Automatic coronary artery calcium scoring from unenhanced-ECG-gated CT using deep learning

• A three-dimensional (3D) deep learning-based model can be trained effectively to compute automatically the coronary artery calcium score from a CT examination. The proposed is based on U-Net architecture designed for 3D segmentation of calcifications. It would useful test and validate this method wider range acquisitions, including non-ECG-gated ones, performed routine examinations lung disease. purpose study was develop evaluate an algorithm that estimate amount (CAC) unenhanced electrocardiography (ECG)-gated computed tomography (CT) cardiac volume acquisitions by using convolutional neural networks (CNN). used set five CNN with database 783 detect segment calcifications in volume. Agatston score, conventional CAC scoring, then slice resulting mask compared ground truth manually estimated radiologists. quality estimation assessed concordance index (C-index) risk category separate testing 98 independent examinations. final yielded C-index 0.951 set. remaining errors were mainly observed small-size and/or low-density calcifications, or located near mitral valve ring. here unenhanced-ECG-gated fast, robust yields accuracy similar those other artificial intelligence methods, which could improve workflow efficiency, eliminating time spent selecting score.