Image based diagnosis of cortical cataract

An automatic approach to detect cortical opacities and grade the severity of cortical cataract from retro-illumination images is proposed. The spoke-like feature of cortical opacity is employed to separate from other opacity type. The proposed algorithms were tested by images from a community study. The success rate of region of interest (ROI) detection is 98.2% for 611 images. For 466 images tested, the mean error of opacity area detection is 3.15% compared with human grader and 85.6% of exact cortical cataract grading is obtained. The experimental results show that the proposed approach is promising in clinical diagnosis.