Description:Field of the Invention

The present invention relates to the field of medical techniques, specifically to a method for detecting and predicting seizures using electroencephalogram (EEG) data.
Background
The background description includes information that may be useful in understanding the present invention. It is not an admission that any of the information provided herein is prior art or relevant to the presently claimed invention, or that any publication specifically or implicitly referenced is prior art.
In the realm of neurological disorders, seizure management represents a critical area of medical research and patient care. Seizures, characterized by sudden and uncontrolled electrical disturbances in the brain, pose significant risks to individuals, necessitating accurate detection and timely intervention. The use of electroencephalogram (EEG) data for monitoring brain activity has become a cornerstone in the diagnosis and study of seizures.
Traditionally, EEG data interpretation has relied heavily on manual analysis by trained professionals, a process fraught with challenges including subjectivity and the potential for human error. Advances in digital signal processing techniques have led to the development of automated systems designed to improve the accuracy and efficiency of EEG data analysis. Such systems employ various algorithms to preprocess, filter, and analyze the electrical signals recorded during EEG sessions.
Another approach comprises the application of traditional machine learning algorithms to identify patterns indicative of seizure activity. These algorithms, while useful, often require extensive feature engineering and may struggle to handle the complexity and variability inherent in EEG data. As a result, the detection and prediction of seizures remain imperfect, with issues such as false positives and missed detections presenting significant concerns.
The advent of deep learning has introduced new possibilities for enhancing the accuracy of seizure detection and prediction models. Deep learning models, particularly convolutional neural networks (CNNs), have demonstrated exceptional ability in processing complex data structures, including images and time-series data like EEG signals. By leveraging multiple convolutional and pooling layers, these models can autonomously extract relevant features from EEG data, potentially bypassing the limitations of manual feature selection and traditional machine learning approaches.
However, the application of deep learning to seizure detection and prediction is not without its challenges. The selection of appropriate model architectures, hyperparameters, and training data plays a crucial role in the performance of these systems. Furthermore, the integration of real-time feedback from healthcare professionals into the prediction model remains an area ripe for exploration. Such feedback could further refine model accuracy, making these systems more adaptable and responsive to the nuances of individual patient data.
In light of the above discussion, there exists an urgent need for solutions that overcome the problems associated with conventional systems and techniques for the detection and prediction of seizures utilizing EEG data.
Summary
The following presents a simplified summary of various aspects of this disclosure in order to provide a basic understanding of such aspects. This summary is not an extensive overview of all contemplated aspects, and is intended to neither identify key or critical elements nor delineate the scope of such aspects. Its purpose is to present some concepts of this disclosure in a simplified form as a prelude to the more detailed description that is presented later.
The following paragraphs provide additional support for the claims of the subject application.
In an aspect, the present disclosure aims to provide a method for detecting and predicting seizures using electroencephalogram (EEG) data. The method encompasses acquiring EEG data from a subject and preprocessing the acquired data through downsampling to a predefined sampling rate. A sliding window is applied to the preprocessed EEG data with a predetermined time window and time step. The EEG data is then processed using a deep learning model comprising multiple convolutional layers followed by batch normalization and max pooling layers, and a dense layer for feature extraction. A seizure alert is generated based on a threshold confidence level determined by the output of the deep learning model. Furthermore, the method includes a feedback loop for incorporating real-time feedback from healthcare professionals to refine the prediction model. The deep learning model is trained using a binary cross-entropy loss function and an Adam optimizer. Additionally, the sliding window applied has a time window of 8 seconds and a time step of 4 seconds.
In another aspect, the present disclosure provides a seizure detection and prediction system comprising an EEG data acquisition unit configured to collect EEG signals, a preprocessing unit configured to downsample the collected EEG signals and apply a sliding window, and a deep learning processing unit that employs the model for analyzing the EEG data. An alert generation unit is configured to generate real-time alerts upon detecting seizure-indicative patterns, and a feedback interface is provided for receiving and incorporating feedback from healthcare professionals into the model. The preprocessing unit applies a band-pass filter to the EEG signals to remove noise and artifacts before downsampling. The deep learning processing unit employs a model with at least three convolutional neural network layers, each followed by batch normalization and a non-linear activation function. The sliding window applied by the preprocessing unit has a configurable time window and time step, adjusted based on the type of seizure patterns being monitored. The alert generation unit is configured to transmit alerts to an external computing device accessible to healthcare professionals or caregivers and to activate auditory or visual alarms within a healthcare facility. The system is integrated with a centralized medical record system to allow for seamless updating of patient data with seizure occurrence and prediction information.

