AI Advances Could Revolutionize Heart Attack Risk Prediction
By Vishal Khetpal | October 20, 2025
Despite significant advances in cardiology, accurately predicting who will suffer a heart attack remains a major challenge. Many individuals at risk go unscreened, leaving potentially life-saving warning signs undiscovered. Now, emerging startups such as Bunkerhill Health, Nanox.AI, and HeartLung Technologies are harnessing artificial intelligence (AI) to analyze millions of routine chest CT scans for signs of coronary artery calcium (CAC), a key marker linked to heart attack risk. This innovative approach has the potential to transform public health by uncovering hidden risks in patients who might otherwise be overlooked. However, questions remain regarding the large-scale effectiveness and implementation of this technology, as well as its impact on clinical definitions of disease.
Coronary Artery Calcium: A Hidden Clue in Routine Scans
Approximately 20 million Americans undergo chest CT scans annually—often following accidents or lung cancer screenings. While these scans primarily focus on detecting immediate threats like trauma or tumors, they frequently reveal the presence of coronary artery calcium, an indicator not always mentioned in radiology reports. CAC reflects the buildup of calcified plaque within heart arteries, which develops over decades and is associated with a heightened risk for heart attacks.
Heart attacks commonly occur when early-stage, lipid-rich plaque ruptures suddenly, triggering inflammation and clotting that blocks blood flow. Although calcified plaque itself tends to be more stable, detecting CAC is strongly suggestive of the presence of more vulnerable, rupture-prone plaques. Traditionally, CAC scoring requires specialized heart-focused CT scans, a resource-intensive approach that limits widespread screening. AI-powered algorithms now offer the promise of extracting accurate CAC scores from routine chest CT scans, significantly broadening access to this predictive measure.
Potential for Earlier Detection and Intervention
These AI-driven CAC assessments could serve as automatic alerts for physicians and patients about elevated heart attack risk, prompting timely follow-up care. Although currently employed by a limited number of startups, this technology is rapidly gaining traction and could help identify high-risk individuals who fall through the cracks of traditional screening programs.
Historically, CAC scans have faced skepticism regarding their clinical value, often regarded as tests for the “worried well” and lacking insurance coverage. However, expert opinions are beginning to shift, with increasing endorsements for CAC scoring as a tool to refine cardiovascular risk estimates and motivate patients toward preventive treatments such as statins.
Challenges and Considerations Ahead
Despite its promise, AI-based CAC screening raises important practical and ethical issues. A population-based study in Denmark in 2022 found no mortality benefit from universal CAC screening, calling into question whether automatically generated CAC results will alter long-term outcomes. Furthermore, widespread detection of abnormal CAC scores could overwhelm healthcare systems unprepared to manage these incidental findings effectively. As Nishith Khandwala, cofounder of Bunkerhill Health, notes, many health providers lack standardized protocols for following up on incidental calcium findings, risking the creation of excessive workload without clear benefits.
Moreover, uncertainty remains about the best clinical actions following AI-identified high CAC scores. While statins are commonly recommended, the additional value of costly cholesterol-lowering drugs like PCSK9 inhibitors for asymptomatic patients with high CAC is still unclear. There is also a risk that patients could undergo unnecessary interventions, increasing healthcare costs and potential harm.
Current reimbursement frameworks do not cover AI-derived CAC scoring as a distinct service, complicating the financial viability of such technologies. This highlights potential perverse incentives in the emerging market, where business motivations may not align perfectly with improved patient outcomes.
Redefining Disease in the AI Era
The integration of AI into diagnostic workflows may also reshape fundamental concepts of disease. Adam Rodman, an AI and hospital medicine expert at Beth Israel Deaconess Medical Center, likens AI-generated CAC findings to “incidentalomas,” unexpected imaging discoveries that disrupted traditional diagnosis decades ago. However, unlike incidentalomas detected by human radiologists, AI tools autonomously identify and classify abnormalities, ushering in what Rodman describes as “machine-based nosology,” where diseases may be defined on algorithmic terms.
This evolution could potentially create a two-tiered diagnostic system—with wealthier patients accessing premium AI tools and others receiving less sophisticated assessments—raising equity concerns.
The Road Ahead
For individuals lacking traditional risk factors or regular healthcare contact, AI-derived CAC scoring might enable earlier identification of heart disease risk, potentially changing the trajectory of care. Yet many questions remain about how these results will be communicated, who will act on them, and whether they ultimately improve outcomes at a population level.
For now, clinicians remain indispensable intermediaries, interpreting algorithmic outputs in the context of each patient’s unique profile. AI may enhance predictive power, but the nuanced judgment of healthcare providers will continue to play a critical role in guiding heart attack prevention strategies.
Vishal Khetpal is a cardiovascular disease fellow. The views expressed here are his own and do not represent his employers.
[MIT Technology Review © 2025]
