A method and device for recognizing electroencephalogram epileptic waves induced by high-dose transcranial magnetic stimulation

Abstract

The present application belongs to the technical field of computer, and particularly relates to a large-dose transcranial magnetic stimulation induced electroencephalogram epilepsy wave recognition method, device, readable storage medium and program product, wherein the method comprises the following steps: acquiring transcranial magnetic stimulation induced electroencephalogram wave data; processing the electroencephalogram wave signal, extracting a stimulation point, and performing global filtering to obtain filtered data and the stimulation point; dividing the filtered data after the stimulation point into multiple segments of data according to a preset truncation unit, and extracting key features for each segment of data; inputting the key features into a preset epilepsy wave segment detection model for detection to obtain multiple suspected epilepsy segments; and post-processing the multiple suspected epilepsy segments to obtain real epilepsy waves. The transcranial magnetic stimulation induced electroencephalogram epilepsy wave recognition method provided by the present application improves the detection accuracy, and enhances the robustness and adaptability of the model.

Description

Technical Field

[0001] This invention relates to the field of brainwave processing, and more particularly to a method and device for identifying epileptic brainwaves induced by high-dose transcranial magnetic stimulation. Background Technology

[0002] High-dose transcranial magnetic stimulation (TMS)-induced epileptic seizures are an effective treatment for severe mental illnesses. Effective seizure induction can "reset" brain electrical activity patterns, prompting neurons to return to normal firing patterns, thus improving symptoms. Simultaneously, seizure induction can effectively regulate neurotransmitter levels in the brain; for example, serotonin, dopamine, and norepinephrine levels are significantly increased after seizure induction. These neurotransmitters are important in improving symptoms in patients with severe depression and other conditions. However, visual detection of epileptic waves requires relatively high levels of experience and knowledge from clinicians, significantly hindering its widespread clinical application. In contrast, electroencephalography (EEG) is the gold standard for detecting epileptic seizures. Identifying epileptic waves (such as sharp waves and sharp-slow-wave complexes) is crucial for guiding the effectiveness of treatment. Similarly, EEG shows promise in detecting epileptic brain waves induced by high-dose TMS. However, unlike simply detecting abnormal EEG patterns in epilepsy patients, existing algorithms for detecting epileptic waves induced by high-dose transcranial magnetic stimulation still face significant problems and limitations. These problems mainly focus on the sensitivity, specificity, and clinical operability (e.g., insufficient real-time performance and automation), severely impacting the accuracy of epilepsy research and treatment efficacy evaluation. Summary of the Invention

[0003] In view of this, the present invention provides a method, device, readable storage medium and program product for identifying epileptic waves induced by high-dose transcranial magnetic stimulation, in order to solve the technical problems of insufficient accuracy and poor clinical applicability of traditional EEG signal processing methods.

[0004] In a first aspect, the present invention provides a method for identifying epileptic waves induced by high-dose transcranial magnetic stimulation (TMS), comprising the following steps: acquiring EEG data induced by TMS; processing the EEG signals, extracting stimulation points, and performing global filtering to obtain filtered data and stimulation points; dividing the filtered data after the stimulation points into multiple segments according to a preset truncation unit, and extracting key features from each segment; inputting the key features into a preset epileptic waveband detection model for detection to obtain multiple suspected epileptic segments; and post-processing the multiple suspected epileptic segments to obtain the actual epileptic waves.

[0005] Preferably, the training of the preset epilepsy band detection model includes the following steps: dividing the preset data into a training set and a test set; extracting the key features of the data in the training set and the test set; inputting the key features of the training set into the preset original model for training to obtain a model to be verified; and inputting the key features of the test set into the model to be verified for verification to obtain the epilepsy band detection model.

[0006] Preferably, the key features include: mean, maximum absolute value, true maximum value, minimum absolute value, and true minimum value.

[0007] Preferably, the key feature includes a CV value used to describe signal stability.

[0008] Preferably, the key features include the Hurst index for analyzing the complexity and memory of electroencephalogram (EEG) signals.

[0009] Preferably, the key features include wavelet entropy for quantifying the complexity and nonlinear characteristics of the signal.

[0010] Secondly, the present invention provides a device for identifying epileptic waves induced by high-dose transcranial magnetic stimulation (TMS), comprising: a receiving module for acquiring EEG data induced by TMS; a preprocessing module for processing the obtained filtered data and stimulation points; a key feature extraction module for extracting key features; an epileptic band detection module for detecting suspected epileptic bands based on the key features; and a post-processing module for post-processing the suspected epileptic bands to obtain real epileptic waves.

[0011] Preferably, the epilepsy band detection model construction module is used to construct the epilepsy band detection model in the epilepsy band detection module.

[0012] Thirdly, the present invention provides a readable storage medium having computer program instructions stored thereon, which, when executed, perform the steps of the method described above.

[0013] Fourthly, the present invention provides a program product having computer program instructions thereon, which, when executed, implement the steps of the method described above.

