A novel artificial intelligence (AI) system, CLAiR, has demonstrated a strong correlation with standard cardiovascular risk assessments by analyzing retinal images. The findings, presented at an Investigative Horizons session during the American College of Cardiology's Annual Scientific Session (ACC.26) in New Orleans on March 30, suggest a promising future for preventive care.
Integrating AI for heart disease risk screening during routine eye examinations could increase awareness of risk and facilitate referrals for preventive care.
System Overview and Purpose
The CLAiR system, developed by Toku, has received Breakthrough Device designation from the U.S. Food and Drug Administration (FDA). This marks a significant step towards its potential clinical application.
Dr. Michael V. McConnell, a clinical professor of medicine at Stanford University and chief health officer at Toku, stated that images from the back of the eye contain significant health information that AI can analyze to inform individuals of their risk, potentially enabling them to seek guideline-based evaluation and preventive therapy.
This initial prospective evaluation in the U.S. is specifically intended to support the system's FDA submission, bringing it closer to widespread use.
Study Methodology
The study encompassed 874 participants, aged 40-75 years. Crucially, these individuals were not taking lipid-lowering medications and had no known atherosclerosis, ensuring a clear assessment of baseline risk. Participants were recruited from 10 eye care and primary care sites across the United States.
The cohort was diverse, including approximately half women, 19% Black or African American individuals, and 26% Hispanic individuals. High-resolution images of each participant's retina were captured using standard retinal cameras, devices commonly found in eye clinics. The CLAiR system then meticulously analyzed these images.
For comparative analysis, a standard atherosclerotic cardiovascular disease (ASCVD) risk estimator was also applied during the same clinic visit. This estimator utilized participants' age, sex, smoking status, blood pressure, and cholesterol to calculate their 10-year risk. The AI system aimed to identify participants with a 10-year risk of heart disease or stroke of 7.5% or higher, a threshold frequently used to recommend statin therapy.
Key Findings
The results highlight CLAiR's strong agreement with established risk assessment methods. According to the standard ASCVD risk estimator, 26% of the participants had a 10-year risk score of 7.5% or greater.
The CLAiR system demonstrated notable accuracy in identifying these cases:
- It correctly identified positive cases 91.1% of the time (sensitivity).
- It correctly identified negative cases 86.2% of the time (specificity).
Furthermore, the AI system proved highly efficient, able to use 94% of the acquired retinal images for analysis, indicating its robustness in real-world clinic settings.
Potential Applications and Future Steps
Researchers indicated that the AI system shows potential as a noninvasive screening method feasible for implementation in most eye care settings. This could democratize early risk detection significantly.
Dr. McConnell clarified that this AI approach is intended to complement, rather than replace, standard cardiovascular risk evaluations. Its aim is to enhance awareness, particularly for individuals who might benefit from preventive care but have not yet undergone a thorough evaluation.
However, further work is necessary to establish clear pathways for referring at-risk patients for cardiovascular evaluation and treatment in primary care following retinal image screening. This will be crucial for translating screening results into actionable health outcomes.
Operational Considerations
- Retinal imaging typically takes approximately five minutes, with the CLAiR algorithm returning results in about 30 seconds, making it a quick addition to routine visits.
- The CLAiR system is not designed for use in pregnant individuals or those with advanced eye disease.
- While retinal imaging is common in U.S. eye clinics, coverage by vision insurance plans as part of a standard visit may vary, potentially incurring additional fees for patients.
- The study was funded by Toku, the developer of the CLAiR system.