Mapping individual cortico-basal ganglia-thalamo-cortical circuits integrating structural and functional connectome: implications for upper limb motor impairment poststroke

Abstract

This study investigated alterations in functional connectivity (FC) within cortico-basal ganglia-thalamo-cortical (CBTC) circuits and identified critical connections influencing poststroke motor recovery, offering insights into optimizing brain modulation strategies to address the limitations of traditional single-target stimulation. We delineated individual-specific parallel loops of CBTC through probabilistic tracking and voxel connectivity profiles-based segmentation and calculated FC values in poststroke patients and healthy controls, comparing with conventional atlas-based FC calculation. Support vector machine (SVM) analysis distinguished poststroke patients from controls. Connectome-based predictive modeling (CPM) used FC values within CBTC circuits to predict upper limb motor function. Poststroke patients exhibited decreased ipsilesional connectivity within the individual-specific CBTC circuits. SVM analysis achieved 82.8% accuracy, 76.6% sensitivity, and 89.1% specificity using individual-specific parallel loops. Additionally, CPM featuring positive connections/all connections significantly predicted Fugl-Meyer assessment of upper extremity scores. There were no significant differences in the group comparisons of conventional atlas-based FC values, and the FC values resulted in SVM accuracy of 75.0%, sensitivity of 67.2%, and specificity of 82.8%, with no significant CPM capability. Individual-specific parallel loops show superior predictive power for assessing upper limb motor function in poststroke patients. Precise mapping of the disease-related circuits is essential for understanding poststroke brain reorganization.

Keywords: cortico–basal ganglia–thalamo–cortical circuits; motor impairment; stroke.

FIGURE 1

Distribution of stroke lesions in the entire sample of patients. The color scale represents number of participants with lesioned voxel as evaluated by T1 images.

FIGURE 2

Classification performance of SVM using connections derived from the two CBTC circuits mapping methods. Accuracy as a function of the number of connections used in the classification process for the individual mapping CBTC circuits based on VCP‐based segmentation (A) and for the conventional atlas‐based CBTC circuits (B). The connections were ranked according to F scores in descending order. ROC curve of the classifier for the individual mapping CBTC circuits (C) and for the conventional atlas‐based CBTC circuits (D).

FIGURE 3

Prediction performance of CPM using connections derived from the individual‐mapping CBTC circuits based on VCP‐based segmentation. (A) Correlation between observed and predicted UE‐FMA scores in positive (red), negative (blue), and combined (green) connections. (B) The distribution of correlation coefficients by a permutation test of 5000 times. *p < 0.05.

FIGURE 4

The cortical (A) and subcortical (B) atlas‐defined regions of interest.

FIGURE 5

The architecture of the whole pipeline.