< Back to previous page

Project

AI-system for real-time automated polyp detection and characterization during colonoscopy: from initial proof of concept to validation and valorization in routine clinical practice

Colorectal cancer is the third most frequent cancer in the general population. Accurate detection of colorectal polyps during colonoscopy and their characterisation (i.e. differentiation between benign and malignant polyps) is crucial for prevention of cancer related mortality, but can be challenging in clinical practice with highly variable detection rates between different operators and reported operator-dependent miss rates as high as 20%. Resected polyps are generally sent out for pathological examination to differentiate between non-neoplastic (hyperplastic) and neoplastic (adenomas) lesions, which has significant cost implications, taking into account that in about 40-50% of patients polyps are resected. Accurate endoscopic diagnosis of polyp histology based on current clinical guidelines would enable a discard and surveillance strategy. We have developed an initial proof of concept of an automated AI-system that supports real-time polyp detection, delineation and characterisation during ongoing colonoscopy based on histological ground truth. In this project, we aim to validate and refine this system in a real-life multi-center setting to investigate its added value for routine clinical practice and to prepare a business plan for economical valorisation of the validated system.
Date:1 Jan 2021 →  31 Dec 2022
Keywords:colonoscopy, polyp characterization, AI-tool, polyp detection, computer-aided diagnosis
Disciplines:Image-guided interventions, Gastro-enterology