Methods and systems for forecasting seizures
A method of estimating the probability of a seizure in a subject, the method comprising: receiving historical data associated with epileptic events experienced by the subject over a first time period, the historical data comprising physiological data associated with each epileptic event and a time at which each epileptic event occurred; generating a temporal probability model of future epileptic events based on the time of each of the epileptic events, the temporal probability model representing a probability of a future seizure occurrence in each of a plurality of time windows; generating a probabilistic model based on the physiological data associated with each epileptic event; weighting the probabilistic model based on the temporal probability model to generate a weighted probabilistic model of future seizure activity; and outputting an estimate of seizure probability in the subject using the weighted probabilistic model.
1 . A method executed by a portable seizure advisory device comprising at least one EEG sensor, wherein the at least one EEG sensor is an intracranial EEG sensor and/or an EEG sensor positioned external to a patient's skull, a processor in communication with the at least one EEG sensor, and a display in communication with the processor, the method comprising:
receiving, from the at least one EEG sensor, a continuous EEG signal sampled at a frequency that is at least 250 Hz and filtered using a band-pass filter;
computing, in a signal-processing module, line length and multi-band energy features from the EEG signal within time delimited moving windows;
training a logistic regression classifier using the features to estimate a probability of seizure occurrence;
generating a temporal probability model of seizure events for a subject based on circadian and infradian cycles determined from historical seizure times;
combining the logistic regression classifier and the temporal probability model using Bayesian updating to produce a time-weighted probability of future seizure occurrence in successive time delimited forecast windows;
updating the model parameters after each recorded seizure event; and
automatically outputting, a seizure-risk warning using the portable device via the display and/or via a haptic alert when the time-weighted probability exceeds a threshold, wherein the warning identifies both the seizure probability and a corresponding risk-level category.
2 . The method of claim 1 , wherein each of the frequency multi-band energy features comprises a different frequency band.
3 . The method of claim 2 , wherein the logistic regression classifier is trained to output the probability, P(S=1|X) that a feature vector, X, is pre-ictal.
4 . The method of claim 2 , wherein combining the logistic regression classifier and the temporal probability model comprises:
updating weights of the logistic regression classifier based on the temporal probability model.
5 . The method of claim 1 , wherein the threshold is user-configurable.
6 . A system for generating and displaying a seizure risk warning in a portable seizure advisory device, the system comprising:
at least one EEG sensor, wherein the at least one EEG sensor is an intracranial EEG sensor and/or an EEG sensor positioned external to a patient's skull;
a display;
one or more processors in data communication with the at least one EEG sensor and the display; and
a memory comprising a non-transitory computer executable instructions, which when executed by the one or more processors, cause the system to:
receive, from the at least one EEG sensor, a continuous EEG signal sampled at a frequency that is at least 250 Hz and filtered using a band-pass filter;
compute, in a signal-processing module, line length and multi-band energy features from the EEG signal within time delimited moving windows;
train a logistic regression classifier using the features to estimate a probability of seizure occurrence;
generate a temporal probability model of future-seizure events for a subject based on circadian and infradian cycles determined from historical seizure times;
combine the logistic regression classifier and the temporal probability model using Bayesian updating to produce a time-weighted probability of future seizure occurrence in successive time delimited forecast windows;
update the model parameters after each recorded seizure event; and
automatically output a seizure-risk warning using the portable device via the display and/or via a haptic alert when the time-weighted probability exceeds a threshold, wherein the warning identifies both the seizure probability and a corresponding risk-level category.
7 . The system of claim 6 , wherein the threshold is user-configurable.
8 . A non-transitory machine-readable storage medium storing instructions which, when executed by one or more processors, cause the one or more processors to perform operations for generating and displaying a seizure risk warning in a portable seizure advisory device comprising at least one EEG sensor, wherein the at least one EEG sensor is an intracranial EEG sensor and/or an EEG sensor positioned external to a patient's skull, a display, and a processor in data communication with the at least one EEG sensor and the display, the operations comprising:
receiving, from the at least one EEG sensor, a continuous EEG signal sampled at a frequency that is at least 250 Hz and filtered using a band-pass filter;
computing, in a signal-processing module, line length and multi-band energy features from the EEG signal within time delimited moving windows;
training a logistic regression classifier using the features to estimate a probability of seizure occurrence;
generating a temporal probability model of seizure events for a subject based on circadian and infradian cycles determined from historical seizure times;
combining the logistic regression classifier and the temporal probability model using Bayesian updating to produce a time-weighted probability of future seizure occurrence in successive time delimited forecast windows
updating the model parameters after each recorded seizure event; and
automatically outputting a seizure-risk warning using the portable device via the display and/or via a haptic alert when the time-weighted probability exceeds a threshold, wherein the warning identifies both the seizure probability and a corresponding risk-level category.
