Description:The present invention relates to the field of biomedical signal processing and artificial intelligence-based healthcare diagnostics. More specifically, it pertains to an integrated system and method for enhancing seizure detection in electroencephalogram (EEG) signals using advanced denoising techniques, generative data augmentation, feature extraction, deep learning classification models, and reinforcement learning for real-time decision optimization.
BACKGROUND OF THE INVENTION
The following description of related art is intended to provide background information pertaining to the field of the disclosure. This section may include certain aspects of the art that may be related to various features of the present disclosure. However, it should be appreciated that this section be used only to enhance the understanding of the reader with respect to the present disclosure, and not as admissions of prior art.

Epileptic seizures are characterized by abnormal and excessive electrical discharges in the brain, which can be unpredictable and life-threatening. EEG monitoring is the most commonly used non-invasive method for capturing brain activity and identifying seizure events. However, the accuracy and reliability of EEG-based seizure detection still face significant challenges due to inherent limitations in data quality and analysis methods.

One major challenge in existing EEG seizure detection systems is the presence of noise and artifacts in the raw signals. Eye blinks, muscle movements, and electrical interference can distort the EEG waveform, resulting in reduced signal-to-noise ratio (SNR) and compromised detection accuracy. Traditional denoising techniques often fail to distinguish artifacts from actual neural signals effectively.

Another limitation is the scarcity and imbalance of seizure-related data. Seizures are rare events in long-duration EEG recordings, and the types of seizures vary widely among patients. This results in severely unbalanced datasets, where most data points belong to non-seizure classes. As a result, conventional machine learning models become biased and show poor generalization to unseen or underrepresented seizure types.

Existing EEG feature extraction techniques are often inadequate for capturing the complex time-frequency patterns that characterize seizures. Methods such as raw time-series or Fourier transforms alone may not sufficiently highlight short bursts of abnormal oscillatory activity, especially in early seizure onset or atypical presentations.

Furthermore, most existing classification approaches rely solely on either spatial or temporal features of the EEG, failing to capture the full dynamics of seizure evolution. Models that ignore temporal dependencies miss important pre-ictal and post-ictal signals, which can be critical for early detection and intervention.

Static, threshold-based alarm systems also limit real-time performance. These systems trigger alerts only when a fixed probability threshold is exceeded, which can lead to delayed detection or frequent false positives in variable EEG patterns.

To address these problems, there is a clear need for a robust and intelligent system that can clean, augment, analyze, and interpret EEG signals comprehensively while adapting dynamically to evolving seizure patterns in real-time.

OBJECTIVE OF THE INVENTION

Some of the objects of the present disclosure, which at least one embodiment herein satisfies are listed herein below.

The primary objective of the present invention is to develop a comprehensive seizure detection system that significantly improves the accuracy, sensitivity, and real-time responsiveness of EEG-based seizure monitoring.

Another objective is to enhance the quality of raw EEG signals by employing Discrete Wavelet Transform (DWT) denoising, which effectively removes non-neural artifacts such as eye blinks, electromyographic noise, and electrical interference, thereby improving the signal-to-noise ratio by up to 25%.

A further objective is to overcome the challenge of limited and imbalanced seizure data by generating synthetic seizure segments using Generative Adversarial Networks (GANs). This augmentation aims to expand the dataset size, diversify seizure pattern representation, and improve model robustness.

The invention also aims to extract rich spectral and temporal features from the denoised and augmented EEG data using Short-Time Fourier Transform (STFT) and Power Spectral Density (PSD), which help in capturing transient seizure-related brainwave patterns that are often missed by conventional methods.

An important objective is to integrate a hybrid deep learning architecture combining Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks. The CNN component captures spatial patterns in time-frequency matrices, while the LSTM component models temporal dependencies in seizure progression.

Another objective is to introduce a reinforcement learning component using Deep Q-Networks (DQNs) that dynamically adjusts seizure detection thresholds based on real-time EEG signal conditions and seizure likelihood, thereby reducing false alarms and latency.

Ultimately, the invention is intended to serve as a robust, real-time, and clinically applicable seizure detection solution suitable for implementation in wearable EEG devices, hospital monitoring systems, and diagnostic software tools.

SUMMARY OF THE INVENTION
This section is provided to introduce certain objects and aspects of the present disclosure in a simplified form that are further described below in the detailed description. This summary is not intended to identify the key features or the scope of the claimed subject matter.

The present invention provides a multistage, deep learning-driven system for detecting seizures from EEG signals with high accuracy and low latency. The system integrates DWT-based denoising, GAN-based data augmentation, STFT and PSD feature extraction, a CNN-LSTM hybrid classification architecture, and a DQN-based dynamic alert threshold adjustment mechanism. Each module is designed to address a specific limitation of conventional seizure detection systems.

By sequentially improving data quality, representation, feature richness, classification accuracy, and real-time responsiveness, the proposed invention demonstrates superior performance with an F1-score increase of 20%, sensitivity of 87%, and classification accuracy between 85–90%. It further reduces alert latency by 30% and improves early detection capabilities, making it highly suitable for clinical and continuous care applications.

