Diagnostic and prognostic performance of artificial intelligence-based fully-automated on-site CT-FFR in patients with CAD

• DOI: 10.1016/j.scib.2024.03.053
• Publication Date: 2024-03-27T17:09:23Z
• Authors: Bangjun Guo; Mengchun Jiang; Xiang Guo; Chunxiang Tang; Jian Zhong; Mengjie Lu; Chunyu Liu; Xiaolei Zhang; Hongyan Qiao; Fan Zhou; Pengpeng Xu; Yi Xue; Minwen Zheng; Yang Hou; Yining Wang; Jiayin Zhang; Bo Zhang; Daimin Zhang; Lei Xu; Xiuhua Hu; Changsheng Zhou; Jianhua Li; Zhiwen Yang; Xinsheng Mao; Guangming Lu; Longjiang Zhang
• Keywords: Fractional Flow Reserve

Currently, clinically available coronary CT angiography (CCTA) derived fractional flow reserve (CT-FFR) is time-consuming and complex. We propose a novel artificial intelligence-based fully-automated, on-site CT-FFR technology, which combines the automated plaque segmentation luminal extraction model with reduced order 3 dimentional (3D) computational fluid dynamics. A total of 463 consecutive patients 600 vessels from updated China study in Cohort 1 undergoing both CCTA invasive (FFR) within 90 days were collected for diagnostic performance evaluation. For 2, 901 chronic syndromes index clinical outcomes at 3-year follow-up retrospectively analyzed. In 3, association between triple-rule-out CTA major adverse cardiac events acute chest pain emergency department was further evaluated. The accuracy this 0.82 an area under curve on per-patient level. Compared manually dependent techniques, operation time technique substantially shortened by times number clicks about 60 to 1. This has highly successful (> 99%) calculation rate also provides superior prediction value than alone pain. Thus, fully automated, can function as objective convenient tool stenosis functional evaluation real-world setting.