Dual-attention temporal graph neural network on resting-state fMRI dynamic functional connectivity identifies risk-related patterns of migraine chronification.

Migraine chronification is challenging to characterize on an individual basis. There are no robust imaging markers for the conversion from episodic migraine to chronic migraine. This study involved 195 participants, including 95 healthy controls and 100 patients with episodic migraine. All migraine patients underwent resting-state functional MRI during the interictal phase. After 12 months, 70 patients remained episodic migraine and 30 converted to chronic migraine. Researchers developed an interpretable temporal graph neural network to analyze brain connectivity. They identified four recurrent connectivity states, with State 3 showing significant group differences. Migraine groups displayed altered connectivity in sensory, attention, and subcortical systems. Chronic migraine converters had additional abnormalities in sensorimotor and limbic-related circuits. The model achieved good performance, with an overall accuracy of 84%. Findings highlight the need for external validation before clinical application. The study offers insights into patterns related to migraine chronification. Clinical relevance: This research may help identify patterns indicating a patient’s risk of migraine chronification.