Wearable epileptic seizure prediction and alert glasses based on neuromorphic computing

Abstract

A wearable device for real-time monitoring and predicting epileptic seizures. The wearable device comprises a plurality of dry Electroencephalogram (EEG) electrodes to capture EEG signals of a subject. The wearable device further comprises a neuromorphic computing unit configured to analyse the captured EEG signals in real-time using neural network models trained on epileptic seizure data to detect abnormalities in brain activity of the subject. These abnormalities indicate preictal phase of epileptic seizure. The wearable device further comprises an alert mechanism to notify the subject upon detection of abnormalities in brain activity.

Description

WEARABLE EPILEPTIC SEIZURE PREDICTION AND ALERT GLASSES BASED ON NEUROMORPHIC COMPUTINGFIELD OF INVENTION

[0001] The subject matter in general relates to the field of medical diagnostics, and in particular to a wearable device and method for real-time monitoring and prediction of epileptic seizures.BACKGROUND OF THE INVENTION

[0002] In the rapidly advancing field of medical diagnostics, monitoring brain activity is crucial for conditions like epilepsy, where an early epileptic seizure prediction can significantly enhance a patient’s well-being. Epileptic seizures are generally caused by sudden bursts of abnormal electrical activity in the brain. The seizures pose serious risks, including loss of consciousness and long-term neurological damage. While various Electroencephalogram (EEG) related monitoring systems exist, many rely on bulky, stationary equipment or impractical wearable devices that are unsuitable for continuous use and often fail to deliver real-time alerts. Conventional systems, such as hospital-based EEGs, provide reliable detection but require patients to be hospitalised and lack early prediction capabilities. Portable ambulatory EEGs are cumbersome and mainly detect seizures at an ictal stage of the seizure (or a middle stage of a seizure event), whereas the home-use wearable EEGs typically use electrodes that require gels for better conductivity to capture signals and generate alerts based on the detected EEG signals after the onset of seizure. As the placement of gel is crucial for capturing the signals accurately, it always requires professional assistance. Such home-use wearable EEGs are bulkier in design.

[0003] US20230248301A1 discloses a wearable sensor-based device for predicting, monitoring, and controlling epilepsy. The device provides real-time, personalized monitoring by collecting physiological data through multiple sensors, processing it with advanced algorithms, and issuing alerts. The device is equipped with sensors for monitoring blood oxygen, temperature, ECG, acceleration, and glucose positioned on various body parts. The sensor data is used to determine seizure likelihood, the accuracy of such prediction are, in general, not completely reliable as changes in heart rate and glucose levels can be due to any other medical condition like diabetes and not necessarily due to epileptic seizures. The data from specific regions of the brain is proved to more accurate in predicting seizures even during the preictal phase, which would provide the required time to take necessary precautions. Additionally, such techniques in general use gel-based electrodes to detect the signals from various body parts, which tend to leave a residue after every use and are inconvenient for the user. The system is resource intensive as it is usingmultiple sensors’ data, and it mandates the data storage and analysis on the remote server to predict seizures. The system requires a reliable network connection at all times, which in reality is not always a possibility, particularly in case of a wearable device due to its portable nature. A technique to analyse the data locally on the device will be more suitable and reliable for a wearable device configured to predict seizures.

[0004] Another non-patent prior art publication titled “A neuromorphic spiking neural network detects epileptic high frequency oscillations in the scalp EEG”, discloses a neuromorphic Spiking Neural Network (SNN) device designed for immediate detection of seizures by identifying High-Frequency Oscillations (HFOs) in the 80-250 Hz range, which are associated with ictal (seizure) phase. The SNN device discloses a technique for detecting in-the-moment seizure using leaky integrate-and-fire neuron models that are optimized to recognize high-frequency signals for rapid response but is limited to the ictal phase alone. Detecting the onset of seizure will not provide sufficient time for the caregivers to take the required safety measures for the user. The document discloses providing quick response during the onset of seizures but fails to provide the required time to take necessary precautions.

