Improved grading and diagnosis of diabetic macular edema using artificial intelligence

Abstract

Diabetic Macular Edema is a potentially threatening complication characterized by the collection of fluid in the macula. A timely and precise diagnosis of diabetic macular edema is important for effective clinical management and avoidance of vision loss.An innovative methodology for enhancing the grading and diagnosis of diabetic macular edema using convolutional neural network, recurrent neural network, and long short-term memory network. A detailed view of the retina is obtained using the fundus images. They form the fundamental source of information. Traditional grading systems are obtained through a manual process done by ophthalmologists. These are a time-consuming process. The first phase in the detection of DME using deep learning involves the CNN architecture which helps to extract information from the fundus images. They are trained on a large dataset that involves various information including exudates, hemorrhages, and microaneurysms. The introduction of temporal and spatial relationships is achieved through RNN and LSTM layers. This helps to obtain sequential information. They also play a prominent role in detecting the progression of diseases. The fundus images are collected through various sources including ethnic and various demographics. The tentative outcomes validate the dominance of the existing system in terms of improved accuracy, precision, sensitivity, and specificity.