Dr. Satish Chander Gupta, G1468, Dr. Abhishek Dagar
Purpose: To present a new software for automatic detection and grading of NPDR from fundus images.
Methods: This computer vision-based algorithm is based on finding abnormalities, such as exudates and red lesions in digital fundus image. The grading is based on
location of these abnormalities in the image referred with its distance from the macula
Results: Overall accuracy of 90% was achieved in the proposed method and also significantly reduce the computational complexity in this region-based approach.
Conclusion: The computer based automatic detection of abnormalities in fundus image and grading of NPDR is a very useful tool for screening the diabetic population for presence or absence of NPDR. The grading of lesions can be very
effective in tele ophthalmology for selecting the patients for referral to a centre for further treatment.