[0014] This invention provides a method for identifying epileptic waves induced by transcranial magnetic stimulation (TMS), comprising the following steps: acquiring EEG data induced by TMS; processing the EEG signals, extracting stimulation points, and performing global filtering to obtain filtered data and stimulation points; dividing the filtered data after stimulation points into multiple segments according to a preset truncation unit, and extracting key features from each segment; inputting the key features into a preset epileptic wave band detection model for detection to obtain multiple suspected epileptic segments; and post-processing the multiple suspected epileptic segments to obtain the actual epileptic wave.

[0015] Compared with the prior art, the present invention has at least the following beneficial effects:

[0016] 1. Precise Stimulus Point Localization and Filtering: By acquiring EEG data induced by transcranial magnetic stimulation (TMS), processing the EEG signals to extract stimulation points, and then performing global filtering, the stimulation points can be located more precisely and noise removed, resulting in cleaner and more representative filtered data. Compared to traditional methods, this processing approach can effectively reduce noise interference in subsequent analysis, improve data quality, and provide more accurate input signals for subsequent epileptic wave detection.

[0017] 2. Effective Extraction of Key Features: By dividing the filtered data after the stimulus point into multiple segments according to a preset truncation unit and extracting key features from each segment, the characteristic changes of EEG over different time periods can be captured more meticulously. Compared with traditional holistic analysis or simple statistical feature extraction methods, this segmented key feature extraction method can more comprehensively and meticulously reflect the dynamic changes of EEG during epileptic seizures, providing richer feature information for the accurate detection of epileptic waves.

[0018] 3. Improve the accuracy of epileptic wave detection: By inputting the extracted key features into a pre-defined epileptic wave detection model, the model's learning and pattern recognition capabilities can be utilized to more accurately identify suspected epileptic segments. Compared to traditional methods based on human experience or simple threshold judgments, this model-based detection approach can better handle complex EEG signals, reduce false positives and false negatives, and improve the accuracy and reliability of epileptic wave detection.

[0019] 4. Improved Detection Efficiency: The entire detection process is automated through pre-set models and algorithms, avoiding the tedious process of manually analyzing EEG data segment by segment, thus greatly improving detection efficiency. This is of great significance for processing large amounts of EEG data in clinical diagnosis and research, enabling the analysis and detection of large amounts of data in a short time, and providing support for the rapid diagnosis and treatment of epilepsy.

[0020] 5. Assisting Clinical Diagnosis: This invention can provide clinicians with more accurate and reliable epilepsy wave detection results, assisting them in making diagnostic decisions more quickly. Through automated detection methods, it can reduce the time and effort doctors invest in EEG data analysis, improving diagnostic efficiency and accuracy, while also avoiding misdiagnosis or missed diagnosis due to human factors.

[0021] 6. Adaptability to Different Epilepsy Types: Since epileptic waves may differ among different patients and epilepsy types, this invention, through meticulous processing and feature extraction of EEG data and the use of a pre-defined model for detection, possesses a certain degree of adaptability and generalization ability, making it applicable to the detection of various types of epileptic waves. This provides a universal detection method for the diagnosis and research of different types of epilepsy in clinical practice, and has promising clinical application prospects.

[0022] 7. Support for Research: The detection method provided by this invention can more accurately extract epileptic wavelengths, providing more reliable data support for research on the pathogenesis of epilepsy, drug development, and evaluation of treatment methods. Through the analysis and research of a large number of epileptic wavelengths, a better understanding of the neurophysiological mechanisms of epilepsy can be achieved, providing a basis for developing more effective treatment options. Attached Figure Description

[0023] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the embodiments of the present invention will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, and these are all within the protection scope of the present invention.

[0024] Figure 1 This is a flowchart illustrating the first embodiment of the present invention.

[0025] Figure 2 This is a schematic diagram of the original electroencephalogram (EEG) in the first embodiment of the present invention.

[0026] Figure 3 This is a schematic diagram of determining the stimulus point after filtering in the first embodiment of the present invention.

[0027] Figure 4 This is a schematic diagram of a fragment of epileptic wave in the first embodiment of the present invention.

[0028] Figure 5 This is a schematic diagram of the epilepsy wave detection results in the first embodiment of the present invention.

[0029] Figure 6 This is a schematic diagram of the structure of the EEG epilepsy wave recognition device provided in the second embodiment of the present invention.

[0030] Figure 7 This is a schematic diagram of the structure of the program product provided in the fourth embodiment of the present invention.

[0031] Figure 8 This is a schematic diagram of the structure of an electronic device provided in the fifth embodiment of the present invention.

[0032] Explanation of icon numbers:

[0033] 1. Method; 2. Apparatus; 3. Process product; 4. Electronic equipment;

[0034] 20. Receiving module; 21. Preprocessing module; 22. Key feature extraction module; 23. Epilepsy band detection module; 24. Postprocessing module;

[0035] 30. Computer program instructions; 40. Processor; 41. Memory; 42. Bus; 43. Communication interface. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0037] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. Please refer to [link to documentation]. Figure 1 The first embodiment of the present invention provides a method 1 for identifying epileptic waves induced by high-dose transcranial magnetic stimulation, comprising the following steps:

[0038] Acquire EEG data induced by transcranial magnetic stimulation; process the EEG signals, extract stimulation points, and perform global filtering to obtain filtered data and stimulation points; divide the filtered data after stimulation points into multiple segments according to preset truncation units, and extract key features from each segment; input the key features into a preset epilepsy band detection model for detection to obtain multiple suspected epilepsy segments; perform post-processing on the multiple suspected epilepsy segments to obtain the actual epilepsy wave.

