AI-enabled cardiac volumetry on non-contrast calcium scoring CT for predicting atrial fibrillation and mortality.
Authors
Affiliations (8)
Affiliations (8)
- Department of Radiology and Research Institute of Radiology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea; Biomedical Engineering Research Center, Asan Institute for Life Sciences, Asan Medical Center, Seoul, Republic of Korea. Electronic address: [email protected].
- Biomedical Engineering Research Center, Asan Institute for Life Sciences, Asan Medical Center, Seoul, Republic of Korea; Department of Biomedical Engineering, AMIST, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea. Electronic address: [email protected].
- Biomedical Engineering Research Center, Asan Institute for Life Sciences, Asan Medical Center, Seoul, Republic of Korea. Electronic address: [email protected].
- Department of Radiology, Chonnam National University Medical School, Chonnam National University Hospital, Gwangju, Republic of Korea. Electronic address: [email protected].
- Biomedical Engineering Research Center, Asan Institute for Life Sciences, Asan Medical Center, Seoul, Republic of Korea; Department of Biomedical Engineering, AMIST, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea. Electronic address: [email protected].
- Department of Radiology and Research Institute of Radiology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea; Biomedical Engineering Research Center, Asan Institute for Life Sciences, Asan Medical Center, Seoul, Republic of Korea. Electronic address: [email protected].
- Department of Radiology and Research Institute of Radiology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea; Biomedical Engineering Research Center, Asan Institute for Life Sciences, Asan Medical Center, Seoul, Republic of Korea. Electronic address: [email protected].
- Department of Radiology and Research Institute of Radiology, Asan Medical Center, University of Ulsan College of Medicine, Seoul, Republic of Korea; Biomedical Engineering Research Center, Asan Institute for Life Sciences, Asan Medical Center, Seoul, Republic of Korea. Electronic address: [email protected].
Abstract
The prognostic value of cardiac volumetry derived from non-contrast coronary calcium scoring CT (CSCT) remains uncertain. This study evaluated whether artificial intelligence (AI)-enabled cardiac volumetry from CSCT improves prediction of incident atrial fibrillation (AF) and all-cause mortality. We analyzed 4402 adults (median age, 55.8 years; 68.6% men) who underwent CSCT at two centers between 2007 and 2014. A deep-learning model automatically quantified four cardiac chamber volumes, left ventricular (LV) mass, and CAC. AI-enabled volumetric measurements were validated against human expert-validated manual measurements using concordance correlation coefficients (CCC) and Spearman correlation. Associations with incident AF and all-cause mortality were evaluated using multivariable Cox regression, and incremental predictive value was assessed using Harrell's C-index, integrated discrimination improvement, and net reclassification improvement. AI-enabled cardiac volumetry showed excellent agreement with manual CSCT measurements (CCC range, 0.80-0.98). During a median follow-up of 14 years, AF occurred in 102 individuals (2.3%), and all-cause mortality occurred in 299 individuals (6.8%). Enlarged left atrial (LA) and right atrial (RA) volumes independently predicted incident AF (hazard ratios [HRs], 7.77 and 9.61; both p < 0.001). Enlarged LA volume and increased LV mass were independently associated with all-cause mortality (HR, 1.61; p = 0.012 and HR, 1.73; p = 0.032, respectively). AI-enabled cardiac volumetry significantly improved discrimination for AF prediction (C-index, 0.74 to 0.83; p < 0.001), whereas its incremental prognostic value for all-cause mortality beyond CAC and clinical variables was modest and not statistically significant. AI-enabled cardiac volumetry from CSCT significantly enhances prediction of incident AF, while its additional value for mortality prediction beyond CAC remains limited.