Skip to main content

Construction of a dense 3d probabilistic anatomical atlas for segmentation registration and labeling of multimodal images of the brain


Research objectives and content
The need for medium or high-level 3D models for interpreting and comparing data from different medical imaging modalities, such as magnetic resonance (MR), computed tomography (CT) and positron/single photon emission computed-/tomography (PET, SPECT), has lead to the recent development of digital anatomical atlases. The goal of this project is to create a probabilistic anatomical atlas of the brain, coding information on anatomical structures and on biological variability of anatomy among a given population. 3D statistical deformable models will be considered to this end, for a dense 3D representation of the brain structures. As a first clinical application, the atlas will be used to the segmentation, registration, labeling and interpretation of epileptic areas in 3D multimodal image sequences of the brain. The clinical interest is to improve the analysis of epileptic seizures, providing a reliable pre-surgical diagnostic tool. The software for this approach will be implemented in C and it will be integrated within a multi-platform application for image processing (MEDIMAX), available for the scientific community.
Training content (objective, benefit and expected impact)
This project will improve the candidate's knowledge in the discipline of medical image processing, in 3D multimodal image registration (MR/SPECT). It will compliment the candidate's experience, whose speciality is segmentation of CT images for RTP (Radiotherapy Treatment Planning). Her expertise will benefit to the project solving the segmentation problems related to the construction and matching of the atlas. Moreover, the applicant would like to see the techniques applied to MR/CT image registration at Walsgrave Hospital (Coventry), where the existing work on RTP has been carried out. In general, besides the professional benefit, experiencing a different culture is also a personal enrichment. Links with industry / industrial relevance (22)
Both associated institutions have several research programmes, related to image analysis, include:
- Contractual links of LSIIT with several European manufactures (via Esprit and Eureka programmes),carrying out researches in computer vision, remote sensing and automation. - Industrial collaboration of IPB with Brucker Spectrospin Company (Wissembourg), for the development of Nuclear MR applications for brain imaging.

Funding Scheme

RGI - Research grants (individual fellowships)


Boulevard Sébastien Brant - Pole Api
67400 Illkirch