COHERENCE-CORRELATION-BASED ALIGNED EEG AND ECoG-CONCATENATED EPILEPTIC SEIZURE DETECTION SYSTEM AND METHOD USING DLEWT AND LTR-LSUVN

Abstract

The present invention relates to a coherence-correlation-based aligned EEG and ECoG concatenated epileptic seizure detection system and method using DLEWT and LTR-LSUVN. The proposed system through the combined analysis of EEG and ECoG signals, capturing both subtle and complex seizure activity, isolating it from background noise, and reducing the impact of artifacts, performs precise detection of epileptic seizure. The proposed system enhances epileptic seizure detection by combining EEG and ECoG signals using coherence, correlation, and wavelet transform techniques(DLEWT). The system isolates seizure activity, reduces noise with SIPKA, simplifies data with API-CA, and detects complex interactions using Mutual Information. The system performs epileptic seizure classificationvia LTR-LSUVN ensures accurate detection of epileptic seizure. The proposed invention greatly improves the detection of epilepsy seizure, leading to better treatment outcomes and quality of life for patients.

Specification

Description:FIELD OF THE INVENTION
The present disclosure relates to an epileptic seizure detection system, specifically to a coherence-correlation-based aligned EEG and ECoG-concatenated epileptic seizure detection system and method using DLEWT and LTR-LSUVN.

BACKGROUND OF THE INVENTION
Epileptic seizure detection remains a critical area of research within the field of neurological disorders, given the chronic nature of epilepsy and its widespread impact on millions of individuals globally. The unpredictable nature of epileptic seizures, coupled with the associated cognitive and emotional distress, emphasizes the need for accurate and early detection methods. Electroencephalogram (EEG) and electrocorticography (ECoG) are the predominant bio-signals used for detecting epileptic seizures, as both provide valuable insight into the electrical activity of the brain. While EEG records brain signals non-invasively via electrodes placed on the scalp, ECoG provides a more invasive approach by recording signals directly from electrodes implanted inside the skull. Despite their usefulness, limitations in current EEG and ECoG-based seizure detection systems hinder the effectiveness of these methods in clinical practice.
The inherent differences in spatial and temporal resolution between EEG and ECoG present a significant challenge in accurately detecting epileptic seizures. EEG offers a broad view of brain activity but lacks the spatial precision to localize seizure foci, while ECoG provides more localized data but is highly invasive. Existing solutions has not sufficiently addressed the potential of combining EEG and ECoG signals to leverage their complementary strengths for improved seizure detection. Existing methods fail to integrate these signals effectively, neglecting the discrepancies in signal characteristics and resolution, which results in suboptimal diagnostic accuracy.
Furthermore, the signals associated with epileptic seizures are often low in amplitude and exhibit non-stationary behavior, making it difficult to differentiate them from normal brain activity. This leads to a low signal-to-noise ratio, where the weak seizure signals are masked by background brain activity. As a result, seizure events, especially those of short duration or subtle intensity, may go undetected or be misclassified. Temporal resolution limitations of EEG, combined with artifacts and noise contamination in both EEG and ECoG recordings, further degrade the quality of the captured signals. These issues make it difficult to accurately localize the epileptogenic zone, increasing the likelihood of incorrect or incomplete diagnoses.
In addition, the high-dimensional nature of ECoG data, owing to the large number of electrodes involved, poses a significant challenge in data management and analysis. Current methods struggle to handle this vast amount of data efficiently, further complicating the process of feature extraction and seizure detection. Moreover, individual variability in ECoG signals, both between different patients and within the same patient over time, introduces another layer of complexity, making it difficult to extract consistent and reliable diagnostic features.
To address these limitations, there is a need for a novel approach that can align and combine EEG and ECoG signals, taking into account their differences in spatial-temporal resolution and signal characteristics. The present invention proposes an epileptic seizure detection system, wherein the system uses Delayed Time Windowing (DELTW) and Long-Term Robust Localized Signal Variability Normalization (LTR-LSUVN) techniques to enhance epileptic seizure detection. By concatenating and aligning EEG and ECoG signals, the proposed system enables the coherent analysis of these two bio-signals, improving the overall diagnostic accuracy and reliability of seizure detection. The DELTW and LTR-LSUVN techniques are specifically designed to address the signal variability, noise, and artifact contamination, providing a robust solution to the challenges faced in existing systems.

