A Study on Diabetic Retinopathy using Deep Learning Algorithms
Shobhana Khanapur, Lakshmi Patil
Abstract
The Diabetic Retinopathy (DR) is a widespread difficulty of diabetes mellitus, which begins lesions on retina and it affects a vision which leads to blindness. A physical color fundus images screening detects DR at an early stage is computationally expensive and consumed a much time. Consistently, a diagnosis of automated DR has become a basis of research recently because of immense growth of diabetic patients. And it is very difficult to identify the disease features in the images at the early stages of the disease. Machine learning based medical image analysis has come up with the good results of fundus images and early diagnosis of Diabetic Retinopathy (DR) is possible with the application of Deep Leaning Techniques. The paper discusses more about the Detection, Segmentation and Classification of Diabetic Retinopathy with available datasets. The main moto of the paper is not just early identification of diabetic Retinopathy, but also takes a major role in detecting the stage of defect like normal, mild, moderate o severe. Many papers have proposed different methods to study the diabetic retinopathy but no paper has presented the case of limited training dataset. It has become the major challenge of this paper. The segmentation of eye’s vasculature can be done manually [21] with the help of expertise but it is very tedious and time consuming and also requires extra attention. Study of research gaps in the field of DR Segmentation and Classification leads to the challenges and investigation has also included in this paper.