Artificial intelligence programs, particularly one called CatBoost, can predict gallstone disease with 79% accuracy using routine body measurements and blood tests, according to a 2026 study of 319 patients. The AI identified inflammation markers, vitamin D levels, and body composition as the strongest predictors. While ultrasound remains the definitive diagnostic tool, this AI approach could help doctors screen high-risk patients during routine checkups.
Researchers tested artificial intelligence computer programs to see if they could predict who has gallstones using simple health measurements and blood work instead of ultrasound. According to Gram Research analysis, a program called CatBoost correctly identified gallstone patients 79% of the time by analyzing body composition, inflammation markers, and vitamin levels from 319 people. While ultrasound remains the gold standard for diagnosis, this AI approach could help doctors decide which patients need imaging and might catch gallstone disease earlier in routine health checkups.
Key Statistics
A 2026 study of 319 patients found that the CatBoost artificial intelligence program correctly identified gallstone disease 79% of the time using routine health measurements and blood tests, with an AUC score of 0.862.
Body composition measurements alone achieved an AUC of 0.797 for predicting gallstones, nearly matching the performance of blood test markers at 0.829, suggesting that how your body is composed carries important diagnostic information.
C-reactive protein, vitamin D, and bone mass were identified as the three most influential health factors for predicting gallstone disease in the artificial intelligence analysis of 319 patients.
The Quick Take
- What they studied: Whether artificial intelligence computer programs could predict gallstone disease using routine health measurements like body fat, inflammation markers, and vitamin D levels instead of requiring an ultrasound scan.
- Who participated: 319 people from a hospital in Ankara, Turkey: 161 with confirmed gallstones and 158 without gallstones. The study used their existing health records and measurements.
- Key finding: The CatBoost AI program correctly identified gallstone patients 79% of the time and showed strong accuracy (AUC score of 0.862 out of 1.0), outperforming three other AI programs tested.
- What it means for you: Doctors might eventually use AI to flag which patients should get ultrasound screening during routine checkups, potentially catching gallstone disease earlier. However, ultrasound remains the definitive diagnostic tool and this AI is not a replacement.
The Research Details
Researchers gathered health information from 319 patients at a Turkish hospital, including 161 people with gallstones and 158 without. They collected routine measurements like body composition (how much fat and muscle each person had), blood test results (inflammation markers and vitamin levels), and basic demographics like age and gender. They then fed this information into four different artificial intelligence programs designed to learn patterns and predict who had gallstones. The AI programs were tested using a method called cross-validation, which means the computer was trained on some patient data and tested on different patient data to ensure honest results. The researchers also used special techniques to understand which health measurements were most important for making predictions.
This research approach matters because gallstone disease affects millions of people but many cases go undiagnosed until symptoms appear. If AI could identify high-risk patients using routine blood work and body measurements already collected during regular checkups, doctors could catch the disease earlier. The study also shows that body composition measurements (like visceral fat and muscle mass) contain useful information for predicting gallstones, suggesting that metabolic health is connected to gallstone risk.
The study used a rigorous testing method called repeated stratified cross-validation to prevent the AI from just memorizing patterns. The researchers tested four different AI programs to ensure their findings were not dependent on one specific approach. They also used SHAP analysis, a technique that explains which health factors the AI considered most important, making the results transparent rather than a black box. The main limitation is that this was a retrospective study using existing patient records from one hospital, so results may not apply to all populations.
What the Results Show
The CatBoost artificial intelligence program achieved the best performance, correctly classifying gallstone status 79% of the time with an AUC score of 0.862 (where 1.0 is perfect and 0.5 is random guessing). This performance was statistically significantly better than the three other AI programs tested. The program’s predictions were reliable across different patient groups, meaning it worked consistently well. When doctors use this type of AI in practice, the decision curve analysis showed it would provide clinical benefit across a wide range of risk thresholds, suggesting it could be useful for different clinical scenarios.
Body composition measurements alone (like visceral fat and muscle mass from bioimpedance analysis) achieved an AUC of 0.797, which is surprisingly good considering they were used without blood test information. Blood test markers alone achieved an AUC of 0.829. The most important health factors for predicting gallstones were C-reactive protein (a marker of inflammation), vitamin D levels, and bone mass. Interestingly, the relationship between body fat and gallstone risk was not straightforward, suggesting that where you carry fat matters more than total fat percentage.
