Novel radiomics features from CCTA images for the functional evaluation of significant ischaemic lesions based on the coronary fractional flow reserve score

• Published in: The International Journal of Cardiovascular Imaging Vol. 36; no. 10; pp. 2039 - 2050
• Main Authors: Hu, Wenchao, Wu, Xiangjun, Dong, Di, Cui, Long-Biao, Jiang, Min, Zhang, Jibin, Wang, Yabin, Wang, Xinjiang, Gao, Lei, Tian, Jie, Cao, Feng
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 01.10.2020; Springer Nature B.V
• Subjects: Aged; Angiography; Cardiac Imaging; Cardiology; Computed tomography; Computed Tomography Angiography; Confidence intervals; Coronary Angiography; Coronary Artery Disease - diagnostic imaging; Coronary Artery Disease - physiopathology; Coronary Stenosis - diagnostic imaging; Coronary Stenosis - physiopathology; Coronary Vessels - diagnostic imaging; Coronary Vessels - physiopathology; Diagnostic systems; Diameters; Feasibility Studies; Feature extraction; Female; Fractional Flow Reserve, Myocardial; Humans; Imaging; Ischemia; Lesions; Male; Medicine; Medicine & Public Health; Middle Aged; Original Paper; Predictive Value of Tests; Prognosis; Radiographic Image Interpretation, Computer-Assisted; Radiology; Radiomics; Reclassification; Regression analysis; Reproducibility of Results; Retrospective Studies; Robustness; Severity of Illness Index; Stenosis; Training; Wavelet Analysis; CT angiography; Myocardial ischaemia; Coronary artery disease; Radiomics
• ISSN: 1569-5794; 1573-0743; 1875-8312
• DOI: 10.1007/s10554-020-01896-4

To explore the superiority of radiomics analysis in the diagnostic performance of coronary computed tomography angiography (CCTA) for identifying myocardial ischaemia and predicting major adverse cardiovascular events (MACE). A total of 105 lesions from 88 patients who underwent CCTA and invasive fractional flow reserve measurement were collected as the training set, and another 31 patients with CCTA and clinical outcome information were used as the validation set. Conventional CCTA features included the stenosis diameter, length, Agatston score and high-risk plaque characteristics. After extracting and selecting radiomics features, the robustness of the radiomics features was examined, and then conventional and radiomics models were established using logistic regressions. The area under the receiver operating characteristic (ROC) curve (AUC) and Net Reclassification Index (NRI) were analysed to compare the discrimination and classification abilities between the two models in both the training and validation sets. A total of 1409 radiomics features were extracted, and three wavelet features were finally screened out. The robustness test showed good stability for the refined radiomics features. Compared with the conventional model, the radiomics model displayed a significantly improved diagnostic performance in the training set (AUC 0.762 vs. 0.631, 95% confidence interval [CI] 0.671–0.853 vs . 0.519–0.742, P = 0.058) but a slightly improved diagnostic performance in the validation set (AUC 0.671 vs. 0.592, 95% CI 0.466–0.875 vs. 0.519–0.742, P = 0.448). The NRI of the radiomics model was increased in both the training and validation sets (NRI 0.198 and 0.238, respectively). Quantitative radiomics analysis was feasible and might help to improve the diagnostic performance of CCTA but is still controversial for predicting MACE.