BRIEF DESCRIPTION OF DRAWINGS
The accompanying drawings, which are incorporated herein, and constitute a part of this invention, illustrate exemplary embodiments of the disclosed methods and systems in which like reference numerals refer to the same parts throughout the different drawings. Components in the drawings are not necessarily to scale, emphasis instead being placed upon clearly illustrating the principles of the present invention. Some drawings may indicate the components using block diagrams and may not represent the internal circuitry of each component. It will be appreciated by those skilled in the art that invention of such drawings includes the invention of electrical components, electronic components or circuitry commonly used to implement such components.

FIG. 1 illustrates an exemplary system for enhancing seizure detection from EEG signals, in accordance with an embodiment of the present disclosure.

DETAILED DESCRIPTION OF THE INVENTION

In the following description, for the purposes of explanation, various specific details are set forth in order to provide a thorough understanding of embodiments of the present disclosure. It will be apparent, however, that embodiments of the present disclosure may be practiced without these specific details. Several features described hereafter can each be used independently of one another or with any combination of other features. An individual feature may not address all of the problems discussed above or might address only some of the problems discussed above. Some of the problems discussed above might not be fully addressed by any of the features described herein.

The ensuing description provides exemplary embodiments only and is not intended to limit the scope, applicability, or configuration of the disclosure. Rather, the ensuing description of the exemplary embodiments will provide those skilled in the art with an enabling description for implementing an exemplary embodiment. It should be understood that various changes may be made in the function and arrangement of elements without departing from the spirit and scope of the disclosure as set forth.

Specific details are given in the following description to provide a thorough understanding of the embodiments. However, it will be understood by one of ordinary skill in the art that the embodiments may be practiced without these specific details. For example, circuits, systems, networks, processes, and other components may be shown as components in block diagram form in order not to obscure the embodiments in unnecessary detail. In other instances, well-known circuits, processes, algorithms, structures, and techniques may be shown without unnecessary detail to avoid obscuring the embodiments.

Also, it is noted that individual embodiments may be described as a process that is depicted as a flowchart, a flow diagram, a data flow diagram, a structure diagram, or a block diagram. Although a flowchart may describe the operations as a sequential process, many of the operations can be performed in parallel or concurrently. In addition, the order of the operations may be re-arranged. A process is terminated when its operations are completed but could have additional steps not included in a figure. A process may correspond to a method, a function, a procedure, a subroutine, a subprogram, etc. When a process corresponds to a function, its termination can correspond to a return of the function to the calling function or the main function.

The word “exemplary” and/or “demonstrative” is used herein to mean serving as an example, instance, or illustration. For the avoidance of doubt, the subject matter disclosed herein is not limited by such examples. In addition, any aspect or design described herein as “exemplary” and/or “demonstrative” is not necessarily to be construed as preferred or advantageous over other aspects or designs, nor is it meant to preclude equivalent exemplary structures and techniques known to those of ordinary skill in the art. Furthermore, to the extent that the terms “includes,” “has,” “contains,” and other similar words are used in either the detailed description or the claims, such terms are intended to be inclusive in a manner similar to the term “comprising” as an open transition word without precluding any additional or other elements.

Reference throughout this specification to “one embodiment” or “an embodiment” or “an instance” or “one instance” means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present disclosure. Thus, the appearances of the phrases “in one embodiment” or “in an embodiment” in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. As used herein, the singular forms “a”, “an”, and “the” are intended to include the plural forms as well, unless the context indicates otherwise. It will be further understood that the terms “comprises” and/or “comprising,” when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. As used herein, the term “and/or” includes any and all combinations of one or more of the associated listed items.

The present invention discloses a multistage system and method for enhanced seizure detection using EEG signals. The system begins with the acquisition of raw EEG data from a multi-channel EEG recording device. This raw signal is often contaminated with noise and artifacts caused by eye movements, muscle contractions, and external interference. To improve the quality of the signal, a Discrete Wavelet Transform (DWT)-based denoising module is applied. DWT decomposes the signal into various frequency sub-bands, allowing targeted removal of noise without compromising essential neurological information. This module improves the Signal-to-Noise Ratio (SNR) by 15–25%, thereby preserving vital patterns relevant to seizure onset.

Once denoised, the EEG data is passed to a Generative Adversarial Network (GAN)-based data augmentation module. This module is specifically designed to tackle class imbalance in seizure datasets by generating synthetic EEG segments that mimic underrepresented seizure types. The GAN is trained using minority-class samples and generates new data instances with similar characteristics, increasing the data size by up to 200%. This not only enhances class representation but also reduces overfitting in the deep learning model and improves its ability to detect a wide range of seizure phenotypes.

Following augmentation, the system proceeds to feature extraction, which plays a pivotal role in identifying meaningful signal patterns. Two complementary techniques are employed: Short-Time Fourier Transform (STFT) and Power Spectral Density (PSD) analysis. STFT converts the one-dimensional EEG time series into a time-frequency representation, effectively capturing transient oscillations indicative of seizure onset. PSD quantifies the power distribution across different frequency bands, allowing the identification of epileptiform abnormalities in specific channels or time windows. These extracted features are structured into 2D matrices that are used as inputs to the deep learning model.