Brief Description of the Drawings

The features and advantages of the present disclosure would be more clearly understood from the following description taken in conjunction with the accompanying drawings in which:
FIG. 1 illustrates a method (100) for the detection and prediction of seizures utilizing electroencephalogram (EEG) data, in accordance with the embodiments of the present disclosure.
FIG. 2 illustrates a block diagram of a seizure detection and prediction system (200), in accordance with the embodiments of the present disclosure.
FIG. 3 illustrates an exemplary architectural design of disclosed DeepConvEEGNet model for analysis of electroencephalogram (EEG) data for the prediction of seizure events, in accordance with embodiment of present disclosure.
FIG. 4 illustrates a graphical representation of seizure prediction performance on CHB-MIT dataset sample, in accordance with embodiment of present disclosure.
FIG. 5 illustrates a workflow of the system utilizing the EEG-based model, in accordance with embodiment of present disclosure.

Detailed Description
In the following detailed description of the invention, reference is made to the accompanying drawings that form a part hereof, and in which is shown, by way of illustration, specific embodiments in which the invention may be practiced. In the drawings, like numerals describe substantially similar components throughout the several views. These embodiments are described in sufficient detail to claim those skilled in the art to practice the invention. Other embodiments may be utilized and structural, logical, and electrical changes may be made without departing from the scope of the present invention. The following detailed description is, therefore, not to be taken in a limiting sense, and the scope of the present invention is defined only by the appended claims and equivalents thereof.
The use of the terms “a” and “an” and “the” and “at least one” and similar referents in the context of describing the invention (especially in the context of the following claims) are to be construed to cover both the singular and the plural, unless otherwise indicated herein or clearly contradicted by context. The use of the term “at least one” followed by a list of one or more items (for example, “at least one of A and B”) is to be construed to mean one item selected from the listed items (A or B) or any combination of two or more of the listed items (A and B), unless otherwise indicated herein or clearly contradicted by context. The terms “comprising,” “having,” “including,” and “containing” are to be construed as open-ended terms (i.e., meaning “including, but not limited to,”) unless otherwise noted. Recitation of ranges of values herein are merely intended to serve as a shorthand method of referring individually to each separate value falling within the range, unless otherwise indicated herein, and each separate value is incorporated into the specification as if it were individually recited herein. All methods described herein can be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The use of any and all examples, or exemplary language (e.g., “such as”) provided herein, is intended merely to better illuminate the invention and does not pose a limitation on the scope of the invention unless otherwise claimed. No language in the specification should be construed as indicating any non-claimed element as essential to the practice of the invention.
Pursuant to the "Detailed Description" section herein, whenever an element is explicitly associated with a specific numeral for the first time, such association shall be deemed consistent and applicable throughout the entirety of the "Detailed Description" section, unless otherwise expressly stated or contradicted by the context.
FIG. 1 illustrates a method (100) for the detection and prediction of seizures utilizing electroencephalogram (EEG) data, in accordance with the embodiments of the present disclosure. In the context of the present disclosure, the method relates to a series of steps undertaken for the detection and prediction of seizures utilizing electroencephalogram (EEG) data. In step (102), EEG data is acquired from a subject, involving the collection of electrical activity from the brain through the use of electrodes placed on the scalp. In step (104), following acquisition, the EEG data undergoes preprocessing, which comprises downsampling to a predefined sampling rate to reduce the data size while retaining the relevant information. A sliding window is then applied to the preprocessed EEG data in step (106), characterized by a predetermined time window and time step, facilitating the segmentation of data for further analysis. The method proceeds with the step (108), processing of EEG data using a deep learning model. Step (110) is composed of multiple convolutional layers, each succeeded by batch normalization and max pooling layers, in addition to a dense layer for the extraction of features significant to seizure detection and prediction. Upon processing, a seizure alert is generated based on a threshold confidence level determined by the output of the deep learning model. This alert signifies the detection or prediction of a seizure event, prompting appropriate response actions. Additionally, in step (112) incorporates a feedback loop for integrating real-time feedback from healthcare professionals. Such feedback is utilized to refine the prediction model, enhancing its accuracy and reliability. The integration of professional insight allows for the continuous improvement of the model based on practical experience and observations, thereby contributing to the effectiveness of the seizure detection and prediction method.