SUMMARY OF THE INVENTION
The present disclosure relates to to a coherence-correlation-based aligned EEG and ECoG-concatenated epileptic seizure detection system and method using DLEWT and LTR-LSUVN. The proposed system performs collection of EEG and ECoG signal, preprocessing and normalization of signals; channel-wise concatenation, removal of artificats, segmentation, coherence and correlation analysis, dimensionality reduction, wavelet transformation, feature extraction, feature selection, mutual information extraction, and classification. The proposed system performs a coherence and correlation-based analysis of aligned concatenated EEG and ECoG signals for effective epileptic seizure detection. The system captures subtle seizure activity by measuring the degree of synchrony or phase-locking between EEG and ECoG channels, as well as analyzing the strength and direction of the linear relationship between signals, thereby reflecting functional connectivity. To further distinguish seizure events from normal brain activity, the system applies a Discrete Lyapunov Exponents Wavelet Transform (DLEWT), which decomposes the signals into multiple frequency components at varying scales, effectively isolating seizure activity from background noise. In order to minimize the impact of noise and artifacts on the detection process, the system integrates Skewness Independent Percentile Kurtosis Analysis (SIPKA), ensuring the preservation of essential neural signals. To handle the high-dimensional nature of ECoG data, the system employs Akaike Principal Information Component Analysis (API-CA), which identifies key patterns and reduces computational complexity. Finally, the system detects both linear and non-linear relationships between variables by extracting Mutual Information (MI) to analyze complex interactions between signals. Epileptic seizure classification is then performed using the LogT Recurrent Layer-Sequential Unit-Variance Network (LTR-LSUVN), ensuring accurate and robust detection of seizures.
The present disclosure seeks to provide a coherence-correlation-based aligned EEG and ECoG concatenated epileptic seizure detection system using DLEWT and LTR-LSUVN. The system comprises: a signal acquisition unit configured to collect aligned EEG (Electroencephalography) and ECoG (Electrocorticography) signals from a plurality of public data sources;a preprocessing unit to augment signal to increase training data volume, sample signal to enhance temporal resolution, and align signal using Dynamic Time Warping (DTW) and Exponential Log Transformation (ELT) to handle time and amplitude distortions between EEG and ECoG signals;a normalization unit performing Z-score normalization to standardize signal amplitude;a channel-wise concatenation unit to fuse spatial and global data from EEG and ECoG signals;an artifact removal unit utilizing Independent Component Analysis (ICA) combined with statistical properties (Percentiles of Kurtosis and Skewness) to remove artifacts like eye and muscle movements;a segmentation unit employing K-Means Clustering (KMC) to segment signals based on similarity in feature space;a coherence-correlation analysis unit to compute coherence (e.g., magnitude of coherence, phase coherence) and correlation (e.g., Pearson correlation, lagged correlation) between EEG and ECoG segments;a dimensionality reduction unit using Akaike Information Criterion (AIC) for optimal component selection;a Discrete Lyapunov Exponent Wavelet Transform (DLEWT) unit to perform wavelet transformation of signals;a feature extraction unit to extract a set of features, selected from Mean, Variance, Kurtosis, Entropy, Granger Causality, Phase Locking Value, Peak Amplitude, Lyapunov Exponent, Fractal Dimension, Power Spectral Density (PSD), and others from the signals, coherence-correlation outputs, and wavelet-transformed signals;a feature selection unit applying the Wilcoxon Rank-Sum Test (WRST) to filter features;a mutual information extraction unit to identify significant features; anda classification unit based on LTR-LSUVN (LogT Mapping with Recurrent Neural Network (RNN) hyperparameter tuning and LSUV weight initialization), configured to classify epileptic seizures with GPU-based parallel processing.
In an embodiment, the system further comprises a signal alignment unitwhich uses DTW to align EEG and ECoG signals with varying speeds and frequencies, followed by ELT to adjust amplitude differences, ensuring accurate concatenation.
In an embodiment, the channel-wise concatenation unit combines the spatial resolution of ECoG with the global scope of EEG for enhanced neural activity analysis.
In an embodiment, the artifact removal unit applies ICA to decompose signals into independent components, while the skewness and kurtosis percentiles differentiate neural activity from artifacts.
In an embodiment, the coherence-correlation analysis unit calculates multiple coherence and correlation metrics across EEG and ECoG segments to detect seizure patterns.
In an embodiment, the dimensionality reduction unit employs Principal Component Analysis (PCA) and AIC to reduce the signal complexity while maintaining essential patterns, minimizing computational load.
In an embodiment, the DLEWT unit performs wavelet transformation on reduced-dimensionality signals to identify transient epileptic events selected from spikes and waves.
In an embodiment, the feature extraction unit extracts multiple features, selected from a group of mean, variance, skewness, kurtosis, Power Spectral Density (PSD), band power, Lyapunov exponents, and coherence variability.
In an embodiment, the classification unit uses a Recurrent Neural Network (RNN) optimized by LogT mapping for hyperparameter tuning, and LSUV for stabilizing weight initialization across layers, ensuring stable training convergence.
The present disclosure also seeks to provide a method for detecting epileptic seizures using concatenated EEG and ECoG signals. The method comprises: collecting EEG and ECoG signals from public data sources;preprocessing collected signals by performing signal augmentation to increase training data, enhancing temporal resolution through signal sampling, and aligning the signals using DTW for temporal consistency and ELT for amplitude normalization;applying Z-score normalization to standardize signal amplitudes;concatenating EEG and ECoG channels to combine spatial and global information;removing artifacts using ICA and quantifying statistical properties selected from kurtosis and skewness to distinguish artifacts from neural signals;segmenting the artifact-removed signals using a K-Means Clustering;performing coherence and correlation analysis on the segmented data;reducing the dimensionality of the signal data using PCA and AIC for optimal component selection;applying wavelet transformation using the DLEWT method;extracting features from wavelet-transformed and coherence-analyzed signals, including both spectral and statistical features;selecting relevant features using the Wilcoxon Rank-Sum Test (WRST);extracting mutual information from the selected features; andclassifying the processed signals using a Recurrent Neural Network (RNN) with LogT mapping for hyperparameter tuning and LSUV weight initialization.
An objective of the present disclosure is to provide to a coherence-correlation-based aligned EEG and ECoG-concatenated epileptic seizure detection system and method using DLEWT and LTR-LSUVN.
Another objective of the present disclosure is to provide a framework that effectively combines and aligns EEG and ECoG signals for improved epileptic seizure detection by accounting for differences in spatial and temporal resolution.
Another objective of the present disclosure is to enhance the signal-to-noise ratio in seizure detection by employing Delayed Time Windowing (DELTW) to better differentiate weak seizure signals from background brain activity.
Another objective of the present disclosure is to mitigate the impact of artifacts and noise in EEG and ECoG recordings, ensuring more accurate localization of the epileptogenic zone.
Another objective of the present disclosure is to manage and analyze high-dimensional ECoG data efficiently using Long-Term Robust Localized Signal Variability Normalization (LTR-LSUVN) to extract consistent features.
Yet, another objective of the present disclosure is to provide a robust and reliable seizure detection system capable of handling individual variability in ECoG signals, improving diagnostic accuracy over time.
To further clarify advantages and features of the present disclosure, a more particular description of the invention will be rendered by reference to specific embodiments thereof, which is illustrated in the appended drawings. It is appreciated that these drawings depict only typical embodiments of the invention and are therefore not to be considered limiting of its scope. The invention will be described and explained with additional specificity and detail with the accompanying drawings.

