Supervision : Bertrand GRANADO, Xavier DRAY
Co-supervision : PINNA Andrea
CRC is the second highest cause of death by cancer worldwide with 880,792 deaths in 2018 and a mortality rate of 47.6%. 95% of CRC cases begin with the presence of a growth on the inner lining of the colon or the rectum, called a polyp. Multiple types of polyps exist; among them, adenoma polyps, which can degenerate into CRC. CRC is treatable in 90% of the cases if it is detected early enough. This is a real public health problem where it is necessary to implement early detection policies to reduce the mortality rate of this cancer.
The endoscopic capsule was invented by Paul Swain in 1990. It is a pill incorpo- rating a camera and a radio communication system that the patient swallows and transmit images from the gastrointestinal tract through the body in a workstation. Once all images are transmitted, a astroenterologist downloads them to perform a visual analysis and detect abnormalities and tumors. Using this device doctors can detect polyps, at least 5 mm, with sensitivity and specificity respectively of 68.8% and 81.3%. This endoscopic capsule presents some limitations and weaknesses re- lated to the spatial and temporal resolution of images, its energy autonomy and the number of images transmitted to be analyzed by the gastroenterologist.
We studied the design of an embedded system containing a processing chain capable of detecting polyps to be integrated into an endoscopic capsule, creating a new medical device: an intelligent endoscopic capsule. To realize this device, we took into account all the non-functional constraints related to the integration into an endoscopic capsule. This device must be a new tool for early detection of precancerous colorectal lesions: polyps. The first part of this thesis is dedicated to the development of a processing chain for the automatic detection of polyps that can be integrated into a capsule. The second part of this work is dedicated to the design of an embedded system to be integrated into the capsule and containing the most critical part of our chain: the Hough transform. The results obtained in this thesis are at the state of the art and contribute to the development of the new generation of the endoscopic capsule. They demonstrate the computational capacity that can be effectively integrated into a highly constrained environment (time, surface and energy).
Keywords: Automatic polyp detection, intelligent endoscopic capsule, Hoguh transform, fuzzy trees, fuzzy forest, artificial vision and hardware accelerator.
Defence : 11/22/2019 - 10h - Campus Jussieu 55-65/211
FRESSE Virginie (Université de Saint-Etienne) [Rapporteur]
BOUTILLON Emmanuel (Université de Bretagne Sud) [Rapporteur]
ROMAIN Olivier (Université de Cergy-Pontoise)
BOL David (Université Catholique de Louvain)
VITRANI Marie Aude (Sorbonne Université)
GRANADO Bertrand (Sorbonne Université)
DRAY Xavier (Sorbonne Université)
PINNA Andrea (Sorbonne Université)