A systematic review on Machine learning and Neural Network based models for disease prediction

Abstract

Over the recent years, conventional artificial intelligence (AI) has witnessed a significant infusion of machine learning and neural networks, marking a substantial evolution in various domains due to their autonomous capacity for feature acquisition and remarkable efficiency. Particularly in the medical field, machine learning-based models have outperformed physicians, exhibiting greater accuracy. Diseases such as cancer, Alzheimer's, dyslexia, skin diseases, and heart diseases have become focal points in medical research. Several deep learning methods, including Convolutional Neural Networks (CNN), Support Vector Machines (SVM), Random Forest, Logistic Regression, Decision Tree, and Recurrent Neural Networks (RNN), play crucial roles in disease prediction. This survey emphasizes the critical analysis of which deep learning models achieve higher accuracy in predicting specific diseases. The objective is to shed light on existing shortcomings in disease prediction and propose potential remedies for future improvements. Results indicate that Convolutional Neural Networks excel in predicting heart and Alzheimer's diseases, as well as breast cancer. Support Vector Machines demonstrate effectiveness in cancer prediction, while logistic regression proves adept at predicting dyslexia, and decision trees emerge as a favorable choice for skin diseases. Looking ahead, the integration of digital twins for predictive analytics, facilitating the simulation and modeling of disease progression based on individual patient characteristics, and leveraging blockchain for secure storage and sharing of health data represent promising avenues for future developments.