In an embodiment, the deep learning model employed in the method for seizure detection and prediction using electroencephalogram (EEG) data is trained using a binary cross-entropy loss function and an Adam optimizer. The binary cross-entropy loss function, also known as log loss, measures the performance of a classification model whose output is a probability value between 0 and 1. The loss increases as the predicted probability diverges from the actual label, making it ideal for binary classification tasks such as seizure detection, where the outcomes are either the presence or absence of a seizure. This choice of loss function ensures that the deep learning model is accurately penalized for incorrect predictions, thereby improving its ability to distinguish between seizure and non-seizure events. Additionally, the Adam optimizer is utilized for updating the network weights iteratively based on training data. The Adam optimizer, known for its efficiency in handling sparse gradients and its adaptiveness to large datasets, enhances the model's learning process. By adjusting the learning rate dynamically, the Adam optimizer facilitates faster convergence of the deep learning model, improving the efficiency and accuracy of the seizure prediction and detection method. This embodiment highlights the importance of selecting an appropriate loss function and optimizer for training the deep learning model, ensuring the model's effectiveness in accurately predicting and detecting seizure events.
In another embodiment, the sliding window applied to the preprocessed EEG data within the method for seizure detection and prediction is characterized by a time window of 8 seconds and a time step of 4 seconds. The application of a sliding window is a critical step in segmenting the continuous EEG data into manageable portions for analysis by the deep learning model. The choice of an 8-second time window provides a substantial duration for capturing relevant EEG patterns associated with seizures, allowing the model to analyze enough data to identify potential seizure activity accurately. Moreover, the 4-second time step ensures that the windows overlap, allowing for a continuous and comprehensive analysis of the EEG data without missing critical information between segments. This overlap increases the chances of detecting seizures early and accurately by ensuring that no potential seizure-indicative patterns are overlooked due to segmentation. The specific configuration of the sliding window, with its predetermined time window and step, plays a crucial role in optimizing the seizure detection process, balancing the need for detailed data analysis with the practicalities of processing time and computational resources. This embodiment demonstrates the careful consideration required in selecting the parameters of the sliding window to maximize the efficacy of the seizure detection and prediction method.
The term "seizure detection and prediction system" as used throughout the present disclosure refers to an apparatus designed for the monitoring and analysis of electroencephalogram (EEG) signals to detect and predict seizures. The system comprises an EEG data acquisition unit, a preprocessing unit, a deep learning processing unit, an alert generation unit, and a feedback interface.
The EEG data acquisition unit is configured to collect EEG signals from an EEG helmet worn by a user. This unit ensures the continuous and accurate recording of electrical activity generated by the user’s brain, facilitating the detection of abnormal patterns indicative of seizures. The collection of EEG signals is the first critical step in the process of seizure detection and prediction, providing the raw data necessary for further analysis.
Following acquisition, the preprocessing unit is responsible for the downsampling of the collected EEG signals to a predefined sampling rate and the application of a sliding window. Downsampling reduces the data volume, making it more manageable for processing while retaining the essential information needed for accurate analysis. The sliding window segments the continuous EEG signals into fixed intervals, enabling the systematic analysis of the data over time. This step is crucial for preparing the EEG signals for detailed examination by the deep learning processing unit.
The deep learning processing unit employs a sophisticated model for analyzing the preprocessed EEG data. This unit leverages advanced algorithms to identify patterns within the EEG signals that are indicative of seizure activity. By processing the data through multiple layers of analysis, the deep learning model is capable of extracting relevant features and making accurate predictions regarding the likelihood of a seizure.
Upon the detection of seizure-indicative patterns by the deep learning processing unit, the alert generation unit is activated to generate real-time alerts. This unit is configured to promptly inform healthcare professionals or caregivers of the potential or occurring seizure, enabling immediate intervention. The generation of timely alerts is vital for the effective management of seizures, potentially reducing the risk of harm to the user.
Additionally, the system includes a feedback interface for receiving and incorporating feedback from healthcare professionals into the model. This interface allows for the continuous improvement of the system’s accuracy and reliability. By integrating real-world insights and observations, the feedback interface ensures that the model remains adaptive and responsive to the nuanced needs of users and healthcare providers.
FIG. 2 illustrates a block diagram of a seizure detection and prediction system (200), in accordance with the embodiments of the present disclosure. In said system, an EEG data acquisition unit (202) is tasked with the collection of EEG signals. Said unit ensures the capture of brain activity data necessary for the subsequent detection and prediction of seizures. A preprocessing unit (204) is connected to the EEG data acquisition unit (202). Said preprocessing unit (204) is configured to downsample the EEG signals to a predetermined sampling rate and apply a sliding window with fixed time parameters, facilitating the preparation of the data for analysis. A deep learning processing unit (206) is included, which employs a sophisticated model comprising multiple convolutional layers. Each layer in the model is followed by batch normalization and max pooling layers, and includes a dense layer dedicated to the extraction of features from the EEG data. An alert generation unit (208) is configured to issue real-time alerts when the model detects patterns indicative of seizures, based on a predefined threshold confidence level. Additionally, a feedback interface (210) is incorporated for the purpose of receiving and integrating real-time feedback from healthcare professionals. Said feedback interface (210) is integral for refining the deep learning model to enhance its predictive accuracy. Collectively, the components of the seizure detection and prediction system (200) interact to provide a comprehensive solution for the monitoring and management of seizure-prone patients.
