Systems and methods for seizure detection using compression-enabled joint entropy estimation
View Patent ↗Systems and methods for seizure detection in accordance with embodiments of the invention are illustrated. One embodiment includes an automated seizure treatment system, including an electroencephalogram (EEG) device configured to record neural activity from a brain of a patient, a treatment device, a processor, and a memory, the memory containing a seizure detection application that configures the processor to obtain an EEG signal from the EEG device, calculate an inverse compression ratio based on the EEG signal, and when the inverse compression ratio is greater than a classification threshold, deliver treatment capable of stopping the seizure to the patient using the treatment device.
1 . An automated seizure treatment system, comprising:
an electroencephalogram (EEG) device configured to record neural activity from a brain of a patient;
a treatment device;
a processor; and
a memory, the memory containing a seizure detection application that configures the processor to:
obtain an EEG signal from the EEG device;
calculate an inverse compression ratio based on the EEG signal, wherein calculating the inverse compression ratio comprises:
measuring an uncompressed size of the EEG signal;
compressing the EEG signal using a lossless compression algorithm;
measuring a compressed size of the EEG signal; and
dividing the compressed size by the uncompressed size; and
when the inverse compression ratio is greater than a classification threshold, automatically deliver treatment capable of stopping the seizure to the patient using the treatment device.
2 . The system of claim 1 , wherein the EEG device is a stereo EEG device.
3 . The system of claim 1 , wherein the EEG device is a scalp EEG device.
4 . The system of claim 1 , wherein the EEG device is an electrocorticogram (ECoG) device.
5 . The system of claim 1 , wherein the treatment device is an implantable neurostimulator configured to produce neurostimulation to counteract the spread of seizure activity in the brain of the patient.
6 . The system of claim 1 , wherein the treatment device is a drug pump configured to deliver an anti-seizure drug.
7 . The system of claim 1 , wherein the treatment device further comprises a warning device capable of producing an alert indicating seizure activity.
8 . The system of claim 1 , wherein the classification threshold is a value selected to maximize an F1 score for a precision recall curve generated using historical EEG signals from the brain of the patient.
9 . The system of claim 1 , wherein the lossless compression algorithm is the DEFLATE algorithm.
10 . An automated seizure treatment method, comprising:
obtaining an electroencephalogram (EEG) signal from an EEG device configured to record neural activity from a brain of a patient;
calculating, using a processor, an inverse compression ratio based on the EEG signal using a seizure detector, wherein calculating the inverse compression ratio comprises:
measuring an uncompressed size of the EEG signal;
compressing the EEG signal using a lossless compression algorithm;
measuring a compressed size of the EEG signal; and
dividing the compressed size by the uncompressed size; and
when the inverse compression ratio is greater than a classification threshold, automatically delivering treatment capable of stopping the seizure to the patient using the treatment device.
11 . The method of claim 10 , wherein the EEG device is a stereo EEG device.
12 . The method of claim 10 , wherein the EEG device is a scalp EEG device.
13 . The method of claim 10 , wherein the EEG device is an electrocorticogram (ECoG) device.
14 . The method of claim 10 , wherein the treatment device is an implantable neurostimulator configured to produce neurostimulation to counteract the spread of seizure activity in the brain of the patient.
15 . The method of claim 1 , wherein the treatment device is a drug pump configured to deliver an anti-seizure drug.
16 . The method of claim 10 , wherein the treatment device further comprises a warning device capable of producing an alert indicating seizure activity.
17 . The method of claim 10 , wherein the classification threshold is a value selected to maximize an F1 score for a precision recall curve generated using historical EEG signals from the brain of the patient.
18 . The method of claim 10 , wherein the lossless compression algorithm is the DEFLATE algorithm.