Study to integrate CNN inside a WCE to realize a screening tool

Orlando Chuquimia (1) , Thomas Garbay (1) , Weiqin Xu (1) , Andrea Pinna (1) , Xavier Dray (2) , Hichem Sahbi (3) , Bertrand Granado (1)

1 SYEL - Systèmes Electroniques
2 CHU Saint-Antoine [AP-HP]
3 MLIA - Machine Learning and Information Access

Screening is a method to improve the earlydetection of colorectal cancer. Now, screening is basedon an immunochemical test that look for blood in faecalsamples, but image is the best modality to detect themarker of colorectal cancer : polyps. In 2003 WirelessCapsule Endoscopy was introduced and opened a way tointegrate automatic image processing to realize a screen-ing tool. In parallel Convolutionnal Neural Networkshave demonstrated their high capacity to detect polyps inmany scientific studies, but fail to be integrable. In thisarticle we present our works to integrate CNN or imageprocessing based on a CNN inside a WCE to realize apowerful screening too.

Keywords: Convolutional Neural Network; Wireless Capsule Endoscopy; Colorectal Cancer; Image Processing

Domains: Automatic; Electromagnetism; Signal and Image processing; Electronics

Orlando Chuquimia, Thomas Garbay, Weiqin Xu, Andrea Pinna, Xavier Dray, et al.. Study to integrate CNN inside a WCE to realize a screening tool. Journées d'Etude sur la TéléSanté, Sorbonne Universités, May 2019, Paris, France. ⟨hal-02161071⟩