Dysregulated Cerebellar–Subcortical–Cortical Interactions in Cluster Headache: Evidence From Edge Analysis Across Three Independent Datasets
Wei Dai, Mingjie Zhang, Shuhua Zhang, Hui Su, Yun Chen, Yuhan Wu, Yun‐Xia Wang, Yueqi Guo, Xiangbing Bian, Xiao Zang, Xinbo Xing, Chenghui Pi, Xiting Nie, Zhe Yu, Guoen Yao, Catherine D Chong, Anders Hougaard, Lixia Tian, Xin Lou, Shengyuan Yu, Zhao DongObjective
The objective of this study was to apply edge‐centric functional connectivity and motif analyses to capture co‐fluctuation communication patterns and test whether altered dynamic network integration underlies clinical manifestations of cluster headache (CH).
Methods
This cross‐sectional study included 100 participants with episodic CH imaged during the in‐bout period outside acute attacks and 116 healthy controls (HCs) using functional magnetic resonance imaging (MRI). Robustness of the findings was assessed via pooled and separate analyses of 3 independent datasets using brain parcellation, edge graph construction, k‐means clustering, community entropy calculations, motif, and clinical correlation analyses.
Results
Participants with CH showed lower entropy in the cerebellar vermis ( β = −0.493, 95% confidence interval [CI] = −0.683 to −0.302, p false discovery rate [ p FDR ] = 3.02 × 10 −6 ) and cerebellar network (mean difference = −0.043, 95% CI = −0.068 to −0.018, t = −3.361, p FDR = 0.009) compared with HCs. In individuals with CH, longer disease duration was accompanied by a greater reduction in entropy of the cerebellar vermis ( r = −0.390, 95% CI = −0.550 to −0.196, p FDR = 0.004). Motif analysis revealed that participants with CH exhibited (i) higher fractions of forked triads with vermis coordinating subcortical and cortical nodes belonging to sensorimotor, visual, default‐mode, and cerebellar networks; and (ii) lower fractions of diverse triads linking vermis to both sensorimotor and cognitive‐ control networks and to both sensorimotor and default‐mode networks (all p FDR < 0.05).
Interpretation
We identified reorganization of the cerebellar–subcortical–cortical network in CH. The cerebellar vermis exhibited a paradoxical pattern: retaining hub‐like connectivity while losing flexible network integration. This pattern worsened with disease duration, supporting its role in symptomatology and potential as a candidate region for future neuromodulation studies. ANN NEUROL 2026