METHODS AND SYSTEMS FOR MULTI-CLASS CLASSIFICATION OF SEIZURE
Described herein are methods and systems for the classification of seizure in a subject. The systems may include a data module configured to obtain a plurality of electroencephalography (EEG) signals collected from a subject. The systems may also include a processing module in communication with the data module. The processing module may be configured to process the data to detect and monitor seizures or related symptoms that the subject is experienced or is experiencing. The processing module may also generate indications or assessments for seizure at an individual level.
1 . A method for classifying seizures comprising:
obtaining data comprising electroencephalography (EEG) signals recorded from a subject over a plurality of channels;
pre-processing the data by:
dividing the data into a plurality of temporal segments, and
extracting a plurality of features from each of the temporal segments;
generating a preliminary seizure classification for each of the temporal segments based on the plurality of extracted features; and
determining an overall seizure classification for the subject based on the preliminary seizure classification for each of the temporal segments.
2 . The method of claim 1 , wherein the overall seizure classification is a ternary classification of electrographic seizure-like activity, highly pathological EEG with high likelihood of epileptiform activity, or normal electrographic activity.
3 . The method of claim 1 , wherein the overall seizure classification is a seizure probability or a seizure severity value.
4 . The method of claim 1 , wherein generating the preliminary seizure classification comprises generating a first probability of seizure under a first comparison and generating a second probability of seizure under a second comparison.
5 . The method of claim 4 , wherein generating the first probability of seizure comprises comparing a first likelihood of a temporal segment as normal electrographic activity to a second likelihood of the temporal segment as highly pathological EEG with high likelihood of epileptiform activity.
6 . The method of claim 5 , wherein determining the overall seizure classification comprises calculating a moving average of the first probability of seizure over the time window.
7 . The method of claim 4 , wherein generating the second probability of seizure comprises comparing a first likelihood of a temporal segment as electrographic seizure to a second likelihood of the temporal segment as highly pathological EEG with high likelihood of epileptiform activity.
8 . The method of claim 4 , wherein generating the first and second probabilities of seizures comprises filtering the each of the temporal segments with cascaded convolutional filters.
9 . The method of claim 4 , wherein generating the preliminary seizure classification comprises combining the first and second seizure probabilities and classifying a corresponding temporal segment based on the combined seizure probabilities.
10 .- 11 . (canceled)
12 . The method of claim 1 , wherein determining the overall seizure classification comprises determining the overall seizure classification for a time window.
13 . (canceled)
14 . The method of claim 12 , wherein determining the overall seizure classification comprises calculating a moving average of the preliminary seizure classification over the time window.
15 . The method of claim 12 , wherein each temporal segment corresponds to at least one epoch, and wherein the time window comprises one or more epochs.
16 . The method of claim 15 , wherein pre-processing the data further comprises extracting a plurality of multi-channel features that quantify a degree of correlation between pairs of temporal segments from different EEG signals corresponding to a given time epoch.
17 . The method of claim 15 , further comprising using a multichannel machine learning model to generate a multi-channel seizure classification for each time epoch based on the plurality of multi-channel features.
18 .- 23 . (canceled)
24 . The method of claim 1 further comprising providing a trace of the overall seizure classification over time.
25 . The method of claim 24 further comprising determining a trendline of the trace.
26 .- 29 . (canceled)
30 . The method of claim 1 , wherein the EEG signals are recorded from a plurality of electrodes incorporated into a headband worn by the subject.
31 . (canceled)
32 . The method of claim 1 , further comprising treating the subject for seizure if one or both of electrographic seizure and highly pathological EEG with high likelihood of epileptiform activity and is detected.
33 . A system for detecting seizure comprising:
a data module configured to receive data comprising a plurality of electroencephalography (EEG) signals recorded during a time window and over a plurality of channels from a subject; and
a seizure detection module comprising a memory storing a set of instructions and one or more processors that are configured to, responsive to the set of instructions:
pre-process the data received by the data module by:
dividing the EEG signal into a plurality of temporal segments, and
extracting a plurality of features from each of the plurality of temporal segments;
generating a preliminary seizure classification for each of the temporal segments based on the plurality of extracted features; and
determining an overall seizure classification for the subject based on the preliminary seizure classification for each of the temporal segments.
34 . The system of claim 33 , wherein the overall seizure classification is a ternary classification of electrographic seizure-like activity, highly pathological EEG with high likelihood of epileptiform activity, or normal electrographic activity.
35 .- 39 . (canceled)
40 . The system of claim 33 , further comprising a headband, the headband comprising a plurality of electrodes from which the plurality of electroencephalography (EEG) signals is recorded.
41 .- 55 . (canceled)