The integration of artificial intelligence (AI) into cardiovascular medicine is no longer a speculative endeavor but an inevitable and rapidly advancing transformation. Recently in the Journal of American College of Cardiology, Harlan M. Krumholz, MD, MSc, articulates a framework for guiding this revolution responsibly, underscoring both its potential and its perils. AI possesses the capacity to enhance diagnostic accuracy, personalize treatment, predict disease onset before clinical manifestation, and accelerate translational research across the entire cardiovascular continuum. However, the trajectory of AI’s impact will depend on deliberate, principled guidance.
Krumholz proposes five guiding principles essential for ensuring AI advances patient care safely and equitably:
- Start With Problems, Not Technology. Sustainable innovation must target well-defined clinical challenges rather than deploying algorithms in search of application. The PRESENT-SHD study exemplifies this approach by developing an ensemble deep learning model to enable scalable, low-cost screening for structural heart disease using widely available ECG data formats.
- Insist on Rigorous Evaluation. AI systems must meet evidentiary standards comparable to pharmacologic and device interventions, including external validation, calibration, and decision-curve analysis across diverse populations. The gap between technical performance and clinical utility remains a barrier to adoption, necessitating robust, reproducible real-world validation.
- Protect Against Harm and Bias—And Actively Reduce Them. AI trained on biased datasets risks perpetuating health inequities. Equitable performance and deployment are mandatory, with AI’s potential to correct systemic biases highlighted by studies demonstrating sex-specific recalibration of hypertrophic cardiomyopathy diagnostic thresholds, improving detection in women.
- Emphasize Implementation. Translational impact hinges on integrating AI seamlessly into clinical workflows, ensuring interoperability, usability, and maintainability. Thoughtful design—such as making tools compatible with existing hardware and data formats—enables broad deployment, particularly in resource-limited settings.
- Build Systems of Trust and Governance. Trust is foundational for adoption. Transparency, explainability, formal performance monitoring, and alignment with emerging global ethical and regulatory frameworks (e.g., the EU AI Act, WHO guidance) are required to embed AI within a trustworthy quality infrastructure.
Krumholz concludes that if the cardiovascular community adheres to these principles, AI will not merely represent another technological advance but a transformative partner in the quest to eradicate cardiovascular disease. The author stresses that the current moment is a critical inflection point—one that requires active stewardship to ensure AI becomes a force for equitable progress rather than a source of unintended harm.
Original article: https://www.jacc.org/doi/10.1016/j.jacc.2025.08.021
