Researchers at the University of Geneva have developed an AI-powered stool test that detects 90% of colorectal cancers, nearly matching colonoscopy's 94% detection rate and outperforming all existing non-invasive screening methods. The breakthrough, published in Cell Host & Microbe on April 9, was led by Professor Mirko Trajkovski and PhD student Matija Tričković.
The team created the first detailed catalogue of all human gut bacteria at the subspecies level using machine learning, revealing subtle microbial patterns linked to cancer that broader species-level analysis misses. By analyzing simple stool samples, the AI identifies cancer-associated bacterial signatures with unprecedented precision, offering a low-cost, non-invasive alternative to colonoscopy.
Current non-invasive tests like FIT and Cologuard typically achieve 70–80% sensitivity. The Geneva team's approach closes much of that gap, and researchers believe the model could match colonoscopy accuracy with more training data. The future could see this test become a routine screening tool, with colonoscopies reserved only for confirming positive results.
