Résumé : Rapid and important cerebral developmental changes occur between the third trimester of gestation and the first postnatal months (Sidman and Rakic, 1982). Assessment of these changes in term and preterm infants is of great interest, as it provides insights into early brain development but also how early birth may affect normal brain development (Mewes et al., 2006).

Conventional brain magnetic resonance imaging (MRI) is a useful technique to provide structural information on brain development, and several studies have correlated brain structure modifications with specific learning or behavioral problems (Peterson et al., 2003, Woodward et al., 2005, Kapellou et al., 2006, Woodward et al., 2006). Nevertheless, this technique is not sensitive enough to evidence subtle microstructural changes.

Diffusion tensor imaging (DTI), which assesses and quantifies water diffusion in biological tissues at a microstructural level, may provide unique clues to the structure and geometric organization of the cerebral tissues (Le Bihan et al., 2001). DTI takes advantage of the fact that, in the brain, water molecules diffuse more easily in the direction of the fibers than orthogonally to study cortex and white matter (WM). DTI indices like fractional anisotropy (FA), which expresses the fraction of the magnitude of the diffusion tensor attributable to anisotropic diffusion, mean diffusivity (MD), which corresponds to the directionally averaged magnitude of water diffusion, and longitudinal and transverse diffusivity (λ// and λ⊥), which express respectively the parallel and perpendicular diffusion of water molecules, are used to indirectly quantify brain microstructure and evaluate brain damage (Hüppi et al., 1998, Miller et al., 2002, Ment et al., 2009, Liu et al., 2012).

Most previously published studies in neonates limited their analysis to particular zones of the WM, using regions of interest (ROI) to select regions where DTI values are expected to change. Approaches on the basis of ROIs have well-known limitations because strong a priori hypotheses about localization and extent of the effects of interest have to be made (Giuliani et al., 2005). Voxel-based methods of neuroimaging data analysis, such as statistical parametric mapping (SPM), do not have such limitations and have been successfully applied to study age-related DTI changes in adults, DTI differences between preterm and infants at term equivalent age, and brain structural asymmetries in infants (Ashtari et al., 2007, Snook et al., 2007, Gimenez et al., 2008, Dubois et al., 2010).

Studies correlating DTI indices at term equivalent age with later neurodevelopment are scarce and their analysis is limited to the WM, without exploring the cortex (Arzoumanian et al., 2003). Moreover, they use neuropsychological testing where language evaluation is combined with cognitive and motor scales to give an overall cognitive score (Krishnan et al., 2007, Rose et al., 2007, Rose et al., 2009).

The aims of this work were, using a voxel-based analysis of DTI sequences, 1) to evidence new brain regions that experience microstructural modifications along post-menstrual age (PMA) during early development of the human brain, and 2) to correlate regional brain microstructure at term equivalent age with subsequent cognitive, motor and language development at two years corrected age in a population of preterm infants.

We first investigated DTI changes in a population of 22 healthy preterm and 6 term infants covering the life period between 34 and 41 weeks PMA, and found that, besides the already-evidenced FA increase in the corticospinal tract (CST) and callosal radiations, the thalami and the thalamic radiations experienced linear microstructural changes. These changes were interpreted as a marker of regression of cytoplasmic arborization and proliferation of immature oligodendrocytes that wrapped around the axons well before the appearance of myelin (Aeby et al., 2009). Then we looked for nonlinear DTI changes, considering that many of biological processes that occur during development follow a nonlinear course. This yielded negative results, probably due to the small sample size. Therefore, in a second study, we searched for regional linear and nonlinear microstructural changes with PMA throughout the brain in a larger population (65 patients) composed exclusively of preterm neonates scanned between 35 and 43 weeks PMA. This study confirmed the linear FA changes with age previously described and, more importantly, evidenced nonlinear changes in brain structures around the right posterior superior temporal sulcus (STS) and in the right lateral occipitotemporal gyrus (LOTG), with FA decrease between 34 and 39 weeks followed by FA increase from 40 weeks to 43 weeks. The right STS belongs to the speech-processing network and is implicated in prosody but also in inter-individual communicative behavior and face processings in close association with the right LOTG. We suggest that the microstructural modifications in brain structures around the right STS and in the LOTG observed between 35 and 43 weeks of gestation in preterm infants could contribute to the functional maturation of these brain regions with increasing age, in a period of life where voices, prosody and faces represent extremely salient stimuli (Aeby et al., 2012).

In the second part of the thesis, we tested the hypothesis that abnormal local brain microstructure of preterm infants at term equivalent age would affect neurodevelopmental abilities at age 2 years. Therefore, we searched throughout the whole brain to correlate changes of the Bayley-III scores (cognitive, motor and language composite scores) with the regional distribution of MD, FA, λ// and λ⊥. We found that language abilities are negatively correlated to MD, λ// and λ⊥ in the left superior temporal gyrus (STG) in preterm infants. These findings suggest that higher MD, λ// and λ⊥ values at term- equivalent age in the left STG are associated with poorer language scores in later childhood. Consequently, this highlights the key role of the left STG for the development of language abilities in children and suggests that brain DTI might be an interesting tool to assess on an individual basis the development of language in the preterm.

To sum up, in this thesis, we showed that, besides the already-evidenced FA increase in the CST and corpus callosum, the thalami and the thalamic radiations experience linear microstructural changes in the early development of the human brain. We further showed that FA changes nonlinearly with age in brain structures around the right STS and in the right LOTG, which are key regions in verbal and non-verbal communicative behavior. We also showed that voxel-based DTI analysis is able to evidence microstructural changes in the lSTG that are negatively correlated with language development at two years in the preterm at the group level. These results highlight the key role of the lSTG in the development of language in the preterm and suggest that brain DTI might be an interesting tool to predict the development of language on an individual basis.