A deep learning model for predicting the outcome of persistent type 2 endoleaks after endovascular abdominal aortic aneurysm repair

Persistent type 2 endoleaks (pT2ELs) require long-term follow-up to prevent life-threatening complications. This study aimed to create a model to predict the outcome of pT2ELs after endovascular abdominal aortic aneurysm repair (EVAR) using deep learning (DL). We retrospectively reviewed 94 patients with pT2ELs treated between January 2010 and December 2019 at Zhongshan Hospital Fudan University. The median follow-up was 38.2 months, and 21 patients (22.3%) had pT2ELs-related severe adverse events (SAEs). ITK-SNAP software was used to draw the region of interest (ROI). Pre-processing of the images and creation of the DL model were performed using MATLAB. Of the total, 80% of the patients were randomly classified as the training set and 20% as the test set. The area under the curve (AUC) was used to evaluate the predictive power of the model. Visualisation techniques were used to better understand the DL model-prediction process. The number of patients in the training set was 75 (including 17 with SAEs) and the number of patients in the test set was 19 (including 4 with SAEs). By training 10240 computed tomography angiography images (n = 75), the DL model achieved encouraging predictive performance in the test set with an AUC of 0.917, accuracy of 0.842, and F1 score of 0.897. Visualisation techniques improved the interpretability of the model. An end-to-end DL model can be used as an additional tool to predict the outcomes of pT2ELs after EVAR.