The Human Globus Pallidus: Parcellation, Connectivity Pattern, and Behavioral Function
Poster No:
2162
Submission Type:
Abstract Submission
Authors:
Kexin Wang1, Tiantian Liu1, Dingjie Suo1, Chuyang Ye1, Tianyi Yan1
Institutions:
1Beijing Institute of Technology, Beijing, China
First Author:
Co-Author(s):
Introduction:
The human globus pallidus (GP) is regarded as an important role in the motor circuits [1], and an effective target in deep brain stimulation (DBS) for patients with motor conditions, such as Parkinson's disease (PD) [2]. However, as the motor symptoms reduce, cognitive function in PD patients can deteriorate mildly [3] while stimulating the GP internus in DBS treatment. This suggests the need to obtain the function-separated parcellation of GP to separate the cognitive and motor circuits in GP for precision treatment. This study used diffusion MRI and functional MRI to obtain the parcellation of GP, explore the structural and functional specificity of GP subregions, and describe the relationship between GP subregions and various behavioral measures. This manuscript provides a new perspective on the organization and function of GP.
Methods:
7T resting state fMRI data and 3T diffusion data of 170 participants (age: 22-36, 68 male) were used from the Human Connectome Project, which was randomly split into dataset 1 (n = 80) for parcellation and dataset 2 (n = 90) for validation. Besides the standard pipeline in the HCP preprocessing, diffusion images were further processed with nonlinear co-registered and probabilistic tractography to obtain structural similarity. Functional images were processed with the first 15 frames removed, 4-mm FWHM smoothed [4], band-pass filtered (0.01~0.08 Hz) [5] and η2-coefficient calculated to obtain the functional similarity.
Shared-nearest-neighbor-based density peak clustering (SNN-based DPC) algorithm [6,7] was used to avoid round parcels, spatially disconnected and unstable one-step allocation strategy. There were three kinds of feature input for the SNN-based DPC algorithm: the similarity of diffusion images, the similarity of functional images, and the concatenated similarity of the first two kinds. Then, the connectivity patterns were defined to compare the specificity of GP subregions by diffusion MRI and functional MRI [8]. Finally, kernel ridge regression (KRR) [9] was used to predict each behavioral phenotype in individuals by functional connectivity to evaluate the function of GP subregions [10].
Shared-nearest-neighbor-based density peak clustering (SNN-based DPC) algorithm [6,7] was used to avoid round parcels, spatially disconnected and unstable one-step allocation strategy. There were three kinds of feature input for the SNN-based DPC algorithm: the similarity of diffusion images, the similarity of functional images, and the concatenated similarity of the first two kinds. Then, the connectivity patterns were defined to compare the specificity of GP subregions by diffusion MRI and functional MRI [8]. Finally, kernel ridge regression (KRR) [9] was used to predict each behavioral phenotype in individuals by functional connectivity to evaluate the function of GP subregions [10].
Results:
First, the three parcellations of GP by diffusion images, functional images, and the concatenated similarity matrices were compared by the dice coefficient of reliability (diffusion-based: 0.91, function-based: 0.93, fusion-based: 0.95). Then, the parcellation based on the hybrid similarity was chosen, which divided the GP into 3 subregions (see Fig. 1A) in each hemisphere. It was significantly more homogeneous than random parcellations (homogeneity: 24.44%, p<0.01).
Second, connectivity profiles and fingerprints were analyzed in structure and function (see Fig. 1B, C, D, E and Fig. 2A, B, C). Results showed that aGP (anterior) had stronger connectivity to the limbic network. mGP (medial) was extensively connected to cortical networks, and more relevant to higher cognitive networks (such as DMN and FPN). pGP (posterior) was connected to the salience network and sensorimotor network.
Finally, the prediction accuracy for each behavioral phenotype is illustrated in Figure 2D. aGP showed stronger correlations with emotion-related behaviors, mGP was more relevant to cognitive tasks, while pGP was more correlated with motor and personality.
Second, connectivity profiles and fingerprints were analyzed in structure and function (see Fig. 1B, C, D, E and Fig. 2A, B, C). Results showed that aGP (anterior) had stronger connectivity to the limbic network. mGP (medial) was extensively connected to cortical networks, and more relevant to higher cognitive networks (such as DMN and FPN). pGP (posterior) was connected to the salience network and sensorimotor network.
Finally, the prediction accuracy for each behavioral phenotype is illustrated in Figure 2D. aGP showed stronger correlations with emotion-related behaviors, mGP was more relevant to cognitive tasks, while pGP was more correlated with motor and personality.
·Figure 1. Parcellation of the GP based on diffusion and functional MRI and the structural analysis of GP subregions
·Figure 2. The functional analysis of GP subregions and the prediction accuracy for each behavioral phenotype
Conclusions:
This study obtained the organization of the GP according to the structural and functional connectivity and explored the difference in structure and function between different subregions. Results showed that the GP parcellation obtained by fusing structural and functional information is more stable between participants. Furthermore, from posterior to anterior, GP showed a trend from low-level networks to high-level networks, and from motor and personality to cognition to emotion tasks. This study will help to reveal the multiple roles of GP in complex circuits and provide convincing support for DBS at GP in PD patients.
Modeling and Analysis Methods:
Connectivity (eg. functional, effective, structural)
Segmentation and Parcellation 2
Neuroanatomy, Physiology, Metabolism and Neurotransmission:
Subcortical Structures 1
Neuroinformatics and Data Sharing:
Brain Atlases
Keywords:
Basal Ganglia
FUNCTIONAL MRI
Sub-Cortical
WHITE MATTER IMAGING - DTI, HARDI, DSI, ETC
Other - Globus Pallidus
1|2Indicates the priority used for review
Provide references using author date format
[2] Ramirez-Zamora.(2018), 'Globus pallidus interna or subthalamic nucleus deep brain stimulation for Parkinson disease: a review', JAMA neurology, vol. 75, no. 3, pp. 367-372.
[3] Combs H L. (2015), 'Cognition and depression following deep brain stimulation of the subthalamic nucleus and globus pallidus pars internus in Parkinson’s disease: a meta-analysis', Neuropsychology review, vol. 25, pp. 439-454.
[4] Tian Y. (2020), 'Topographic organization of the human subcortex unveiled with functional connectivity gradients', Nature neuroscience, vol. 23, no. 11, pp. 1421-1432.
[5] Luo Z. (2020), 'Functional parcellation of human brain precuneus using density-based clustering', Cerebral cortex, vol. 30, no. 1, pp. 269-282.
[6] Liu R. (2018), 'Shared-nearest-neighbor-based clustering by fast search and find of density peaks', Information sciences, vol. 450, pp. 200-226.
[7] Peng L. (2023), 'Parcellating the human brain using resting-state dynamic functional connectivity', Cerebral cortex, vol. 33, no. 7, pp. 3575-3590.
[8] Yeo B T T. (2011), 'The organization of the human cerebral cortex estimated by intrinsic functional connectivity', Journal of neurophysiology, vol. 106, no. 3, pp. 1125-1165.
[9] Kong R. (2019), 'Spatial topography of individual-specific cortical networks predicts human cognition, personality, and emotion', Cerebral cortex, vol. 29, no. 6, pp. 2533-2551.
[10] Li K. (2022), 'The human mediodorsal thalamus: Organization, connectivity, and function', NeuroImage, vol. 249, 118876.