[0039] For example, an epilepsy detection method based on high-dose transcranial magnetic stimulation (TMS) induced electroencephalogram (EEG) aims to quickly and accurately detect epileptic wave segments by automating the processing of EEG data, thereby assisting in clinical diagnosis and research.

[0040] Specific implementation steps: 1. Acquiring EEG data induced by transcranial magnetic stimulation: Using a transcranial magnetic stimulation (TMS) device and a high-precision electroencephalogram (EEG) device; recruiting a group of epilepsy patients and a group of healthy controls, and conducting TMS-induced experiments on each. During the experiment, the TMS device stimulates specific areas of the brain at a specific frequency and intensity, while the EEG device records the brainwave signals; the EEG device records brainwave signals at a high sampling rate (e.g., 1000Hz), for a duration ranging from several minutes to more than ten minutes. The acquired brainwave data includes multiple channels (e.g., 64 channels), with each channel recording the electrical activity of different brain regions.

[0041] 2. Combining Figure 2 and Figure 3The brainwave signals are processed to extract stimulation points and perform global filtering: First, the acquired brainwave data is denoised to remove power frequency interference (such as 50Hz or 60Hz), eye movement artifacts, and electromyography artifacts. Independent component analysis (ICA) and other methods can be used to separate and remove artifact components. By analyzing the timestamps of the brainwave signals and combining them with the stimulation records from the TMS instrument, the exact time points of each TMS stimulation are extracted. For example, the TMS instrument emits stimulation at the 1st, 3rd, and 5th seconds, and these time points are the stimulation points. Bandpass filtering is then performed on the brainwave signals to remove frequency components below 1Hz and above 30Hz. This is because epileptic waves are usually concentrated in the frequency range of 1Hz to 30Hz, and the filtered data can more clearly reflect the brain electrical activity related to epilepsy. After filtering, the filtered data and corresponding stimulation points are obtained.

[0042] 3. Divide the filtered data after the stimulation point into multiple segments according to a preset truncation unit, and extract key features from each segment: Starting from the stimulation point, divide the filtered data into multiple segments according to a preset time unit (e.g., 100ms). For example, starting from the stimulation point, extract a segment every 100ms until the end of the EEG signal. Assuming the total duration of the filtered data is 10 seconds, then 100 segments can be obtained; extract key features from each segment, including but not limited to:

[0043] Temporal characteristics: such as mean, variance, peak-to-trough ratio, energy, etc. For example, the mean of each data segment is calculated to reflect the average potential level of the brain waves, and the variance is calculated to reflect the degree of fluctuation of the brain waves.

[0044] Frequency domain characteristics: Each data segment is transformed from the time domain to the frequency domain using Fast Fourier Transform (FFT), and the energy distribution of each frequency band is calculated, such as the energy values ​​of delta waves (1-4Hz), theta waves (4-8Hz), alpha waves (8-13Hz), and beta waves (13-30Hz). Epilepsy waves typically show abnormal energy increases in the theta and delta wave frequency bands; therefore, the energy characteristics of these frequency bands are particularly important for the detection of epilepsy waves.

[0045] Time-frequency domain features: Using wavelet transform and other methods to extract time-frequency domain features can simultaneously reflect the changes in brain waves in time and frequency, and capture the characteristics of epileptic waves more comprehensively.

[0046] Nonlinear characteristics, such as sample entropy and Lyapunov exponent, are used to reflect the complexity and degree of chaos in brain waves. During an epileptic seizure, the complexity of brain waves may change, and these nonlinear characteristics help distinguish between normal brain waves and epileptic waves.

[0047] 4. Input key features into a pre-defined epileptic wave segment detection model for detection, obtaining multiple suspected epileptic segments. Select a suitable machine learning model for epileptic wave detection, such as Support Vector Machine (SVM), Convolutional Neural Network (CNN), or Long Short-Term Memory Network (LSTM). Train the model using labeled epileptic wave data and normal EEG data as training sets. During training, the model learns how to distinguish between epileptic waves and normal EEG based on the input feature vectors. Input the extracted key features into the trained epileptic wave segment detection model. The model classifies each data segment according to the learned patterns, determining whether it is an epileptic wave segment. For example, the model outputs a probability value; when the probability value is higher than a set threshold (e.g., 0.8), the data segment is considered a suspected epileptic segment. After detection, multiple suspected epileptic segments are obtained.

[0048] 5. Post-processing multiple suspected epileptic segments to obtain the true epileptic wave: Due to potential overlap during data segmentation, the same epileptic wave segment may be detected as a suspected epileptic segment multiple times. By comparing the temporal position and feature similarity of adjacent suspected epileptic segments, duplicate segments are removed, and independent epileptic wave segments are retained. For adjacent suspected epileptic segments, if they are temporally continuous and feature-similar, they can be spliced ​​into a complete epileptic wave segment. Simultaneously, based on the degree of matching between the spliced ​​segment and known epileptic wave patterns, the start and end times of the segment are fine-tuned, and its boundaries are corrected to more accurately reflect the true epileptic wave.

