Abstract 13475: Epicardial Adiposity Measured by a Deep Learning Network, Predicts Mortality and Cardiovascular Events in Patients Undergoing Cardiac CT

Abstract

Introduction: Epicardial adipose tissue (EAT) is a visceral fat deposit within the pericardial sac. The automated quantification of EAT volume is possible from routine CCTA scans via deep-learning. The use of automated EAT quantification for the assessment of cardiovascular disease (CVD) risk in addition to standard measures of obesity like BMI has not been fully explored. Purpose: To use deep-learning for automated segmentation of EAT from routine CCTA scans to assess the long-term CVD risk conveyed by EAT. Methods: A deep-learning automated EAT segmentation tool using a 3D Residual-U-Net neural network architecture for 3D volumetric segmentation of CCTA data was created and trained on over 2500 consecutive CCTAs from within the Oxford Risk Factors And Non Invasive Imaging (ORFAN) Study. External validation in 817 patients demonstrated excellent correlation between machine and human expert (CCC = 0.972). The prognostic value of deep-learning derived EAT volume was assessed against 5 years outcomes from the SCOTHEART trial (n=1588), with adjustment for CVD risk factors. An optimal cutoff was selected by identifying the EAT value that maximized the Youden’s J index (sum of sensitivity and specificity) for the three outcomes of interest - high risk was deemed to be EAT ≥ 170.5cm 3 . Results: There were 35 deaths (all-cause mortality), 35 non-fatal myocardial infarctions and 8 non-fatal strokes during the 5 years follow up period. By using multi-variable cox-regression, EAT volume was predictive of all-cause mortality (Figure 1A), non-fatal MI (Figure 1B), and non-fatal stroke (Figure 1C) independently from CVD risk factors. Conclusions: Automatically segmented EAT volume measured using a deep learning network, predicts 5-year all-cause mortality, heart attacks and stroke independently of BMI and clinical risk profile of the patients. This suggests that measures of visceral obesity will be of value in the interpretation of cardiovascular computed tomography.