A new clinical trial is exploring whether artificial intelligence can detect diseases unrelated to the heart by analyzing cardiac ultrasound images, potentially transforming how physicians use routine imaging to catch hidden health problems earlier. Researchers at Kaiser Permanente Northern California have received a $5 million grant from the American Heart Association to study how AI can identify missed patterns in echocardiograms, or heart ultrasounds, that might reveal conditions like liver disease that cardiologists aren't trained to assess .

What Problem Are Researchers Trying to Solve?

The core insight driving this research is deceptively simple: medical images contain far more information than clinicians typically extract from them. When a cardiologist performs a heart ultrasound, the images often capture surrounding organs, including the liver, but these structures fall outside the cardiologist's area of expertise. Dr. David Ouyang, the research scientist and cardiologist leading the study, explained that the goal is to leverage AI to identify this hidden diagnostic value .

The concept is called "opportunistic screening," and it represents a shift in how healthcare systems might use existing imaging data. Rather than ordering additional tests, physicians could use AI to flag potential health issues visible in images patients already receive as part of routine care. This approach could reduce wait times for specialist referrals and enable earlier intervention for conditions that might otherwise go undetected .

How Will the AI Technology Work in Practice?

The research team is training large-scale AI models on millions of echocardiographic videos to analyze multiple cardiac features and diagnoses simultaneously. These advanced models can generate detailed reports that may help physicians improve diagnosis for heart failure, valve disease, and other cardiac abnormalities . The key innovation is that these models don't stop at cardiac analysis; they're designed to identify patterns in non-cardiac structures visible within the imaging data.

The clinical trial, which runs through the end of 2028, will test whether AI screening for liver disease actually changes patient outcomes by enabling earlier diagnosis and specialist referral. This represents a critical step that many AI healthcare projects skip: moving from laboratory validation to real-world clinical impact .

Steps to Implement AI Screening in Your Healthcare System

• Establish Clinical Trials First: Before deploying AI screening tools, conduct blinded randomized clinical trials to generate evidence that algorithms work effectively in actual clinical workflows, not just in research settings.
• Build Multi-Stakeholder Support: Engage physicians, patients, hospital administrators, and specialists across your health system to build trust in the technology and ensure it integrates smoothly into existing care pathways.
• Train on Large, Diverse Datasets: Develop AI models trained on millions of medical images to ensure the algorithms can recognize patterns across different patient populations and imaging equipment.
• Design for Augmentation, Not Replacement: Position AI as a safety net that augments physician decision-making rather than replacing clinical judgment, providing thoughtful suggestions to improve care quality.

Why Is This Different From Other AI Healthcare Projects?

Many AI healthcare initiatives stop at publishing research papers or completing small pilot studies, but this Kaiser Permanente project is explicitly designed to bridge the gap between research and clinical deployment. Dr. Ouyang noted that thousands of manuscripts on AI in healthcare are published annually, but very few examine how these technologies actually impact patient care in real-world settings .

"There are thousands of manuscripts being published on new AI technologies in health care, but very few of these papers actually look at how these AI technologies will impact patients in clinical practice. Our trial is designed to look at how the algorithms we are studying in the lab can be used in the clinic to study their impact on patient care," said Dr. Ouyang.

Dr. David Ouyang, Research Scientist and Cardiologist at Kaiser Permanente Northern California Division of Research

The research team previously designed one of the first blinded randomized clinical trials of AI in medicine, and they hope their trial design and experience will help other physicians and researchers implement AI in clinical settings . This methodological rigor is essential for building the evidence base that healthcare systems need to confidently adopt new AI tools.

What Does This Mean for Cardiologists and Patients?

The vision for AI in cardiology is not autonomous decision-making, but rather augmentation of physician expertise. Dr. Ouyang described AI as "a safety net, another set of eyes to augment physicians to improve quality of care," emphasizing that AI is not ready to function like a fully autonomous system . Instead, it can provide thoughtful suggestions and assistance across multiple areas of clinical care.

Dr. Ouyang

For patients, the potential benefit is significant: conditions like liver disease could be identified earlier through routine cardiac imaging, allowing physicians to refer patients to specialists before symptoms develop or disease progresses. Earlier diagnosis typically leads to better treatment outcomes and may reduce the need for more invasive interventions later .

The trial will assess not only whether AI screening of liver disease changes disease trajectory through earlier diagnosis, but also what steps are needed to prepare this type of AI technology for widespread use in clinical care. By the end of the project, researchers hope to report on both the effectiveness of the technology and the practical requirements for implementation across different healthcare systems .

As this research unfolds through 2028, it will provide critical insights into how AI can extract hidden diagnostic value from imaging data that patients already receive, potentially transforming routine ultrasounds into more comprehensive screening tools without requiring additional tests or patient burden.