To improve the accuracy achieved in the diabetic retinopathy detection neural network models and implement it on a portable mobile device

Abstract

Diabetes is one of the leading causes of deaths across the world. Prolonged patients of diabetes often face complications in other organs as well such as eyes, heart, and limbs. The growing population combined with increasing urbanization and the associated lifestyle has led to increase in the cases of diabetes as well all across the world. The along with the increase in the number of diabetes patients the cost of healthcare is also steadily increasing which will prove to be a big problem for the lower sections of society in the future in the present state of healthcare. In light of the above the need of the hour a scalable solution which can aid the providing healthcare and is economical. The project aims to tackle the problem of screening of people who suffer from diabetic retinopathy (DR), which is a complication in the eyes caused by diabetes. Retinal images are fed into a convolutional neural network to classify the images as positive or negative for DR. The moonshot for the project is to develop a mobile application running the model to detect diabetic retinopathy using the phone camera and the processor to provide a scalable and easily accessible solution