Machine Learning-Based Prediction of Distal Aortic Enlargement after Thoracic Endovascular Aortic Repair of Stanford Type B Aortic Dissection: A Multicenter, Retrospective Cohort Study

Abstract

Abstract Background: Thoracic endovascular aortic repair (TEVAR) is an important treatment for patients with acute complicated Stanford type B aortic dissection. However, postoperative distal aortic enlargement is a severe complication of TEVAR. This study aimed to construct a prediction model for postoperative distal aortic enlargement using machine learning algorithms and explore high-risk factors that accelerate the progression of postoperative distal aortic enlargement. Methods: Study participants were selected from the multicenter, open cohort ROBUST (Registry Of type B aortic dissection with the Utility of STent graft) study. The least absolute shrinkage and selection operator regression method was employed to select variables. Seven machine learning algorithms (k-nearest neighbor, logistic regression, decision tree, random forest, support vector machine, sklearn neural network, and extreme gradient boosting) were applied to construct prediction models. Further, to explore the important factors that affected the progression of postoperative distal aortic enlargement, we applied three different regression algorithms (random forest, extreme gradient boosting, and light gradient boosting machine) to analyze the data of patients with postoperative distal aortic enlargement. Results: We retrospectively analyzed the data of 184 patients who underwent thoracic endovascular aortic repair for type B aortic dissection at four medical centers. The median follow-up time was 12.37 months (IQR, 8.07–18.17 months). In total, 26 variables were identified using the least absolute shrinkage and selection operator regression. The model constructed using the random forest algorithm exhibited the best prediction performance among the seven models. The regression model constructed by the light gradient boosting machine showed better performance than the other two models. Accordingly, TAT_2.0 (partial thrombosis of the thoracic aorta), SVS (scored according to the comorbidity grading system of the Society for Vascular Surgery), and VRAOFL (visceral or renal aorta originating from a false lumen) were the top three high-risk factors that accelerated the progression of PDAE. Conclusions: The random forest prediction model can improve the prediction accuracy for postoperative distal aortic enlargement, and the light gradient boosting machine regression model can identify high-risk factors that accelerate the progression of postoperative distal aortic enlargement.