Methods and systems for forecasting seizures

1. 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 future time windows;

generating a probabilistic model based on the physiological data associated with each epileptic event;

combining the probabilistic model and the temporal probability model to generate a weighted probabilistic model representing a time-weighted probability of future seizure activity in each of the plurality of future time windows; and

outputting an estimate of seizure probability in the subject for one or more of the plurality of future time windows using the weighted probabilistic model.

2. The method of claim 1 , wherein the physiological data comprises an electroencephalography (EEG) signal received from a brain of the subject.

3. The method of claim 1 , wherein the probabilistic model is a logistic regression model.

4. The method of any one of claim 3 , wherein generating the logistic regression model comprises:

determining, for a logistic regression classifier, a set of features in the physiological data; and

training the logistic regression model using the set of features.

5. The method of claim 4 , wherein the set of features comprises:

a) a line length feature; and

b) one or more energy features, each of the one or more energy features being at a different frequency band.

6. The method of claim 4 , further comprising applying a moving average filter to the set of features.

7. The method of claim 4 , wherein the logistic regression model is trained to output the probability, P(S=|X), that a feature vector, X, is pre-ictal.

8. The method of claim 4 , wherein combining the probabilistic model and the temporal probability model comprises:

updating the weights of the logistic regression classifier based on the temporal probability model.

9. The method of claim 1 , wherein generating the temporal probability model comprises:

estimating a probability density function based on the times at which each epileptic event occurred.

10. The method of claim 9 , wherein the probability density function is estimated using kernel density estimation.

11. The method of claim 1 , wherein the weighted probabilistic model is given by

P

⁡

(

S

=

1

❘

X

)

=

1

1

+

exp

⁢

⁢

(

log

⁢

1

-

P

S

P

S

+

log

⁢

P

⁡

(

X

❘

S

=

0

)

P

⁡

(

X

❘

S

=

1

)

)

where P S represents the prior probability of a seizure occurring in one of the plurality of future time windows.

12. The method of claim 1 , wherein the physiological data comprises one or more of:

heart rate data;

blood pressure data;

sweat data;

movement data measured by an accelerometer worn by the patient.

13. The method of claim 1 , wherein the time at which each epileptic event occurred is measured from the time of a previous epileptic event.

14. The method of claim 1 , wherein the historical data further comprises values of one or more environmental variables associated with each epileptic event, and wherein the method further comprises:

generating an environmental probability model of future epileptic events based on the environmental variables associated with each epileptic event, the environmental probability model representing a probability of a future seizure occurrence over a range of values of the one or more environmental variable; and

combining the probabilistic model with the environmental probability model to generate the weighted probabilistic model of future seizure activity in each of the plurality of future time windows.

15. The method of claim 1 , wherein the historical data further comprises values of one or more variables of sleep associated with each epileptic event, and wherein the method further comprises:

generating a sleep probability model of future epileptic events based on the sleep variables associated with each epileptic event, the sleep probability model representing a probability of a future seizure occurrence over a range of values of the one or more sleep variable; and

combining the probabilistic model with the sleep probability model to generate the weighted probabilistic model of future seizure activity in each of the plurality of future time windows.

16. A seizure advisory system, comprising:

an input for 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;

a processor configured to:

generate a temporal probability model of future epileptic events based on the time of each of the epileptic event, the temporal probability model representing a probability of a future seizure occurrence in each of a plurality of future time windows;

generate a probabilistic model based on the physiological data associated with each epileptic event;

combine the probabilistic model and the temporal probability model to generate a weighted probabilistic model representing a time-weighted probability of future seizure activity in each of the plurality of time windows; and

output an estimate of seizure probability in the subject for one or more of the plurality of future time windows using the weighted probabilistic model.

17. The system of claim 16 , further comprising one or more measurement devices configured to record the physiological data.

18. The system of claim 17 wherein the one or measurement devices comprises an EEG sensor comprising a lead having at least one electrode for recording neural activity at a neural structure in a brain of the subject, wherein the physiological data comprises the recorded neural activity.

19. 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 an environmental variable associated with each epileptic event;

generating an environmental probability model of future epileptic events based on the environmental variable associated with each epileptic event, the environmental probability model representing a probability of a future seizure occurrence for each of a plurality of values of the environmental variable;

generating a probabilistic model based on the physiological data associated with each epileptic event;

combining the probabilistic model and the environmental probability model to generate a weighted probabilistic model representing a time-weighted probability of future seizure activity for each of the plurality of values of the environmental variable in one or more future time windows; and

outputting an estimate of seizure probability in the subject for one or more of the plurality of values of the environmental variable for one or more future time windows using the weighted probabilistic model.

20. The method of claim 19 , wherein the environmental variable is temperature, humidity, wind speed, barometric pressure, rainfall or a combination thereof.