9 . The non-transitory machine-readable storage medium of claim 8 , wherein the threshold is user-configurable.
10 . A method executed by a portable seizure advisory device comprising at least one EEG sensor, wherein the at least one EEG sensor is an intracranial EEG sensor and/or an EEG sensor positioned external to a patient's skull, a display, and a processor in data communication with the at least one EEG sensor and the display, the method comprising:
receiving, from the EEG sensor, a continuous EEG signal sampled at a frequency that is at least 250 Hz and filtered using a band-pass filter and an environmental variable associated with epileptic events;
computing, in a signal-processing module, line length and multi-band energy features from the EEG signal within time delimited moving windows;
training a logistic regression classifier using the features to estimate a probability of seizure occurrence;
generating an environmental probability model of seizure events for a subject based on the environmental variable associated with each epileptic event, wherein the environmental probability model represents a probability of a seizure occurrence for each of a plurality of values of the environmental variable;
combining the logistic regression classifier and the environmental probability model using Bayesian updating to produce a time-weighted probability of future seizure occurrence in successive time delimited forecast windows;
updating the model parameters after each recorded seizure event; and
automatically outputting a seizure-risk warning using the portable device via the display and/or via a haptic alert when the time-weighted probability exceeds a threshold, wherein the warning identifies both the seizure probability and a corresponding risk-level category.
11 . The method of claim 10 , wherein the environmental variable is temperature, humidity, wind speed, barometric pressure, rainfall or a combination thereof.
12 . The method of claim 10 , wherein the threshold is user-configurable.
13 . A system for generating and displaying a seizure risk warning in a portable seizure advisory device, the system comprising:
at least one EEG sensor, wherein the at least one EEG sensor comprises an intracranial EEG sensor and/or an EEG sensor positioned external to a patient's skull;
a display;
one or more processors in data communication with the at least one EEG sensor and the display; and
memory comprising computer executable instructions, which when executed by the one or more processors, cause the system to:
receive, from the at least one EEG sensor, a continuous EEG signal sampled at a frequency that is at least 250 Hz and filtered using a band-pass filter and an environmental variable associated with epileptic events;
compute, in a signal-processing module, line length and multi-band energy features from the EEG signal within time delimited moving windows;
train a logistic regression classifier using the features to estimate a probability of seizure occurrence;
generate an environmental probability model of seizure events for a subject based on the environmental variable associated with each epileptic event, wherein the environmental probability model represents a probability of a seizure occurrence for each of a plurality of values of the environmental variable;
combine the logistic regression classifier and the environmental probability model using Bayesian updating to produce a time-weighted probability of future seizure occurrence in successive time delimited forecast windows;
update the model parameters after each recorded seizure event; and
automatically output a seizure-risk warning using the portable device via the display and/or via a haptic alert when the time-weighted probability exceeds a threshold, wherein the warning identifies both the seizure probability and a corresponding risk-level category.
14 . The system of claim 13 , wherein the threshold is user-configurable.
15 . A non-transitory machine-readable storage medium storing instructions which, when executed by one or more processors, cause the one or more processors to perform operations for generating and displaying a seizure risk warning in a portable seizure advisory device comprising at least one EEG sensor, wherein the at least one EEG sensor is an intracranial EEG sensor and/or an EEG sensor positioned external to a patient's skull, a display, and a processor in data communication with the at least one EEG sensor and the display, the operations comprising:
receiving, from the at least one EEG sensor, a continuous EEG signal sampled at a frequency that is at least 250 Hz and filtered using a band-pass filter and an environmental variable associated with each epileptic events;
computing, in a signal-processing module, line length and multi-band energy features from the EEG signal within time delimited moving windows;
training a logistic regression classifier using the features to estimate a probability of seizure occurrence;
generating an environmental probability model of seizure events for a subject based on the environmental variable associated with each epileptic event, wherein the environmental probability model represents a probability of a seizure occurrence for each of a plurality of values of the environmental variable;
combining the logistic regression classifier and the environmental probability model using Bayesian updating to produce a time-weighted probability of future seizure occurrence in successive time delimited forecast windows;
updating the model parameters after each recorded seizure event; and
automatically outputting a seizure-risk warning using the portable device via the display and/or via a haptic alert when the time-weighted probability exceeds a threshold, wherein the warning identifies both the seizure probability and a corresponding risk-level category.
16 . The non-transitory machine-readable storage medium of claim 15 , wherein the threshold is user-configurable.