System and method for identifying ictal states in a patient
A system and method for identifying ictal states in a patient, the system including a plurality of sensors for sensing different non-electroencephalographic signals from the patient, and a processing unit. The processing unit has a processor and memory with instructions for classifying data from the plurality of sensors to determine probability of the patient being ictal, when the probability is high asking the patient if the patient is in an ictal state, reporting to a caregivers if the patient is ictal, and updating the classifier based upon sensor data, probability, and the response. The method includes sensing, using a plurality of non-electroencephalographic sensors, determining, using a classifier trained using a training dataset, probability of the patient being in an ictal state, and if probability is high, asking the patient if the patient is in ictal state, logging the occurrence of an ictal state, and updating the classifier.
1. A system for identifying ictal states in a patient, comprising:
a sensor group consisting of a plurality of non-electroencephalographic sensors selected from the group consisting of photoplethysmographic (PPG) sensors, one or more single or tri-axial accelerometers, electrocardiogram (ECG) sensors, electromyography (EMG) sensors, microphones, temperature sensors, eve-motion sensors, facial movement sensors, galvanic skin response (GSR) sensors, and limb acceleration sensors:
the plurality of non-electroencephalographic sensors being configured to sense a plurality of non-electroencephalographic signals from the patient;
a signal processing unit coupled to receive time-series data from a sensor group consisting of only the plurality of non-electroencephalographic sensors and comprising:
a processor; an output device; and an input device;
a memory storing machine-readable instructions that when executed by the processor are capable of:
classifying, using a classifier based upon a training set and
implemented by the machine readable instructions, the time-series data from the sensor group to determine a probability of the patient being in an ictal state; and when the probability is greater than a threshold:
asking the patient if the patient is in an ictal state using the output device;
if the patient fails to respond to the asking using the input device or enters a positive reply, logging occurrence of a seizure; and
updating the classifier based upon the time-series data from the sensor group, the probability, and the patient response.
2. The system of claim 1 , wherein the classifier implements ensembles of two or more classifiers selected from the group consisting of discriminative classifiers and generative classifiers.
3. The system of claim 1 , the memory further comprising machine readable instructions that when executed by the processor are capable of updating at least one of the discriminative classifiers and the generative classifiers based upon the time-series data from the sensor group, the probability, and the patient response.
4. The system of claim 3 , the sensor group comprising an ECG sensor and an accelerometer.
5. The system of claim 4 , the non-electroencephalographic sensors comprising a sensor selected from the group consisting of a photoplethysmographic (PPG) sensor, a GSR sensor, a microphone, a temperature sensor, a facial movement sensor, and an eye-motion sensor.
6. The system of claim 5 wherein at least one sensor of the sensor group is coupled to the signal processing unit through a short range digital radio.
7. The system of claim 4 , wherein the classifier incorporates machine readable code for machine learning to improve reliability of identifying the ictal state for the patient over time.
8. The system of claim 1 , wherein the signal processing unit comprises a smartphone.
9. The system of claim 1 wherein the processor generates a report to a caregiver that the patient is in the ictal state if one of a positive reply and no reply is received in response to the asking.
10. The system of claim 1 , the memory further comprising machine readable instructions that when executed by the processor are capable of processing the time-series data from the sensor group to determine at least one feature vector.