Understanding the human subcortex using ultra-high field MRI and computational cognitive models

In this thesis, novel approaches for understanding the human subcortex using Ultra-High field MRI (UHF-MRI) were explored. The approaches were also applied to empirically validate the tripartite model of the subthalamic nucleus (STN). In Chapter 2, current state-of-the-art UHF-MRI methods and their relevance for the study of the human subcortex were reviewed. In Chapter 3, the iron distribution of the STN was visualized with quantitative susceptibility mapping (QSM) in both living subjects, as well as post-mortem tissue. We showed that the spatial distribution of iron is heterogeneous throughout the STN, but that there are no hard borders or clear subdivisions in the spatial distribution of iron across the STN. In Chapter 4 we investigated post-mortem STN samples using multiple MRI sequences, as well as immunocytochemical stainings. The histochemical structure of the STN is complex, gradual, and shows no clear tripartite subdivision. In Chapter 5 we critically reviewed the methods used in the empirical neuroimaging literature on the STN. Conventional methods turn out to be inadequate for distinguishing between signals from adjacent nuclei. We promote a ROI-approach and cautioned against spatial smoothing. In Chapter 6, a set of 7 T fMRI protocols was compared, to assess which spatial resolution is feasible for functionally imaging iron-rich subcortical nuclei at UHF. We show that also at 7 T, some modesty in terms of resolution is necessary for functional imaging of subcortex and that 3 T fMRI shows inferior BOLD sensitivity. In Chapter 7 we reviewed different approaches for linking cognitive computational models to neural measurements. Such approaches might aid in the interpretation of BOLD measurements in subcortex. Finally, in Chapter 8, we tested the tripartite model of the STN with an empirical study on how BOLD activity in different parts of the STN is related to different aspects of perceptual decision-making. The STN is involved in implementing response biases, there was no evidence for functional differentiation across different parts of the STN.