From data to diagnosis: evaluation of machine learning models in predicting kidney stones

Iparraguirre-Villanueva O.; Paucar-Palomino G.; Paulino-Moreno C. · Neural Computing and Applications (2025) · DOI: 10.1007/s00521-025-11049-0

Resumen

Kidney stones are one of the diseases affecting a large part of the world’s population, and it is getting worse and worse with dietary changes, obesity, medical conditions etc. This disease causes high morbidity and has a high cost for patients’ health, as it can cause severe complications such as hydronephrosis, reduced kidney function and ultimately kidney failure. The aim of this research is to detect kidney stones using Machine Learning (ML) algorithms. For which, the performance of Logistic Regression (LR), K-nearest neighbors (KNN), Decision tree (DT), Support vector machines (SVM), Gaussian naive bayes (GNB) and Random forests (RF) models were evaluated. The results showed that the LR model obtained the best accuracy rate with 0.78 in detecting kidney stones, significantly outperforming the other models, such as KNN, DT, SVM, GNB and RF, which to some extent also achieved high accuracy rates of 0.61, 0.66, 0.68, 0.70 and 0.76, respectively. The result obtained by the LR model provides us with a more accessible approach to develop solutions in environments, where computational resources are limited. Also, the correlation of variables (gravity, pH, osmolality) highlights their relevance in kidney stone formation. This approach may help motivate future research on variable integration. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2025.

Instituciones: Universidad Tecnológica del Perú; Universidad Nacional Agraria de la Selva; Universidad Nacional Ciro Alegría