AI Tool CADDIE Boosts Colonoscopy Accuracy in Landmark European Study

A new AI-assisted tool called CADDIE has shown promising results in improving colonoscopy accuracy. The system, tested across eight European countries, helped detect more precancerous lesions than standard procedures. The clinical trial, known as the EAGLE Study, involved 841 patients and 22 endoscopists from Germany, France, Italy, Spain, the Netherlands, Poland, the UK, and Sweden. It compared traditional colonoscopy methods with CADDIE-assisted procedures. The AI tool was trained on a dataset focused on hard-to-spot lesions, including flat sessile serrated lesions (SSLs) and large polyps. Results showed a 7.3% absolute increase in adenoma detection rates when using CADDIE. The system particularly improved detection of high-risk lesions: large adenomas over 10 mm rose by 93%, non-polypoid adenomas by 57%, and SSLs by 230%. These findings highlight CADDIE's potential to catch lesions that standard colonoscopies often miss. The application also demonstrated real-time performance and efficiency across different testing environments. Its cloud-based deployment allows hospitals to adopt it without heavy hardware investments, offering a subscription-based model instead. Rawen Kader, the study's lead investigator and a GI researcher at University College London, called the results a turning point for AI in endoscopy. The trial confirms that CADDIE enhances lesion detection during colonoscopies, particularly for difficult-to-spot subtypes. Hospitals may now consider the tool as a scalable solution, given its cloud flexibility and proven performance. This could lead to fewer missed cancers and more effective screenings in the future.