[0049] Through the above embodiments, the present invention enables automated processing of EEG data induced by high-dose transcranial magnetic stimulation and detection of epileptic waves. Compared with traditional methods, it has at least the following specific advantages:

[0050] Precise localization and filtering: By accurately extracting stimulus points and performing global filtering, noise interference is effectively removed, improving the quality of EEG data and providing a more reliable basis for subsequent detection.

[0051] Comprehensive feature extraction: Key features are extracted from multiple perspectives, including time domain, frequency domain, time-frequency domain, and nonlinearity, which can more comprehensively reflect the characteristic changes of EEG waves and improve the accuracy of epileptic wave detection.

[0052] Efficient Detection and Post-processing: Automated detection is achieved using a pre-defined epileptic wave segment detection model. Post-processing removes repetitive segments, spliced ​​fragments, and corrects boundaries, improving detection efficiency and result reliability. Furthermore, manual review further ensures the accuracy of the detection results.

[0053] In some embodiments, training a preset epilepsy band detection model includes the following steps: dividing preset data into a training set and a test set; extracting key features from the data in the training set and the test set; inputting the key features of the training set into a preset original model for training to obtain a model to be verified; and inputting the key features of the test set into the model to be verified for verification to obtain an epilepsy band detection model.

[0054] For example, EEG data from epilepsy patients and healthy controls were acquired, and then preprocessed and labeled.

[0055] Specific Implementation Steps: 1. Data Preparation: Data Collection: EEG data were collected from 100 epilepsy patients and 50 healthy controls. Each subject underwent multiple TMS-induced experiments, each lasting approximately 10 minutes. The collected EEG data included multiple channels (e.g., 64 channels) at a sampling rate of 1000Hz. Data Labeling: The EEG data of epilepsy patients were labeled, marking the start and end times of the epileptic wavebands. The EEG data of the healthy controls were labeled as normal EEG wavebands. Data Preprocessing: All EEG data underwent noise reduction to remove power line interference, eye movement artifacts, and electromyography artifacts. Then, bandpass filtering (1Hz to 30Hz) was applied to the data to preserve epilepsy-related EEG activity.

[0056] 2. Data Splitting: Divide the preprocessed data into training and test sets randomly. For example, use 70% of the data as the training set and 30% as the test set. The training set is used for model training, and the test set is used to validate the model's performance. Data Formatting: Format the training and test set data into a format suitable for model input. Assuming we choose a Convolutional Neural Network (CNN) model, we need to convert the EEG data into a two-dimensional matrix (e.g., number of channels × number of time points) and extract the key features of each data segment.

[0057] 3. Feature Extraction: Key Feature Extraction: Key features are extracted from each segment of EEG data in the training and test sets, including time-domain, frequency-domain, and nonlinear features. Time-domain Features: The mean, variance, energy, peak-to-trough ratio, etc., of each data segment are calculated. Frequency-domain Features: The energy values ​​of each frequency band (e.g., delta wave, theta wave, alpha wave, beta wave) are calculated using Fast Fourier Transform (FFT). Nonlinear Features: Sample entropy, Lyapunov exponent, etc., are calculated to reflect the complexity and chaos of the EEG. Feature Vector Construction: The extracted features are combined into a feature vector, which serves as the input to the model. For example, the feature vector of each data segment may contain 20 feature values, which comprehensively reflect the characteristics of the EEG.

[0058] 4. Model Training: Selecting the Original Model: A Convolutional Neural Network (CNN) is chosen as the original model. CNNs have powerful feature learning capabilities and are suitable for processing time series data. Training Process: Inputting Training Data: Key features from the training set are input into the CNN model. The model adjusts its parameters by learning the relationship between the features in the training set and the epileptiform labels. Loss Function and Optimization: The Cross-Entropy Loss function is used to measure the difference between the model's predictions and the true labels. Backpropagation and optimization algorithms (such as Adam) are used to update the network parameters to minimize the loss function. Training Iterations: Multiple training iterations (e.g., 100 epochs) are performed until the model's loss function converges, resulting in the model to be validated.

[0059] 5. Model Validation: Input Test Data: Input the key features of the test set into the model to be validated. Performance Evaluation: Prediction Results: The model classifies each data segment in the test set and outputs the probability value of the epileptic band. Evaluation Metrics: Use metrics such as accuracy, recall, precision, and F1 score to evaluate the model's performance. For example: Accuracy: The proportion of samples correctly classified by the model out of the total number of samples. Recall: The proportion of epileptic bands detected by the model out of the actual epileptic bands. Precision: The proportion of epileptic bands detected by the model that are actually epileptic bands. F1 Score: The harmonic mean of recall and precision, which comprehensively evaluates the model's performance. Model Optimization: Optimize the model based on the evaluation results of the test set. For example, adjust the network structure, add regularization terms, adjust the learning rate, etc., to improve the model's performance.

[0060] 6. Obtaining the Epilepsy Wavelength Detection Model: Final Model Determination: After multiple training and validation cycles, the model with the best performance is selected as the final epilepsy wavelength detection model. Model Saving and Application: The final model is saved for subsequent epilepsy wavelength detection tasks. For example, in clinical diagnosis, new EEG data can be input into the model to quickly detect epilepsy wavelengths, assisting doctors in diagnosis.

