Abstract
Increasing evidence has suggested that plaque characteristics are closely associated with ischemia, and coronary computed tomography (CT) angiography-derived fractional flow reserve (FFR(CT)) based on deep machine learning algorithms has also been used to identify lesion-specific ischemia. Therefore, the aim of the present study was to explore the predictive ability of plaque characteristics in combination with deep learning-based FFR(CT) for lesion-specific ischemia. To meet this end, invasive FFR was used as a reference standard, with the joint aims of the early prediction of ischemic lesions and guiding clinical treatment. In the present study, the plaque characteristics, including non-calcified plaque (NCP), low-density NCP (LD-NCP), plaque length, total plaque volume (TPV), remodeling index, calcified plaque, fibrous plaque and plaque burden, were obtained using a semi-automated program. The FFR(CT) values were derived based on a deep machine learning algorithm. On the basis of the data obtained, differences among the values between the atopic ischemia and the non-significant lesions groups were analyzed to further determine the predictive value of independent predictors for atopic ischemia. Of the plaque features, FFR(CT), LD-NCP, NCP, TPV and plaque length differed significantly when comparing between the lesion-specific ischemia and no hemodynamic abnormality groups, and LD-NCP and FFR(CT) were both independent predictors for ischemia. Additionally, FFR(CT) combined with LD-NCP showed a greater ability at discriminating ischemia compared with FFR(CT) or LD-NCP alone. Taken together, the findings of the present study suggest that the combination of FFR(CT) and LD-NCP has a synergistic effect in terms of predicting ischemia, thereby facilitating the identification of specific ischemia in patients with coronary artery disease.