AI-CVD-AF: A Novel Atrial Fibrillation Prediction Model Based on Coronary Artery Calcium Scans.
Authors
Affiliations (21)
Affiliations (21)
- HeartLung.AI, Houston, TX, 77021, USA.
- Department of Electrical and Computer Engineering, Cornell University, Ithaca, NY 14853, USA.
- Clinic of Radiology and Nuclear Medicine, University Hospital Basel, Petersgraben 4, CH-4031 Basel, Switzerland.
- Tustin Teleradiology, Tustin, CA 92780, USA.
- Icahn School of Medicine at Mount Sinai, New York, NY 10029, USA.
- Department of Radiology, University of California Irvine, CA 92697, USA.
- Kravis Center for Clinical Cardiovascular Health, Mount Sinai Fuster Heart Hospital, New York, NY 10029 USA.
- Houston Methodist Hospital, Houston, TX 77030, USA.
- BioMedical Engineering and Imaging Institute. Icahn School of Medicine at Mount Sinai, New York, NY 10029, USA.
- Stanford Prevention Research Center, Stanford University School of Medicine, Stanford, CA, 94305, USA.
- Division of Cardiology, Kaiser Permanente Oakland Medical Center, California, USA.
- Department of Radiology, University Medical Center Groningen, GZ 9713, Netherlands.
- University of Houston, Houston, TX, 77030, USA.
- The Lundquist Institute, CA 90502, USA.
- Population Sciences Branch, Division of Intramural Research, National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, MD, 20824, USA.
- University of Louisville, Louisville, KY, 40202, USA.
- Cedars-Sinai Medical Center, Los Angeles, CA 90048, USA.
- Herbert Wertheim College of Medicine, Department of Emergency Medicine and Critical Care, Miami Beach, Florida; Mount Sinai Medical Center, Department of Emergency Medicine, Miami Beach, Florida.
- Department of Medicine, Herbert Wertheim College of Medicine, Florida International University, Miami, Florida, USA.
- Huntington Medical Rese Boston Medical Center and Chobanian & Avedisian School of Medicine and School of Public Health, Boston University, Boston, MA 02215, USAarch Institutes, Pasadena, CA, 91105, USA and Keck School of Medicine of University of Southern California, Los Angeles, CA 90033.
- Heart Disease Prevention Program, Mary and Steve Wen Cardiovascular Division, University of California Irvine School of Medicine, CA 92697, USA.
Abstract
The AI-CVD initiative seeks to extract actionable insights from coronary artery calcium (CAC) scans beyond the traditional CAC score. We previously demonstrated that AI-derived cardiac chamber volumes from CAC scans predict incident heart failure (HF). We aimed to evaluate whether left-to-right cardiac chamber volume ratios outperform chamber volumes in predicting HF. We used AI-CVD cardiac chambers volumetry data from CAC scans of 5,732 asymptomatic Multi-Ethnic Study of Atherosclerosis (MESA) participants (age 62.2±10.3 years; 47.7% male). Left-to-right ventricular (LV/RV), atrial (LA/RA), and left atrial-to-right ventricular (LA/RV) volume ratios were evaluated using multivariable Cox models and feature selection techniques. External validation was performed in the Framingham Heart Study Offspring (FHS-O) cohort (N=1,052, age:58.3±8.3, 42.9% male). During a median follow-up of 17.7 years in MESA, 369 participants (6.3%) developed HF. Elevated ratios (≥75th & ≥95th percentile) of LV/RV, LA/RA, and LA/RV were strongly associated with incident HF: hazard ratio (HR) for ≥95th percentile were 4.04 (95% CI:2.89-5.65), 2.90 (95% CI:2.07-4.06), and 2.61 (95% CI:1.87-3.46), respectively. Among participants with normal LV sizes (interquartile-range), LV/RV ≥95th significantly predicted HF (HR:2.34; 95% CI:1.29-4.25). In FHS-O (median follow-up 14.4 years), 56 HF events (5.3%) occurred. LV/RV ≥75th percentile was significantly associated with HF (HR:2.23; 95% CI:1.16-4.30), whereas LA/RA was not (HR:1.22; 95% CI:0.65-2.29). Feature selection techniques identified LV/RV as the strongest predictor. In these two prospective cohorts, AI-derived LV/RV ratio from CAC scans strongly predicted HF. New clinical trials guided by these imaging biomarkers are warranted to establish their clinical utility.