Predict the endoleak risk after EVAR based on multi-dimensional anatomical features before AAA surgery

Endovascular Aortic Aneurysm Repair (EVAR) can effectively prevent the continuous growth and rupture of Abdominal Aortic Aneurysms (AAA). It is widely used in the treatment of abdominal aortic aneurysm. Endoleak is a serious complication after EVAR operation. How to predict the occurrence of endoleak is an urgent problem to be solved. In this study, local multi-dimensional anatomical features of 381 AAA patients were calculated. Combined with morphological features and different machine learning (ML) methods, the occurrence of endoleak was predicted. The results showed that the model using Gaus Naive Bayes classfier had the best comprehensive prediction performance for the occurrence of endoleak, with AUC of0.735 and accuracy of 0.771. The use of ML combined with multi-modal local anatomical features can preliminarily predict the occurrence of endoleak events before surgery, avoid excessive follow-up, and improve long-term management of patients in clinical practice.