Diagnosis of migraine via expert system
Inventors
Doidge, Mark S. • GARINGO, MARIO • SAHBA, Farhang
Assignees
Publication Number
US-10932725-B2
Publication Date
2021-03-02
Expiration Date
2038-02-26
Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Abstract
Systems and methods are provided for identifying, monitoring, and treating migraines and migraineurs. An electroencephalogram (EEG) of a patient is obtained. The EEG comprises a plurality of EEG signals. At least two features are extracted of a network feature across at least one pair of the plurality of EEG signals in the alpha frequency band, a feature derived from a signal decomposition of at least one EEG signal, and a feature representing the power spectrum density of at least one EEG signal. The patient is classified into one of a plurality of classes, each representing one of the presence of migraine symptoms, a response to migraine treatment, a type of migraineur, a current stage of a migraine, and a likelihood that the patient is a migraineur, according to the extracted at least two features.
Core Innovation
The invention provides systems and methods for identifying, monitoring, and treating migraines by analyzing electroencephalogram (EEG) data of a patient. The EEG signals are processed to extract features including network features across pairs of EEG signals in the alpha frequency band, features derived from signal decomposition of EEG signals, and features representing the power spectrum density of EEG signals. These features are used to classify the patient into various classes such as presence of migraine symptoms, response to treatment, type of migraineur, current migraine stage, and likelihood of being a migraineur.
The problem being addressed is the lack of an objective diagnostic test for migraines, which are currently diagnosed mostly by questionnaires. Although EEG has been studied for many decades, an effective objective method was not previously achieved. This invention fills the unmet need in medical diagnostics by providing a method to analyze EEG signals to objectively identify migraineurs and characterize their conditions.
Claims Coverage
The patent contains one independent method claim covering a multi-feature EEG analysis and classification method for diagnosing migraines.
Feature extraction from EEG signals
Obtaining an EEG comprising multiple signals and extracting at least two features from three categories: a network feature across EEG signal pairs, a feature from signal decomposition of EEG signals, and a feature representing the power spectrum density of EEG signals.
Network feature based on phase synchronization
Extracting network features as phase synchronization across pairs of EEG signals in a frequency band of interest by filtering signals, applying Hilbert transform to obtain analytic signals, comparing phase differences, and summing differences over time, including averaging over multiple pairs mainly in occipital and parietal regions.
Signal decomposition feature extraction
Generating coefficients from a wavelet decomposition of EEG signals, deriving scale-dependent and scale-independent features from these coefficients, primarily from electrodes in frontal and central regions.
Power spectrum density feature extraction using autoregression
Dividing EEG signals into frequency bands, calculating autoregression coefficients for selected bands (alpha, delta, gamma), and determining descriptive statistics representing signal power from these coefficients.
Classification of patient into multiple migraine-related classes
Classifying the patient into one of multiple classes representing presence of migraine symptoms, response to treatment, migraine type, migraine stage, or likelihood of migraineur status, using the extracted features.
Training classifiers with hyperplane boundaries
Extracting features from EEGs of known patients, defining hyperplanes separating classes in a multidimensional feature space to maximize margins, and classifying new patients by locating their feature vectors relative to these hyperplanes.
Monitoring treatment response via feature changes over time
Obtaining EEGs from a patient at different times and extracting differences in network features, signal decomposition features, and power spectrum density features to monitor response to migraine treatment.
The claims cover a method of obtaining EEG data, extracting multiple features capturing network interactions, signal decompositions, and power spectral density, and classifying patients into various migraine-related categories using trained classifiers, including monitoring changes over time for treatment assessment.
Stated Advantages
Provides an objective, EEG-based diagnostic test for migraines addressing the unmet need for objective migraine diagnostics.
Allows classification into multiple clinically relevant classes such as presence of symptoms, migraine types, stages, treatment response, and likelihood of being a migraineur.
Enables monitoring of patient response to treatment through comparative analysis of EEG features over time.
Reduces feature complexity via composite features and selective electrode pairing, improving classifier efficiency and reducing overfitting risk.
Documented Applications
Diagnosing migraine presence and classification into subtypes and stages.
Monitoring patient response to migraine treatment through changes in EEG features over time.
Providing treatment recommendations such as biofeedback, relaxation training, psychotherapy, cognitive behavioral therapy, mindfulness training, sleep treatments, and pharmaceutical interventions based on diagnostic outputs.
Interested in licensing this patent?