An optimal prediction of cardiovascular disease using deep learning techniques in diabetes mellitus

Abstract

This research thesis presents the development and evaluation of a newlinedeep learning model aimed at predicting early-stage cardiovascular disease newlineamong individuals with diabetes. By incorporating complications such as foot newlineulcers, glaucoma, and chronic kidney diseases, this study seeks to enhance the newlineaccuracy of risk assessment through advanced deep learning techniques. The newlineresearch is organized around three distinct objectives. The first objective involves the creation of a deep learning model that leverages images of foot ulcers in diabetic patients to detect cardiovascular disease risk. This model integrates a Deep Recurrent Neural Network for effective feature extraction and a Pre-trained Fast Convolutional Neural Network for subsequent classification. The second objective focuses on exploring the potential of utilizing glaucoma fundus images from diabetic patients to predict cardiovascular disease. This objective includes an investigation into network pruning techniques to extract relevant features by eliminating less significant ones. Additionally, the use of a genetically optimized neural network enhances the classification accuracy newline newline