Radiomics Combined with CT-FFR for Detecting Myocardial Ischemia: Prospective Cohort Study (Preprint)

Le résumé fourni par la source

BACKGROUND Current coronary CT angiography (CCTA)-based methods have limited accuracy in identifying myocardial ischemia. Radiomics may help to provide a more accurate assessment. OBJECTIVE To investigate the diagnostic and reclassification value of myocardial radiomic signatures combined with CCTA-derived fractional flow reserve (CT-FFR) and stenosis quantification in detecting hemodynamically significant coronary artery disease (CAD). METHODS Consecutive symptomatic patients clinically referred for CCTA and invasive coronary angiography (ICA) from five medical centers were prospectively recruited. The datasets were randomly divided into a training and internal validation cohort in a 7:3 ratio from three centers, the remaining patients were enrolled in the external testing cohort. The radiomics features were extracted from the vessel-related myocardium, Boruta and Random-forest was applied for further feature selection and model construction. A traditional model was established combining CT-FFR with CCTA quantification and a hybrid model was further developed by adding radiomics signatures. RESULTS A total of 504 vessel-based myocardial territories from 226 patients underwent analysis. Compared with traditional model, the hybrid model showed superior discrimination of flow-limiting CAD in the training or validation cohort (AUC: 0.81 vs. 0.73 and 0.79 vs. 0.74; all P<.05). In the testing cohort, with radiomics added, it exhibited significant reclassification performance (NRI=0.46, P<.001), especially in non-flow-limiting CAD discrimination within the CCTA positive group. (NRI=0.21, P<.001). CONCLUSIONS Combined with radiomics, the hybrid model outperformed the traditional approach for detecting myocardial ischemia. The radiomics further improved reclassification performance in patients with obstructive stenosis in CCTA.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.