Replico (IMAG2-Endométriose)

Extending automatic segmentation algorithms developed on children's pelvic area to adult females suffering from endometriosis. Anatomical 3D model combined with pelvic nerves modeling provides important information to understand the origin of the pain felt by the patients. It also improves high precision surgery planning, giving better results while avoiding damage on highly sensitive areas. Close collaboration with surgeons and radiologists since 2015, to create the most precise patient specific 3D anatomical models of the pelvic area, including the nerves going through this region. This experience gives us a serious advantage compared to competitors.

• Two MRI sequences needed as input: a cube T2-weighted sequence for anatomical segmentation and a diffusion weighted MRI sequence for nerve modeling.
• State-of-the-art deep learning models for 3D segmentation and model creation of all anatomical structures and organs (bones, bladder, artery, etc.).
• Use of tractography algorithms on Diffusion MR Imaging to extract large sets of fibers.
• Tract nerve modeling, segmentation and recognition from these sets using our symbolic IA method based on fuzzy modeling and spatial reasoning.

• Our automatic and deep-learning based segmentation methods are faster and more accurate than the ones of competitors.
• Nerves modeling and segmentation: key and original feature of our project. Nerves hold precious information for surgical planning to avoid any after-effect on patient’s motricity.