Deep learning to predict the histology of polyps per colonoscopy, offers a promising solution to reduce variation in colonoscopy performance. Meanwhile, the concept of 'optical biopsy' where in vivo classification of polyps based on enhanced imaging replaces histopathology has not been incorporated into routine practice, largely limited by inter-observer variability and generally meeting accepted standards only in expert settings. Real-time decision support software has been developed to detect and characterise polyps, whilst also offering feedback on the technical quality of inspection.
This study will evaluate the performance of the CAD EYE automatic characterization system for the histology compared to histological analysis. And secondary aims : the diagnostic performance of the CAD EYE automated detection device compared to a standardized video recording with blind independent review.
Procedure: The screening colonoscopy will be performed by an investigator. The automatic detection and characterization system will be activated at the time of descent of the colonoscopy (after caecal intubation), with video recording (image without CAD EYE and image with CAD EYE). The investigator performing the colonoscopy will be blinded by the results of the CAD EYE.
Follow-up: no specific follow-up is planned after colposcopy
