DMERCNET: A Quirky Deep Learning Classifier for Diabetic Retinopathy and Macular Edema Risk Using CNN

Since past two decades the research in Deep Learning is highly accelerative. The power of deep learning is being unleashed by using it widely in numerous applications and domains. Health care is not an exception. Besides pharmaceutical operations and Clinical diagnosis are also far and wide deploying the technology. May it be pediatric or heart related; from fever to Cancer diseases; deep learning is potentially being used and the yield seems to be promising. Diabetic patients are primarily prone to Diabetic Retinopathy (DR) in which Diabetic Macular Edema (DME) plays a major role. In this research splinter we used convolution neural networks a subsidiary of deep learning to classify the DME grade by predicting the level of DR as well the risk factor of DME. The novel deep neural network architecture in the form of CNN we implemented, exhibited notable performance resulting in 87.38% accuracy and around 1% loss. We made this research on the Indian Diabetic Retinopathy Database (IDRID) which consists of fundus images of eye.