Cross-sectional angle prediction of lipid-rich and calcified tissue on computed tomography angiography images

• DOI: 10.1007/s11548-024-03086-2
• Publication Date: 2024-03-13T08:55:41Z
• Authors: Xiaotong Zhang; Alexander Broersen; Hessam Sokooti; Anantharaman Rama...; Pieter Kitslaar; Ramya Parasa; Medeni Karaduman; Amear Souded Ali ...; Christos V. Boura...; Jouke Dijkstra
• Keywords: Hounsfield scale; Computed Tomography Angiography

Abstract Purpose The assessment of vulnerable plaque characteristics and distribution is important to stratify cardiovascular risk in a patient. Computed tomography angiography (CTA) offers promising alternative invasive imaging but limited by the fact that range Hounsfield units (HU) lipid-rich areas overlaps with HU fibrotic tissue calcified plaques contrast within contrast-filled lumen. This paper investigate whether can be detected more accurately on cross-sectional CTA images using deep learning methodology. Methods Two (DL) approaches are proposed, 2.5D Dense U-Net Mask-RCNN, which separately perform detection Cartesian polar domain. spread-out view used evaluate show prediction result regions. accuracy F1-score calculated lesion level for DL conventional methods. Results For plaques, median mean values two proposed methods 91 lesions were approximately 6 3 times higher than those method. was comparable U-Net-based method 3% Conclusion this contribute finer predictions compared studies focusing only longitudinal prediction. angular performance outperforms convincing plaque.