Every year, roughly 19 million Americans get a chest CT scan — for a broken rib, a cough that won't quit, a routine checkup. The radiologist reads the images, clears the lungs, and sends the patient home. What the radiologist may not have mentioned: there was calcium in the arteries feeding that heart, and it matters.
Coronary artery calcium, or CAC, is a buildup of plaque in the walls of the heart's blood vessels. It doesn't cause symptoms. It shows up as an incidental finding on chest CTs designed to look at lungs. And radiologists miss it somewhere between 20% and 40% of the time, according to STAT News. That is 4 million to 8 million people a year who paid for a scan, got told their lungs were fine, and left with a heart risk nobody quantified.
A new artificial intelligence tool could change that. The system reads the same chest CT images after the radiologist has finished, detects any CAC present, and assigns a score reflecting how much calcium has accumulated. A CAC score above 400 — roughly the top 10% of scores — signals very high risk: those patients face a 3.5-fold higher chance of dying from any cause over the next decade compared to people with no calcium in their coronary arteries, according to research published in Science Daily that was trained and tested across the Department of Veterans Affairs hospital system. In the VA study, involving more than 8,000 CT scans from 98 medical centers, the AI correctly identified whether a scan contained CAC in 89.4% of cases.
For most of those high-risk patients, the fix is straightforward. In the same VA study, 99.2% of patients with CAC scores above 400 would benefit from lipid-lowering therapy — a cheap, widely available, effective intervention. The AI doesn't prescribe. It surfaces the risk so a doctor can.
The barrier until now was not clinical merit. It was reimbursement. CMS, the federal agency that sets payment rates for Medicare and Medicaid, established a new billing code — HCPCS G0680, effective April 1, 2026 — that pays hospitals for running an AI algorithm specifically designed to detect coronary artery calcium on chest CTs performed for other reasons. The code assigns the service to APC 1492 with status indicator S, meaning it is paid as a standalone service in the hospital outpatient setting, according to HeartLung AI, one of the companies that lobbied for the code. Before G0680, there was no dedicated pathway for hospitals to bill for opportunistic CAC screening — it existed in a billing gray zone.
"We need to find more of these patients," said Dr. Ami Bhatt, chair of the FDA's Digital Health Advisory Committee and chief innovation officer of the American College of Cardiology, in the STAT News report. That imperative — find the patients who are already in the scanner — is what the new reimbursement code is designed to act on.
That CMS decision transforms the unit economics. A hospital running 10,000 chest CTs a year that previously generated zero revenue from opportunistic CAC detection can now run an AI tool on every scan and collect a facility fee. Nanox.AI, whose HealthCCSng algorithm is one of the products cleared for use in the United States, reported that moderate to severe CAC was present in 58% of patients it analyzed across a large dataset — a far higher rate than many clinicians would have guessed. If that figure holds across the broader population of 17 million eligible individuals who get non-cardiac chest CTs each year, the upstream opportunity for preventive cardiology is substantial.
The access question is real. Not every hospital has the IT infrastructure to integrate an AI tool into an existing radiology workflow, and the companies selling these systems are still building out their commercial teams. HeartLung AI, which focuses specifically on cardiovascular AI, is a small company competing against imaging giants like Nanox.AI and a handful of other players. Whether G0680 pays enough to justify the integration cost is still being worked out in hospital contracting. The code sets a price, but what hospitals actually receive depends on their payer mix and negotiating leverage.
There is also a downstream question the data doesn't fully answer: what happens to the patient who gets a CAC score they never asked for? A chest CT ordered for one reason now produces a finding that mandates a cardiology referral. That is clinically appropriate for high-risk scores, but it also loads additional appointments, costs, and potential anxiety onto patients who may have been expecting a simple lung read. The health system benefits are real. The distribution of those benefits across different patient populations is less studied.
Radiologists are not being replaced by this workflow — the AI runs after their read, not alongside it. That matters for adoption: a tool that slows down the radiologist or creates liability ambiguity would face resistance. A tool that runs in the background and delivers a CAC finding when the primary read is already filed is easier to absorb into practice. Several commercial systems have FDA clearance for this specific use case, and the regulatory pathway for software that analyzes existing images rather than producing a new diagnostic image is relatively well established.
The 2026 effective date for G0680 means hospitals are just beginning to operationalize the code. Real-world utilization data — how many scans actually get analyzed, whether cardiologists follow up appropriately, whether the expected downstream reduction in cardiovascular events materializes — will take years to accumulate. But the structural pieces are in place: a reimbursement pathway, multiple FDA-cleared vendors, and a clinical evidence base that is large enough to be credible, if not yet definitive.
The patient with a CAC score over 400 and no idea it exists is the starting point of this story. Whether 19 million annual chest CTs become a de facto cardiovascular screening opportunity depends on whether hospitals actually deploy the tools, whether cardiologists have the capacity to absorb the referrals, and whether the people most at risk are the ones getting scans in the first place.
