Computer Aided Diagnosis for Diabetic Retinopathy based on Fundus Image

Diabetic Retinopathy (DR) is an eye abnormality caused by long term diabetes, which can lead to vision defects or even blindness. Performing retinal screening examinations on all diabetic patients is a hard work due to the limited number of specialists cannot keep up with the increasing prevalence of diabetes, and hence there are many undiagnosed and untreated cases of DR. Computer aided diagnosis (CAD) is a good way to save the patient's vision and to help the ophthalmologists in mass screening of diabetes sufferers. The main purpose of the proposed study is to design an automated grading approach for DR screening, using a publicly available database of retinal images, which can evaluate the fundus images like human experts while achieving a high sensitivity for the detection of DR. This paper gives a summary of the results obtained by our previous studies. Together with these results and the DR grading schemes, we provide an automatic analysis approach for the risk of macular edema and the retinopathy grade. Experimental results indicate that the performance of our approach on this database is comparable to that of human experts.