Identifying Migraine: Building Classifiers Using Resting-State fMRI Data

OBJECTIVE: To classify migraine by applying a novel machine-learning approach to resting-state functional connectivity magnetic resonance neuroimaging (rs-fMRI) data.BACKGROUND: Migraine is a prevalent neurological disorder that is associated with structural and functional brain alterations in regions associated with the perception and modulation of pain. This study explored the use of machine-learning techniques to develop discriminative brain-connectivity biomarkers from rs-fMRI data that distinguish between individual migraine patients and healthy controls.DESIGN/METHODS: This study included 58 interictal migraine patients and 50 age-matched healthy controls (migraine patients: mean age= 36.1; SD=11.2; healthy controls: mean age=36.4; SD=11.0; p=.85). Ten minutes of resting state data were collected from each patient using a 3T scanner. Whole-brain functional connectivity to pain-related regions was interrogated. A 10-fold cross-validation classification algorithm was used to construct classifiers that determine if an individual brain MRI belongs to someone with migraine or to a healthy control.RESULTS: Overall classification accuracy was 77[percnt] (best accuracy of 86[percnt]). Amongst multiple regions that contributed to the classification, connectivity of the right middle temporal area with large clusters located in the contralateral parietal lobe had the highest discriminative power. Additionally, a post-hoc analysis showed that migraine patients who had longer disease durations were more accurately classified than migraine patients with shorter disease durations.CONCLUSIONS: Alterations in the resting functional connectivity of brain regions in people with migraine allow for discrimination of the migraine brain from that of a healthy control. Migraine patients with longer disease duration were classified more accurately than migraine patients with shorter disease burden, potentially indicating that migraine leads to reorganization of brain circuitry. Classification of migraine using rs-fMRI provides insights into pain-circuits that are altered in migraine and could potentially contribute to the development of a new non-invasive migraine biomarker. Disclosure: Dr. Chong has nothing to disclose. Dr. Gaw has nothing to disclose. Dr. Fu has nothing to disclose. Dr. Li has nothing to disclose. Dr. Wu has nothing to disclose. Dr. Schwedt has received personal compensation in an editorial capacity for Headache and Pain Medicine.