BRIEF DESCRIPTION OF FIGURES
These and other features, aspects, and advantages of the present disclosure will become better understood when the following detailed description is read with reference to the accompanying drawings in which like characters represent like parts throughout the drawings, wherein:
Figure 1 illustrates a block diagram of a coherence-correlation-based aligned EEG and ECoG concatenated epileptic seizure detection system using DLEWT and LTR-LSUVN, in accordance with an embodiment of the present disclosure;
Figure 2 illustrates a flow chart of a method for detecting epileptic seizures using concatenated EEG and ECoG signals, in accordance with an embodiment of the present disclosure; and
Figure 3 illustrates a block diagram of the proposed invention, in accordance with an embodiment of the present disclosure.
Further, skilled artisans will appreciate that elements in the drawings are illustrated for simplicity and may not have been necessarily been drawn to scale. For example, the flow charts illustrate the method in terms of the most prominent steps involved to help to improve understanding of aspects of the present disclosure. Furthermore, in terms of the construction of the device, one or more components of the device may have been represented in the drawings by conventional symbols, and the drawings may show only those specific details that are pertinent to understanding the embodiments of the present disclosure so as not to obscure the drawings with details that will be readily apparent to those of ordinary skill in the art having benefit of the description herein.

DETAILED DESCRIPTION:
For the purpose of promoting an understanding of the principles of the invention, reference will now be made to the embodiment illustrated in the drawings and specific language will be used to describe the same. It will nevertheless be understood that no limitation of the scope of the invention is thereby intended, such alterations and further modifications in the illustrated system, and such further applications of the principles of the invention as illustrated therein being contemplated as would normally occur to one skilled in the art to which the invention relates.
It will be understood by those skilled in the art that the foregoing general description and the following detailed description are exemplary and explanatory of the invention and are not intended to be restrictive thereof.
Reference throughout this specification to "an aspect", "another aspect" or similar language 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, appearances of the phrase "in an embodiment", "in another embodiment" and similar language throughout this specification may, but do not necessarily, all refer to the same embodiment.
The terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process or method that comprises a list of steps does not include only those steps but may include other steps not expressly listed or inherent to such process or method. Similarly, one or more devices or sub-systems or elements or structures or components proceeded by "comprises...a" does not, without more constraints, preclude the existence of other devices or other sub-systems or other elements or other structures or other components or additional devices or additional sub-systems or additional elements or additional structures or additional components.
Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The system, methods, and examples provided herein are illustrative only and not intended to be limiting.
Embodiments of the present disclosure will be described below in detail with reference to the accompanying drawings.
The functional units described in this specification have been labeled as devices. A device may be implemented in programmable hardware devices such as processors, digital signal processors, central processing units, field programmable gate arrays, programmable array logic, programmable logic devices, cloud processing systems, or the like. The devices may also be implemented in software for execution by various types of processors. An identified device may include executable code and may, for instance, comprise one or more physical or logical blocks of computer instructions, which may, for instance, be organized as an object, procedure, function, or other construct. Nevertheless, the executable of an identified device need not be physically located together, but may comprise disparate instructions stored in different locations which, when joined logically together, comprise the device and achieve the stated purpose of the device.
Indeed, an executable code of a device or module could be a single instruction, or many instructions, and may even be distributed over several different code segments, among different applications, and across several memory devices. Similarly, operational data may be identified and illustrated herein within the device, and may be embodied in any suitable form and organized within any suitable type of data structure. The operational data may be collected as a single data set, or may be distributed over different locations including over different storage devices, and may exist, at least partially, as electronic signals on a system or network.
Reference throughout this specification to "a select embodiment," "one embodiment," or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the disclosed subject matter. Thus, appearances of the phrases "a select embodiment," "in one embodiment," or "in an embodiment" in various places throughout this specification are not necessarily referring to the same embodiment.
Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided, to provide a thorough understanding of embodiments of the disclosed subject matter. One skilled in the relevant art will recognize, however, that the disclosed subject matter can be practiced without one or more of the specific details, or with other methods, components, materials, etc. In other instances, well-known structures, materials, or operations are not shown or described in detail to avoid obscuring aspects of the disclosed subject matter.
