A picu seizure detection system and storage medium
By acquiring EEG, ECG, and EMG signals in real time in a PICU setting, and combining them with feature extraction and scoring mechanisms, a highly sensitive and low-false-detection-rate epileptic seizure detection method was achieved, overcoming the shortcomings of real-time detection in existing technologies.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- PEKING UNIVERSITY FIRST HOSPITAL (PEKING UNIVERSITY FIRST CLINICAL MEDICAL COLLEGE)
- Filing Date
- 2022-08-31
- Publication Date
- 2026-06-02
AI Technical Summary
Existing epileptic seizure detection technologies cannot achieve real-time detection in PICU settings, have a high false detection rate, and lack effective differentiation between seizure waveforms and artifact waveforms.
The system uses a data acquisition module to acquire EEG, ECG, and EMG signals in real time. Through suspicious fragment detection, feature extraction, and epilepsy detection modules, combined with the features of EEG, ECG, and EMG signals, it uses preprocessing, sliding rectangular window segmentation, and feature scoring mechanisms to detect epileptic seizures in real time.
It enables real-time seizure detection in PICU settings with a sensitivity of up to 90%, a low false detection rate, and approximately 2 detections per hour.
Smart Images
Figure CN117653021B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of biomedical technology, and in particular to a detection system and storage medium for epileptic seizures in a PICU (Pediatric Intensive Care Unit). Background Technology
[0002] Epilepsy, commonly known as "seizures," is a common neurological disorder characterized by recurrent seizures caused by abnormal electrical activity in the brain. It affects 200 to 300 people per 100,000. Because seizures involve sudden, generalized convulsions and loss of consciousness, they pose various risks. Childhood epilepsy is common in children, and pediatric intensive care units (PICUs) encompass critically ill patients from many clinical departments. Therefore, PICUs provide close monitoring of various brain function systems in critically ill patients, and EEG is the best method for evaluating changes in brain function due to its high sensitivity and non-invasiveness. However, because the conditions of critically ill patients in the ICU are complex, EEG results are influenced by many factors, leading to interpretations and understanding that differ from standard EEG results. Currently, there is no specific epilepsy detection technology for PICU patients.
[0003] Currently, the diagnosis of epileptic seizures, besides clinical observation, is mainly achieved through the analysis of electroencephalograms (EEGs). Long-term EEG analysis is still performed manually by doctors, which is time-consuming and labor-intensive. Computer-aided detection can reduce the workload of doctors and improve efficiency. Existing seizure detection technologies have a high false positive rate, failing to effectively distinguish between seizure-phase waveforms and artifacts. Furthermore, seizures in the PICU are prolonged, and there is currently no specific real-time seizure detection technology designed for the PICU setting. Summary of the Invention
[0004] The technical problem solved by the solution provided in the embodiments of the present invention is the inability to achieve real-time epileptic seizure detection in PICU scenarios.
[0005] A detection system for PICU epileptic seizures according to an embodiment of the present invention includes:
[0006] The data acquisition module is used to collect EEG, ECG, and EMG signal data of epilepsy subjects in real time.
[0007] The suspicious segment detection module is used to divide the EEG signal data into multiple EEG signal data segments according to time, and to obtain suspicious epileptic EEG signal data segments by detecting each EEG signal data segment separately.
[0008] The suspicious segment feature extraction module is used to obtain electrocardiogram (ECG) and electromyogram (EMG) signal data segments corresponding to the suspicious epilepsy EEG signal data segments based on the suspicious epilepsy EEG signal data segments, and to obtain EEG suspicious segment features, ECG suspicious segment features, and EMG suspicious segment features by performing feature extraction on the suspicious epilepsy EEG signal data segments, ECG signal segments, and EMG signal segments respectively.
[0009] The epilepsy detection module is used to detect whether the suspected epileptic EEG signal data fragments are actual epileptic seizure fragments by utilizing the suspected EEG fragment features, suspected ECG fragment features, and suspected EMG fragment features.
[0010] Preferably, it further includes:
[0011] The data preprocessing module is used to preprocess the electroencephalogram (EEG) signal data, the electrocardiogram (ECG) signal data, and the electromyogram (EMG) signal data respectively to obtain preprocessed EEG signal data, preprocessed ECG signal data, and preprocessed EMG signal data.
[0012] The preprocessing includes: channel filtering, power frequency noise removal, baseline offset and high frequency noise removal, and rereference.
[0013] Preferably, the suspicious fragment detection module includes:
[0014] The segmentation unit is used to segment the EEG signal data after power frequency noise removal using a sliding rectangular window of fixed width to obtain multiple EEG signal data segments of fixed length.
[0015] The suspicious segment detection unit is used to acquire the signal amplitude of each EEG signal data segment and determine whether the absolute value of the signal amplitude is within a preset amplitude range. When the absolute value of the signal amplitude is determined to be within the preset amplitude range, the line length of the EEG signal data segment is calculated. When the line length reaches the line length reference value, the EEG signal data segment is regarded as a suspicious epileptic EEG signal data segment.
[0016] Preferably, calculating the line length of the EEG signal data segment includes:
[0017]
[0018] Among them, L l N represents the line length. ll S(t) represents the window width used in the online length calculation, S(t) represents the signal of the EEG signal data segment, t represents the time sampling point, and abs[] is the function for calculating the absolute value.
[0019] Preferably, the suspicious fragment feature extraction module includes:
[0020] The acquisition unit is used to acquire time period information of the suspected epileptic electroencephalogram (EEG) signal data segment, and to acquire an ECG signal data segment corresponding to the time period from the preprocessed ECG signal data according to the time period information, and to acquire an electromyography (EMG) signal data segment corresponding to the time period from the preprocessed EMG signal data according to the time period information.
