Machine learning for real-time prediction of complications induced by flexible uretero-renoscopy with laser lithotripsy

Abstract

It is not always easy to predict the outcome of a surgery. Peculiarly, when talking about the risks associated to a given intervention or the possible complications that it may bring about. Thus, predicting those potential complications that may arise during or after a surgery will help minimize risks and prevent failures to the greatest extent possible. Therefore, the objectif of this article is to propose an intelligent system based on machine learning, allowing predicting the complications related to a flexible uretero-renoscopy with laser lithotripsy for the treatment of kidney stones. The proposed method achieved accuracy with 100% for training and, 94.33% for testing in hard voting, 100% for testing and 95.38% for training in soft voting, with only ten optimal features. Additionally, we were able to evaluted the machine learning model by examining the most significant features using the shpley additive explanations (SHAP) feature importance plot, dependency plot, summary plot, and partial dependency plots.