Coronary artery stenosis and vulnerable plaque quantification on CCTA by deep learning methods

Abstract Introduction Coronary computed tomography angiography (CCTA) has emerged as a reliable non-invasive modality to assess coronary artery stenosis (CAS) severity and vulnerable plaque (VP). However, comprehensive CCTA assessment, especially VP, is time-consuming dependent on reader expertise, limiting CCTA's true potential. Purpose In this study, we aim develop validate deep learning (DL) based system capable of evaluating CAS characterising VP CCTA. Methods A DL was trained 3909 expert annotated vessels. subset 824 vessels used train the model for presence VP. The 2D U-Net 3D convolutional neural network architecture. automatically performed vessel tracking segmentation quantify characterise categorised 0%, 1–49% ≥50%. defined as: low attenuation (LAP; ≤30 Hounsfield units), positive remodelling (PR; ≥10% diameter) spotty calcification (SC; <3mm). then tested 1435 (mean calcium score 197±502) 365 419±551), its diagnostic performance compared with readers. Results testing data had prevalence 75% (1080/1435), 18% (257/1435) 6.8% (98/1435) ≥50% stenosis, respectively. present in 20% (72/365) respective dataset 44% (32/72) LAP, 19% (14/72) PR 36% (26/72) SC. Average analysis time 3.7±2.0s 3.5±1.8s, Diagnostic our summarised Tables 1 (CAS severity) 2 (VP characteristics). Conclusions We developed an rapidly Our demonstrated high specificity accuracy both quantification when Funding Acknowledgement Type funding sources: None.