Using machine learning methods (MLMs) to predict stone-free status after percutaneous nephrolithotomy (PCNL). We compared the performance of this system with Guy’s stone score and S.T.O.N.E score system.
Materials and Methods
Data from 222 patients (90 females, 41%) who underwent PCNL at our center were used. Twenty-six parameters, including individual variables, renal and stone factors, surgical factors were used as input data for MLMS. We evaluate the efficacy of four different techniques: Lasso-logistic (LL), random forest (RF), support vector machine (SVM) and Naive Bayes. Model performance was evaluated using area under curve (AUC) and compared with Guy’s stone score and S.T.O.N.E score system.
Overall stone free rate was 50% (111/222). To predict the stone-free status, all receiver operating characteristic curves of the four MLMs were above the curve for the Guy’s stone score. The AUCs of LL, RF, SVM and Naive Bayes were 0.879, 0.803, 0.818, 0.803 respectively. Those values were higher than the AUC of the Guy’s score system, 0.800. The accuracies of the MLMs (0.803–0.818%) were also superior to S.T.O.N.E score system (0.788%). Among the MLMs, Lasso-logistic showed the most favorable AUC.
Machine learning methods can predict stone-free rate with AUCs no inferior to those of Guy’s stone score and S.T.O.N.E score system.