The feature matrices are fed into a hybrid Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) model. The CNN layers are responsible for detecting spatial features and local seizure signatures in the EEG spectrograms, such as rhythmic bursts or spike-wave discharges. The output from CNN layers is then passed to LSTM layers, which capture temporal dependencies and transitions between different seizure phases (pre-ictal, ictal, post-ictal). This hybrid architecture ensures that both spatial and temporal seizure characteristics are adequately captured, improving classification accuracy by 10–15%.

To enhance the decision-making process in real time, a Deep Q-Network (DQN)-based dynamic thresholding mechanism is integrated. Instead of using a fixed probability threshold for seizure detection, the DQN module evaluates the output probabilities from the CNN-LSTM model and dynamically adjusts the decision threshold based on environmental context, signal stability, and seizure probability trends. This leads to a 30% reduction in alert latency and a 15% increase in early seizure sensitivity. The result is a responsive and reliable seizure alert system suitable for real-time clinical applications.

The entire pipeline is optimized for deployment in different environments such as bedside monitors, wearable EEG devices, or cloud-based diagnostic platforms. The modularity of each stage allows for customization based on patient-specific conditions, hardware constraints, and clinical requirements.

In one embodiment, the entire seizure detection pipeline is integrated into a wearable EEG headset designed for ambulatory monitoring of epilepsy patients. The headset includes dry or semi-dry electrodes for non-invasive data acquisition, a compact signal processor capable of performing DWT denoising and real-time inference using the CNN-LSTM-DQN architecture, and a wireless module for alert notifications.

The GAN module is pre-trained offline and embedded in the device's firmware to simulate rare seizure events during training phases. A dedicated mobile application receives alerts and visualizes seizure probability trends, allowing caregivers or physicians to monitor patients remotely. The system enables continuous at-home monitoring, making it especially beneficial for children, elderly, or rural patients with limited access to hospitals.

In another embodiment, the invention is implemented as a cloud-integrated platform for in-hospital patient monitoring. EEG signals from intensive care unit (ICU) patients are streamed to a centralized server equipped with high-performance GPUs. The denoising, augmentation, feature extraction, and deep learning inference steps are performed in real-time.

Hospital staff are notified via dashboards and alarms when seizures are predicted with high probability. The DQN module fine-tunes alert sensitivity according to clinical requirements (e.g., higher threshold in low-risk patients, more sensitive in high-risk cases). The platform also supports retrospective seizure analysis, helping neurologists review patient data and evaluate treatment efficacy over time.

In yet another embodiment, the invention is implemented as a standalone software application installed on desktop computers in neurology clinics. Neurologists can upload pre-recorded EEG datasets from Holter devices or long-term monitoring units into the software.

The DWT-GAN-CNN-LSTM-DQN pipeline is executed locally, and the software generates a comprehensive diagnostic report, highlighting probable seizure segments, statistical metrics (sensitivity, F1-score, seizure duration), and visualization of extracted features. This offline tool is particularly useful in settings with limited internet access or where data privacy regulations prevent cloud storage.

While considerable emphasis has been placed herein on the preferred embodiments, it will be appreciated that many embodiments can be made and that many changes can be made in the preferred embodiments without departing from the principles of the invention. These and other changes in the preferred embodiments of the invention will be apparent to those skilled in the art from the disclosure herein, whereby it is to be distinctly understood that the foregoing descriptive matter to be implemented merely as illustrative of the invention and not as limitation.
, Claims:1. A system for enhancing seizure detection from EEG signals, comprising:
o a denoising module employing Discrete Wavelet Transform (DWT) for artifact removal;
o a data augmentation module utilizing Generative Adversarial Networks (GANs) to synthesize underrepresented seizure data;
o a feature extraction module applying Short-Time Fourier Transform (STFT) and Power Spectral Density (PSD) to extract time-frequency features;
o a classification module comprising a hybrid deep learning architecture of Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM) for spatial-temporal seizure feature recognition;
o and a dynamic thresholding module using a Deep Q-Network (DQN) for real-time alert optimization.
2. The system of claim 1, wherein the DWT denoising increases signal-to-noise ratio by 15–25%.
3. The system of claim 1, wherein the GAN augmentation module increases training dataset size by 200% and reduces class imbalance.
4. The system of claim 1, wherein the STFT captures transient oscillations associated with seizure onset.
5. The system of claim 1, wherein the CNN processes spectral EEG images to extract spatial features from seizure regions.
6. The system of claim 1, wherein the LSTM identifies temporal relationships across EEG sequences during different seizure phases.
7. The system of claim 1, wherein the DQN module adapts the seizure detection threshold in real-time to minimize false positives and alert latency.
8. The system of claim 1, wherein the overall model achieves 85–90% accuracy, 87% sensitivity, and a 20% improvement in F1-score compared to conventional systems.
9. The system of claim 1, wherein the system is implemented on wearable or bedside EEG monitoring hardware.