2026-10-30 · 4 min read · healthcarewithai.com Editorial Team

AI heart checks: how an ECG or CT scan can now reveal hidden heart disease

Many heart attacks strike people who felt fine. AI is getting better at finding the warning signs years earlier, from a ten-second ECG or a single CT scan.

AI can now spot a weak heart pump from a routine ECG, flag hidden atrial fibrillation, measure inflammation around the heart arteries and calculate total plaque from a CT scan. Several tools are FDA-cleared and some are paid for by insurers and the NHS. Here is what they do and where the limits are.

Key points

FDA-cleared AI reads a standard 12-lead ECG, or a digital stethoscope recording, to flag a weak heart pump (ejection fraction 40% or less) before symptoms appear.
In a 45-clinic randomised trial at Mayo Clinic, AI-ECG alerts led to 32% more diagnoses of low ejection fraction.
CaRi-Heart, which measures inflammation around the coronary arteries on a CT scan, received FDA De Novo authorisation in July 2026.
AI plaque analysis (Cleerly, HeartFlow and others) has a dedicated US billing code from January 2026; HeartFlow FFR-CT is funded across the NHS.
These tools support doctors and can produce false alarms. Smartwatches can detect atrial fibrillation but not heart attacks.

Why standard tests miss early heart disease

Routine heart tests are designed to catch problems that are already large: a blocked artery on a stress test, calcium on a CT scan, or clear changes on an ECG after damage has occurred. Yet many heart attacks start from plaques that narrow the artery only slightly but are inflamed and unstable. And a heart can lose pumping strength for months before a person feels unwell.

What AI can now detect

A weak heart pump from a routine ECG

An ECG records the heart's electrical signal. AI models trained on tens of thousands of ECGs paired with heart ultrasound scans learned to recognise tiny electrical patterns linked to a weak pump. Anumana's algorithm, developed with Mayo Clinic, was FDA-cleared in September 2023. Tempus received clearance for a similar tool in July 2025. Eko's digital stethoscope, cleared in March 2024, combines a short ECG with heart sounds and gives a result in about 15 seconds during a normal check-up.

Hidden atrial fibrillation

Atrial fibrillation (AFib) comes and goes, so a person may have a normal rhythm at the clinic while being at risk of stroke. AI can look at a normal-rhythm ECG for subtle signs of a strained upper heart chamber and estimate the likelihood of intermittent AFib, prompting longer monitoring with a patch or Holter device.

Inflammation around the heart arteries

Inflamed arteries release chemical signals into the fat that surrounds them, changing how that fat looks on a CT scan. Caristo's CaRi-Heart measures this as a Fat Attenuation Index. In the ORFAN study of more than 40,000 patients published in The Lancet in 2024, high inflammation predicted heart attacks and cardiac deaths even in people whose arteries were not significantly narrowed. CaRi-Heart received FDA De Novo authorisation in July 2026.

Total plaque and blood flow from one CT scan

AI plaque analysis tools such as Cleerly, HeartFlow and CaRi-Plaque trace every coronary artery and measure the total volume of plaque, separating hard calcified plaque from soft plaque that is more likely to rupture. HeartFlow can also simulate blood flow (FFR-CT) to show whether a narrowing really limits flow, which can avoid an invasive angiogram.

Tool	What it finds	US status	Europe / UK
Anumana Low EF AI-ECG	Weak heart pump from 12-lead ECG	FDA-cleared, September 2023	No official data
Eko Low EF tool	Weak heart pump via digital stethoscope	FDA-cleared, March 2024	MHRA registered
Tempus ECG-Low EF	Weak heart pump from 12-lead ECG	FDA-cleared, July 2025	No official data
CaRi-Heart (Caristo)	Coronary inflammation and cardiac risk from CT	FDA De Novo, July 2026	CE marked; under NICE early value assessment
Cleerly	Total plaque, plaque type and likely ischaemia from CT	FDA-cleared	CE marked
HeartFlow FFR-CT and Plaque	Blood-flow limitation and plaque from CT	FDA-cleared	Funded across the NHS (MedTech Funding Mandate)
AIRE (Imperial College)	Risk of death, heart failure, heart block from ECG	Research	Being trialled in NHS hospitals

What the evidence shows

EAGLE trial (Mayo Clinic): in 45 primary care clinics, doctors who received AI-ECG alerts diagnosed low ejection fraction 32% more often than those without.
CONFIRM2: an international registry presented in 2024 found that AI-measured total plaque volume predicted heart events better than the usual visual grading of narrowing or calcium scores. Risk rose clearly once plaque exceeded about 250 cubic millimetres.
Women: a sub-study found women tend to have less plaque but more reduced blood flow per unit of plaque than men, a reminder that thresholds based on mostly male data can underestimate risk in women. Imperial's AIRE model also identified women at higher risk whose ECGs showed no classic warning signs.

Note on numbers: published effect sizes for coronary inflammation vary widely between studies, depending on how risk groups are defined. Treat any single headline figure with caution.

