Abstract
BACKGROUND: The course of patients with type B aortic intramural hematoma (IMH) is unstable, and different studies have shown that the evolution of this type of IMH is highly heterogeneous. This study sought to explore the value of radiomics in predicting the prognosis of type B aortic IMH, and to develop and validate a prediction model of type B aortic IMH progression. METHODS: A total of 119 patients with type B aortic IMH who had not undergone surgical or thoracic endovascular aortic repair treatment were enrolled in this study. These patients were divided into the progressive group (n=61) and stable group (n=58) based on re-examination aortic computed tomography angiography (CTA) imaging. The patients were then randomly divided into the training cohort (n=95) and the validation cohort (n=24). The uAI Research Portal (URP) was used to perform the radiomics feature extraction of the intensity, shape, texture, and gradient features. Next, the least absolute shrinkage and selection operator (LASSO) logistic regression (LR) method was used for feature selection, and prediction models were constructed based on clinical features, CTA imaging features, and radiomic features. Different machine-learning algorithms were used to build the models, including random forest (RF), support vector machine (SVM), LR, K-nearest neighbor (KNN), decision tree, and stochastic gradient descent (SGD) algorithms. The area under the curve (AUC) of the receiver operating characteristic (ROC) curve, sensitivity, specificity, accuracy, and F1 score were used to evaluate the efficacy of the prediction models. RESULTS: After the application of the LASSO method, 12 radiomic features were selected from an initial pool of 1,004 radiomic features, 12 features were selected from the 21 clinical features, and 11 features were selected from the 15 CTA imaging features. Five predictive models were then constructed using distinct combinations of feature sets. For the test set, the AUC of the SVM algorithm in the radiomics model was the highest (0.833), that of the KNN algorithm in the clinical model was the highest (0.701), that of the RF algorithm in the CTA imaging model was the highest (0.806), and those of the LR and SGD algorithms in the clinical + CTA imaging model were the highest (both 0.792). The combined radiomics + clinical + CTA model had the highest AUC value (0.917), which was higher than that of the single radiomics model (0.833), CTA model (0.806), clinical + CTA model (0.792), and clinical model (0.701). The sensitivity, specificity, accuracy, precision and F1 scores of the combined radiomics + clinical + CTA model were all >0.75. CONCLUSIONS: The comprehensive model that incorporated clinical, CTA imaging, and radiomic features performed the best and accurately predicted the progression of type B aortic IMH. This model could help clinicians make optimal treatment decisions.