Systems and methods for seizure detection and closed-loop neurostimulation
Embodiments described herein relate to systems, devices, and methods for monitoring brain activity and delivering electrical brain stimulation to a patient. In some embodiments, a system can deliver responsive electrical stimulation in a closed-loop manner and can offer real-time or near real-time monitoring of induced neurophysiological effects. After detecting a targeted brain pattern (i.e., an epileptic seizure), the system may deliver high-intensity ultra-short electrical stimulation impulses non-invasively or minimal-invasively to diminish or stop the neural oscillations underlying the epileptic seizure. The stimuli are delivered in time and space, relative to the emerging seizure rhythm patterns, such that they diminish or terminate the seizure. The system may include an implantable device including one or more electrodes electrically coupled to one or more processors. The processor(s) may be operatively coupled to a memory, one or more communication modules, and may optionally be coupled to a battery and one or more additional sensor(s).
1 . A system, comprising:
a plurality of electrodes configured for implantation in a patient and configured to measure a brain activity of the patient;
a memory; and
one or more processors operatively coupled to the memory and the plurality of electrodes, the one or more processors configured to:
receive brain activity data from the plurality of electrodes;
detect an onset of a seizure based on the brain activity data;
identify a pattern in the brain activity data based on mutual information between one or more pairs of electrodes from the plurality of electrodes, the mutual information determined, at least in part, by:
computing covariances between signals from the one or more pairs of electrodes from the plurality of electrodes,
generating a first value based on the covariances,
normalizing the first value to generate a second value r sq , and
calculating a logarithm of (1−r sq );
determine, based on the pattern in the brain activity data, a timing with which to deliver current pulses to a brain of the patient to disrupt at least one oscillation in brain activity contributing to the seizure; and
activate a subset of electrodes from the plurality of electrodes to deliver the current pulses to a target region of the brain of the patient according to the timing.
2 . The system of claim 1 , wherein the brain activity data includes electroencephalography (EEG) data.
3 . The system of claim 1 , wherein a subset of electrodes from the plurality of electrodes is implanted in one of a subgaleal space of the patient, a subdural space of the patient, an epidural space of the patient, or the brain of the patient.
4 . The system of claim 1 , wherein the plurality of electrodes is configured to deliver Intersectional Short-Pulse (ISP) stimulation to disrupt the at least one oscillation in brain activity contributing to the seizure.
5 . The system of claim 1 , wherein the one or more processors is configured to determine the timing based on at least one of a phase or a frequency of the brain activity data, and to activate the subset of electrodes to deliver the current pulses one of immediately or after a predetermined delay, and with a predefined frequency.
6 . The system of claim 1 , wherein the one or more processors is further configured to activate the subset of electrodes to deliver the current pulses one of immediately or after a predetermined delay, the one or more current pulses configured to align with an inherent rhythmicity of the brain activity data.
7 . The system of claim 1 , further comprising:
a sensor configured to measure biosignal data associated with the patient, the one or more processors configured to detect a precursor activity leading to a seizure, the onset of the seizure, or a presence of the seizure further based on the biosignal data.
8 . The system of claim 7 , wherein the sensor is configured to measure at least one of electromyography (EMG) data, electrocardiogram (ECG) data, or heart rate.
9 . The system of claim 1 , wherein the one or more processors is further configured to quantify the pattern in the brain activity data by calculating a measure of rhythmicity at predetermined frequency components of the brain activity data.
10 . The system of claim 1 , further comprising:
a communication interface configured to transfer information between the one or more processors and an external device,
the external device configured to train a model for detecting a precursor activity leading to a seizure, the onset of the seizure, or a presence of the seizure, the model configured to be executed by the one or more processors.
11 . The system of claim 10 , wherein the model is trained using datasets including ictal EEG data and non-ictal EEG data from at least one of the patient or another patient.
12 . An implantable neurostimulator device, comprising:
a memory; and
a processor operatively coupled to the memory, the processor configured to be electrically coupled to a plurality of electrodes implanted in a patient, the processor configured to:
receive brain activity data from the plurality of electrodes;
detect a precursor activity leading to a seizure, an onset of the seizure, or a presence of the seizure based on the brain activity data;
identify a pattern in the brain activity data based on mutual information between one or more pairs of electrodes from the plurality of electrodes, the mutual information determined, at least in part, by:
computing covariances between signals from the one or more pairs of electrodes from the plurality of electrodes,
generating a first value based on the covariances,
normalizing the first value to generate a second value r sq , and
calculating a logarithm of (1−r sq );
determine a timing with which to deliver current pulses to a brain of the patient to interfere with at least one oscillation in brain activity contributing to the seizure based on the pattern in the brain activity data; and
activate a subset of electrodes from the plurality of electrodes to deliver the current pulses to a target region of the brain of the patient based on the timing.
13 . The implantable neurostimulator device of claim 12 , wherein the brain activity data includes electroencephalography (EEG) data.
14 . The implantable neurostimulator device of claim 12 , wherein at least a subset of electrodes from the plurality of electrodes is implanted in one of a subgaleal space of the patient, a subdural space of the patient, an epidural space of the patient, or the brain of the patient.
