AUA 2023: A Novel Machine-Learning Algorithm to Predict Stone Recurrence with 24-hour Urine Data

(UroToday.com) Dr. Kevin Shee presented on his team’s efforts to utilize machine learning as a tool to determine the utility of a 24-hour urine test in predicting stone recurrence. The team gathered their training data from the Registry for Stones of the Kidney and Ureter to gather a set of 423 kidney stone patients who had stone event data and 24-hour urine samples. With this dataset, they were able to feed this information to seven prediction classification methods (Decision Tree, Logistic regression with ElasticNet, Extra Trees, kNeighbors, LightGBM, Logistic Regression, and Random Forest) to assess for the most optimal model. In doing so, they used feature importance analyses distills to a 4 variable-model that is easy to use clinically. Additionally, a web interface was created for clinical translation. Upon developing this novel algorithm, they were able to use machine-learning to identify patients who are more likely to experience stone events.

This study certainly sparked lots of discussion in the potential this model holds; however, Dr. Shee acknowledges that their model requires external validation, and “another interesting thing to do would be to do some combinatory modeling that does include clinicodemographic characteristics, radiologic findings, and other known risk factors for stone recurrence”. With this new information, Dr. Shee shares that the model will prove to be beneficial in boosting the efficiency of clinical trials to get a better sense of which patients recurrence will occur.

1. Shee, K., Chan, C., Liu, A., Yang, H., Sui, W., Ho, S., Chi, T., & Stoller, M. (2023). PD34-05 A NOVEL MACHINE-LEARNING ALGORITHM TO PREDICT STONE RECURRENCE WITH 24-HOUR URINE DATA. In Journal of Urology (Vol. 209, Issue Supplement 4). Ovid Technologies (Wolters Kluwer Health). https://doi.org/10.1097/ju.0000000000003327.05