Resting facial behavioural signatures associated with migraine phenotype, clinical state, and disease burden

Migraine lacks an objective, scalable signature that captures its heterogeneous phenotypes and fluctuating clinical states. We investigated whether spontaneous resting facial activity reflects migraine phenotype, headache phase, preventive treatment status, and clinical burden. Facial activity was extracted from 10-s resting videos and quantified as the magnitude and temporal variability of 31 facial action units. Healthy controls (HCs) and patients with episodic (EM) or chronic migraine (CM) were examined, with the patients stratified by preventive treatment and interictal or ictal phase. Group differences were evaluated using age- and sex-adjusted rank-based permutation analysis of covariance with false discovery rate correction. Behavioural partial least squares (PLS) analysis identified facial patterns covarying with disease burden, emotional distress, and current headache intensity. We consecutively recruited 842 participants: 90 HCs, 252 patients with CM and 500 with EM. Migraine was characterised by a multiregional pattern dominated by diminished bilateral lip-corner pulling, which persisted across phenotypes, phases, and treatment groups. Drug-naïve patients exhibited broader reductions in cheek and lower-face activity, whereas treated patients retained the core perioral alteration with greater periocular, brow, and lower-face involvement. Direct comparisons between EM and CM revealed few statistically significant differences. Increased bilateral lid-tightening magnitude and variability represented the clearest ictal feature, particularly in drug-naïve EM. PLS revealed coordinated patterns associated with disease burden and emotional distress. Greater burden covaried with reduced cheek and perioral activity and increased brow activity; emotional distress was associated with reduced lip-corner pulling, cheek raising and lip stretching, together with greater lip-corner depression. Pain intensity categories, but not current headache intensity, was associated with a predominantly periocular pattern in the overall group and drug-naïve subgroups. Emotional-distress and current-pain models were not significant in treated patients. Brief resting facial recordings revealed complementary group-level patterns associated with migraine status, current clinical state, and clinical burden. A consistent perioral signature accompanied ictal periocular activation and treatment-status differences, but these comparisons cannot establish preventive-therapy effects. Resting facial activity may represent a contactless digital phenotype, although longitudinal studies with disease controls and external validation are required to establish its diagnostic or monitoring utility. Not applicable.