Vestibular cortex-related effective connectivity signatures: characterization and differentiation of vestibular migraine patients via magnetic resonance imaging and machine learning.

BACKGROUND

Neuroimaging studies of vestibular migraine (VM) have revealed abnormal functional connectivity in the central vestibular system. However, it remains to be determined whether effective connectivity (EC) of the vestibular cortex is specifically disrupted in VM and whether this brain measure can aid in the differential diagnosis of VM patients.

METHODS

In 56 VM patients, 58 episodic migraine (EM) patients, and 63 healthy controls (HCs), the directional influences between the vestibular cortex and the whole brain were examined via resting-state functional magnetic resonance imaging and Granger causality analysis, with bilateral parietal operculum cortex 2 (OP2) as the seed regions. Statistical analyses were performed to investigate the group differences and the associations of directional influences with clinical variables. Classification models based on linear support vector machine (SVM) analysis were established to assess the performance of effective connectivity in discriminating VM patients from HCs and EM patients.

RESULTS

Relative to the EM group and HC group, the VM group presented elevated positive influence from the left OP2 to the right middle frontal gyrus (MFG), as well as increased negative influence from the right MFG and inferior parietal lobule (IPL) to the left OP2. Furthermore, the EM group exhibited increased positive influence from the bilateral medial superior frontal gyrus (SFGmed) to the left OP2 compared with the VM group and HC group. The EC-derived SVM models showed favorable and moderate performance for distinguishing VM patients from EM patients (area under the curve = 0.8264, p < 0.001) and HCs (area under the curve = 0.7764, p < 0.001).

CONCLUSIONS

The causal flow from and to the left OP2 was altered in the prefrontal and inferior parietal lobes, regions closely implicated in cognitive control and multisensory integration. Abnormal EC associated with the right MFG and IPL appears to be specific to VM, whereas altered EC linked to SFGmed suggests a resilience trait in EM. The EC of the OP2 may serve as a neuroimaging biomarker for VM identification.