Unveiling state-specific neural dynamics in migraine with and without depressive symptom: a hidden Markov model and interpretable machine learning approach.

Researchers have discovered that migraine patients with depression experience different brain activity patterns than those without depression, findings that could help acupuncturists better understand and treat these distinct patient groups. This study examined brain scans from 204 migraine patients—100 with depressive symptoms and 104 without—plus 90 healthy individuals. Using advanced computer analysis, scientists identified six different "brain states" and found significant differences between the two migraine groups. Patients with both migraine and depression showed more unstable brain patterns, with frequent switching between different states and heightened activity in areas controlling movement, thinking, and emotions. Their brains appeared to activate both pain-processing and mood-regulating networks simultaneously. In contrast, migraine patients without depression showed less flexible brain patterns, spending prolonged time in one particular state with persistent activity mainly in pain-related regions. The computer model could distinguish between these two groups with 86% accuracy based on how much time patients spent in certain brain states. These findings suggest that migraine with depression represents a fundamentally different condition requiring distinct treatment approaches. For acupuncture practitioners, this research supports the traditional Chinese medicine principle of treating the whole person rather than isolated symptoms. The brain instability seen in migraine-depression patients may respond to acupuncture's regulatory effects on both pain and emotional pathways, while the rigidity seen in migraine-only patients might benefit from treatments targeting flexibility and circulation. When considering acupuncture for migraine, it's essential to find a licensed practitioner who conducts thorough assessments including mood evaluation.

This cross-sectional neuroimaging study utilized Hidden Markov Model analysis of resting-state fMRI data from 204 migraine patients (100 with depressive symptoms [dMIG], 104 without [ndMIG]) and 90 healthy controls to identify dynamic brain network biomarkers. Six distinct brain states were identified. The dMIG group demonstrated higher transition probabilities between states 4, 5, and 6, with enhanced activity in sensorimotor, dorsolateral prefrontal, and temporal regions, suggesting brain state instability and co-activation of pain and mood networks. The ndMIG group showed prolonged dwell time in state 3, reduced global transition rates, and heightened sensorimotor activity, indicating functional inflexibility. Machine learning classification using XGBoost achieved test set AUC=0.86 and accuracy=75.81%. SHAP analysis identified fractional occupancy of states 3, 5, and 2 as top discriminative features. Clinical implications suggest dMIG and ndMIG represent neurobiologically distinct conditions requiring differentiated therapeutic approaches targeting either network flexibility or stability respectively.