Early Detection and Prediction of Cataract Using Deep Learning

The lens of the eye becomes dull or cloudy when it develops a cataract. This is among the most prevalent illnesses that can lead to blindness. We suggest using computer science to automatically diagnose cataracts based on how retinal images are categorised in light of the harm that cataracts can cause. This technique concentrates on the retinal image feature extraction process. First, the fundus images are preprocessed using the maximum entropy approach. Next, we automatically extract more distinguishing features from fundus images using a deep learning network built on the Caffe framework. Lastly, to identify automatically derived characteristics, a variety of representative classification techniques are applied. VGG-19 (Visual Geometry Group) is employed for cataract compared using wavelet characteristics derived from the retinal vasculature and deep learning features.In this work, a VGG-19 model is proposed to detect cataracts in retinal fundus pictures. More layers are added to the foundation VGG-16 model to produce the proposed model. A preprocessed dataset has portions for training and validation. The suggested model is further validated after being trained on the training data, and an accuracy of 100% is reported. Following the feeding of the testing data into the model, a high level of testing accuracy of 96.79% is generated.