In an embodiment of the seizure detection and prediction system (200), the preprocessing unit (204) is enhanced with the capability to apply a band-pass filter to the EEG signals prior to downsampling. This band-pass filter is designed to remove noise and artifacts from the EEG signals, thereby ensuring that the data fed into the deep learning processing unit (206) is of the highest possible quality. Noise and artifacts in EEG data, which can arise from various sources including electrical interference or movement by the user, can significantly degrade the accuracy of seizure detection and prediction. By filtering out frequencies that do not correspond to brain activity relevant to seizure detection, the band-pass filter effectively enhances the signal-to-noise ratio. This preprocessing step is crucial for the accurate analysis of EEG signals, as it enables the deep learning model to focus on the most pertinent features of the data that are indicative of seizure activity.
In another embodiment, the system (200) features a deep learning processing unit (206) which employs a model with at least three convolutional neural network (CNN) layers. Each of these layers is followed by batch normalization and a non-linear activation function. The incorporation of multiple CNN layers allows for the hierarchical extraction of features from the EEG signals, with each layer capturing increasingly complex patterns. Batch normalization facilitates the stabilization of the learning process by normalizing the inputs to each layer, thus accelerating the training of the deep learning model. The use of a non-linear activation function introduces the ability to capture non-linear relationships in the data, which is essential for modeling the complex dynamics of brain activity related to seizures. This architecture enhances the system's ability to accurately detect and predict seizures by leveraging deep learning's powerful feature extraction capabilities.
In yet another embodiment, the preprocessing unit (204)'s sliding window is made configurable, with adjustable time window and time step parameters. This adaptability allows the system (200) to be fine-tuned based on the specific type of seizure patterns being monitored. Different seizure types may exhibit distinct temporal characteristics, necessitating different window sizes and steps for optimal detection and analysis. By adjusting these parameters, the system (200) can be optimized for the detection of a wide range of seizure types, improving its versatility and efficacy in different clinical scenarios.
In an embodiment, the alert generation unit (208) of the system is configured to transmit alerts to an external computing device accessible to healthcare professionals or caregivers. This configuration ensures that those responsible for the care of the user can receive timely notifications about potential or actual seizure events, facilitating prompt intervention. The ability to transmit alerts to external devices enhances the practical utility of the system, integrating it into the broader ecosystem of healthcare communication and response mechanisms.
In another embodiment, the alert generation unit (208) is also capable of activating auditory or visual alarms within a healthcare facility. This feature allows for immediate awareness of seizure events among on-site healthcare personnel, ensuring a swift response. The use of auditory and visual alarms serves as an effective method for alerting staff in environments where immediate action can be critical to the well-being of the user.
In yet another, the system (200) is integrated with a centralized medical record system, allowing for the seamless updating of patient data with information on seizure occurrence and prediction. This integration facilitates the comprehensive documentation of seizure events, enhancing patient care through the availability of detailed seizure activity records. The ability to automatically update patient records with seizure data reduces the administrative burden on healthcare professionals and ensures that patient records are always current, supporting informed decision-making in clinical settings.
The system (200) utilizes deep learning model developed to accurately classify seizures by analyzing EEG signals. The deep learning model can be associated with custom-tailored convolutional layers that are highly effective at discerning complex temporal seizure patterns from EEG data. The deep learning model distinguishes between seizure and non-seizure events with high precision, thereby aiding physicians in swift and individualized patient care. The predictive capability of system (200) for early seizure detection, a threshold-based alerting system for timely intervention, and compatibility with existing EEG systems, making easily integrable into clinical workflows. The system enhances seizure detection, improves patient safety by enabling early response, and supports personalized care through adaptable algorithms tailored to individual patient profiles.
FIG. 3 illustrates an exemplary architectural design of disclosed DeepConvEEGNet model for analysis of electroencephalogram (EEG) data for the prediction of seizure events, in accordance with embodiment of present disclosure. Said architecture comprises a sequential convolutional neural network (CNN), which is constituted by three convolutional blocks. Each block contains an increasing number of filters—64, 128, and 256 respectively—optimized for the extraction of features at varying levels of abstraction from EEG signals. Post convolution, each layer applies batch normalization to maintain stable learning dynamics throughout the network. Subsequent to the convolutional layers, max pooling layers serve to reduce spatial dimensionality, thus lowering the computational complexity and processing requirements of the model. Within the disclosed architecture, following the convolutional and pooling layers, a global average pooling layer condenses the feature maps into a single vector. Such reduction in dimensions is purposed to decrease the overall count of parameters within the model, aiding in the prevention of overfitting and enhancing the generalization capability of the system. A densely connected layer, equipped with 256 neurons utilizing the ReLU activation function, follows. This layer facilitates the processing of high-level feature reasoning. Further regularization is achieved through the application of batch normalization and a dropout layer, the latter of which operates with a dropout rate of 0.5 to bolster the network's resistance to overfitting. Concluding the model, a dense output layer comprising a singular neuron utilizing a sigmoid activation function is employed to yield a binary classification output that signifies the likelihood of seizure occurrence within the input EEG data.