15 . The implantable neurostimulator device of claim 12 , wherein the plurality of electrodes is configured to deliver Intersectional Short-Pulse (ISP) stimulation, and each of the current pulses has an amplitude of about 0.1 milliamps (mA) to about 80 mA.
16 . The implantable neurostimulator device of claim 12 , wherein the processor is configured to determine the timing based on one of a measure of rhythmicity in the brain activity data or a feature of the rhythmicity in the brain activity data, and to activate the subset of electrodes to deliver the current pulses one of immediately or after a predetermined delay, and with a predefined frequency.
17 . The implantable neurostimulator device of claim 12 , wherein the processor is further configured to activate the subset of electrodes to deliver the current pulses one of immediately or after a predetermined delay, the current pulses configured to align with an inherent rhythmicity of the brain activity data.
18 . The implantable neurostimulator device of claim 12 , further comprising:
one or more sensors configured to measure biosignal data associated with the patient, the processor configured to detect the precursor activity leading to the seizure, the onset of the seizure, or the presence of the seizure further based on the biosignal data.
19 . A method, comprising:
measuring brain activity data associated with a brain of a patient using a plurality of electrodes implanted in the patient;
detecting a precursor activity leading to a seizure, an onset of the seizure, or a presence of the seizure based on the brain activity data;
identifying a pattern in the brain activity data based on mutual information associated with the plurality of electrodes, the mutual information determined, at least in part, by:
computing covariances between signals associated with the plurality of electrodes,
generating a first value based on the covariances,
normalizing the first value to generate a second value r sq , and
calculating a logarithm of (1−r sq );
determining, based on the pattern in the brain activity data, a timing with which to deliver electrical stimulation to the brain of the patient to disrupt oscillations in brain activity contributing to the seizure; and
in response to detecting the precursor activity leading to a seizure, the onset of the seizure, or the presence of the seizure, causing delivery of electrical stimulation to the brain of the patient via at least a subset of electrodes from the plurality of electrodes and according to the timing.
20 . The method of claim 19 , wherein the brain activity data includes electroencephalography (EEG) data collected from a plurality of EEG channels.
21 . The method of claim 20 , further comprising:
applying a filter to each EEG channel from the plurality of EEG channels; and
generating a virtual channel by calculating a weighted average of the plurality of EEG channels.
22 . The method of claim 21 , further comprising:
applying a sliding window to the plurality of EEG channels and the virtual channel to produce windowed EEG channels and a windowed virtual channel; and
calculating features for each of the windowed EEG channels and the windowed virtual channel based on a predefined time interval.
23 . The method of claim 22 , wherein the features include at least one of time domain features, frequency domain features, spatial features, or temporal dynamic features.
24 . The method of claim 22 , wherein the features include at least one of a root mean square, a line length, a variance, a kurtosis, a Hjorth mobility, a Hjorth complexity, a wavelet transform, a mutual information, a mean coherence, a standard deviation of mean phase delay, a recurrence rate, a determinism, an entropy, or an averaged diagonal line length.
25 . The method of claim 24 , further comprising:
inputting the features to a regressive decision tree trained using EEG data from the patient or EEG data from at least one other patient;
averaging outputs of the regressive decision trees to produce an averaged output; and
comparing the averaged output to a dynamic threshold that is updated based on previous outputs of a classification model.
26 . A method, comprising:
calculating one or more features based on brain activity recorded from each electrode of a plurality of electrodes implanted in a patient, the one or more features including mutual information between one or more pairs of electrodes from the plurality of electrodes, the mutual information identified, at least in part, by:
computing covariances between signals associated with the one or more pairs of electrodes,
generating a first value based on the covariances,
normalizing the first value to generate a second value r sq , and
calculating a logarithm of (1−r sq );
inputting at least one of (1) the one or more features, or (2) at least a portion of the recorded brain activity into a model trained using brain activity data from at least one of the patient or at least one other patient,
determining a range of effective prediction probability threshold values and selecting a threshold value from the range of effective prediction probability threshold values;
comparing an output of the model to the threshold value;
in response to determining the output crosses the threshold value, determine a timing and a subset of electrodes with which to deliver electrical pulses; and
causing delivery of the electrical pulses to a brain of the patient via at least the subset of electrodes and according to the timing.
27 . The method of claim 26 , wherein the determining the range of effective prediction probability threshold values is based on balancing between performance metrics from at least one pair of performance metrics, each pair of performance metrics from the at least one pair of performance metrics having the property that improving one performance metric from that pair of performance metrics results in a diminishing of the other.
28 . The method of claim 26 , wherein the computing the one or more features includes:
performing recurrence quantification analysis (RQA) without using a recurrence plot matrix.
29 . The method of claim 26 , wherein the one or more features includes cross-channel coherence, the determining the cross-channel coherence includes, for each frequency component:
applying a frequency-specific sine kernel convolution or applying Discrete Fourier Transform (DFT) to a brain activity signal of each electrode from the plurality of electrodes to generate frequency-specific coherence values or a frequency spectrum of the brain activity signal;
calculating relevant cross-coherence values from the frequency-specific coherence values or via cross spectrum calculation; and
storing only relevant cross-channel coherences in an array before moving to a subsequent frequency.