[0061] Through the above embodiments, the present invention enables effective training and validation of the epilepsy waveband detection model. Specific advantages are as follows: Data-driven model training: By training the model with a large amount of labeled EEG data, the model can learn the characteristic differences between epileptic wavebands and normal EEG wavebands, exhibiting strong generalization ability. Multi-dimensional feature extraction: Key features are extracted from multiple perspectives, including the time domain, frequency domain, and nonlinearity, providing rich input information for the model and improving its detection accuracy. Rigorous model validation: The model is validated using independent test sets to ensure good performance on unseen data, avoiding overfitting.

[0062] In some embodiments, key features include: mean, maximum absolute value, true maximum value, minimum absolute value, and true minimum value.

[0063] The following key features were extracted from the EEG data of each channel: mean, maximum absolute value, true maximum value, minimum absolute value, and true minimum value.

[0064] The following are the specific steps of feature extraction: Mean, calculate the average value of each data segment, reflecting the overall trend or baseline level of the signal.

[0065]

[0066] Where x_i is the i-th sampling point of the signal, and N is the total number of sampling points.

[0067] Maximum Absolute Value:

[0068]

[0069] This indicator is highly sensitive to detecting transient and dramatic changes in signals, such as spikes in electroencephalogram (EEG) signals.

[0070] TrueMaximumValue:

[0071]

[0072] In EEG signal analysis, the true maximum value can help capture the maximum amplitude of the signal, reflecting the intensity of abnormal discharges.

[0073] Minimum Absolute Value:

[0074]

[0075] This indicator reflects the smallest fluctuations in a signal and is often used to evaluate the resting state characteristics of a signal.

[0076] TrueMinimumValue:

[0077]

[0078] In electroencephalogram (EEG) signals, this indicator helps to capture abnormally low-amplitude discharges.

[0079] Feature vector construction combines the extracted features into a feature vector, which serves as the input to the model. For example, for each data segment in each channel, the feature vector can be represented as follows: assuming each channel has 5 features, and the feature vector length for each channel is 5. If the EEG data has 64 channels, then the total feature vector length for each data segment is 64 × 5 = 320.

[0080] Selecting the Original Model: Support Vector Machine (SVM) is selected as the original model. SVM is a commonly used classifier suitable for handling high-dimensional feature data. Training Process: Input Training Data: Input the feature vectors of the training set into the SVM model. The model learns the relationship between the features in the training set and the epilepsy wave labels, adjusting the model parameters accordingly. Loss Function and Optimization: The cross-entropy loss function is used to measure the difference between the model's predictions and the true labels. The model parameters are updated using an optimization algorithm (such as Sequence Minimum Optimization, SMO) to minimize the loss function. Training Iteration: Multiple training iterations are performed until the model's loss function converges, resulting in the model to be validated. Inputting Test Data: The feature vectors of the test set are input into the model to be validated. The model classifies each data segment in the test set and outputs the probability value of the epilepsy wave segment. Evaluation of Model Performance: The model's performance is assessed using metrics such as accuracy, recall, precision, and F1 score. For example: Accuracy: The proportion of correctly classified samples out of the total number of samples. Recall: The proportion of epilepsy wave segments detected by the model out of actual epilepsy wave segments. Precision: The proportion of epilepsy wave segments detected by the model that are actually epilepsy wave segments. F1 score: The harmonic mean of recall and precision, used to comprehensively evaluate the model's performance. After multiple training and validation cycles, the model with the best performance was selected as the final epilepsy band detection model.

[0081] Compared with existing technologies, this solution has at least the following beneficial effects:

[0082] Key feature extraction: By extracting key features such as mean, maximum absolute value, true maximum value, minimum absolute value, and true minimum value, the characteristics of EEG signals can be comprehensively reflected, providing rich input information for the model and improving the detection accuracy of the model.

[0083] Data-driven model training: The model is trained using a large amount of labeled EEG data, enabling it to learn the characteristic differences between epileptic EEG segments and normal EEG segments, thus exhibiting strong generalization ability. Rigorous model validation: The model is validated using independent test sets to ensure good performance on unseen data and avoid overfitting.

[0084] In some embodiments, key features also include a CV value for describing signal stability, a Hurst exponent for analyzing the complexity and memory of EEG signals, and wavelet entropy for quantifying the complexity and nonlinear characteristics of the signal.

[0085] For example, CV value (coefficient of variation):

[0086]

[0087] The standard deviation (CV) is used to describe the stability of a signal. It is calculated as the ratio of the standard deviation to the mean, where the standard deviation represents the degree of signal fluctuation, and the mean represents the average level of the signal. The smaller the CV value, the more stable the signal.

[0088] Hurst index: Used to measure the long-term correlation and trend of time series. The Hurst index value ranges from 0 to 1.

[0089] H<0.5: This indicates that the signal has anti-persistence, meaning that the future value trend is likely to reverse.

[0090] H=0.5: This indicates that the signal is random.

[0091] H>0.5: indicates that the signal is persistent, meaning that the future value trend is likely to continue.