In accordance with the exemplary embodiments, the disclosed computer programs or modules can be executed in many exemplary ways, such as an application that is resident in the memory of a device or as a hosted application that is being executed on a server and communicating with the device application or browser via a number of standard protocols, such as TCP/IP, HTTP, XML, SOAP, REST, JSON and other sufficient protocols. The disclosed computer programs can be written in exemplary programming languages that execute from memory on the device or from a hosted server, such as BASIC, COBOL, C, C++, Java, Pascal, or scripting languages such as JavaScript, Python, Ruby, PHP, Perl or other sufficient programming languages.
Some of the disclosed embodiments include or otherwise involve data transfer over a network, such as communicating various inputs or files over the network. The network may include, for example, one or more of the Internet, Wide Area Networks (WANs), Local Area Networks (LANs), analog or digital wired and wireless telephone networks (e.g., a PSTN, Integrated Services Digital Network (ISDN), a cellular network, and Digital Subscriber Line (xDSL)), radio, television, cable, satellite, and/or any other delivery or tunneling mechanism for carrying data. The network may include multiple networks or sub networks, each of which may include, for example, a wired or wireless data pathway. The network may include a circuit-switched voice network, a packet-switched data network, or any other network able to carry electronic communications. For example, the network may include networks based on the Internet protocol (IP) or asynchronous transfer mode (ATM), and may support voice using, for example, VoIP, Voice-over-ATM, or other comparable protocols used for voice data communications. In one implementation, the network includes a cellular telephone network configured to enable exchange of text or SMS messages.
Examples of the network include, but are not limited to, a personal area network (PAN), a storage area network (SAN), a home area network (HAN), a campus area network (CAN), a local area network (LAN), a wide area network (WAN), a metropolitan area network (MAN), a virtual private network (VPN), an enterprise private network (EPN), Internet, a global area network (GAN), and so forth.
Figure 1 illustrates a block diagram of a coherence-correlation-based aligned EEG and ECoG concatenated epileptic seizure detection system (100) using DLEWT and LTR-LSUVN, in accordance with an embodiment of the present disclosure.
Referring to Figure 1, the system (100) includesa signal acquisition unit (102) configured to collect aligned EEG (Electroencephalography) and ECoG (Electrocorticography) signals from a plurality of public data sources;a preprocessing unit (104) to augment signal to increase training data volume, sample signal to enhance temporal resolution, and align signal using Dynamic Time Warping (DTW) and Exponential Log Transformation (ELT) to handle time and amplitude distortions between EEG and ECoG signals;a normalization unit (106) performing Z-score normalization to standardize signal amplitude;a channel-wise concatenation unit (108) to fuse spatial and global data from EEG and ECoG signals;an artifact removal unit (110) utilizing Independent Component Analysis (ICA) combined with statistical properties (Percentiles of Kurtosis and Skewness) to remove artifacts like eye and muscle movements;a segmentation unit (112) employing K-Means Clustering (KMC) to segment signals based on similarity in feature space;a coherence-correlation analysis unit (114) to compute coherence (e.g., magnitude of coherence, phase coherence) and correlation (e.g., Pearson correlation, lagged correlation) between EEG and ECoG segments;a dimensionality reduction unit (116) using Akaike Information Criterion (AIC) for optimal component selection;a Discrete Lyapunov Exponent Wavelet Transform (DLEWT) unit (118) to perform wavelet transformation of signals;a feature extraction unit (120) to extract a set of features, selected from Mean, Variance, Kurtosis, Entropy, Granger Causality, Phase Locking Value, Peak Amplitude, Lyapunov Exponent, Fractal Dimension, Power Spectral Density (PSD), and others from the signals, coherence-correlation outputs, and wavelet-transformed signals;a feature selection unit (122) applying the Wilcoxon Rank-Sum Test (WRST) to filter features;a mutual information extraction unit (124) to identify significant features; anda classification unit (126) based on LTR-LSUVN (LogT Mapping with Recurrent Neural Network (RNN) hyperparameter tuning and LSUV weight initialization), configured to classify epileptic seizures with GPU-based parallel processing.
In an embodiment, the system (100)further comprises a signal alignment unit (128) which uses DTW to align EEG and ECoG signals with varying speeds and frequencies, followed by ELT to adjust amplitude differences, ensuring accurate concatenation.
In an embodiment, the channel-wise concatenation unit (108) combines the spatial resolution of ECoG with the global scope of EEG for enhanced neural activity analysis.
In an embodiment, the artifact removal unit (110) applies ICA to decompose signals into independent components, while the skewness and kurtosis percentiles differentiate neural activity from artifacts.
In an embodiment, the coherence-correlation analysis unit (114) calculates multiple coherence and correlation metrics across EEG and ECoG segments to detect seizure patterns.
In an embodiment, the dimensionality reduction unit (116) employs Principal Component Analysis (PCA) and AIC to reduce the signal complexity while maintaining essential patterns, minimizing computational load.
In an embodiment, the DLEWT unit (118) performs wavelet transformation on reduced-dimensionality signals to identify transient epileptic events selected from spikes and waves.
In an embodiment, the feature extraction unit (120) extracts multiple features, selected from a group of mean, variance, skewness, kurtosis, Power Spectral Density (PSD), band power, Lyapunov exponents, and coherence variability.
In an embodiment, the classification unit (126) uses a Recurrent Neural Network (RNN) optimized by LogT mapping for hyperparameter tuning, and LSUV for stabilizing weight initialization across layers, ensuring stable training convergence.
In an embodiment, signal acquisition unit (102), preprocessing unit (104), normalization unit (106), channel-wise concatenation unit (108), artefact removal unit (110), segmentation unit (112), coherence-correlation analysis unit (114), dimensionality reduction unit (116), DLEWT unit (118), feature extraction unit (120), feature selection unit (122), mutual information extraction unit (124) and classification unit (126) may be implemented in programmable hardware devices such as processors, digital signal processors, central processing units, field programmable gate arrays, programmable array logic, programmable logic devices, cloud processing systems, or the like.