[0005] CN114550907A discloses an epilepsy detection system using multimodal data. The epilepsy detection system uses a data acquisition system and a cloud server for gathering and analysing real-time data. The system employs a recurrent neural network to analyse dynamic data, focusing on time-series correlations between different signals (heart rate, PPG signals, etc.). Feedback from users about warning accuracy (true / false alarms) is used by the cloud server to fine-tune the model for individual users. Similar to US20230248301A1, the system is resource intensive as it is using multiple sensors’ data, and it mandates the data storage and analysis on the cloud server to predict seizures. The use of recurrent neural network slows down the process of analysis, since real-time data cannot be analysed parallelly in recurrent neural networks. The system requires a reliable network connection at all times, which in reality is not always a possibility, particularly in case of a wearable device due to its portable nature. A technique to analyse the data locally on the device will be more suitable and reliable for a wearable device configured to predicting seizures.

[0006] The Non-EEG wearables, like motion sensors, capture physical symptoms alone and are ineffective for non-motor seizures. Overall, the existing solutions do not meet the critical need for accurate monitoring, real-time analysis and prediction of seizure in a practical, user-friendly and non-invasive form. Additionally, these techniques fail to predict early seizure i.e. during the preictal phase which could significantly reduce the risks associated with seizures by providing early warnings and allowing individuals to take preventative actions. Further, existing systems failto seamlessly integrate with advanced prediction algorithms, making them ineffective for real-time use in everyday life for predicting seizures in the preictal phase instead of ictal phase of the epileptic seizure.

[0007] In view of the foregoing, there is a need for a portable, user-friendly, and discrete device to detect preictal biomarkers and accurately predict the seizure onset well in advance. There is a need to focus on preictal phases instead of ictal or interictal biomarkers using less resources for real-time local processing of sensor data in order to provide a valuable window for preventive interventions. The device must be able to integrate seamlessly into daily life, offering continuous monitoring and timely alerts while ensuring user’s comfort. Additionally, the device must be capable of providing physical alerts to the subject, ensuring immediate attention to potential seizure activity. This way of sending notifications to both the subject as well as their respective caregivers helps in enhancing safety and responsiveness in comparison to conventional EEG monitoring.SUMMARY

[0008] A wearable device for real-time monitoring and predicting epileptic seizures is disclosed. The wearable device comprises a plurality of dry Electroencephalogram (EEG) electrodes to capture EEG signals of a subject. The wearable device further comprises a neuromorphic computing unit configured to analyse the captured EEG signals in real-time using neural network models trained on epileptic seizure data to detect abnormalities in brain activity of the subject. These abnormalities indicate preictal phase of epileptic seizure. The wearable device further comprises an alert mechanism to notify the subject upon detection of abnormalities in brain activity.BRIEF DESCRIPTION OF DIAGRAMS

[0009] Exemplary embodiments of the present invention will be understood and appreciated more fully from the following detailed description, taken in conjunction with the drawings in which,

[0010] FIG. 1 illustrates a block diagram of a wearable device 100, in accordance with an embodiment;

[0011] FIG. 2 A illustrates perspective view of the wearable device 100, in accordance with an embodiment;

[0012] FIG. 2B-2D illustrates exploded view of the wearable device 100, in accordance with an embodiment;

[0013] FIG. 3 illustrates an exemplary representation of a 10-10 electrode placement system according to the International Federation of Clinical Neurophysiology (IFCN);

[0014] FIG. 4 is a flowchart 400 depicting a method of working of the wearable device 100, in accordance with an embodiment; and

[0015] FIG. 5 is a flowchart 500 depicting a method of generating EEG dataset, in accordance with an embodiment.DETAILED DESCRIPTION OF THE INVENTION