FIG. 4 illustrates a graphical representation of seizure prediction performance on CHB-MIT dataset sample, in accordance with embodiment of present disclosure. The DeepConvEEGNet achieved exemplary precision, recall, and F1-score, each with a weighted average of 0.99, as well as a total accuracy of 0.99. Such metrics manifest an admirable balance between the model's sensitivity and specificity, accrediting the network with the successful categorization of 99% of seizure and non-seizure instances. A Receiver Operating Characteristic (ROC) score of 0.9 further elucidates the model's adeptness at discriminating between seizure and non-seizure phases. A graphical representation delineates the predictions of the DeepConvEEGNet in a simulated real-time operation, with the model's prediction confidence indicated by orange markers when exceeding the predefined threshold of 0.5, a key component for the system's function as an early warning apparatus.
FIG. 5 illustrates a workflow of the system utilizing the EEG-based model, in accordance with embodiment of present disclosure. The workflow of the system utilizing the EEG-based model for real-time seizure anticipation begins with the data acquisition phase, wherein EEG data is collected via a sensor array situated upon the scalp. Subsequent preprocessing involves the downsampling of the signals to a frequency of 128Hz, followed by the application of a sliding window technique that utilizes an 8-second temporal window with a 4-second step for sequential data analysis. The DeepConvEEGNet then proceeds to analyze the EEG data during the model inference stage, resulting in predictions of potential seizure activity. The threshold-based alerting phase ensues, whereby the system emits warnings if the model's predictive confidence repeatedly surpasses the preset threshold, thereby signaling an impending seizure. Conversely, the absence of such threshold exceedance implies a lack of imminent seizure risk. Alert generation mechanisms then notify relevant healthcare entities in the event of seizure indication. The final stage involves a feedback loop, allowing for real-time input from healthcare practitioners aimed at the enhancement and optimization of both the predictive model and its corresponding alert parameters. Said technology is designed for the prompt and precise detection of seizures to facilitate swift medical response.
Example embodiments herein have been described above with reference to block diagrams and flowchart illustrations of methods and apparatuses. It will be understood that each block of the block diagrams and flowchart illustrations, and combinations of blocks in the block diagrams and flowchart illustrations, respectively, can be implemented by various means including hardware, software, firmware, and a combination thereof. For example, in one embodiment, each block of the block diagrams and flowchart illustrations, and combinations of blocks in the block diagrams and flowchart illustrations can be implemented by computer program instructions. These computer program instructions may be loaded onto a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions which execute on the computer or other programmable data processing apparatus create means for implementing the functions specified in the flowchart block or blocks.
Throughout the present disclosure, the term ‘processing means’ or ‘microprocessor’ or ‘processor’ or ‘processors’ includes, but is not limited to, a general purpose processor (such as, for example, a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, a microprocessor implementing other types of instruction sets, or a microprocessor implementing a combination of types of instruction sets) or a specialized processor (such as, for example, an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP), or a network processor).
The term “non-transitory storage device” or “storage” or “memory,” as used herein relates to a random access memory, read only memory and variants thereof, in which a computer can store data or software for any duration.
Operations in accordance with a variety of aspects of the disclosure is described above would not have to be performed in the precise order described. Rather, various steps can be handled in reverse order or simultaneously or not at all.
While several implementations have been described and illustrated herein, a variety of other means and/or structures for performing the function and/or obtaining the results and/or one or more of the advantages described herein may be utilized, and each of such variations and/or modifications is deemed to be within the scope of the implementations described herein. More generally, all parameters, dimensions, materials, and configurations described herein are meant to be exemplary and that the actual parameters, dimensions, materials, and/or configurations will depend upon the specific application or applications for which the teachings is/are used. Those skilled in the art will recognize, or be able to ascertain using no more than routine experimentation, many equivalents to the specific implementations described herein. It is, therefore, to be understood that the foregoing implementations are presented by way of example only and that, within the scope of the appended claims and equivalents thereto, implementations may be practiced otherwise than as specifically described and claimed. Implementations of the present disclosure are directed to each individual feature, system, article, material, kit, and/or method described herein. In addition, any combination of two or more such features, systems, articles, materials, kits, and/or methods, if such features, systems, articles, materials, kits, and/or methods are not mutually inconsistent, is included within the scope of the present disclosure.