This study builds on existing research showing that metabolic factors like obesity and inflammation are connected to gallstone disease. Previous studies relied mainly on ultrasound for diagnosis. This research demonstrates that machine learning can extract predictive information from routine health measurements, offering a new approach to risk stratification. The finding that vitamin D and inflammation markers are key predictors aligns with emerging research on metabolic causes of gallstones.
The study included only 319 patients from one hospital in Turkey, so results may not apply to other populations with different genetic backgrounds or healthcare practices. The study was retrospective, meaning researchers looked at past medical records rather than following patients forward in time. The AI was trained and tested on the same general patient population, so it may not work as well on completely different groups of people. The study does not prove that using this AI would actually improve patient outcomes in real clinical practice, only that it can identify patterns in existing data.
The Bottom Line
This research provides proof-of-concept that AI could help identify gallstone risk using routine health measurements. However, ultrasound remains the gold standard for diagnosing gallstones and should not be replaced. The AI approach could potentially be used to flag high-risk patients for ultrasound screening during routine checkups or to help doctors decide which patients with unclear symptoms need imaging. Confidence level: Moderate. This is promising research but needs testing in larger, more diverse populations before widespread clinical use.
This research matters most to doctors and healthcare systems looking for better ways to screen for gallstone disease. People with risk factors for gallstones (obesity, high inflammation, low vitamin D) might benefit from earlier screening using this approach. This does not replace the need for ultrasound if you have symptoms like abdominal pain. People without gallstone symptoms or risk factors do not need to change their behavior based on this research.
If this AI approach is eventually adopted in clinical practice, it would likely be used as a screening tool during routine checkups. Any benefits would depend on whether earlier identification of gallstones leads to better health outcomes, which would require additional research to determine.
Frequently Asked Questions
Can artificial intelligence replace ultrasound for diagnosing gallstones?
No. This AI approach achieved 79% accuracy and is designed for screening and risk assessment, not diagnosis. Ultrasound remains the gold standard for confirming gallstones. The AI could help doctors decide which patients need ultrasound imaging.
What health measurements does the AI use to predict gallstones?
The AI analyzes body composition (visceral fat and muscle mass), blood test markers (especially inflammation and vitamin D levels), and basic demographics. These routine measurements are already collected during standard health checkups.
Should I get screened for gallstones if I have low vitamin D?
Low vitamin D was identified as a risk factor for gallstones in this study, but this single finding does not mean you need screening. Discuss your overall risk factors with your doctor, who can recommend appropriate testing based on your symptoms and health history.
How accurate is this AI compared to ultrasound?
The AI achieved 79% accuracy in identifying gallstone status. Ultrasound is more accurate for diagnosis but requires scheduling and cost. The AI’s strength is using existing health data for quick risk assessment and deciding who should get ultrasound.
Could this AI help catch gallstones earlier?
Potentially. By identifying high-risk patients during routine checkups using existing blood work and body measurements, doctors could recommend ultrasound screening before symptoms develop. However, this benefit needs to be proven in real clinical practice.
Want to Apply This Research?
- Track your body composition metrics (visceral fat percentage, muscle mass), vitamin D levels, and inflammation markers (C-reactive protein) from routine blood work every 6 to 12 months to monitor gallstone risk factors over time.
- Use the app to log your vitamin D intake and sun exposure, monitor your waist circumference and body composition, and track inflammation-reducing habits like regular exercise and anti-inflammatory foods. Share these trends with your doctor during checkups.
- Create a dashboard showing your key risk factors (vitamin D, inflammation markers, body composition) alongside lifestyle factors like diet quality and exercise frequency. Set quarterly reminders to review trends and discuss results with your healthcare provider.
This research demonstrates that artificial intelligence can identify patterns associated with gallstone disease using routine health measurements. However, this AI approach is not a replacement for medical diagnosis. Gallstones must be confirmed by ultrasound or other imaging ordered by a healthcare provider. If you experience symptoms like abdominal pain, fever, or nausea, seek immediate medical attention. Do not use this information to self-diagnose or delay seeking professional medical care. Always consult with your doctor before making health decisions based on research findings.
This research translation is published by Gram Research, the science division of Gram, an AI-powered nutrition tracking app.