Figure 2 illustrates a flow chart of a method (200) for detecting epileptic seizures using concatenated EEG and ECoG signals, in accordance with an embodiment of the present disclosure.
Referring to Figure 2, the method (200) includes a plurality of steps as described under:
At step (202), the method (200) includes collecting EEG and ECoG signals from public data sources.
At step (204), the method (200) includes preprocessing collected signals by performing signal augmentation to increase training data, enhancing temporal resolution through signal sampling, and aligning the signals using DTW for temporal consistency and ELT for amplitude normalization.
At step (206), the method (200) includes applying Z-score normalization to standardize signal amplitudes.
At step (208), the method (200) includes concatenating EEG and ECoG channels to combine spatial and global information.
At step (210), the method (200) includes removing artifacts using ICA and quantifying statistical properties selected from kurtosis and skewness to distinguish artifacts from neural signals.
At step (212), the method (200) includes segmenting the artifact-removed signals using a K-Means Clustering.
At step (214), the method (200) includes performing coherence and correlation analysis on the segmented data.
At step (216), the method (200) includes reducing the dimensionality of the signal data using PCA and AIC for optimal component selection;
At step (218), the method (200) includes applying wavelet transformation using the DLEWT method.
At step (220), the method (200) includes extracting features from wavelet-transformed and coherence-analyzed signals, including both spectral and statistical features.
At step (222), the method (200) includes selecting relevant features using the Wilcoxon Rank-Sum Test (WRST).
At step (224), the method (200) includes extracting mutual information from the selected features.
At step (226), the method (200) includes classifying the processed signals using a Recurrent Neural Network (RNN) with LogT mapping for hyperparameter tuning and LSUV weight initialization.
Figure 3 illustrates a block diagram of the proposed invention, in accordance with an embodiment of the present disclosure.
Referring to Figure 3, the proposed a coherence-correlation-based aligned EEG and ECoG concatenated epileptic seizure detection system that uses DLEWT and LTR-LSUVN, performs the following steps: EEG and ECoG signal collection, Preprocessing, Normalization, Channel-wise concatenation, Artifacts removal, Segmentation, Coherence and correlation analysis, Dimensionality reduction, Wavelet transformation, Feature extraction, Feature selection, Mutual information extraction, and classification.
The proposed system starts by collecting both the EEG and ECoG signals from the publically accessible data sources for the effective prediction of epileptic seizures. At first, both the signals undergo preprocessing to enhance the signal quality. Pre-processing of signals includes: signal augmentation, signal sampling, and signal alignment, whereinsignal augmentation increases the number of signals for effective training of the deep learning models,signal sampling increases the temporal resolution of the signals for better concatenation, and signal alignment in terms of DELTW transforms both signals into a form that is suitable for concatenation. The proposed system employs Dynamic Time Warping (DTW) which aligns two-time series that may vary in speed or length by finding an optimal alignment path. DTW can handle non-linear time distortions and varying speeds between signals. As collected EEG and ECoG signals have variations in timing or different frequencies, DTW aligns these signals even if they are not perfectly synchronized. However, if the signals have significant amplitude or magnitude variations, DTW may misalign signals, because DTW aligns the signals based on temporal similarity without considering amplitude differences. Therefore, the proposed system utilizes an Exponential Log Transformation (ELT) function to compress the large values and expand the small ones to reduce the impact of varying scales, wherein said function also makes signals more comparable and facilitates better alignment. The proposed system further performs Zscore normalization, in order to convert the signals into a desired range of values, and then both signals are fused based on their channels, wherein Channelwise fusion combines the fine spatial detail from ECoG with the broader global information from EEG, leading to enhanced spatial resolution in the fused data, improving the accuracy of source localization and neural decoding.
The proposed system performs the removal of artifacts such as, such as eye movements and muscle movements, from the concatenated signals by means of the SIPKA technique. Independent Component Analysis (ICA) decomposes the EEG or ECoG signals into components that are statistically independent, which helps in isolating the neural signals from artifacts, such as eye blinks, muscle movements, or electrical noise. It effectively reduces noise and artifacts without significantly affecting the neural signals of interest. This leads to cleaner data that better reflects the brain's true activity. In order to perform accurate and effective removal of artificats, the proposed system is configured to measure the Percentiles of Kurtosis and Skewness distributions, to differentiate neural signals and artifacts by quantifying their statistical properties, wherein the differentiation is carried out on the basis of the fact that Neural signals often have distributions close to Gaussian with low kurtosis and skewness, while artifacts may deviate from these patterns. Further, the system is configured to perform the segmentation, where individual components obtained during artifact removal are segmentedvia the KMC technique, wherein K-Means clusters segment the signal based on similarity in feature space, which helps in identifying homogeneous regions or patterns within the signal. Segmentation minimizes within-cluster variance and maximizes between-cluster variance, which can lead to clear and distinct segment boundaries. Next, the system analyses the coherence and correlation between each segment.
