AI-measured epicardial adipose tissue may improve CV risk prediction
Click Here to Manage Email Alerts
A deep learning model trained to quantify epicardial adipose tissue volume measured via routine coronary CT angiography scans can improve risk assessment for CV and non-CV outcomes, independent of other risk factors, researchers reported.
“Body weight is not a reliable measure of obesity, as the fat distribution in the human body is much more important than its absolute quantity,” Charalambos Antoniades, MD, PhD, FRCP, FESC, professor and chair of cardiovascular medicine at the University of Oxford and chair of the British Atherosclerosis Society, told Healio. “With routine cardiac CT, we can measure the volume of epicardial fat in a fully automated way using artificial intelligence, and this can become a routine metric in standard cardiac CT reporting. Epicardial adipose tissue volume has predictive value not only for strokes, atrial fibrillation and heart attacks, but most importantly for all-cause mortality driven by noncardiac causes.”
Data from routine scans
Antoniades and colleagues trained and validated a deep learning network to automatically extract epicardial adipose tissue volume in 3,720 coronary CTA scans from the Oxford Risk Factors and Noninvasive Imaging Study (ORFAN) cohort. The model was tested in patients with challenging anatomy and scan artifacts and applied to a longitudinal cohort of 253 patients after cardiac surgery and 1,558 patients from the SCOT-HEART trial to investigate its prognostic value.
The findings were published in JACC: Cardiovascular Imaging.
External validation of the deep learning network yielded a concordance correlation coefficient of 0.97 for machine vs. human.
Epicardial adipose tissue volume was associated with CAD, with an OR per standard deviation increase in epicardial adipose tissue volume of 1.13 (95% CI, 1.04-1.3; P = .01), and was also associated with AF (OR = 1.25; 95% CI, 1.08-1.4; P = .03). Results persisted after adjustment for risk factors including BMI.
Epicardial adipose tissue volume also predicted all-cause mortality, with an HR per standard deviation of 1.28 (95% CI, 1.1-1.37; P = .02), as well as MI (HR = 1.26; 95% CI, 1.09-1.38; P = .001) and stroke (HR = 1.2; 95% CI, 1.09-1.38; P = .02) independent of risk factors in SCOT-HEART (5-year follow-up).
The model also predicted in-hospital AF (HR = 2.67; 95% CI, 1.26-3.73; P < .01) and long-term post-cardiac surgery AF (7-year follow-up; HR = 2.14; 95% CI, 1.19-2.97; P < .01).
The researchers noted that they did not have detailed adiposity data or mortality data available within the SCOT-HEART trial population to investigate the exact causes of noncardiac mortality that could be driving the finding of elevated risk for all-cause mortality conveyed by epicardial adipose tissue volume.
“Large-scale clinical studies are needed to further explore the links between visceral obesity and particularly epicardial adipose tissue volume with non-CV mortality, a striking finding of this study,” Antoniades told Healio. “Also, the impact of weight loss on epicardial adipose tissue and its prognostic value should also be investigated.”
AI for risk stratification
In a related editorial, Cardiology Today Editorial Board Member Daniel S. Berman, MD, and Andrew Lin, MBBS, PHD, from the departments of imaging and medicine at Cedars-Sinai, wrote the data highlight the promise of AI in CV CT to enhance risk stratification.
“The provision of rapid, automated measurements of epicardial adipose tissue could potentially be incorporated into these clinical approaches,” Berman and Lin wrote. “By using AI methods, integration of epicardial adipose tissue measures with other parameters, including coronary artery stenosis, plaque characteristics, CT fractional flow reserve, and pericoronary adipose tissue attenuation may improve the global assessment of cardiovascular risk from coronary CTA. These AI assessments, including measurements of epicardial adipose tissue, are likely to play an important role in the use of imaging to guide clinical decisions regarding patient management in the future.”
Reference:
For more information:
Charalambos Antoniades, MD, PhD, FRCP, FESC, can be reached at charalambos.antoniades@cardiov.ox.ac.uk; Twitter: @charis_oxford.
Collapse
Click Here to Manage Email Alerts