[0016] The following detailed description includes references to the accompanying drawings, which form part of the detailed description. The drawings show illustrations in accordance with example embodiments. The numerals in the figure represent like elements throughout the several views, exemplary embodiments of the present disclosure are described. For convenience, only some elements of the same group may be labelled with numerals. The purpose of the drawings is to describe exemplary embodiments and not for production. Therefore, features shown in the figures are chosen for convenience and clarity of presentation only. Moreover, the language used in this disclosure has been principally selected for readability and instructional purposes and may not have been selected to delineate or circumscribe the inventive subject matter, resort to the claims being necessary to determine such inventive subject matter. These example embodiments are described in enough detail to enable those skilled in the art to practice the present subject matter. However, it will be apparent to a person of ordinary skill in the art that the present invention may be practiced without these specific details. The embodiments can be combined, other embodiments can be utilized, or structural and logical changes can be made without departing from the scope of the invention. The following detailed description is, therefore, not to be taken in a limiting sense.

[0017] In this document, the terms “a” or “an” are used, as is common in patent documents, to include one or more than one. In this document, the term “or” is used to refer to a non-exclusive “or” such that “a or b” includes “a but not b,” “b but not a,” and “a and b,” unless otherwise indicated.

[0018] Reference in the specification to "one embodiment" or to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiments is included in at least one embodiment of the invention, and multiple references to "one embodiment" or "an embodiment" should not be understood as necessarily all referring to the same embodiment.

[0019] FIG. 1 illustrates a block diagram of a wearable device 100. The wearable device 100 is an eyewear or any other non-invasive device like a frame for wearing glasses. The wearable device 100 may be a light weight, everyday wearable frame designed as standard eyewear. The wearable device 100 may be discrete, portable and ergonomically suitable for extended wear. Thewearable device 100 is designed to be suitable for continuous use without posing any discomfort or visibility issues to a subject using the wearable device 100. The wearable device 100 is configured to predict epileptic seizure at preictal stage and provide an alert for the subject based on a neuromorphic computation conducted locally on the device in real-time.

[0020] The wearable device 100 may comprise a plurality of dry Electroencephalogram (EEG) electrodes 102, a signal acquisition module 104 and a Microcontroller unit (MCU) 106. The plurality of dry Electroencephalogram (EEG) electrodes 102, the signal acquisition module 104 and the Microcontroller unit (MCU) 106 are communicably connected through the bus 108. The plurality of dry EEG electrodes 102 may be configured to capture EEG signals from a brain of the subject. The signal acquisition module 104 may be configured to acquire the EEG signals through the plurality of dry EEG electrodes 102. The MCU 106 may comprise a signal processing unit 1060, a communication unit 1062, a power supply 1064, a charging interface 1066, an alert mechanism 1068 and a neuromorphic computing unit 1069. The MCU 106 may be configured to process the captured EEG signal for the purpose of predicting an epileptic seizure at a preictal stage.

[0021] FIG. 2A illustrates a perspective view of a wearable device 100. Referring to the figure, the wearable device 100 may comprise the plurality of dry Electroencephalogram (EEG) electrodes 102, a first part 202, a second part 204 and a third part 206 connecting the first part 202 and the second part 204.

[0022] The first part 202, the second part 204 and the third part 206 may be arranged in such a manner that the wearable device 100 may be used as an eyewear. The third part 206 may be configured to connect the first part 202 with the second part 204 to allow the wearable device 100 (or eyewear) to be comfortably worn by the subject as regular eyeglasses.

[0023] The plurality of dry EEG electrodes 102 may be integrated into the first part 202 and second part 204 of the wearable device 100. The placement of the plurality of dry EEG electrodes 102 may be optimized for capturing seizure-related brainwave patterns while maintaining a non- invasive design. Each of the plurality of dry EEG electrodes 102 may be made from biocompatible materials that can maintain stable contact with the skin while conducting the electrical signals associated with the captured EEG signals. The plurality of dry EEG electrodes 102 may be made from any one of Silver / Silver Chloride (Ag / AgCl), Graphite or Carbon-Based Materials, stainless steel, gold, or titanium. Alternatively, conductive Polymers like polypyrrole, polyaniline, and poly (3,4-ethylenedioxythiophene)(PEDOT) may be coated or printed onto regular electrodes to improve conductivity without using gels to thereby obtain dry EEG electrodes 102. The plurality of dry EEG electrodes 102 may be of high sensitivity and precision while capturing the EEG signals. The dry EEG electrodes 102 may be configured to eliminate the need for conductive gelthereby improving the subjects comfort and enabling long-term use. The dry EEG electrodes 102 may be configured to allow the wearable device 100 to obtain bipolar montage EEG readings from the brain of the subject.

