Interested in licensing this patent?
MTEC can help explore whether this patent might be available for licensing for your application.
Publication Number
US-9173610-B2
Publication Date
2015-11-03
Expiration Date
Abstract
Automated seizure detection from within an electroencephalogram (EEG) by instrumental means employs novel algorithms within software, using specific measurements of individual waves in trains, rather than any “bulk” process. The acquired signal is filtered, the wave shapes are individually described within a number of parallel runs using a variety of parameters, then criteria (including regularity criteria) are calculated and applied in order to create raw detection results. Finally the raw results are “integrated” for display. As a result, reported seizures closely follow the incidence and duration of seizures detected by trained clinicians. The invention is useful in intensive-care monitoring of EEGs from neonates and in EEG monitoring in general.
Core Innovation
The invention relates to a computer-implemented method for detecting seizures in an electroencephalogram (EEG) signal obtained from electrodes applied to an animal’s cranium. The method generates an EEG signal and obtains an inverted version of the EEG signal. Representative numerical data are obtained by comparing the EEG signal to an averaged EEG signal to locate intersection positions, and the representative numerical data from both the EEG signal and inverted version are combined.
The method characterises the waveform shape of the EEG signal at a plurality of times within the EEG signal. It carries out at least one calculation on at least some of the representative numerical data to produce related numerical data indicative of at least two individual oscillations identified within the EEG signal, and the related numerical data are compared to predetermined criteria.
A seizure event is determined to take place at particular time or times in the EEG signal when the related numerical data indicative of at least two individual oscillations at the particular time or times are within the predetermined criteria. The disclosed approach includes gating logic that rejects seizure determinations based on waveform-shape regularity and noise or artefact conditions, and an example implementation is described in the Brain Rescue Monitor (BRM).
Claims Coverage
Independent claims covered: clm-00001, clm-00026, and clm-00027. Across these claims, the core framework is combined dual-polarity representative numerical data extraction using intersection positions against an averaged EEG signal, followed by calculation to produce oscillation-indicative numerical data, comparison to predetermined criteria, and seizure event determination at particular time or times.
Dual-polarity representative numerical data from averaged EEG intersection positions
Obtain representative numerical data by comparing an EEG signal to an averaged EEG signal to locate intersection positions, obtain representative numerical data on an inverted version of the EEG signal, and combine representative numerical data from both the EEG signal and inverted version.
Waveform-shape characterisation at plurality of times
Characterise the waveform shape of the EEG signal at a plurality of times within the EEG signal.
Oscillation-indicative numerical data from representative numerical data
Carry out at least one calculation on at least some of the representative numerical data to produce related numerical data indicative of at least two individual oscillations identified within the EEG signal.
Predetermined criteria comparison and seizure event determination at particular time(s)
Compare the related numerical data indicative of at least two individual oscillations identified to predetermined criteria and determine that a seizure event has taken place at a particular time or times in the EEG signal if the related numerical data at said particular time or times are within the predetermined criteria.
Noise and waveform-shape gating
Reject seizure determinations based on waveform-shape regularity and noise or artefact conditions, including periods in which noise exceeds a predetermined threshold and waveform-shape variability and a central dip in amplitude between adjacent peaks.
The independent claims broadly cover a computer-implemented seizure detection framework that derives representative waveform-shape numerical data using intersection positions between the EEG signal and an averaged EEG signal on both the original and inverted EEG signal, combines these results, computes oscillation-indicative numerical data, and determines seizure events by comparing oscillation-related numerical data to predetermined criteria at particular time(s).
Stated Advantages
Not explicitly described in patent.
Documented Applications
Example implementation in the Brain Rescue Monitor (BRM).
Clinical display and recall/highlighting of detected seizures, including real-time or substantially real-time processing.
Interested in licensing this patent?