[0092] Wavelet entropy: Used to quantify the complexity and nonlinear characteristics of a signal. The calculation steps for wavelet entropy are as follows: 1. Perform wavelet decomposition on the signal to obtain a set of wavelet coefficients. 2. Calculate the energy probability distribution of each component. 3. Calculate the entropy value based on the energy distribution.

[0093]

[0094] Where, p i It represents the energy ratio of each frequency band. Wavelet entropy can be used in EEG signal analysis to quantify the complexity and nonlinear characteristics of signals, and is particularly suitable for the detection of epileptic waves.

[0095] Please see Figure 4 For example, Figure 4 There are two peaks, located at 38 seconds and 110 seconds respectively. However, the peak at 110 seconds is too small compared to the overall area. Therefore, the peak at 110 seconds is determined to be a false peak. Filtering out the shorter, continuous epileptic waves yields the true epileptic waves, such as... Figure 5 As shown.

[0096] Compared with existing technologies, the present invention introduces key features such as CV value, Hurst exponent, and wavelet entropy into epilepsy band detection. The addition of these features brings the following beneficial effects:

[0097] 1. More comprehensive signal characteristic description. CV value (Coefficient of Variation): Signal stability analysis: The CV value describes the relative volatility of a signal and quantifies its stability. By calculating the ratio of the signal's standard deviation to its mean, the CV value can intuitively reflect the magnitude of the signal's relative volatility. In epileptic wave detection, EEG signals during epileptic seizures typically exhibit large fluctuations, while normal EEG waves are relatively stable. The CV value can effectively distinguish between these two states.

[0098] Dimensionlessness: The CV value is a dimensionless indicator that can be used to compare data of different units or magnitudes. This makes it more flexible in processing EEG signals from different channels and enables a unified assessment of signal stability.

[0099] Hurst Index: Long-Term Memory Analysis: The Hurst index measures the long-term memory or self-similarity of time series signals. In electroencephalogram (EEG) signals, the Hurst index can reveal the dynamic characteristics of the signal and help identify long-term dependencies. During a seizure, the self-similarity of EEG signals may change, and the Hurst index can capture this change. Complexity Assessment: The Hurst index also reflects the complexity of the signal. A lower Hurst index generally indicates a more complex signal, while a higher Hurst index indicates a smoother signal. This complexity assessment helps distinguish between normal EEG waves and epileptic waves.

[0100] Wavelet Entropy: Quantifying Complexity and Nonlinear Features: Wavelet entropy quantifies signal complexity by analyzing the distribution of wavelet coefficients. It captures the complexity and nonlinear features of signals at different time scales, which is particularly important for analyzing non-stationary EEG signals. Multi-scale Analysis: Wavelet entropy combines the multi-scale properties of wavelet transform, enabling the analysis of signal complexity at different scales. This multi-scale analysis can more comprehensively reflect the dynamic changes of signals, helping to more accurately detect epileptic waves.

[0101] 2. Improve detection accuracy and reliability. Enrich the feature set: By introducing features such as CV value, Hurst exponent, and wavelet entropy, the model can analyze EEG signals from multiple perspectives, thereby more accurately identifying epileptic wavebands. These features not only include the statistical properties of the signal (such as mean, maximum, etc.), but also cover the signal's complexity, stability, and long-term memory, enabling the model to understand the signal more comprehensively. Enhance the model's generalization ability: The addition of these features allows the model to better adapt to different types of epileptic wavebands, improving its generalization ability across different datasets. For example, wavelet entropy can capture the nonlinear characteristics of the signal, while the Hurst exponent can reflect the signal's long-term memory; the combination of these features allows the model to more accurately identify various types of epileptic waves.

[0102] 3. Enhanced Model Robustness. Increased Noise Resistance: Features such as CV value and wavelet entropy exhibit robustness to signal noise. The CV value assesses signal stability through relative volatility, effectively ignoring the impact of noise to some extent; wavelet entropy extracts the complexity features of the signal through multi-scale analysis, effectively filtering noise. Adaptability to Complex Signals: Hurst exponent and wavelet entropy can effectively handle complex and non-stationary signals. These features enable the model to extract useful information more accurately and reduce misjudgments when faced with complex EEG signals.

[0103] 4. Enhanced Clinical Application Value. Diagnostic Aid: The introduction of these features allows the model to more accurately detect epileptic wavebands, providing clinicians with more reliable diagnostic evidence. For example, the Hurst index can reveal long-term memory changes in EEG signals, helping doctors more accurately determine the pattern of epileptic seizures. Research Support: These features not only aid in clinical diagnosis but also support research into the pathogenesis of epilepsy. By analyzing features such as CV values, Hurst index, and wavelet entropy during epileptic seizures, researchers can better understand the neurophysiological mechanisms of epilepsy.

[0104] 5. Synergistic Effects with Other Features. Complementarity: Features such as CV value, Hurst exponent, and wavelet entropy are complementary to other traditional features (such as mean, maximum, and minimum values). These features describe the characteristics of the signal from different perspectives, and combining them can provide more comprehensive signal information. Enhanced Feature Representation: By combining these features with other features, the model can more accurately represent the characteristics of the signal, thereby improving detection accuracy.