[0024] FIG. 3 illustrates an exemplary representation of a 10-10 electrode placement system according to International Federation of Clinical Neurophysiology (IFCN) standards. The plurality of dry EEG electrodes 102 may be positioned according to the conventional 10-10 electrode placement system of IFCN standards for positioning EEG electrodes. The use of the 10- 10 electrode placement system may enable the dry EEG electrodes 102 to capture signals from target areas on a scalp of the subject that may be relevant for seizure detection such as a frontal lobe and a temporal lobe of the scalp of the subject.

[0025] FIG. 2B-2D illustrates exploded view of the wearable device 100. The first part 202 may comprise a first hollow space 208 and a first cover 210 (see FIG. 2C). The first hollow space 208 may be configured to house the signal processing unit 1060, the communication unit 1062, the alert mechanism 1068 and the neuromorphic unit 1069. The first cover 210 may comprise at least one first opening 220 to accommodate the plurality of dry EEG electrodes 102. The plurality of dry EEG electrodes 102 may be disposed on the first cover 210 such that the at least one first opening 220 may enable the plurality of dry EEG electrodes 102 to access the neuromorphic computing unit 1069. The first cover 210 may be configured to cover the first hollow space 208 such that the alert mechanism 1068 and the neuromorphic computing unit 1069 may be embedded therein.

[0026] The second part 204 may comprise a Second hollow space 212 and a second cover 214. The second hollow space 212 may be configured to house the power supply 1064 and the charging interface 1066. The power supply unit 1064 and the charging interface 1066 may be connected to each using at least one of the plurality of flexible cables 224. The second cover 214 may comprise at least one second opening 222 to accommodate the plurality of dry EEG electrodes 102. The plurality of dry EEG electrodes 102 may be disposed on the second cover 214 such that the at least one second opening 222 may enable plurality of dry EEG electrodes 102 to be connected to the neuromorphic computing unit 1069. The second cover 214 may be configured to cover the Second hollow space 212 such that the power supply 1064 and the charging interface 1066 may be embedded therein.

[0027] Referring to FIG. 2C, the third part 206 may comprise a third hollow space 216 and a third cover 218. The third hollow space 216 may be configured to house at least one of the plurality of flexible cables 224 that may be configured to establish an electrical connection between the components embedded within in the first part 202 and the components embedded within the second part 204. Specifically, the plurality of flexible cables 224 disposed in the third part 206 may beconfigured to enable the power supply 1064 in the second part 204 to supply electrical power to the rest of the hardware components in the wearable device 100 via the flexible cables 224.

[0028] In an embodiment, the power supply 1064 may be an inbuilt battery to power the wearable device 100. The inbuilt battery may be a rechargeable battery such as Lithium Polymer (Li-Po) battery, Li-Ion Battery and nickel metal hydride (Ni-MH) battery, among other battery compositions. The power supply 1064 may support continuous EEG monitoring and neuromorphic processing for extended periods of time on a single charge. The power supply 1064 may be configured to be electrically recharged via the charging interface 1066.

[0029] In an embodiment, the charging interface 1066 may include but not limited to a USB - C charging port, a wireless charging or pogo pins. The charging interface 1066 may be configured to charge the rechargeable power supply 1064 when connected to an external electrical supply unit via the charging port.

[0030] Furthermore, the third cover 218 may be configured to cover the third hollow space 216 such that at least one of the flexible cables 224 may be embedded seamlessly within the third hollow space 216. The third cover 218 may be an aesthetic appeal design to enhance the subject’s comfort while using the wearable device 100 for longer periods.