In a similar manner as stated above, the proposed system, from the segmented result, reduces the dimensionality of the signal using the API-CA algorithm, eliminating redundant features, making the analysis more manageable and focused on meaningful components. The system employs Principal Component Analysis (PCA) for extracting principal components that capture the most significant variance in the concatenated data. This can simplify the interpretation of complex data by focusing on key patterns and trends, which can significantly decrease computational complexity for further analysis or modeling. In order to determine the optimal number of principal components, the proposed system utilizes the Akaike Information Criterion (AIC) for selecting a number of components that adequately represent the data while avoiding the inclusion of unnecessary components, thus making the detection more interpretable and efficient. Then, the proposed system is configured to perform the wavelet transformation which is carried out based on the DLEWT technique.The Discrete Wavelet Transform (DWT) is effective at analyzing non-stationary signals, it struggles to capture highly complex or subtle patterns in EEG or ECoG signals that are associated with certain types of epileptic activity. In order to avoid this and capture the subtle signal patterns, the Lyapunov Exponents are introduced, which measure the rate of separation of infinitesimally close trajectories in phase space. This can help in analyzing the stability and chaotic behavior of EEG or ECoG signals. In epilepsy, abnormal chaotic behavior can precede seizures. By analyzingLyapunov exponents, it is possible to detect changes in the dynamics of brain activity that might signal an impending seizure.
After the wavelet transformation, the proposed system extracts the features, wherein features are extracted from the dimensionality reduced signal (Mean, variance, Skewness, Kurtosis, Entropy, Granger Causality, Phase Locking Value (PLV), Peak Amplitude, Latency, Lyapunov Exponent¸ Fractal Dimension, Signal Power¸ Root Mean Square (RMS), Zero Crossing Rate, Band Power, Spectral Entropy, Spectral Centroid, Spectral Flatness, and Hurst Exponent), coherence and correlation output (Magnitude of Coherence, Frequency Band Coherence, Peak Coherence Frequency, Phase Coherence, Coherence Variability, CrossFrequency Coherence, Pearson Correlation Coefficient, Lagged Correlation, Partial Correlation, Cross-Correlation, Correlation Coefficient Variability, Mean Correlation), and the wavelet transformed signal (Power Spectral Density (PSD), Band Power Ratios, Dominant Frequency, Wavelet Coefficients, Wavelet Entropy, Harmonic Ratio, Detailed Coefficients, Approximation Coefficients¸ Mean and Variance, Skewness and Kurtosis, Peak-to-Peak Amplitude) followed by feature selection via Wilcoxon Rank-Sum Test (WRST), mutual information extraction, and classification. The proposed system is configured to classify the normal and seizures by using the LTR-LSUVN classification model with a Graphic Processing Unit (GPU) for parallel processing as well as to handle large datasets with better memory and ease of processing. The proposed system is incorporated with Recurrent Neural Network (RNN) in order to handle temporal dependencies and sequences effectively. The concatenated signals are inherently sequential and have temporal patterns that evolve over time. RNNs can model these temporal dependencies, capturing the sequence of events that may indicate epileptic seizures. The RNNs have several hyperparameters that need tuning, such as the number of layers, hidden units, and learning rate, wherein the hyperparameters are optimizedusing the LogT mapping function, wherein the sensitivity ofLogT map to initial conditions allows for dynamic and adaptive search behavior, which helps in converging to better hyperparameter settings for RNNs. Moreover, weight values of RNN are initialized using the Layer-Sequential UnitVariance (LSUV) technique, which ensures that the variance of activations is consistent across layers, preventing the gradients from vanishing or exploding as they propagate through the network. This leads to more stable and faster convergence.
The proposed invention provides significant benefits by offering a more accurate and robust system for detecting epileptic seizures through the combined analysis of EEG and ECoG signals. By capturing both subtle and complex seizure activity, isolating it from background noise, and reducing the impact of artifacts, the system enhances diagnostic precision. Its ability to handle high-dimensional data and analyze both linear and non-linear relationships improves computational efficiency and adaptability across different patients. In real-world applications, this invention can significantly improve the early detection and monitoring of epilepsy, leading to better treatment outcomes and quality of life for patients.
The drawings and the forgoing description give examples of embodiments. Those skilled in the art will appreciate that one or more of the described elements may well be combined into a single functional element. Alternatively, certain elements may be split into multiple functional elements. Elements from one embodiment may be added to another embodiment. For example, orders of processes described herein may be changed and are not limited to the manner described herein. Moreover, the actions of any flow diagram need not be implemented in the order shown; nor do all of the acts necessarily need to be performed. Also, those acts that are not dependent on other acts may be performed in parallel with the other acts. The scope of embodiments is by no means limited by these specific examples. Numerous variations, whether explicitly given in the specification or not, such as differences in structure, dimension, and use of material, are possible. The scope of embodiments is at least as broad as given by the following claims.
Benefit s, other advantages, and solutions to problems have been described above with regard to specific embodiments. However, the benefits, advantages, solutions to problems, and any component(s) that may cause any benefit, advantage, or solution to occur or become more pronounced are not to be construed as a critical, required, or essential feature or component of any or all the claims.
 