[0105] Compared with existing technologies, the present invention, by introducing key features such as CV value, Hurst exponent, and wavelet entropy, can more comprehensively describe the characteristics of EEG signals, improve the accuracy and reliability of epilepsy band detection, enhance the robustness and generalization ability of the model, and provide stronger support for clinical diagnosis and research. The addition of these features enables the model to analyze signals from multiple perspectives, thereby more accurately identifying epilepsy bands.

[0106] Please see Figure 6 The second embodiment of the present invention also provides a device 2 for recognizing epileptic waves induced by high-dose transcranial magnetic stimulation, used to implement the above-described method. The device 2 includes:

[0107] Receiver module 20: Used to acquire brainwave data induced by transcranial magnetic stimulation;

[0108] Preprocessing module 21: used to process the obtained filtered data and stimulus points;

[0109] Key feature extraction module 22: used to extract key features;

[0110] Epilepsy band detection module 23: used to detect suspected epileptic bands based on the key features;

[0111] Post-processing module 24: used to post-process the suspected epileptic segment to obtain the real epileptic wave.

[0112] Specifically, the receiving module 20 is the foundation of the entire device, and its core function is to acquire brainwave data induced by transcranial magnetic stimulation (TMS). In practice, high-precision electroencephalography (EEG) equipment is typically used, with electrodes placed on the patient's scalp to capture the weak electrical signals generated by brain activity. These signals change with the application of TMS, thus providing a data basis for subsequent analysis. For example, after receiving TMS, epileptic patients may exhibit abnormal brainwave characteristics such as sharp waves and spikes. This data will be accurately acquired by the receiving module 20 for processing and analysis by subsequent modules.

[0113] The preprocessing module 21 performs preliminary processing on the raw EEG data acquired by the receiving module 20 to remove noise and interference, and extract useful stimulation point information. EEG signals are often affected by various noises during acquisition, such as electrooculography (EOG), electromyography (EMG), and environmental electromagnetic interference. These noises can mask or interfere with the true characteristics of epileptic waves, thus requiring preprocessing operations such as filtering. For example, a bandpass filter can be used to limit the EEG signal to a specific frequency range, removing high-frequency noise and low-frequency drift. Furthermore, the preprocessing module 21 can also identify the specific points of action of transcranial magnetic stimulation (TMS), i.e., stimulation points, using algorithms. This is crucial for subsequent analysis of the relationship between epileptic waves and stimulation, as the occurrence of epileptic waves may be related to the location and intensity of the stimulation point. The key feature extraction module 22 is one of the core components of the identification device. Its purpose is to extract feature information related to epileptic waves from the preprocessed EEG data. These key features reflect the unique properties of epileptic waves, thus providing a basis for subsequent detection. For example, time-domain features of brain waves, such as amplitude, frequency, and waveform complexity, can be extracted; frequency-domain features, such as energy distribution in different frequency bands, can also be extracted. Taking spikes, a common feature during epileptic seizures, as an example, their amplitude is usually high and their frequency is fast. The feature extraction module can accurately extract these features for further analysis and judgment by subsequent modules. The role of the epilepsy band detection module 23 is to identify suspected epileptic bands based on the features extracted by the key feature extraction module 22. This module usually uses advanced algorithms, such as machine learning or deep learning algorithms, to analyze and classify the extracted features. For example, a support vector machine (SVM) classifier can be trained, and the extracted features can be input into the classifier. The learned model can then be used to determine whether a brain wave segment belongs to an epileptic band. If a segment of brain wave is detected to be highly similar to the features of an epileptic wave, then that segment of brain wave will be marked as a suspected epileptic band. It should be noted that due to the complexity of brain waves and individual differences, the epilepsy band detection module 23 may have a certain false positive rate, so post-processing is required for further confirmation. The post-processing module 24 is the final stage of the identification device. Its main task is to further analyze and process the suspected epileptic segments identified by the epilepsy band detection module 23 to ultimately determine the true epileptic wave. The post-processing module 24 can perform verification and correction in various ways. For example, it can combine the patient's clinical symptoms and medical history to make a comprehensive judgment on suspected epileptic segments; it can also use statistical analysis methods to compare and screen multiple suspected epileptic segments, removing those that may be misjudged due to noise or artifacts. Finally, after rigorous screening and confirmation by the post-processing module 24, accurate true epileptic waves are obtained, providing a reliable basis for clinical diagnosis and treatment.

[0114] In a real clinical application, during transcranial magnetic stimulation (TMS) treatment, the receiving module 20 acquires the patient's electroencephalogram (EEG) data using a high-precision EEG device. The preprocessing module 21 first filters this data to remove interfering signals such as electrooculography (EOG) and electromyography (EMG), and accurately identifies the stimulation points for TMS. The key feature extraction module 22 then extracts features related to epileptic waves from the filtered data, such as amplitude, frequency, and energy distribution. The epileptic wave segment detection module 23 uses a trained deep learning model to identify several suspected epileptic segments based on the extracted features. Finally, the post-processing module 24 combines the patient's medical history and clinical manifestations to comprehensively analyze the suspected epileptic segments, ultimately determining one segment as the actual epileptic wave, and feeding the result back to the doctor so that the doctor can develop a more precise treatment plan based on the information of this actual epileptic wave.