[0031] In an embodiment, a first set of the plurality of dry EEG electrodes 102 may be disposed on the frontal lobe of the brain of the subject, and a second set of the plurality of dry EEG electrodes 102 may be disposed on the temporal lobe of the brain of the subject.

[0032] Referring to FIG. 3, the first set of EEG electrodes 102 disposed on the frontal lobe may be positioned at points F7 and F8 of the 10-10 EEG electrode placement system. The points F7 and F8 may enable each of the plurality of dry EEG electrodes 102 to capture EEG signal from the frontal lobe of the brain. The position of the points F7 and F8 on the subject’s brain may be associated with activity from the inferior frontal gyrus and parts of the prefrontal cortex. The location of points F7 and F8 may be configured to provide critical insights into frontal lobe activity, which is essential in detecting frontal lobe seizures, often characterized by motor phenomena.

[0033] The second set of the plurality of dry EEG electrodes 102 disposed on the temporal lobe may be positioned at points T7 and T8 of the 10-10 EEG electrode placement system. The points T7 and T8 may enable the second set of the plurality of dry EEG electrodes 102 to capture EEG signals from the temporal lobe of the brain. The position of the points T7 and T8 on the subject’s brain may be associated with superior temporal gyrus and other temporal structures of the brain. The position of the points T7 and T8, covering the temporal lobe regions may be associated with temporal lobe epilepsy (TLE), which is among the most prevalent types of focal epilepsy. The temporal lobes may contribute to emotional processing, social behavior, memory, and language, and often exhibit early seizure signatures, particularly during preictal periods. Earlyshifts in EEG activity associated with preictal brain dynamics may be captured by monitoring EEG signals obtained from the plurality of dry EEG electrodes 102, thereby enhancing prediction accuracy of the wearable device 100.

[0034] FIG. 4 is a flowchart 400 depicting a method of working of the wearable device 100. At step 402, the plurality of dry EEG electrodes 102 are configured to capture EEG signals from the subject’s brain based on the position of each of the plurality of dry EEG electrodes 102 according to the 10-10 system of IFCN standards for EEG monitoring. The signal acquisition module 104 may comprise an analog to digital convertor (ADC) for capturing the EEG signal. The ADC may provide high-resolution, multi-channel EEG signal processing. The ADC may interface directly with the plurality of dry EEG electrodes 102. The ADC may ensure reliable real-time data acquisition despite environmental challenges, physiological changes and motion. Also, the ADC may amplify and digitize the captured EEG signals with high accuracy while minimizing noise.

[0035] At step 404, the MCU 106 may be configured to prepare an EEG dataset based on the captured EEG signals. The captured EEG signals may comprise noise from various sources including muscle movements, eye blinks, and environmental interference in the form of artifacts.

[0036] FIG. 5 represents a flowchart for preparing the EEG dataset. At step 502, the signal processing unit 1060 of the MCU 106 may be configured to remove artifacts in the captured EEG signal. Further, the signal processing unit 1060 may be configured to separate the EEG signal into independent components to thereby isolate the noise sources like muscle movements or eye blinks from the brain's electrical activity. At step 504, the signal processing unit 1060 may apply bandpass filtering to eliminate low-frequency drifts and high-frequency noise from the captured EEG signal. At step 506, the signal processing unit 1060 may be configured to extract features of the captured EEG signal from the filtered EEG signal for identifying important characteristics associated with seizure activity. The feature extraction process may include usage of Short-Time Fourier Transform (STFT). The STFT may be used to transform the EEG signal from a time domain to a time-frequency domain. At step 508, a Hanning window may be applied to divide the EEG data in the time-frequency domain into small segments. The Fourier transform for each segment may be computed for generating EEG dataset in the form of a spectrogram. This spectrogram may reveal how the signal's frequency components evolve over time.

