Rationale and design of a multicenter, prospective, diagnostic clinical study: A study protocol for evaluating the diagnostic validation of deep learning-based noninvasive CT-FFR for in-stent restenosis.

INTRODUCTION: Currently, there are many diagnostic strategies for in-stent restenosis (ISR) used clinically, including invasive coronary angiography (ICA), coronary computed tomography angiography (CCTA), and Fractional Flow Reserve (FFR). CCTA is not recommended for post-stent implantation patients owing to suboptimal image quality caused by artifacts. The FFR application is limited by its procedural complexity. Precise evaluation may be achieved by using computed tomography-derived fractional flow reserve (CT-FFR), which combines computational fluid dynamics (CFD) with CCTA. Anatomical and functional assessments of ISR lesions could be integrated effectively as well in this way. However, the computational complexity and prolonged processing time may hinder its utility in clinical use.

METHODS: This study is a multicenter, prospective, diagnostic study, aiming to establish a deep learning-based CT-FFR model for the accurate assessment of ISR and to validate its diagnostic performance using invasive FFR as the reference standard. This study will be carried out in Beijing Anzhen Hospital and 6 subcenters in China. We planned to prospectively enroll 331 post-stent implantation patients with available CCTA data since June 2022, and invasive FFR will be performed within 3 months when clinically indicated. Patient recruitment is currently ongoing. Among them, 250 patients from Beijing Anzhen Hospital will be used to adapt and extend the existing DEEPVESSEL model, a deep learning-based CT-FFR computational software designed for the non-invasive functional assessment of coronary artery disease and previously validated in de novo coronary lesions, for application in the assessment of in-stent restenosis (ISR), and 81 patients from the other 6 subcenters will be used in external validation. Sensitivity, specificity, accuracy, positive predictive value, and negative predictive value with their corresponding 95% confidence intervals (CIs) were calculated for CT-FFR. The receiver operating characteristic (ROC) curve was analyzed, and the area under the curve (AUC) was calculated. The McNemar test and Bland-Altman plot will be used to examine the diagnostic consistency between CT-FFR and invasive FFR. The correlation was analyzed by Spearman's correlation coefficient.

TRIAL REGISTRATION NUMBER: ChiCTR2200058822.