AI-Powered Tool TARGET-AI Improves Cardiovascular Disease Diagnosis

Artificial intelligence is rapidly transforming cardiology, offering the potential to improve diagnosis, treatment, and patient care. However, simply having powerful AI tools isn’t enough. A key challenge lies in determining when to deploy these tools effectively within the complex, real-world clinical environment. Researchers at Yale School of Medicine are addressing this issue with a new approach, developing AI systems designed to guide the use of other AI technologies.

Evangelos Oikonomou, MD, DPhil, assistant professor of medicine (cardiovascular medicine) at Yale, and his colleagues recently published a paper in NEJM AI detailing TARGET-AI, a clinical decision support tool designed to help clinicians and health systems use AI more strategically. The tool aims to bridge the gap between the controlled settings where AI models are trained and the often-messy reality of clinical practice.

The Challenge of AI Implementation

“We are witnessing a wave of artificial intelligence tools in cardiology that can effectively help clinicians diagnose different heart conditions,” says Oikonomou. “However, many of these tools are not being used in real life, because real life is different from the controlled environment in which a model is trained. The question now becomes: how can we actually use AI effectively in real clinical settings?”

The core problem, Oikonomou explains, is that AI models, while highly accurate in specific scenarios, can generate numerous false positives when applied broadly to the general population. Overuse of AI in low-risk individuals can lead to unnecessary testing, increased healthcare costs, and heightened patient anxiety. TARGET-AI seeks to address this by providing guidance on when the benefits of using an AI tool outweigh the potential drawbacks.

“We have many effective traditional tests in medicine, but we don’t use them for everyone,” Oikonomou notes. “Instead, we look at the patient in front of us, consider their history, symptoms, and goals of care, and determine which tests are appropriate. The same principle applies to AI tools. However, AI models are still very new, and as a field, we have not evolved our thinking about when and how to use them.”

How TARGET-AI Works: Recognizing Clinical Trajectories

TARGET-AI leverages the principles of large language models (LLMs), but instead of analyzing text, it analyzes sequences of clinical events from a patient’s electronic health record. Like LLMs predicting the next word in a sentence, TARGET-AI predicts the likelihood of a patient progressing towards a specific diagnosis based on their medical history.

By analyzing deidentified patient data, the algorithm identifies patterns that suggest an increased risk of a particular cardiovascular condition. The researchers developed and validated the algorithm across the Yale New Haven Health system and then confirmed its performance with datasets from the U.S. And the U.K., demonstrating its generalizability beyond a single patient population.

“We hope that health systems will begin using TARGET-AI to help determine when and how to deploy AI throughout their systems,” Oikonomou says. “Our goal is to help build the guardrails to maximize the precision of other AI detection tools.”

The Future of AI in Cardiology: Human-AI Partnership

For those entering the field of AI-related research in cardiology, Oikonomou offers practical advice. He encourages researchers to identify areas within their own clinical practice or laboratories that could benefit from optimization through AI. He emphasizes the importance of focusing on real-world problems and developing solutions that address specific clinical needs.

However, he also stresses the need to move beyond simply developing AI tools and to focus on understanding the complex interplay between humans and AI. “There is much more we need to learn about the partnership between humans and AI, including how they work together and where the pain points lie,” he says. “This is the next phase of AI research.”

The development of TARGET-AI represents a significant step towards responsible and effective AI implementation in cardiology. By providing clinicians with guidance on when and how to use AI tools, researchers are paving the way for a future where AI enhances, rather than complicates, the delivery of cardiovascular care.