[0037] Seizure activity may often exhibit distinct frequency patterns, such as increased power in the delta (0.5-4 Hz) and theta (4-8 Hz) bands, or sharp spikes. The STFT may enable real-time tracking of the distinct frequency patterns by differentiating between seizure frequency pattern and non-seizure frequency pattern, thereby enabling the MCU 106 to detect when abnormal brain activity which represents begins of the preictal phase of seizure.

[0038] Further, the MCU 106 is configured to calculate power in specific frequency bands and compare it to a baseline brain activity. Sudden changes in power may be detected to indicate the onset of a seizure.

[0039] Referring to Fig. 4, At step 406, the neuromorphic computing unit 1069 may be configured to process the EEG dataset corresponding to the captured EEG signal using neural network training model. The neural network models may be trained on epileptic seizure data to detect abnormalities in brain activity of the subject. The abnormalities may indicate preictal phase of the epileptic seizure.

[0040] In an embodiment, the neuromorphic computing unit 1069 may include a neuromorphic chip 1069a. The neuromorphic computing unit 1069 may be configured to implement a neural network training model (NNTM) on the neuromorphic chip 1069a. The neuromorphic chip 1069a may enable low-power, real-time processing of neural signals.

[0041] In an embodiment, the neuromorphic computing unit 1069 may be configured for training a classification Convolutional Neural Network (CNN) model based on the generated EEG datasets. The CNN model may be configured to classify the EEG dataset based on the time domain features as well as the frequency domain features. The EEG dataset may be subjected to further adapt the trained CNN model into a Spiking Neural Network (SNN) model using a cross -platform model.

[0042] In an embodiment, the SNN model may be configured to detect seizures at the preictal phase by identifying subtle shifts in signal amplitude, phase or connectivity across F7, F8, T7 and T8. The neuromorphic chip’s 1069a’s capabilities may be harnessed to process the EEG dataset efficiently in real-time. The neuromorphic chip 1069a may be configured to mimic the functioning of neurons of the subject’s brain (like the way the neurons in the brain fire in response to stimuli). The neuromorphic chip 1069a may include a machine learning model such as the SNN model, that is trained with seizure frequency patterns and non- seizure frequency patterns. This pretraining may enable the machine learning model to recognize complex seizure patterns with high accuracy while consuming minimal power. The SNN may continuously analyze the incoming EEG dataset, searching for brainwave patterns that correspond to the onset of a seizure at any instant. The SNN model may assess gradual changes in the frequency pattern of the EEG dataset. The gradual changes may include shifts in phase coherence as well as synchrony between any two of the plurality of dry EEG electrodes 102. The SNN model may be configured to assess patterns over an extended window before seizure onset to detect preictal biomarkers in real-time. The neuromorphic chip 1069a may be configured to provide rapid processing, while operating in realtime and may help in maintaining low-power profile that is critical for wearables.

[0043] At Step 408, the neuromorphic chip 1069a may be configured to apply Artificial Intelligence (Al) techniques to check for any abnormalities in brain activity. The Al techniques used may be at least one of Bayesian Networks, Expert Systems, Computer Vision or Neuro- Symbolic Al. The Al technique may assist in analysing EEG dataset of each subject by adapting to the unique patterns and nuances of an individual's seizure precursors. In the event, the neuromorphic chip 1069a detects an abnormal brain activity by matching the current EEG dataset with previously learned seizure patterns, it may notify the subject through the alert mechanism 1068 at step 410. The neuromorphic chip 1069a enables prediction of a potential epileptic seizure, and at step 410, the alert mechanism 1068 notifies the subject and / or the caregiver during the preictal phase, i.e. up to an hour in advance. The alert mechanism 1068 may be configured to alert the subject in real-time.

[0044] In an embodiment, the alert mechanism 1068 may comprise at least one of a haptic feedback, visual indicator, and mobile alerts. For example, the haptic feedback may be configured to provide vibrations near the back of the ear of the subject when an abnormality in brain activity is detected. The visual indicator may be one or more LED lights 226 that may be configured to glow when an abnormality in brain activity is detected. The alert mechanism 1068 may be configured to use the haptic feedback to physically notify the subject upon detection of the abnormalities in brain activity. The intensity of the haptic feedback is proportionally increased based on at least one of the degree of abnormalities and as the seizures approach an ictal phase from the preictal phase.

