1. Motivations

The explosive growth in brain imaging technologies (such as Magnetic Resonance Imaging, or MRI) has only been matched by a tremendous increase in the number of investigations focusing on the structural and functional organization of the human brain. However, the formidable complexity and variability of cerebral structures present a substantial challenge to their effective analysis and modelisation, and require the conjunction of a number of techniques drawn from a large panel of scientific domains: computer vision, image analysis or statitical modeling to name just a few. A natural way to condense, analyse and communicate brain data, digital brain atlases consequently stem from the combination of many algorithms whose underlying hypotheses must be validated a posteriori.

The INRIA EPIDAURE team focuses on the design and development of tools adapted to the analysis of multidimensional and multimodal medical images, with the aim to enhance the available diagnosis. A non-exhaustive selection of the main research axes would encompass the extraction of quantitative parameters for automated/aided diagnosis (shape, texture, displacement), the spatial and temporal registration of 2-D and 3-D medical images, multimodal image fusion problems, the analysis of deformable displacements, or the design and manipulation of digital atlases with, in the longer run, the goal to automate the morphometric and functional analysis of the brain.

A member of the UCLA School of Medicine, the LONI laboratory is dedicated to the development of scientific approaches for the comprehensive mapping of brain structure and function. Use of fluid warping approaches has enabled the development of disease-specific digital brain atlases (for Alzheimer and schizophrenia, among others). Based on large human populations, and containing thousands of 3D structure models, these atlases encode patterns of anatomical variation and can be used to detect group-specific patterns of anatomic or functional alterations.

The association of both teams within this program enables a better factorisation of the experience and knowledge they have both acquired, and permit to jointly tackle the restauration, segmentation or image registration issues that the development of digital brain atlases invariably raises. Consequently, the EPIDAURE and LONI teams share both the algorithms and the databases required to build and validate these atlases. The developped tools are carefully studied, and their performances assessed with the hope to design and develop ever more complete, precise and robust atlases. Part of these tasks has been implemented during the PhD thesis of Alain Pitiot, co-tutored by N. Ayache (EPIDAURE) and P. Thompson (UCLA). The collaboration is now being continued with the PhD theses of V. Arsigny and P. Fillard.