[0115] It should be noted that although several modules or units for executing the process have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0116] Furthermore, although the steps of the method in this disclosure are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps may be omitted, multiple steps may be combined into one step, and / or a step may be broken down into multiple steps.

[0117] From the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, mobile terminal, or network device, etc.) to execute the methods according to the embodiments of this disclosure.

[0118] The third embodiment of the present invention also provides a readable storage medium having computer program instructions stored thereon, which, when executed, implement the steps of the method described above.

[0119] In some possible embodiments, various aspects of this disclosure may also be implemented as a program product comprising program code that, when the program product is run on a terminal device, causes the terminal device to perform the steps described in the foregoing “” section of this specification according to various exemplary embodiments of this disclosure.

[0120] Please see Figure 7 The fourth embodiment of the present invention also provides a program product 3, which includes computer program instructions 30, which, when executed, implement the steps of the method described above.

[0121] Please see Figure 8 The fifth embodiment of the present invention also provides an electronic device 4, specifically, the electronic device 4 includes a processor 40 and a memory 41; the memory 41 stores a computer program, and the computer program executes the method of any of the above embodiments when run by the processor.

[0122] Furthermore, the processor 40, the communication interface 43, and the memory 41 are connected via a bus 42; the processor 40 is used to execute executable modules, such as computer programs, stored in the memory 41.

[0123] The memory 41 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 43 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc.

[0124] Bus 42 can be an ISA bus, PCI bus, or EISA bus, etc. Buses can be divided into address buses, data buses, control buses, etc. For ease of representation, Figure 4 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0125] The memory 41 is used to store the program. After receiving the execution instruction, the processor 40 executes the program. The method executed by the device for the flow process definition disclosed in any of the foregoing embodiments of the present invention can be applied to the processor 40 or implemented by the processor 40.

[0126] Processor 40 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 40 or by instructions in software form. Processor 40 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 41, and the processor 40 reads the information from memory 41 and, in conjunction with its hardware, completes the steps of the above method.

[0127] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, electronic device, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0128] It should be noted that similar reference numerals and letters in the accompanying drawings indicate similar items. Therefore, once an item is defined in one accompanying drawing, it does not need to be further defined and explained in subsequent accompanying drawings. In addition, the terms "first," "second," "third," etc. are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0129] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this invention.

Claims

1. A method for identifying epileptic waves induced by high-dose transcranial magnetic stimulation, characterized in that: Includes the following steps: Acquire EEG data induced by transcranial magnetic stimulation; The brainwave signal is processed to extract stimulation points and perform global filtering to obtain filtered data and stimulation points. The filtered data after the stimulation point is divided into multiple segments according to a preset truncation unit, and key features are extracted from each segment. The key features are input into a preset epilepsy band detection model for detection, resulting in multiple suspected epilepsy segments. The suspected epileptic segments were post-processed to obtain the actual epileptic waves.

2. The method for identifying epileptic waves induced by high-dose transcranial magnetic stimulation as described in claim 1, characterized in that: The training of the pre-defined epilepsy band detection model includes the following steps: The pre-defined data is divided into a training set and a test set; Extract the key features from the data in the training set and the test set; The key features of the training set are input into a preset original model for training to obtain a model to be verified. The key features of the test set are input into the model to be verified to obtain the epilepsy band detection model.

3. The method for identifying epileptic waves induced by high-dose transcranial magnetic stimulation as described in claim 2, characterized in that: The key features include: Mean, maximum absolute value, true maximum value, minimum absolute value, and true minimum value.

4. The method for identifying epileptic waves induced by high-dose transcranial magnetic stimulation as described in claim 2, characterized in that: The key features include the CV value, which describes signal stability.

5. The method for identifying epileptic waves induced by high-dose transcranial magnetic stimulation as described in claim 2, characterized in that: The key features include the Hurst index, used to analyze the complexity and memory of EEG signals.

6. The method for identifying epileptic waves induced by high-dose transcranial magnetic stimulation as described in claim 2, characterized in that: The key features include wavelet entropy, used to quantify the complexity and nonlinear characteristics of the signal.

7. A device for identifying epileptic brain waves induced by high-dose transcranial magnetic stimulation (TMS), used to implement the method for identifying epileptic brain waves induced by high-dose TMS as described in any one of claims 1-6, characterized in that: include: Receiver module: Used to acquire EEG data induced by transcranial magnetic stimulation; Preprocessing module: used to process the obtained filtered data and stimulus points; Key feature extraction module: used to extract key features; Epilepsy band detection module: used to detect suspected epileptic bands based on the key features; Post-processing module: used to post-process the suspected epileptic segments to obtain the true epileptic waves.

8. The device for recognizing epileptic waves induced by high-dose transcranial magnetic stimulation as described in claim 7, characterized in that: It also includes an epilepsy band detection model construction module, which is used to construct the epilepsy band detection model in the epilepsy band detection module.

9. A readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed, they implement the steps of the method as described in any one of claims 1-6.

10. A program product comprising computer program instructions, characterized in that: When the computer program instructions are executed, they implement the steps of the method as described in any one of claims 1-6.

Images