[0045] In an embodiment, the wearable device 100 may be configured to connect seamlessly via Bluetooth to the subject’s smartphone, where a dedicated application may alert the subject when a "pre-seizure" stage is detected, potentially during the preictal phase i.e. up to an hour in advance. For subjects susceptible to Photosensitive Epilepsy and Photo Paroxysmal Response, the wearable device 100 may be equipped with a forward-facing sensor that may detect flashing lights and promptly notify the subject of such stimuli. Additionally, integrated haptics may provide physical alerts to the subject, ensuring immediate attention to potential seizure activity. Notifications may be sent to both the subject and designated caregivers, enhancing safety and responsiveness.

[0046] At step 408, if the abnormal brain activity is not detected, then the neuromorphic chip 1069a continues to monitor the EEG signals for any abnormal activities in real time.

[0047] The wearable device 100 is advantageous in its low-power design to ensure extended battery life in real-time monitoring and prediction of epileptic seizures. The high-sensitivity dry EEG electrodes 102, placed according to the 10-10 system, may enable non-invasive, comfortable, and continuous EEG monitoring without disrupting the subjects’ daily activities or causinginconvenience like traditional wet electrodes. With its ability to predict seizures in real-time, the wearable device 100 provides subjects and their caregivers with valuable early warnings, enabling timely intervention and potentially reducing the risk of seizure-related injuries. The ergonomical design of the wearable device 100 and machine learning algorithms ensure comfort throughout the day, allowing for continuous monitoring without disrupting daily activities. The use of USB Type- C for charging and firmware updates further enhances the usability and flexibility of the device, making it a practical solution for individuals living with epilepsy. The wearable device 100 represents a significant advancement in seizure management, combining cutting-edge technology with subject-centric design to improve the quality of life for people at risk of epileptic seizures.

[0048] The various embodiments have been described using detailed descriptions that are provided by way of example and are not intended to limit the scope of the invention. The described embodiments comprise different features, not all of which are required in all embodiments. Some embodiments utilize only some of the features or possible combinations of the features. Many other ramification and variations are possible within the teaching of the embodiments comprising different combinations of features noted in the described embodiments.

[0049] It will be appreciated by any person skilled in the art that the various embodiments are not limited by what has been particularly shown and described herein above. Rather the scope of the invention is defined by the claims that follow.

Claims

AMENDED CLAIMS received by the International Bureau on 02 June 2026 (02.06.2025)What is claimed:

1. A wearable device for real-time monitoring and predicting epileptic seizures, the wearable device comprises: a plurality of dry Electroencephalogram (EEG) electrodes to capture EEG signals from the brain of a subject; a neuromorphic computing unit configured to analyse the captured EEG signals in realtime using neural network models trained on epileptic seizure data by performing spectral analysis to track seizure-related frequency patterns up to one hour in advance to detect abnormalities in brain activity of the subject, wherein the abnormalities indicate preictal phase of epileptic seizures; an alert mechanism to notify the subject upon detection of abnormalities in brain activity.

2. The wearable device as claimed in claim 1, wherein a first set of the plurality of dry EEG electrodes are configured to capture the EEG signals from a frontal lobe of the brain and a second set of the plurality of dry EEG electrodes are configured to capture the EEG signals from a temporal lobe of the subject.

3. The wearable device as claimed in claim 2, wherein the frontal lobe includes F7 and F8 positions of a 10-10 electrode placement system on the subject’ s brain for capturing the EEG signals, and the temporal lobe includes T7 and T8 positions of the 10-10 electrode placement system on the subject’s brain for capturing the EEG signals.

4. The wearable device as claimed 1, further comprises a signal processing unit configured to: remove artifacts in the captured EEG signals; eliminate low-frequency drifts and high frequency noise in the captured EEG signals; and extract features of the EEG signals for generating the EEG dataset of the captured EEG signals.

