Unsupervised Fiber Bundles Registration Using Weighted Measures Geometric Demons

Abstract

Abstract. Brain image registration aims at reducing anatomical variability across subjects to create a common space for group analysis. Multi-modal approaches intend to minimize cortex shape variations along with internal structures, such as fiber bundles. A difficulty is that it requires a prior identification of these structures, which remains a challenging task in the absence of a complete reference atlas. We propose an extension of the log-Geometric Demons for jointly registering images and fiber bundles without the need of point or fiber correspondences. By representing fiber bundles as Weighted Measures we can register subjects with different numbers of fiber bundles. The efficacy of our algorithm is demonstrated by registering simultaneously T1 images and between 37 and 88 fiber bundles depending on each of the ten subject used. We compare results with a multi-modal T1 + Fractional Anisotropy (FA) and a tensor-based registration algorithms and obtain superior performance with our approach.