, Claims:1. A coherence-correlation-based aligned EEG and ECoG concatenated epileptic seizure detection system using DLEWT and LTR-LSUVN, comprising:
a signal acquisition unit configured to collect aligned EEG (Electroencephalography) and ECoG (Electrocorticography) signals from a plurality of public data sources;
a preprocessing unit to augment signal to increase training data volume, sample signal to enhance temporal resolution, and align signal using Dynamic Time Warping (DTW) and Exponential Log Transformation (ELT) to handle time and amplitude distortions between EEG and ECoG signals;
a normalization unit performing Z-score normalization to standardize signal amplitude;
a channel-wise concatenation unit to fuse spatial and global data from EEG and ECoG signals;
an artifact removal unit utilizing Independent Component Analysis (ICA) combined with statistical properties (Percentiles of Kurtosis and Skewness) to remove artifacts like eye and muscle movements;
a segmentation unit employing K-Means Clustering (KMC) to segment signals based on similarity in feature space;
a coherence-correlation analysis unit to compute coherence (e.g., magnitude of coherence, phase coherence) and correlation (e.g., Pearson correlation, lagged correlation) between EEG and ECoG segments;
a dimensionality reduction unit using Akaike Information Criterion (AIC) for optimal component selection;
a Discrete Lyapunov Exponent Wavelet Transform (DLEWT) unit to perform wavelet transformation of signals;
a feature extraction unit to extract a set of features, selected from Mean, Variance, Kurtosis, Entropy, Granger Causality, Phase Locking Value, Peak Amplitude, Lyapunov Exponent, Fractal Dimension, Power Spectral Density (PSD), and others from the signals, coherence-correlation outputs, and wavelet-transformed signals;
a feature selection unit applying the Wilcoxon Rank-Sum Test (WRST) to filter features;
a mutual information extraction unit to identify significant features; and
a classification unit based on LTR-LSUVN (LogT Mapping with Recurrent Neural Network (RNN) hyperparameter tuning and LSUV weight initialization), configured to classify epileptic seizures with GPU-based parallel processing.