5. The wearable device as claimed in claim 1, wherein the neural network model is trained with seizure frequency patterns and non-seizure frequency patterns associated with the subject to generate trained seizure patterns, and wherein the trained seizure patterns and the captured EEG signals are analyzed to detect abnormalities in brain activity.

6. The wearable device as claimed in claim 5, wherein the neuromorphic computing unit comprises a neuromorphic chip configured to: assess gradual change in the frequency patten of the captured EEG signals, that includes:assess shifts in phase coherence; assess synchrony between the captured EEG signals obtained from the plurality of dry EEG electrodes; and enable real-time detection of preictal biomarkers.

7. The wearable device as claimed in claim 6, wherein the SNN is configured to process captured EEG signals in real-time to predict the epileptic seizure up to one hour in advance.

8. The wearable device as claimed in claim 1, wherein the alert mechanism uses a haptic feedback to physically notify the subject upon detection of abnormalities in brain activity.

9. The wearable device as claimed in claim 8, wherein the intensity of the haptic feedback is proportionally increased based on at least one of the degree of abnormalities and as the seizures approach an ictal phase from the preictal phase.

10. The wearable device as claimed in claim 1, wherein the wearable device is an eyewear.

11. A method for real-time monitoring and predicting epileptic seizures using a wearable device, the method comprising: capturing Electroencephalogram (EEG) signals from a subject using a plurality of dry electrodes ; analyzing the captured EEG signals in real-time using a neuromorphic computing unit comprising a neural network training model trained on epileptic seizure data to detect abnormalities in brain activity of the subject, wherein the abnormalities indicate preictal phase of epileptic seizure; and notifying the subject, upon detection of abnormalities in brain activity using an alert mechanism.

12. The method as claimed in claim 11, wherein capturing Electroencephalogram (EEG) signals includes: capturing EEG signals by positioning a first set of the dry EEG electrodes on the frontal lobe of the subject’s brain; and capturing EEG signals by positioning a second set of the dry EEG electrodes on a temporal lobe of the subject’s brain.

13. The method as claimed in claim 12, wherein the first set of EEG electrodes are positioned at F7 and F8 positions corresponding to the frontal lobe of the subject in a 10-10electrode placement system, and the second set of EEG electrodes are positioned at T7 and T8 positions corresponding to the temporal lobe of the subject in the 10-10 electrode placement system.

14. The method as claimed in claim 11, further comprising: removing artifacts, using a signal processing unit, and eliminating low-frequency drifts and high frequency noise in the captured EEG signals; extracting features of the EEG signals, using the signal processing unit, for generating the EEG dataset of the EEG signals.

15. The method as claimed in claim 14, wherein preparing the EEG dataset further comprises transmitting the EEG dataset from the signal processing unit to the neuromorphic computing unit.

16. The method as claimed in claim 11 , wherein analyzing the captured EEG signals in real-time using the neuromorphic computing unit comprises: conducting a spectral analysis of the EEG dataset for tracking a frequency pattern of the EEG dataset in real-time; training a machine learning model with seizure frequency patterns and non-seizure frequency patterns associated with the subject; and detecting abnormalities in brain activity based on an analysis of the trained seizure patterns and the captured EEG signals.

17. The method as claimed in claim 16, wherein training a machine learning model with seizure frequency patterns and non-seizure frequency patterns includes assessing gradual changes in the frequency pattern of the EEG dataset, and enabling real-time signal processing using spike neural networks by detecting preictal biomarkers, and wherein assessing the gradual changes in the frequency pattern comprises assessing shifts in phase coherence, and assessing synchrony between the EEG signals captured by the plurality of dry EEG electrodes.

18. The method as claimed in claim 11, wherein the alert mechanism uses haptic feedback to physically notify the subject upon detection of abnormalities in brain activity.

19. The method as claimed in claim 18, wherein the intensity of the haptic feedback is proportionally increased based on at least one of the degree of abnormalities and as the seizures approach an ictal phase from the preictal phase.

Images