AI-quantified Myosteatosis at CAC CT for Prediction of Atrial Fibrillation and Heart Failure: The Multi-Ethnic Study of Atherosclerosis.
Authors
Affiliations (18)
Affiliations (18)
- HeartLung.AI, 2450 Holcombe Blvd, Houston, TX 77021.
- Department of Electrical and Computer Engineering, Cornell University, Ithaca, NY.
- University of Basel, Basel, Switzerland.
- Tustin Teleradiology, Tustin, Calif.
- Icahn School of Medicine at Mount Sinai, New York, NY.
- Kravis Center for Clinical Cardiovascular Health, Mount Sinai Fuster Heart Hospital, New York, NY.
- Department of Radiology, University of California, Irvine, Calif.
- The Lundquist Institute, Torrance, Calif.
- Houston Methodist Hospital, Houston, Tex.
- Stanford Prevention Research Center, Stanford University School of Medicine, Stanford, Calif.
- University of Houston, Houston, Tex.
- Division of Cardiology, Michigan State University, East Lansing, Mich.
- University Medical Center Groningen, Groningen, the Netherlands.
- University of Louisville, Louisville, Ky.
- Cedars-Sinai Medical Center, Los Angeles, Calif.
- Population Sciences Branch, Division of Intramural Research, National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, Md.
- School of Public Health, Boston University, Boston, Mass.
- Heart Disease Prevention Program, Mary and Steve Wen Cardiovascular Division, University of California, Irvine, Calif.
Abstract
Purpose To evaluate whether the artificial intelligence (AI)-quantified mean thoracic skeletal muscle (TSM) attenuation from coronary artery calcium (CAC) scans predicts incident cardiovascular disease (CVD), with a focus on atrial fibrillation (AF) and heart failure (HF). Materials and Methods Data from the Multi-Ethnic Study of Atherosclerosis, including participants without baseline CVD who underwent CAC scanning, were retrospectively analyzed. Myosteatosis was defined by sex-specific, AI-quantified mean TSM attenuation cutoffs. The Cox proportional hazards model was used to compare total CVD, AF, and HF risks between the bottom and top quartiles of TSM attenuation after adjusting for CVD risk factors, inflammatory markers, insulin resistance, Agatston score, TSM volume, and social determinants of health. Results Among 5739 participants (mean age, 62.1 years ± 10.3 [SD]; 3002 [52.3%] female), 1826 CVD events occurred over 19 years, including 1139 AF and 359 HF events. Myosteatosis was independently associated with increased risks of total CVD (hazard ratio [HR], 1.48 [95% CI: 1.25, 1.75]; <i>P</i> = .001), AF (HR, 1.68 [95% CI: 1.37, 2.07]; <i>P</i> < .001), and HF (HR, 1.61 [95% CI: 1.18, 2.19]; <i>P</i> < .002). Participants with both myosteatosis and high Agatston scores had markedly higher cumulative incidences compared with those with high Agatston scores alone (total CVD: 84.6% vs 68.6%; AF: 52.9% vs 42.4%; HF: 22.6% vs 16.1%). Adding myosteatosis to the Agatston score significantly improved prediction (time-dependent area under the receiver operating characteristic curve, total CVD: 0.74 vs 0.80, <i>P</i> < .001; AF: 0.68 vs 0.76, <i>P</i> < .001; HF: 0.73 vs 0.78, <i>P</i> = .007). Conclusion AI-quantified mean TSM attenuation on CAC scans independently predicted AF and HF and enhanced the Agatston score's predictive value. <b>Keywords:</b> Myosteatosis, Coronary Artery Calcium Scan, Atrial Fibrillation, Heart Failure, Artificial Intelligence, Applications-CT, Cardiac, Thorax, Muscular, Heart ClinicalTrials.gov NCT00005487 <i>Supplemental material is available for this article.</i> © RSNA, 2026.