Ask AI
Back Back
Actionable AI in cardiology
Actionable AI in Cardiology: Addressing Current Limitations and Future Directions

Released: October 28, 2025

Pierre Elias
Pierre Elias, MD
Sammy Elmariah
Sammy Elmariah, MD, MPH
Francisco Lopez-Jimenez
Francisco Lopez-Jimenez, MD, MSc, MBA
Evangelos K. Oikonoumou
Evangelos K. Oikonoumou, MD, DPhil
Louise Sun
Louise Sun, MD, SM, FRCPC, FAHA
Activity Information

Activity

Progress
1
Course Completed
Continue
Key Takeaways
  • Artificial intelligence (AI) is transforming healthcare delivery, especially in cardiology practice, due to its ability to predict outcomes and inform actionable clinic workflows.
  • A 2024 JACC state-of-the-art review on AI provides a structured roadmap for developing equitable and reliable algorithms that are validated for real-world cardiology practice.
  • The future of AI requires robust regulation and the integration of cardiac imaging, genomics and biomarkers, as well as electronic health record data to assist healthcare professionals in delivering multimodal, precision cardiovascular care.

Artificial intelligence (AI) continues to revolutionize how healthcare professionals (HCPs) detect, stratify, and manage cardiovascular disease (CVD), as the emphasis shifts from innovation to implementation. The 2024 Journal of the American College of Cardiology (JACC) state-of-the-art review by Rohan Khera, Evangelos K. Oikonomou, and colleagues, argues that actionable AI in cardiology requires more than just high-performing algorithms. It demands a system that connects AI-driven insights to care decisions with real-world effectiveness, is equitable in its performance and deployment, and is overseen by robust and novel regulatory pathways. In addition, the review identifies 3 distinct stages of AI integration in cardiology:

  • Discovery—AI algorithms generate new insights and/or knowledge (ie, biomarkers, drug targets)
  • Translation—AI insights are validated across populations and linked to clinical workflows
  • Deployment—AI becomes an embedded, auditable part of cardiology practice

This 3-tiered approach highlights the need for alignment across all staff, including the developers, impacted HCPs, and regulators, from the first step in its creation to the final step in its implementation and overall patient impact. The JACC review cautions that without built-in fairness, explainability, and continuous monitoring, AI may further amplify health inequities. Therefore, the authors call for standardized reporting, interoperable AI infrastructure, and novel regulatory pathways that accommodate evolving AI models to ensure their equitability and safety when used.

From Theoretical Promise to Practical Utility
During CCO’s “Heart-ificial Intelligence” Think Tank, leading experts in cardiology echoed the principles laid out in the JACC review, while describing their own experiences. Dr Oikonomou served as the chair for this Think Tank discussion and brought valuable insights as the coauthor of the JACC review. In the Think Tank, Dr Francisco Lopez-Jimenez reminded HCPs that the technical architecture of AI is less important than its clinical relevance. For frontline HCPs, trust stems from reliability, accuracy, and demonstrable benefit. This pragmatism aligns closely with Khera and Dr Oikonomou’s assertion in the review that actionable AI should influence a clinical decision, not just predict a probability.

Dr Pierre Elias expanded on this idea by emphasizing the innate variability in adoption dynamics. In his view, HCP engagement lands on a spectrum—from early adopters who are eager to pilot innovations themselves to skeptics who need repeated evidence of reliability for their buy-in. AI implementation, therefore, depends on transparency and HCP education as much as it does on the science behind it. That is because 1 inaccurate alert can derail confidence among HCPs and across an entire department.

Dr Louise Sun then illustrated how this principle plays out in high-stakes environments. Early in the COVID-19 pandemic, her team rapidly built and deployed predictive AI models to triage cardiac surgery cases and balance intensive care unit utilization. The success of these AI tools relied on their clinical immediacy for solving an urgent problem and were codeveloped with technology staff, hospital administrators, and HCP users. This example embodies the translation stage of AI integration: AI that moves from experimental accuracy to measurable and operational value that is embedded in already established clinical workflows.

AI as a Cardiology Systems Tool
Both the JACC review and CCO’s expert discussion converge on the view that AI’s influence in cardiology extends beyond enhancing diagnostics alone. Dr Sammy Elmariah described how AI-driven 3D modeling that uses CT scans is now integral to transcatheter valve planning for simulating valve deployment and predicting complications like annular rupture or coronary obstruction. This procedural foresight improves patient outcomes, which is another tangible example of actionable AI. 

At the population level, AI also offers solutions to systemic inefficiencies. Automated electronic health record (EHR)–based referral systems can identify untreated patients with aortic stenosis or reduced ejection fraction who otherwise might fall through the cracks. This bridges a crucial gap between CVD detection and treatment, thereby turning AI prediction into action. This example also realizes the JACC review’s call for closing the loop between algorithmic insight and patient outcomes.

Ethics, Equity, and Adaptability
Perhaps the most forward-looking element of the review is the authors’ insistence on novel and adaptable regulatory pathways. Traditional regulatory models treat AI algorithms as static medical devices, yet healthcare data and populations continuously evolve. The authors advocate for robust frameworks that accelerate innovation and dissemination as well as allow for continuous and safe AI model updates.

In the Think Tank, Dr Elias supported this direction, noting that AI, in general, should adapt as human behavior and patient demographics shift. For example, facial recognition on smartphones continually learns about its user’s features to operate correctly. The JACC review further underscores this point. Its authors argue that maintaining good AI performance requires continuous monitoring and validation.

Equity is another defining theme. Both the review and Dr Sun warned that underrepresentation in datasets can produce biased AI models. Her examples of using generative AI illustrated how seemingly neutral algorithms can further reinforce harmful gender and racial stereotypes. Such examples strengthen the review’s argument that inclusivity is foundational to building trustworthy AI.

The Future of AI in Cardiology
The path forward requires AI to be woven into daily cardiology practice, with future tools integrating cardiovascular imaging, genomics and biomarkers, as well as EHR data together to enable multimodal, precision care. As Dr Oikonomou concluded in the Think Tank, the goal of using AI in healthcare is not to automate care but to augment it. AI can help discover new insights HCPs cannot easily perceive alone, not replace their clinical intuition.

In practice, actionable AI will succeed if it helps HCPs make faster, better, and fairer clinical decisions. For cardiology, that means integrating AI systems that more quickly identify high-risk patients, streamline referrals, and personalize therapy while maintaining HCP oversight. The vision shared by the JACC review and CCO’s expert Think Tank is clear: AI should transform cardiovascular care into a continuously learning, patient-centered ecosystem that is measurable, equitable, and enduring. 

Your Thoughts
How are you currently engaging with AI tools or systems in your cardiology-related practice? You can get involved in the conversation by answering the poll question and posting a comment below.

Continue

Poll

1.

How are you currently engaging with AI tools or systems in your cardiology-related practice?

Submit