Heart scan features better predicted atrial fibrillation
Scan features scored 0.83 versus 0.74 for a clinical risk score in distinguishing who developed the condition; 1 would mean perfect separation.
Our read
Biomarkers- Studied:
- Cohort study of 4,226 asymptomatic patients without known cardiac disease, 4.0 ± 2.9 years
- What changed:
- Radiomics alone exceeded the predictive performance of CHARGE-AF at 10 years: AUC 0.83 versus 0.74
- The catch:
- Average follow-up was four years despite the 10-year prediction target
- Why it’s aging research:
- It bears on staying healthy in midlife and later life.
Editor of record: Boris Djordjevic · AgingResearch.News
Scans from 6,036 symptom-free patients without known cardiac disease were analyzed in a retrospective cohort study. The computed tomography scans measured calcium in the heart’s arteries. Researchers built prediction models from features of the heart and nearby blood vessels.
Testing involved 4,226 patients, averaging 59 years old; 42% were women. During an average four years of follow-up, 254 developed atrial fibrillation, an irregular heartbeat. The team identified cases using diagnosis codes. Scan features better predicted risk at 10 years than CHARGE-AF, a clinical risk score. Combining scan features with that score yielded a prediction-performance score of 0.84, versus 0.74 for the clinical score alone.
Why it matters
Predicting future heart rhythm problems is relevant to health as people grow older. The authors suggested that analyzing existing scans may improve prediction without additional radiation.
Caveats
Average follow-up was four years despite the 10-year prediction target, and diagnosis codes defined cases. These conference-abstract results do not establish predictive performance in patients with known cardiac disease.
The paper
Cardiac Structure Radiomics from CAC CT Outperforms CHARGE-AF for Atrial Fibrillation Prediction
Radiology & Imaging Sciences; Emory Univ
AHA Scientific Sessions 2026 Abstract, Nov 2, 2026
Presented at AHA Scientific Sessions 2026, abstract 14597
Comments
Be the first to comment.