Current State of Artificial Intelligence in Assessing Cardiac Function.
Authors
Affiliations (5)
Affiliations (5)
- David Geffen School of Medicine, University of California, Los Angeles, USA.
- Department of Cardiology, Smidt Heart Institute, Cedars-Sinai Medical Center, Los Angeles, CA, USA.
- Division of Research, Kaiser Permanente Northern California, Pleasanton, CA, USA.
- Division of Research, Kaiser Permanente Northern California, Pleasanton, CA, USA. [email protected].
- Department of Cardiovascular Medicine, Stanford University School of Medicine, Stanford, CA, USA. [email protected].
Abstract
Accurate, timely quantification of cardiac function is central to the diagnosis, management, and monitoring of cardiovascular disease. This review synthesizes recent advances in artificial intelligence (AI) applications across the major data modalities used in cardiovascular medicine, spanning electrocardiography (ECG), echocardiography, cardiac CT/MRI, and clinical text in electronic health records (EHR). State-of-the-art deep learning algorithms now enable highly accurate assessment of cardiac disease across a range of modalities. These models excel in detecting subclinical cardiovascular disease, occult disease etiologies, and ventricular dysfunction that may elude conventional interpretation. Recent randomized controlled trials demonstrate that AI models can match or even outperform clinicians in identifying myocardial infarction from ECGs, occult atrial fibrillation from sinus rhythm ECGs, and in quantifying left ventricular ejection fraction from echocardiography. Concurrently, the emergence of foundation models and multimodal architectures is accelerating label-efficient learning, enabling automated report generation, and facilitating scalable population-level screening across diverse clinical settings. AI is poised to transition from proof-of-concept to indispensable clinical partner in cardiology. Robust multicenter validation, open-source code transparency, and prospective trials are essential to confirm generalizability and to quantify patient-level benefit. As foundation models mature and multimodal learning becomes routine, AI will enable scalable screening, precision phenotyping, and more equitable cardiovascular care-particularly in resource-limited settings-while allowing clinicians to refocus on patient-centered practice.