A Hybrid Deep Learning-Metaheuristic Model for Diagnosis of Diabetic Retinopathy

International Diabetes Federation (IDF) reports that diabetes is a rapidly growing illness. About 463 million adults between 20-79 years have diabetes. There are also millions of undiagnosed patients. It estimated there will be about 578 diabetics by 2030 [1]. reasons different eye diseases. Diabetic retinopathy (DR) one them and the most common vision loss or blindness worldwide. DR progresses slowly has few indicators in early stages. makes diagnosis problematic task. Automated systems promise to support DR. Many deep learning-based models been developed for classification. This study aims ophthalmologists process increase performance through hybrid model. A publicly available Messidor-2 dataset was used this study, comprised retinal images. In proposed model, images were pre-processed, learning namely, InceptionV3, feature extraction, where transfer approach applied. Next, number features obtained vectors decreased with selection Simulated Annealing. Lastly, best representation XGBoost The algorithm gives an accuracy 92.55% binary classification shows pre-trained ConvNet metaheuristic satisfactory result