Related Experiment Video
Updated: Jun 18, 2026
05:49
An Automated Squint Method for Time-syncing Behavior and Brain Dynamics in Mouse Pain Studies
Published on: November 1, 2024
Machine Diagnostics and Machine Phenotyping of Migraine: A HUNT Study
Antonios Danelakis1,2, Håkon Kvisle Abildsnes1,3, Fahim Faisal1,3
1NorHead Norwegian Centre for Headache Research, NTNU Norwegian University of Science and Technology, Trondheim, Norway.
Neurology
|June 16, 2026
Summary
Machine learning accurately diagnosed migraine using multimodal data, identifying distinct patient subgroups with unique genetic profiles. This approach reveals biological markers for migraine, potentially improving future diagnosis and treatment strategies.
Area of Science:
- Neuroscience
- Genetics
- Computational Biology
Background:
- Migraine's biological basis is poorly understood due to a lack of biomarkers.
- Machine learning (ML) offers a promising approach to analyze multimodal data for migraine characterization.
Purpose of the Study:
- To develop ML diagnostic models for migraine using multimodal data.
- To identify data-driven migraine phenotypes and subgroups.
Main Methods:
- Cross-sectional analysis of the Trøndelag Health Study data (1995-2008).
- Developed predictive ML models using genotype and clinical data (excluding headache) for migraine diagnosis.
- Employed unsupervised ML on headache data and predictive features to identify subgroups.
- Compared subgroups using genome-wide association studies, polygenic risk scores (PRS), and ML-based genetic risk scores.
Main Results:
- A gradient boosting machine achieved an Area Under the Receiver Operating Characteristic curve (AUC) of 0.80 for migraine diagnosis.
- Identified two main clusters: one predominantly migraine (94%) and another with non-migraine headaches (71%).
- Subclustered the migraine group into four distinct phenotypes based on demographics, pain characteristics, and comorbidities.
- ML-based genetic risk scores better discriminated subgroups than conventional PRSs.
Conclusions:
- Migraine is biologically definable using combinations of clinical, genetic, and environmental data.
- Data-driven phenotyping reveals migraine subgroups with distinct genetic and phenotypic signals.
- These findings may lead to improved migraine classification and personalized management strategies.
