A Study on Machine Learning Based Diabetic Retinopathy

Diabetic Retinopathy (DR), one of the main causes of blindness in the eyes, is curable in its early stages. The first indicators are red lesions, which include both micro aneurysms (MAs) and hemorrhages (HEs). Diabetic Retinopathy diagnosis by color fundus images is challenging and time-consuming as it takes trained doctors to recognize the presence and significance of many characteristics, as well as a complex classification method. Even though various Machine Learning based automatic Diabetic Retinopathy diagnosis approaches exist, they are still unable to offer point-of-care diagnosis, which would allow one to upload an input image, segment the retinal blood vessels, examine each vessel's expansion, and provide accurate result. Machine Learning and Deep Learning techniques have swiftly gained traction to understand medical images with promising results. Machine Learning can quantify and get more accurate results for the challenges mentioned above, as well as calculate the severity level of the given input eye image as an accuracy rate or predict Diabetic Retinopathy through feature extraction. In this research, a study is conducted to aid in the development of better Diabetic Retinopathy based models, with the aim of enhancing the detection accuracy of Diabetic Retinopathy by evaluating existing Machine Learning models and mobile applications with multiple datasets.