The basal ganglia (BG) are organized in specific anatomo-functional territories including sensorimotor, associative and limbic territories. BG neurons display highly specific movement-related activity modulations [1]. In PD, BG neurons lose their selective movement-related activity profiles and display activity modulations related to a larger range of behavioral events [1]. This loss of specificity may result in reduced specificity and larger overlap of BG anatomo-functional territories. Here, we tested the hypothesis that loss of specificity of the BG in PD would result in increased overlap between anatomo-functional territories. To this aim, we used diffusion tensor imaging and resting state functional MRI (rs-fMRI) to study anatomical and functional connectivity profiles in PD as compared to healthy volunteers.The study included 56 patients with idiopatic PD and 29 healthy volunteers (HV) matched for age. Each PD patient had two MRI acquisitions, one without (PD-Off) and another with L-dopa (PD-On) medication. Magnetic resonance images were acquired using a 3.0T scanner (Siemens Trio) including: T1 weighted MRI and resting state BOLD fMRI (rs-fMRI) with T2*-weighted echo planar images (EPI's) (repetition time: 2.95ms; echo time: 25ms; 48 slices, field of view: 224*224mm², flip angle: 90°; voxel size: 2mm isotropic; acquisition time: 10mn). The method pipeline is illustrated in the flowchart in figure 1. Rs-fMRI were preprocessed using SPM8 [soft1] (slice timing, motion correction, spatial smoothing, low-pass filter at 0.1Hz). Then EPI's were coregistered into the T1 space of each subject. Cortical territories were extracted from T1 MRI using Freesurfer [soft2] based on the Desikan atlas [2] generating 114 ROI: 20 for sensorimotor (SM), 58 for associative (As) and 36 for limbic (Li) cortical areas (see Figure 2). We studied: pallidum, caudate and putamen, segmented with the histological deformable YeB atlas [3]. We computed the functional connectivity profile $$$\mathcal{C}_v$$$ of each voxel $$$v$$$ of BG. After averaging the BOLD signal over all voxels in each cortical region, the correlation $$$r$$$ between BG voxel v and each cortical territory (As,SM,Li) is computed resulting a vector $$$\mathcal{C}_v=[r(v,SM),r(v,As),r(v,Li)]$$$. We postulate that BG can be divided into these seven following territories : the specific ones SM, As, Li and the overlapping ones SM-As, SM-Li, As-Li, SM-As-Li. Thus, we performed a functional profile-based hierarchical clustering into seven clusters of the voxels of each BG. Then for each cluster$$$l$$$, we computed its functional profile $$$\mathcal{C}_l$$$ as the normalized mean of the functional profiles of voxels composing this cluster. Finally, to determine the functional territory of the cluster, we computed the binarized functional profiles $$$\mathring{\mathcal{C}}_l$$$ such as $$\mathring{\mathcal{C}}_{l}(k) = \begin{cases} 1 \: \: \: if \: \: \: {\mathcal{C}}_{l}(k)> t \: * \: \smash{\displaystyle\max_{1 \leq k \leq 3}} {\mathcal{C}}_{l}(k) \\ 0 \: \: \: otherwise \end{cases}$$. The computation of the proportion of each territory implies that the threshold $$$t$$$ has to be fixed resulting a direct consequence on comparative results. For this reason, we analyzed the evolution of the proportion of each territory as a function of $$$t$$$ sampling between 0% to 100%. Thus, for each subject, we computed $$$Ps(t)$$$ and $$$Po(t)$$$ respectively equal to the proportion of specific and overlapping territories in BG as a function of $$$t$$$. Then, we computed the intersection point $$$ti$$$ such as $$$Ps(ti) = Po(ti)$$$. Finally, we performed $$$ti$$$-based population statistics between the three groups using ANOVA. Then for p-values inferior to 0.05, post-hoc comparison using two sample t-test was used between each pair of groups.We studied the specificity of all territories as well as the specificity of SM territories ($$$Ps(t)$$$ equal to the proportion of SM territory and $$$Po(t)$$$ to the proportion of overlapping territories where SM is involved), as well as As and Li territories. Results presented in Figure 3 show that the intersection point was significantly lower in HV than in PD-Off in the SM territories (P= 0.0357). Considering the intersection values, we fixed for each BG the threshold of 95% and compared the proportion of territories between groups. Results are presented in Figure 4. In the right pallidum, the proportion of the SM-As-Li territory was significatively lower in the PD-Off than PD-On subjects whereas the spatial proportion of the SM-As territory was greater in the PD-Off than both PD-On and HV. The spatial proportion of the SM territory in HV was greater than in both the PD-On and PD-Off groups in the left putamen.Increased proportion of the SM-As-Li overlap may indicate reduced specificity of neuronal representations in the BG. SM (specific or overlapping) territories were involved in all between-group differences. This result was in accordance with the known predominance of PD-pathological changes in SM regions of the BG [4].We proposed an innovative method allowing to characterize the functional specificity of territories without hypothesis on the functional correlation threshold. Results are in agreement with the loss of specificity of functional territories in PD and showed predominant involvement of the SM territory. This method can be applied to all modalities which give access to cortico-BG connectivity. Thus the same technique - applied on diffusion MRI in order to get structural territories - would allow investigating the link between structure and function in the BG.