2. The system as claimed in claim 1, further comprising a signal alignment unit (128) which uses DTW to align EEG and ECoG signals with varying speeds and frequencies, followed by ELT to adjust amplitude differences, ensuring accurate concatenation.

3. The system as claimed in claim 1, wherein the channel-wise concatenation unit combines the spatial resolution of ECoG with the global scope of EEG for enhanced neural activity analysis.

4. The system as claimed in claim 1, wherein the artifact removal unit applies ICA to decompose signals into independent components, while the skewness and kurtosis percentiles differentiate neural activity from artifacts.

5. The system as claimed in claim 1, wherein the coherence-correlation analysis unit calculates multiple coherence and correlation metrics across EEG and ECoG segments to detect seizure patterns.

6. The system as claimed in claim 1, wherein the dimensionality reduction unit employs Principal Component Analysis (PCA) and AIC to reduce the signal complexity while maintaining essential patterns, minimizing computational load.

7. The system as claimed in claim 1, wherein the DLEWT unit performs wavelet transformation on reduced-dimensionality signals to identify transient epileptic events selected from spikes and waves.

8. The system as claimed in claim 1, wherein the feature extraction unit extracts multiple features, selected from a group of mean, variance, skewness, kurtosis, Power Spectral Density (PSD), band power, Lyapunov exponents, and coherence variability.

9. The system as claimed in claim 1, wherein the classification unit uses a Recurrent Neural Network (RNN) optimized by LogT mapping for hyperparameter tuning, and LSUV for stabilizing weight initialization across layers, ensuring stable training convergence.

10. A method for detecting epileptic seizures using concatenated EEG and ECoG signals, comprising:
collecting EEG and ECoG signals from public data sources;
preprocessing collected signals by performing signal augmentation to increase training data, enhancing temporal resolution through signal sampling, and aligning the signals using DTW for temporal consistency and ELT for amplitude normalization;
applying Z-score normalization to standardize signal amplitudes;
concatenating EEG and ECoG channels to combine spatial and global information;
removing artifacts using ICA and quantifying statistical properties selected from kurtosis and skewness to distinguish artifacts from neural signals;
segmenting the artifact-removed signals using a K-Means Clustering;
performing coherence and correlation analysis on the segmented data;
reducing the dimensionality of the signal data using PCA and AIC for optimal component selection;
applying wavelet transformation using the DLEWT method;
extracting features from wavelet-transformed and coherence-analyzed signals, including both spectral and statistical features;
selecting relevant features using the Wilcoxon Rank-Sum Test (WRST);
extracting mutual information from the selected features; and
classifying the processed signals using a Recurrent Neural Network (RNN) with LogT mapping for hyperparameter tuning and LSUV weight initialization.

Documents