AI-Powered Tool CADDIE Boosts Colonoscopy Accuracy in Landmark European Trial

A large European trial has tested an AI-powered tool designed to improve colonoscopy accuracy. The study, known as the EAGLE Trial, compared standard procedures with CADDIE-assisted colonoscopies. Results showed a significant boost in detecting hard-to-spot and high-risk lesions.

The trial took place across eight countries: Germany, Italy, France, Spain, the Netherlands, Poland, the UK, and Sweden. It involved 841 patients and 22 endoscopists, making it one of the largest studies of its kind. The CADDIE application, deployed via cloud, demonstrated real-time performance without heavy reliance on local hardware.

Hospitals can adopt the system through subscription models, reducing upfront costs. The AI was trained on a dataset rich in clinically challenging lesions, including flat sessile serrated lesions (SSLs) and large polyps. These are often missed during standard procedures.

Key findings included a 7.3% absolute increase in adenoma detection rates when using CADDIE. Detection of large adenomas (over 10 mm) and non-polypoid adenomas also rose sharply. Most notably, SSLs—considered high-risk—were detected 230% more frequently with AI assistance.

Despite its benefits, CADDIE is not meant to replace a full patient evaluation or serve as the sole basis for diagnosis. Gastroenterologists remain responsible for reviewing AI-flagged areas and making final decisions on polyp presence or treatment. The full study is available at https://doi.org/10.1038/s41746-025-02270-1.

The trial highlights CADDIE's potential to enhance colonoscopy accuracy, particularly for lesions that are easily overlooked. With cloud-based deployment and subscription options, hospitals may adopt the system more flexibly. However, clinical decisions will still depend on the expertise of medical professionals.