Claims

I/We claims:

A method (100) for seizure detection and prediction using electroencephalogram (EEG) data, the method (100) comprising:
a. Acquiring EEG data from a subject;
b. Preprocessing the acquired EEG data through downsampling to a predefined sampling rate;
c. Applying a sliding window to the preprocessed EEG data with a predetermined time window and time step;
d. Processing the EEG data using a deep learning model, which comprises multiple convolutional layers, each followed by batch normalization and max pooling layers, and a dense layer for feature extraction;
e. Generating a seizure alert based on a threshold confidence level determined by the output of the; and
f. Providing a feedback loop for incorporating real-time feedback from healthcare professionals to refine the prediction model.
The method (100) of claim 1, wherein the deep learning model is trained using a binary cross-entropy loss function and an Adam optimizer.
The method (100) of claim 1, wherein the sliding window has a time window of 8 seconds and a time step of 4 seconds.
A seizure detection and prediction system (200), comprising:
a. An EEG data acquisition unit (202) configured to collect EEG signals from an EEG helmet worn by a user;
b. A preprocessing unit (204) configured to downsample the collected EEG signals and apply a sliding window;
c. A deep learning processing unit (206) that employs the model for analyzing the EEG data;
d. An alert generation unit (208) configured to generate real-time alerts upon detecting seizure-indicative patterns; and
e. A feedback interface (210) for receiving and incorporating feedback from healthcare professionals into the model.
The seizure detection and prediction system (200) of claim 4, wherein the preprocessing unit (204) applies a band-pass filter to the EEG signals to remove noise and artifacts before downsampling.
The seizure detection and prediction system (200) of claim 4, wherein the deep learning processing unit (206) employs a model with at least three convolutional neural network layers, each followed by batch normalization and a non-linear activation function.
The seizure detection and prediction system (200) of claim 4, wherein the sliding window applied by the preprocessing unit (204) has a configurable time window and time step, which is adjusted based on the type of seizure patterns being monitored.
The seizure detection and prediction system (200) of claim 4, wherein the alert generation unit (208) is configured to transmit alerts to an external computing device that is accessible to healthcare professionals or caregivers.
The seizure detection and prediction system (200) of claim 4, wherein the alert generation unit (208) is configured to activate auditory or visual alarms within a healthcare facility.
The seizure detection and prediction system (200) of claim 4, wherein the system (200) is integrated with a centralized medical record system to allow for seamless updating of patient data with seizure occurrence and prediction information.

