AI Privacy and Security in Healthcare: A Systematic Literature Review

Background: Artificial intelligence is expanding into telemedicine and telerehabilitation, yet significant privacy and security concerns persist. Scope: To synthesize empirical evidence on privacy and security approaches in health care, particularly those relevant to distributed home care. Methodology: A systematic review identified 80 studies (2019 to 2025), and Latent Dirichlet Allocation (LDA) topic modeling characterized the privacy and security themes. Results: Sixty-six studies addressed privacy, only seventeen addressed security, and three studies addressed both. LDA identified four themes: patient data privacy, federated learning for medical imaging, encrypted training and secure computation, and healthcare data governance. Most studies emphasized privacy-preserving approaches, like federated learning, encryption, and differential privacy. Almost half were conducted outside healthcare environments, limiting insight into real teleclinical and telerehabilitation workflow. Conclusion: Securing healthcare AI will require a multi-layered governance framework, broader global representation, and integration of privacy and security protections into routine clinical workflows.