Efficient local cloud-based solution for diabetic retinopathy detection

Damage of blood vessels in retina due to diabetes is known as diabetic retinopathy. It is one of the one of the important origins of blindness for adults. Loss of vision can be avoided by detecting damage of retina (leaking fluid or blood). Efficient local cloud-based solution for diabetic retinopathy detection is designed in the work, where convolution neural network is used for training and classification module and achieved an accuracy of 86% using kappa metric. Fundus images are used for training and classification. System network architecture is derived from VGGNet. Network is trained using 80,000 images. Since everything is automated, a doctor is only required for treatment, not for diagnosis.