI eye tracking platform and metrics to characterize radiologists’ search patterns. Four board-certified radiologists searched for PCa on mpMRI in 40 patients with known PCa while eye tracking data were collected. Quantitative metrics were developed and included characterizing physical eye movements, 3D and multiparametric search patterns, and search patterns in relation to structures. Saccade magnitudes and fixation durations varied for the different radiologists, while the relative ordering of radiologists’ saccade magnitude and fixation duration remained consistent between MR images. Among the four radiologists, one searched mpMRIs slice-by-slice (scanning), one scrolled through slices (drilling), and two used a combination. More time was spent fixating and covering the prostate defined on the T2-weighted image compared to the other images. Here, a novel platform to enable 3D mpMRI eye tracking was developed and quantitative metrics identified search pattern differences descriptive of radiologists’ PCa search patterns.