Development and deployment of Computer-aided Real-Time feedback for improving quality of colonoscopy in a Multi-Center clinical trial

• Published in: Biomedical signal processing and control Vol. 83; p. 104609
• Main Authors: Tavanapong, Wallapak, Pratt, Jacob, Oh, JungHwan, Khaleel, Mohammed, Wong, Johnny S., de Groen, Piet C.
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.05.2023
• Subjects: Colonoscopy; Computer-aided analysis; Polyp detection rate; Quality; Real-time feedback; CADe; EID; GI; Real-time feedback; OS; EMIS-Proc; PID; FoV; Polyp detection rate; Computer-aided analysis; CNN-Proc; SID; EMIS; CRC; Colonoscopy; Quality; HUD; VCC; IRB; ActHIPAA; EVP
• ISSN: 1746-8094
• DOI: 10.1016/j.bspc.2023.104609

•New visual feedback for intra-procedure quality during live colonoscopy.•Quality metric from automated retroflexion detection.•Convolutional Neural Network models for video frame analysis.•Modular software architecture for adding new analysis models.•Lessons learned from deployment in a multi-center clinical trial. The quality of colonoscopy has been a subject of interest in the gastroenterology community for over two decades. High-quality colonoscopy leads to high polyp detection rates, reducing mortality associated with colorectal cancer. Methods: This paper describes the development and deployment of Endoscopic Multimedia Information System (EMIS), computer-aided software that provides real-time feedback on colonoscopy quality such as the endoscopist’s techniques in inspecting the colon. On the contrary, most other software in this field aims to recognize polyps. The deployed version of EMIS includes new analysis components using Convolutional Neural Networks and new types of visual feedback. EMIS gives feedback only when an important change in the quality of the examination occurs. We present first-hand technical and operational challenges faced during our three-center trial and current solutions. Results: This work provides valuable information for others attempting to implement real-time feedback in routine colonoscopy screening.