Artificial Intelligence Driven Healthcare - A Survey of Machine Learning, Deep Learning and Intelligent Predictive Methods

Abstract

Artificial intelligence has moved from a promising research direction to a working component of modern clinical practice, supporting screening, diagnosis, prognosis, treatment planning and hospital operations. This survey reviews the principal families of methods that drive this transformation: classical machine learning models built on engineered features, deep learning architectures that learn representations directly from images, signals and records, and intelligent predictive frameworks that combine learning with attention, graphs, generative modeling, reinforcement and federated optimization. We trace the methodological evolution from support vector machines and ensemble trees to convolutional, recurrent, transformer and graph neural networks, and we examine landmark clinical results in medical imaging, physiological signal analysis and electronic health record modeling. Comparison tables consolidate representative studies across modalities, tasks, architectures and reported outcomes, and mathematical formulations of the core models are provided to make the survey self-contained. We further review privacy-preserving and communication-efficient federated learning, blockchain-assisted data integrity, explainability techniques and the documented risks of bias and adversarial fragility. Finally, we identify open challenges in generalization, validation, deployment and regulation, and outline research directions toward multimodal foundation models and trustworthy clinical decision support. The survey spans foundational algorithms, landmark clinical evaluations and advances through 2026, and is intended as a structured entry point for researchers and practitioners.