METHOD FOR SEIZURE DETECTION AND PREDICTION UTILIZING EEG DATA

A method for detecting and predicting seizures utilizing electroencephalogram (EEG) data is disclosed. The method involves acquiring EEG data from a subject and preprocessing this data through downsampling to a predefined sampling rate. A sliding window is applied to the preprocessed EEG data with a predetermined time window and step, followed by processing the data using a deep learning model. This model includes multiple convolutional layers, each accompanied by batch normalization and max pooling layers, and a dense layer for feature extraction. A seizure alert is generated based on a threshold confidence level determined by the output of the deep learning model. The method further includes a feedback loop for incorporating real-time feedback from healthcare professionals to refine the prediction model.

Drawings
/
FIG. 1
/
FIG. 2
/
FIG. 3
/
FIG. 4
/
FIG. 5

, Claims:I/We claims:

A method (100) for seizure detection and prediction using electroencephalogram (EEG) data, the method (100) comprising:
a. Acquiring EEG data from a subject;
b. Preprocessing the acquired EEG data through downsampling to a predefined sampling rate;
c. Applying a sliding window to the preprocessed EEG data with a predetermined time window and time step;
d. Processing the EEG data using a deep learning model, which comprises multiple convolutional layers, each followed by batch normalization and max pooling layers, and a dense layer for feature extraction;
e. Generating a seizure alert based on a threshold confidence level determined by the output of the; and
f. Providing a feedback loop for incorporating real-time feedback from healthcare professionals to refine the prediction model.
The method (100) of claim 1, wherein the deep learning model is trained using a binary cross-entropy loss function and an Adam optimizer.
The method (100) of claim 1, wherein the sliding window has a time window of 8 seconds and a time step of 4 seconds.
A seizure detection and prediction system (200), comprising:
a. An EEG data acquisition unit (202) configured to collect EEG signals from an EEG helmet worn by a user;
b. A preprocessing unit (204) configured to downsample the collected EEG signals and apply a sliding window;
c. A deep learning processing unit (206) that employs the model for analyzing the EEG data;
d. An alert generation unit (208) configured to generate real-time alerts upon detecting seizure-indicative patterns; and
e. A feedback interface (210) for receiving and incorporating feedback from healthcare professionals into the model.
The seizure detection and prediction system (200) of claim 4, wherein the preprocessing unit (204) applies a band-pass filter to the EEG signals to remove noise and artifacts before downsampling.
The seizure detection and prediction system (200) of claim 4, wherein the deep learning processing unit (206) employs a model with at least three convolutional neural network layers, each followed by batch normalization and a non-linear activation function.
The seizure detection and prediction system (200) of claim 4, wherein the sliding window applied by the preprocessing unit (204) has a configurable time window and time step, which is adjusted based on the type of seizure patterns being monitored.
The seizure detection and prediction system (200) of claim 4, wherein the alert generation unit (208) is configured to transmit alerts to an external computing device that is accessible to healthcare professionals or caregivers.
The seizure detection and prediction system (200) of claim 4, wherein the alert generation unit (208) is configured to activate auditory or visual alarms within a healthcare facility.
The seizure detection and prediction system (200) of claim 4, wherein the system (200) is integrated with a centralized medical record system to allow for seamless updating of patient data with seizure occurrence and prediction information.

METHOD FOR SEIZURE DETECTION AND PREDICTION UTILIZING EEG DATA