Plate Nº 40 · recorded November 13, 2026
Health & Medicine ResearchReported finding
AI Tool Predicts Kidney Stone Type More Than Twice as Accurately
Mayo Clinic's AI read routine 24-hour urine tests and predicted kidney stone type with 73% accuracy, more than doubling the traditional calculation's 33% — though real-world benefits remain unproven.
By Nathan Brooks3 min read645 words
In brief
- The AI tool predicted main kidney stone type correctly in 73% of patients, versus 33% for the traditional urine-based calculation.
- Researchers trained the system on data from 13,000 patients and validated it on 1,653 patients with known stone types.
- Findings were published in Mayo Clinic Proceedings in 2026.
- The tool still gave incorrect predictions for roughly one in four patients.
- The study measured prediction accuracy, not whether the tool prevents new stones or improves health outcomes.
An AI tool from Mayo Clinic researchers correctly predicted a patient's main kidney stone type 73% of the time, compared with 33% for the standard calculation doctors use today. That's more than double the accuracy, and it worked even when no stone was available for lab testing — a common gap in kidney stone care.
The findings appeared in Mayo Clinic Proceedings in 2026. The research team built and validated the system using data from thousands of real patients, not lab simulations.
Why does stone type matter?
Kidney stones are hard deposits that form when certain substances build up in urine. They cause severe pain, and people who've had one often develop another — roughly a recurring condition rather than a one-off event for many patients.
Different stone types call for different prevention plans. A stone that's mainly calcium oxalate needs different advice than one made of calcium phosphate or uric acid. Knowing the composition helps doctors tailor treatment, especially for people who keep forming new stones.
How did doctors identify stones before?
Usually, doctors examine a stone after it passes out of the body or is surgically removed. That's the gold standard — direct chemical analysis of the stone itself.
The catch: many patients don't have a stone available. The stone may never be caught, or it may not be recovered after a procedure. In those cases, doctors fall back on a traditional calculation based on a 24-hour urine collection, which measures substances linked to stone formation. That method guessed right only a third of the time in this study.
How does the AI approach work?
The new tool uses the same familiar 24-hour urine test patients already undergo. Instead of running the results through a standard formula, the AI searches for patterns across all the measurements.
To build it, the researchers trained the system on information from 13,000 patients, including their ages, sexes and urine test results. They then tested it on a separate group of 1,653 patients whose stone types were already known — a genuine out-of-sample check, not a replay of the training data.
The result: the AI beat the traditional calculation in the overall test group and held its edge across different age and sex groups. That consistency matters, because a tool that only works for one demographic would be far less useful in a general clinic.
What are the limitations?
The tool still got the wrong answer for about one in four patients. So it can't replace direct stone testing, and it can't replace a doctor's judgment. Think of it as a better-informed guess, not a diagnosis.
There's a bigger caveat. The study measured how well the system predicted stone types — not whether those predictions actually prevented new stones or improved patients' health. Those are the outcomes patients care about most, and they remain untested.
The researchers say more research in everyday clinical care is needed before the tool's benefits can be confirmed. Promising as the numbers look, real-world validation is the next hurdle.
What happens next?
The practical appeal is that the tool draws on a test already used routinely in kidney stone care. No new equipment, no new patient procedure — just a smarter reading of data doctors already collect. That could speed adoption if further studies pan out.
For now, patients who want to reduce their risk can look to established lifestyle measures. Related research also explores bacteria linked to kidney stone formation, and diet remains a central lever in prevention.
The bottom line: a 73% accurate prediction from a routine urine test is a meaningful step up from 33%. But until clinical trials show it prevents stones, it's a decision-support aid — one that helps doctors and patients choose smarter prevention, while the stone itself remains the final word.
via knowridge.com (Source)
More from Nathan Brooks
Show full bio
Market editor covering companies, deals and strategy in general science news at SciBeat.
352 articles
Nearby plates
- Half a Million Years Ago, Humans May Have Kept Souvenirs
- New Study Narrows the Search for Stonehenge’s Altar Stone
- FDA-Cleared Urine Test Detects Deadly Mold Infection Within Hours
- Tin deposits come from source rocks, not cooling, study finds
- Nearly 1 in 3 Diabetes Patients Face Kidney, Heart Disease or Death