AUTOMATIC OPACITY DETECTION IN RETRO-ILLUMINATION IMAGES FOR CORTICAL CATARACT DIAGNOSIS Huiqi Li* a , Liling Ko a,b , Joo Hwee Lim a , Jiang Liu a , Damon Wing Kee Wong a , Tien Yin Wong c , Ying Sun b a Institute for Infocomm Research, A*STAR(Agency for Science, Technology and Research), Singapore b Department of Electrical & Computer Engineering, National University of Singapore c National University of Singapore; Singapore Eye Research Institute ABSTRACT Computer aided analysis of medical images, a unique type of non-text media, can facilitate clinical diagnosis. As an example, an automatic opacity detection approach is proposed in this paper to grade cortical cataract more objectively. The automatic pupil detection is performed by detecting the strongest edges on the convex hull and ellipse fitting using nonlinear least square method. The cortical opacity is detected by radial edge detection and post- processing. The automatic grades are assigned following Wisconsin cataract grading protocol. The accuracy of pupil detection is 98.2%. The mean error of opacity area detection is 7 percent compared with the result of human grader. And 86.3% accurate grades of cortical cataract are achieved. This is the first time that the spoke-like feature is utilized in the automatic detection of cortical cataract to separate from other opacity types. The encouraging results show that it is probable to apply the proposed approach to clinical diagnosis later. Index Terms— medical image, opacity detection, cortical cataract. 1. INTRODUCTION Medical images are a unique and important type of non-text media. It is one essential means to facilitate clinical diagnosis. Currently most of the medical images are analyzed by human graders. There are two main issues for such analysis. One issue is that the grading is usually subjective, which means the inter-grader reliability is low. Time-consuming is another problem that bothers human grader for quantitative measurement. Computer aided analysis of medical images can help in some way to resolve these two challenges. In this paper, automatic detection of opacity for cortical cataract grading is proposed as an example of computer aided diagnosis based on medical images. Cataracts are the leading cause of blindness worldwide. It was reported that 47.8% of global blindness is caused by cataract [1]. A cataract is due to opacity or darkening of crystalline lens. Cataracts are classified into three types according to the location of opacity: nuclear cataract, cortical cataract, and posterior subcapsular (PSC) cataract. Cortical cataract which occurs in the cortex (or periphery) of the lens is reported to be the most prevalent type (44.7%) of cataract in some studies [2]. Retro-illumination images are taken for grading cortical and sub-capsular cataracts. There are multiple grading systems established [3-4], which are based on similar principle. A standard set of images with increasing cataract severity are assigned consecutive integer grades. Ophthalmologists compare the picture observed with the standard set to assign a reasonable grade, which is termed as clinical grading or subjective systems. In order to classify the lens opacity more objectively, human graders are trained to classify cataracts based on photographs or digital images [5-6], which is termed as grader’s grading or objective system. But studies showed that the reproducibility of intra-grader and inter- grader measurement is still not high [5]. Some efforts have been put in the development of automatic systems to improve grading objectivity. For nuclear cataract classification, automated systems were proposed using slit-lamp images [7-8]. For cortical cataract and PSC, the current methods employed so far are relatively simple. Nidek EAS-1000 software [9] extracts opacities based on the global threshold principle, with the threshold value picked as 12% from the highest point. There is no distinction between opacity types and its pupil detection is manual. The user may manually select the threshold value if automatic detection is not satisfactory. Opacity detection by global thresholding is often inaccurate due to non-uniform illumination of the lens. There is an upgrade of the software by researchers [10]. First improvement is to detect pupil automatically as a circle of 95% of the maximal radius detected. Second improvement is the opacity detection by contrast based thresholding. This contrast based approach has its limitation too when opacities are so dense that the contrast in the opacified areas is no longer high. Still there is little effort made in the separation of the opacity types. An automatic opacity detection approach is proposed in this paper for the purpose of objective grading of cortical cataract. The workload of graders can be saved as well in 553 978-1-4244-2571-6/08/$25.00 ©2008 IEEE ICME 2008