Research On Automatic Optic Disc Localization And Lesion Detection In Digital Fundus Images

Diabetes is one of the most common chronic diseases. As people's living standards improve, the incidence of diabetes has shown an upward trend year by year. Diabetic retinopathy is both a serious retinal complication associated with diabetes and a main reason for blindness and visual impairment. Retinal screening has become an important means of early detection. In this paper, through the analysis of fundus images, we try to develop an automatic diabetic retinopathy diagnostic system. The optic disc localization is one of the key steps in automatic diagnostic system. Due to the impact of light and lesions, localizing optic disc accurately is difficult. To bypass this difficulty, A fast and relatively robust method for optic disc localization using the main blood vascular network is proposed. In this method, the main blood vessels are first extracted in the H channel (chroma) of the HSV images and their directions are detected, then the pixel with the highest response of a weighted matched filter is chosen as the center of optic disc. On the other hand, the exudate detection is also an important component of the automatic diagnosis system, we present a corresponding method by combining brightness and edge intensity. In this detection method, firstly finding the candidate regions using edge sharpen and area filling and then verifying the regions with the high brightness areas obtained by morphological grayscale reconstruction. Experiments show that the method is not only effective but also can distinguish the hard exudates from the soft exudates.Finally, an automatic diabetic retinopathy diagnostic system is preliminarily implemented using OpenCV(Open Computer Vision) under the framework of MFC. The demonstration system can provide multiple functions of the system such as blood vessel segmentation, optic disc localization, fovea and hard exudates detection etc. The system can assist the doctor's diagnosis and tracking disease. By further improvement, the system can be easily extended to the automatic diabetic retinopathy diagnostic system.