Evaluation of the Risk of Urinary System Stone Recurrence Using Anthropometric Measurements and Lifestyle Behaviors in a Developed Artificial Intelligence Model.

Background/Objectives: Urinary system stone disease is an important health problem both clinically and economically due to its high recurrence rates. In this study, an innovative hybrid approach based on deep learning is proposed to predict the recurrence risk of stone disease. Methods: Patient data were divided into three subsets: anthropometric measurements (Part A), derived body composition indices (Part B), and other clinical and demographic information (Part C). Each data subset was processed with autoencoder models, and low-dimensional, meaningful features were extracted. The obtained features were combined, and the classification process was performed using four different machine learning algorithms: Extreme Gradient Boosting (XGBoost), Cubic Support Vector Machines (Cubic SVM), k-Nearest Neighbor algorithm (KNN), and Decision Tree (DT). Results: According to the experimental results, the highest classification performance was obtained with the XGBoost algorithm. The suggested approach adds to the literature by offering a novel solution that makes early risk calculation for stone disease recurrence easier. It also shows how well structural feature engineering and deep representation can be integrated in clinical prediction issues. Conclusions: Prediction of the stone recurrence risk in advance is of great importance both in terms of improving the quality of life of patients and reducing the unnecessary diagnostic evaluations along with lowering treatment costs.