The objective of this thesis is to identify and characterize human anatomical and functional cortico-subcortical loops, using data from resting-state functional magnetic resonance imaging (fMRI) and diffusion MRI. A loop is a set of cortical, subcortical and cerebellar regions that interact to perform or prepare for a task.We first aim to identify cortico-subcortical functional networks from resting-state fMRI data. We propose a robust statistical method that separates the analysis of cortical regions from that of subcortical structures. A spatial independent component analysis is first performed on individual cortical regions, followed by a hierarchical classification. The associated subcortical regions are then extracted by using a general linear model, the regressors of which contain the dynamics of the cortical regions, followed by a random-effect group analysis. The proposed approach is assessed on two different data sets. An immunohistochemical subcortical atlas is then used to determine the sensorimotor, associative or limbic function of the resulting networks. We finally demonstrate that anatomy is a support for function in healthy subjects.The last part is devoted to the study of the Gilles de la Tourette syndrome, thought to be due to adysfunction of cortico-subcortical loops. Firstly, cortico-subcortical functional loops are characterized using metrics such as integration and graph theory measures, showing differences in terms of connectivity between adult patients and healthy volunteers. Secondly, we show that the cortico-subcortical functional loops in patients are supported by the underlying anatomy.
