ROCHESTER, Minn. — Mayo Clinic researchers developed and tested an artificial intelligence (AI) model that used results from a standard 24-hour urine test to predict kidney stone composition more accurately than a conventional urine-based calculation. Using the AI model, researchers correctly predicted kidney stone composition in 73% of patients, compared with 33% using the conventional calculation alone.
The findings , published in Mayo Clinic Proceedings, could provide doctors and other healthcare professionals with another source of clinical information when a kidney stone is not available for testing. Knowing what a stone is made of can help guide treatment and prevention strategies. The model is intended to supplement, not replace, existing test results and trends.
"Many patients never have a kidney stone available for analysis, so doctors are often working with an incomplete picture," says Patrick Day, senior author of the study and a principal developer in Mayo Clinic's Department of Laboratory Medicine and Pathology . "Our findings show that AI may help us get more clinically useful information from a test that is already part of kidney stone care. For patients, that could mean more informed treatment decisions. For laboratories and healthcare systems, it highlights how existing laboratory test data can be used in new ways to support clinical care."
The study's first author, Eric B. Lantz, Ph.D., a senior data science analyst in Mayo Clinic's Department of Laboratory Medicine and Pathology, developed the AI model using data from more than 13,000 patients, including age, sex and results from 24-hour urine tests. The researchers then evaluated the model using data from 1,653 additional patients.
Kidney stones are common and can be very painful. One in 11 Americans will have a stone at least once in their life, and the lifetime recurrence rate is high at 60% to 80%. Most kidney stones are made of calcium oxalate, though calcium phosphate and uric acid stones are also common.
Identifying the type of stone can help doctors determine treatment and strategies to reduce the risk of another stone. The most direct way to determine composition is to analyze the stone itself, but a stone is not always available. Patients may not collect the stones they pass, and stones may not be available for testing after surgery. In those cases, doctors may rely on clinical information and laboratory results to estimate the stone's composition.
The 24-hour urine test measures substances in the urine that can contribute to kidney stones. The test can help identify a patient's risk factors and guide treatment, but it produces many individual measurements, and the relationships among them can be challenging to interpret.
The researchers developed the AI model to analyze those measurements together, identify patterns and predict the most likely kidney stone type. The study found the model performed better than the conventional calculation alone, including when researchers looked at different age and sex groups.
"Laboratory tests contain useful information not only in the individual results but also in the relationships among them," says Dr. Lantz. "AI can identify those patterns and could help us get more clinical value from data we are already collecting."
For a complete list of authors, disclosures and funding, review the study .
###
About Mayo Clinic
Mayo Clinic is a nonprofit organization committed to innovation in clinical practice, education and research, and providing compassion, expertise and answers to everyone who needs healing. Visit the Mayo Clinic News Network for additional Mayo Clinic news.
Mayo Clinic Proceedings
A Validated Artificial Intelligence Model Versus Urinary Supersaturation Alone for Predicting Kidney Stone Composition
12-Nov-2026