Systems and methods for seizure prediction and detection

Abstract

The present disclosure provides systems and methods for seizure detection. The method for seizure detection may include receiving a plurality of electroencephalography (EEG) signals over a plurality of channels for a subject, preprocessing the plurality of EEG signals by segmenting the plurality of EEG signals for each channel into a plurality of temporal data segments, extracting a plurality of features from each temporal data segment for each channel, and applying a machine learning algorithm to the plurality of features to perform a seizure binary classification for each temporal data segment for each channel. A control policy may be employed to determine a seizure burden on the aggregated seizure binary classifications. When the seizure burden is equal to or exceeds a threshold, a notification may be generated. The notification may be usable by a healthcare practitioner to assess whether the subject may be at risk of having a seizure.

1 . (canceled) 
     
     
         2 . A method for seizure detection, comprising:
 (a) receiving a plurality of electroencephalography (EEG) signals over a plurality of channels for a subject;   (b) preprocessing the plurality of EEG signals by segmenting the plurality of EEG signals for each channel into a plurality of temporal data segments;   (c) extracting a plurality of features from each temporal data segment for each channel, wherein each temporal data segment is associated with a time epoch;   (d) applying a machine learning algorithm to the plurality of features to perform a seizure binary classification for each temporal data segment for each channel, thereby generating a plurality of classifications for the plurality of temporal data segments, wherein the seizure binary classification for each temporal data segment comprises classifying each temporal data segment for each channel as (1) seizure-positive or (2) seizure-negative, wherein the plurality of classifications are compared sequentially across a plurality of time epochs on each channel;   (e) aggregating the classifications for the plurality of temporal data segments for the plurality of channels over a moving time window, wherein the moving time window comprises a fixed period of time between about one minute and about one hour; and   (f) determining a seizure burden as a continuous output measure for the moving time window based on the aggregated classifications, wherein the seizure burden comprises a percentage of the temporal data segments that are classified as seizure-positive, and wherein the seizure burden is a metric that provides a measure of a degree of severity or likelihood of a seizure.   
     
     
         3 . The method of  claim 2 , wherein the continuous output measure is used to generate one or more notifications when the seizure burden is equal to or exceeds one or more thresholds, wherein the one or more notifications are indicative of different seizure activities and are usable for assessing whether the subject is at risk of having a seizure. 
     
     
         4 . The method of  claim 2 , wherein each temporal data segment has a duration ranging from about one second to about twenty seconds. 
     
     
         5 . The method of  claim 4 , wherein the duration of each temporal data segment is about ten seconds. 
     
     
         6 . The method of  claim 2 , wherein the moving time window is about five minutes. 
     
     
         7 . The method of  claim 3 , wherein the one or more notifications are generated in the form of visual, audio, and/or textual alerts. 
     
     
         8 . The method of  claim 3 , wherein a first notification indicative of frequent seizure activity is generated when the seizure burden is equal to or exceeds a first threshold of 10%. 
     
     
         9 . The method of  claim 3 , wherein a second notification indicative of abundant seizure activity is generated when the seizure burden is equal to or exceeds a second threshold of 50%. 
     
     
         10 . The method of  claim 3 , wherein a third notification indicative of continuous seizure activity is generated when the seizure burden is equal to or exceeds a third threshold of 90%. 
     
     
         11 . The method of  claim 2 , wherein the plurality of channels comprises at least three channels. 
     
     
         12 . The method of  claim 11 , wherein the plurality of channels comprises eight channels. 
     
     
         13 . The method of  claim 2 , wherein the plurality of features comprises time and/or frequency domain features that are intrinsic in the plurality of EEG signals. 
     
     
         14 . The method of  claim 13 , wherein the plurality of features comprises at least twenty different time and/or frequency features. 
     
     
         15 . The method of  claim 13 , wherein the plurality of features comprises a plurality of discrete values associated with the time and/or frequency domain features. 
     
     
         16 . The method of  claim 2 , wherein the machine learning algorithm comprises a random forest, a boosted decision tree, a classification tree, a regression tree, a bagging tree, a neural network, or a rotation forest. 
     
     
         17 . The method of  claim 2 , wherein the machine learning algorithm is individually applied to the plurality of features extracted for each channel, such that each channel has a separate iteration of the machine learning algorithm. 
     
     
         18 . The method of  claim 2 , wherein the preprocessing of the plurality of EEG signals further comprises:
 applying a filter to the plurality of EEG signals over the plurality of channels, prior to the segmentation of the plurality of EEG signals.   
     
     
         19 . The method of  claim 18 , wherein the filter comprises a bandpass filter configured to filter the plurality of EEG signals between about one Hertz (Hz) and about thirty-five Hz. 
     
     
         20 . The method of  claim 2 , further comprising:
 classifying a particular time epoch as associated with a potential electrographic seizure, if the temporal data segments for a subset of the plurality of channels are classified as seizure-positive.   
     
     
         21 . The method of  claim 20 , wherein the subset comprises at least half of the plurality of channels. 
     
     
         22 . The method of  claim 2 , wherein sequential periods of time formed by the moving time window are non-overlapping. 
     
     
         23 . The method of  claim 2 , further comprising: outputting the seizure burden as a graphical visual element on a display. 
     
     
         24 . The method of  claim 23 , further comprising: displaying one or more thresholds of the seizure burden in the graphical visual element. 
     
     
         25 . The method of  claim 23 , wherein the graphical visual element comprises a time-series plot, bar graph, or chart. 
     
     
         26 . The method of  claim 24 , wherein the time-series plot is configured to change in color as the seizure burden passes a threshold of the one or more thresholds. 
     
     
         27 . The method of  claim 23 , further comprising:
 using the graphical visual element to (i) assess a condition of the subject, (ii) determine a course of treatment, (iii) monitor an effectiveness of a course of treatment if the course of treatment is provided to the subject, or (iv) monitor a progression of the subject's condition over time.