Cost and insurance

United States: HeartFlow FFR-CT has had a permanent billing code (CPT 75580) since January 2024 and is widely covered for symptomatic patients. AI plaque analysis received its own permanent code (CPT 75577) from January 2026, and several large insurers now cover it. AI-ECG tools are usually built into hospital systems without a separate charge. Purely elective screening scans for people without symptoms are generally self-pay.
United Kingdom: HeartFlow is funded by the NHS for eligible patients. CaRi-Heart and AIRE are available through NHS pilots and trials.
European Union: mostly available in specialist academic centres; public reimbursement is still developing.
Turkey: CT coronary angiography is widely available in private and foundation hospitals. We found no official data on SGK reimbursement for AI plaque or FFR-CT analysis, so access is mostly self-funded or part of private check-up packages.

Limits and risks

False alarms. When AI-ECG screening is used in the general population, a large share of people flagged for low ejection fraction turn out to have a normal pump on ultrasound.
Overtreatment. Detecting small amounts of plaque can lead to unnecessary angiograms, which carry their own risks.
Data quality. Poor electrode contact, movement or fast heart rates on CT can distort results.
Decision support only. Regulators clear these tools to assist doctors, not to diagnose on their own. A normal AI result never rules out a heart attack in someone with symptoms.

Trials to watch

TRANSFORM (NCT06112418): about 7,500 people with diabetes or metabolic syndrome, testing whether AI plaque-staged care prevents more heart attacks and strokes than usual risk-factor care.
DEEP-AF (NCT07173673): whether AI-ECG finds more hidden AFib in people with palpitations or dizziness.
AIRE Chest Pain (NCT07757425): whether AI-ECG can safely speed up and simplify care in rapid-access chest pain clinics.

Questions patients ask

Can AI predict a heart attack before it happens?

It cannot predict the day, but it can find the conditions that lead to heart attacks, such as inflamed arteries or large amounts of soft plaque, years earlier so treatment can start.

What does an AI flag for low ejection fraction mean if I feel fine?

It means the ECG shows patterns linked to a weaker heart pump. It is not a diagnosis; your doctor should arrange an echocardiogram to check.

If AI finds plaque, will I need a stent?

Usually not. Stents are for significant blockages causing symptoms or reduced blood flow. Most plaque is treated with medicines such as statins and lifestyle changes.

Can my smartwatch detect a heart attack?

No. Some smartwatches are cleared to detect atrial fibrillation, but none can diagnose a heart attack. Chest pain or pressure needs emergency care.

How is AI plaque analysis different from a calcium score?

A calcium score sees only hardened plaque on a non-contrast scan. AI plaque analysis uses a contrast CT to measure soft plaque, plaque volume and inflammation as well.

Is AI heart analysis available in Turkey?

CT coronary angiography is widely offered in private hospitals. We found no official data on SGK covering AI plaque or FFR-CT analysis.

Sources

FDA 510(k) K232699, Anumana Low EF AI-ECG, September 2023. https://www.accessdata.fda.gov/cdrh_docs/pdf23/K232699.pdf
FDA 510(k) K233409, Eko Low Ejection Fraction Tool, March 2024. https://www.accessdata.fda.gov/cdrh_docs/pdf23/K233409.pdf
FDA 510(k) K250119, Tempus ECG-Low EF, July 2025. https://www.accessdata.fda.gov/cdrh_docs/pdf25/K250119.pdf
Caristo's CaRi-Heart authorized by US FDA, HCPLive, 30 July 2026. https://www.hcplive.com/view/cari-heart-receives-fda-de-novo-authorization-for-coronary-inflammation
FDA De Novo DEN250042 (CaRi-Heart). https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfpmn/denovo.cfm?id=DEN250042
Chan et al. Inflammatory risk and cardiovascular events (ORFAN), The Lancet 2024. https://pubmed.ncbi.nlm.nih.gov/38823406/
NICE early value assessment of CaRi-Heart, 2024. https://pmc.ncbi.nlm.nih.gov/articles/PMC11299050/
CONFIRM2: AI-guided QCT, ACC, October 2024. https://www.acc.org/latest-in-cardiology/articles/2024/10/24/19/43/mon-312pm-confirm2-tct-2024
AI-ECG allows early diagnosis of low EF (EAGLE trial), TCTMD. https://www.tctmd.com/news/ai-ecg-allows-early-diagnosis-low-ef
AI model can predict health risks including early death from ECGs (AIRE), Imperial College London. https://www.imperial.ac.uk/news/257300/ai-model-predict-health-risks-including/
AI model can identify women at higher risk of heart disease from ECGs, British Heart Foundation, 2025. https://www.bhf.org.uk/what-we-do/news-from-the-bhf/news-archive/2025/march/ai-model-can-identify-women-at-higher-risk-of-heart-disease-from-ecgs-study-suggests
HeartFlow reimbursement resources. https://www.heartflow.com/ccta-advantage/reimbursement-resources/
CMS billing article: AI-enabled CT-based plaque analysis (CPT 75577). https://www.cms.gov/medicare-coverage-database/view/article.aspx?articleid=59733
TRANSFORM trial (NCT06112418), ClinicalTrials.gov. https://clinicaltrials.gov/study/NCT06112418
DEEP-AF trial (NCT07173673), ClinicalTrials.gov. https://clinicaltrials.gov/study/NCT07173673
AIRE Chest Pain Clinic trial (NCT07757425), ClinicalTrials.gov. https://clinicaltrials.gov/study/NCT07757425

This article is for information only and is not medical advice. Talk to your doctor before making any decision about your health or treatment. Approval status and prices change; check with the provider or regulator for the latest position.