Fig. 1: Overall study design.
From: Predicting mortality from AI cardiac volumes mass and coronary calcium on chest computed tomography
We utilized a convolutional long short-term memory (ConvLSTM) model (light gray) which uses computed tomography slices as inputs (blue arrows) to segment coronary artery calcium (CAC). TotalSegmentator (dark gray), which uses a no new-net UNet (nnUNet) architecture was used to segment cardiac chamber volumes and left ventricular (LV) myocardium. We applied these models to patients in the external populations (dark blue) to quantify CAC and cardiac chamber volumes, where we evaluated correlations between measures and associations with outcomes (light blue). CT computed tomography, CV cardiovascular, ECG electrocardiogram, LA left atrium, RA right atrium, RV right ventricle.
