Artificial intelligence in cardiovascular care for internal medicine: from promising algorithms to useful clinical services.

Abstract

Artificial intelligence is expanding rapidly in cardiovascular medicine, but its value in internal medicine depends less on raw model performance than on whether it improves triage, risk targeting, and coordination of care within real-world workflows. Rather than cataloguing applications, this review focuses on three use cases with near-term relevance in internal medicine: AI-augmented electrocardiography, risk prediction for targeted prevention, and AI-enabled clinical decision support, including selected applications of large language models. Across these domains, evaluation should extend beyond discrimination to calibration, positive predictive value at action thresholds, net benefit, alert burden, override rates, and downstream testing. Cardiovascular AI should not be judged as an autonomous decision-maker, but as a supervised component of care delivery. Its usefulness depends on predefined confirmation pathways, integration into existing information systems, actionable outputs, user training, and post-deployment monitoring for drift, safety, and equity. We propose a practical implementation lens - Train, Explain, Integrate, Accompany (TEIA) - to structure deployment in routine care, and summarize governance issues relevant to European practice, including intended use, interoperability, traceability, cybersecurity, and lifecycle oversight. The central question is not whether an AI tool can classify or predict, but whether its use in routine care supports better decisions with acceptable workload, safer pathways, and measurable clinical value.