Identification of Migraine Subtypes Using Functional Near-Infrared Spectroscopy Data: A Domain-Based Feature...

Area of Science:

• Neuroscience
• Medical Technology
• Biomedical Engineering

Background:

• Current migraine diagnosis relies on subjective patient reports and International Headache Society guidelines, often leading to misdiagnoses.
• Objective and reliable diagnostic tools are crucial for accurate clinical practice in migraine management.

Purpose of the Study:

• To develop and validate a framework using functional near-infrared spectroscopy (fNIRS) for objective migraine diagnosis.
• To differentiate between healthy individuals and migraine patients (with and without aura) using prefrontal cortex (PFC) activity.

Main Methods:

• Utilized fNIRS to measure prefrontal cortex (PFC) activity, analyzing hemoglobin dynamics (oxyhemoglobin, deoxyhemoglobin, total hemoglobin).
• Extracted features in time, frequency, and time-frequency domains.
• Applied XGBoost machine learning algorithm to classify participants based on extracted fNIRS features.

Main Results:

• The XGBoost model, using time-frequency features of oxyhemoglobin in the left PFC, achieved 92% balanced accuracy.
• Demonstrated high performance metrics: 89% sensitivity, 95% specificity, and 89% F1 score.
• Successfully distinguished healthy individuals from interictal migraine patients (with and without aura).

Conclusions:

• Non-invasive fNIRS coupled with machine learning presents a cost-effective and promising alternative to traditional migraine diagnostic methods.
• This approach can enhance early and accurate migraine diagnosis, facilitating better-targeted treatments and improved patient outcomes.
• The study lays a strong foundation for future research and clinical applications in objective migraine diagnosis.