Methods and apparatus for detection and imaging of epileptogenicity from scalp high-frequency oscillations
A system and method to facilitate the analysis of long-term scalp EEG recordings and the investigation of underlying epileptogenicity. Pathological high frequency oscillations are identified and extracted from the EEG recordings and electrophysiological source imaging is used to reconstruct cortical activity, which can be used as an aid in surgical resection for the treatment of epilepsy.
1. A method of identifying high-frequency oscillation events in EEG data of brain activity, the EEG data scalp-recorded from a plurality of channels, the method comprising:
eliminating noise from the EEG data by applying a filter to the EEG data;
segmenting the filtered EEG data into time periods centered around a spike to form data samples segmented by time;
detecting frequency events in the segmented samples;
selecting the frequency events that qualify as high frequency oscillations;
extracting features from the selected frequency events;
clustering the selected frequency events into a plurality of clusters;
identifying a putative spike ripple in a cluster of the plurality of clusters;
extracting a portion of data from the filtered EEG data corresponding to the cluster containing the identified putative spike ripple; and
projecting brain activity associated with the identified putative spike ripple onto a brain cortex model.
2. The method of claim 1 , wherein the noise comprises data generated by at least one of a physiological event and a non-physiological event.
3. The method of claim 2 , wherein the physiological event is selected from the group consisting of cardiac pulse, respiration, eye movement, and muscle movement.
4. The method of claim 1 , wherein applying a filter comprises:
applying a high-pass filter above 1 Hz; and
applying a notch filter at 60 Hz with harmonics using a fast Fourier transform Hann filter with a slope of 2 Hz.
5. The method of claim 1 , wherein detecting frequency events in the segmented data samples comprises:
comparing an amplitude for each segmented data sample to a respective threshold.
6. The method of claim 5 , wherein each respective threshold comprises a multiple of a standard deviation of a corresponding amplitude.
7. The method of claim 1 , wherein selecting the frequency events that qualify as high frequency oscillations comprises:
comparing the morphology of each frequency event to a known morphology to identify high frequency oscillations.
8. The method of claim 7 , further comprising:
grouping the selected frequency events from the plurality of channels temporally.
9. The method of claim 1 , wherein extracting features from the selected frequency events comprises:
extracting at least one of time-frequency domain features and spectrogram features.
10. The method of claim 9 , wherein the time-frequency domain features are selected from the group consisting of mean, variance, skewness, median frequency, and spectral centroid.
11. The method of claim 9 , wherein the spectrogram features are selected from the group consisting of entropy, third-order moment of a histogram, directionality, and block-wise power.
12. The method of claim 1 , wherein clustering the selected frequency events comprises:
estimating the center of a feature distribution; and
mapping each selected frequency event into one of the plurality of clusters using the expectation-maximization method.
13. The method of claim 12 , further comprising:
evaluating the homogeneity of the spatial and temporal distributions of the selected frequency events in each cluster of the plurality of clusters.
14. The method of claim 13 , wherein evaluating the homogeneity is based on the following metrics: a scalp spatial extent and a spike time-locking,
wherein a potential pathological cluster is selected as the cluster having a minimum mean of the two metrics.
15. The method of claim 1 , wherein the EEG data originates from an epileptic brain.
16. The method of claim 1 further comprising:
modeling the brain activity by equivalent dipoles.
17. The method of claim 1 further comprising:
modeling the brain activity using a distributed cortical current density model.
18. The method of claim 1 , further comprising:
imaging brain anatomy using a magnetic resonance imaging device; and
comparing the brain anatomy with the projected brain activity.