[0021] The suspicious segment feature extraction unit is used to obtain suspicious EEG segment features by extracting features from the suspicious epileptic EEG signal data segments; to obtain suspicious ECG segment features by extracting features from the ECG signal data segments; and to obtain suspicious ECG segment features by extracting features from the ECG signal data segments.
[0022] Preferably, the features of the suspected EEG segments include the temporal correlation coefficient and the temporal rate of change; the features of the suspected ECG segments include the heart rate value and whether a heart rate burst occurs; and the features of the suspected EMG segments include whether an EMG burst occurs and whether an EMG burst delay occurs.
[0023] Preferably, the epilepsy detection module includes:
[0024] The calculation unit is used to calculate the total feature score of the suspected EEG segment features, suspected ECG segment features, and suspected EMG segment features based on a preset feature scoring mechanism table.
[0025] An epilepsy detection unit is configured to detect the suspicious epilepsy EEG signal data segment as a real epileptic seizure segment when the total feature score is greater than a preset feature threshold, and to detect the suspicious epilepsy EEG signal data segment as a spurious epileptic seizure segment when the total feature score is not greater than the preset feature threshold.
[0026] According to an embodiment of the present invention, a storage medium stores a program, which, when executed by a processor, performs the following steps:
[0027] Real-time acquisition of electroencephalogram (EEG), electrocardiogram (ECG), and electromyogram (EMG) data from subjects with epilepsy;
[0028] The EEG signal data is divided into multiple EEG signal data segments according to time, and each EEG signal data segment is detected separately to obtain suspected epileptic EEG signal data segments.
[0029] Based on the suspected epilepsy EEG signal data segment, obtain the corresponding ECG signal data segment and EMG signal data segment, and obtain the suspected EEG segment, ECG segment, and EMG segment features by performing feature extraction on the suspected epilepsy EEG signal data segment, ECG segment, and EMG segment respectively.
[0030] Using the features of the suspected EEG segments, suspected ECG segments, and suspected EMG segments, it is possible to detect whether the suspected epileptic EEG signal data segments are actual epileptic seizure segments.
[0031] Preferably, when the program is executed by the processor, it further includes:
[0032] The electroencephalogram (EEG) signal data, the electrocardiogram (ECG) signal data, and the electromyogram (EMG) signal data are preprocessed to obtain preprocessed EEG signal data, preprocessed ECG signal data, and preprocessed EMG signal data.
[0033] The preprocessing includes: channel filtering, power frequency noise removal, baseline offset and high frequency noise removal, and rereference.
[0034] Preferably, when the program is executed by the processor, it divides the EEG signal data into multiple EEG signal data segments according to time, and obtains suspected epileptic EEG signal data segments by detecting each EEG signal data segment separately, including:
[0035] The EEG signal data after power frequency noise removal is segmented using a sliding rectangular window of fixed width to obtain multiple EEG signal data segments of fixed length.
[0036] The signal amplitude of each EEG signal data segment is acquired, and it is determined whether the absolute value of the signal amplitude is within a preset amplitude range. When it is determined that the absolute value of the signal amplitude is within the preset amplitude range, the line length of the EEG signal data segment is calculated. When the line length reaches the line length reference value, the EEG signal data segment is regarded as a suspected epileptic EEG signal data segment.
[0037] According to the solution provided in the embodiments of the present invention, different seizure characteristics of different seizure types are taken into account, and relatively good detection effect is achieved for different types of seizures. The sensitivity of seizure detection is as high as 90%, and the false detection rate is about 2 times per hour. Attached Figure Description
[0038] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to understand the invention, but do not constitute an undue limitation of the invention. In the drawings:
[0039] Figure 1 This is a schematic diagram of a PICU epileptic seizure detection system provided in an embodiment of the present invention;
[0040] Figure 2 This is a schematic diagram of the specific structure of a PICU epileptic seizure detection system provided in an embodiment of the present invention;
[0041] Figure 3 This is a flowchart of the detection process for epileptic seizures in the PICU provided in an embodiment of the present invention;
[0042] Figure 4 This is a schematic diagram of an example of an attack detection fragment provided in an embodiment of the present invention. Detailed Implementation
[0043] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the preferred embodiments described below are only for illustration and explanation of the present invention and are not intended to limit the present invention.
[0044] Figure 1 This is a schematic diagram of a PICU epileptic seizure detection system provided in an embodiment of the present invention, as shown below. Figure 1 As shown, the system includes: a data acquisition module 101, used to acquire EEG signal data, ECG signal data, and EMG signal data of epileptic subjects in real time; a suspicious segment detection module 102, used to divide the EEG signal data into multiple EEG signal data segments according to time, and obtain suspicious epileptic EEG signal data segments by detecting each EEG signal data segment separately; a suspicious segment feature extraction module 103, used to obtain ECG signal data segments and EMG signal data segments corresponding to the suspicious epileptic EEG signal data segments based on the suspicious epileptic EEG signal data segments, and obtain EEG suspicious segment features, ECG suspicious segment features, and EMG suspicious segment features by extracting features from the suspicious epileptic EEG signal data segments, ECG suspicious segment features, and EMG suspicious segment features respectively; and an epilepsy detection module 104, used to detect whether the suspicious epileptic EEG signal data segments are actual epileptic seizure segments using the EEG suspicious segment features, ECG suspicious segment features, and EMG suspicious segment features.
[0045] like Figure 2 As shown, the embodiments of the present invention further include: a data preprocessing module, used to preprocess the electroencephalogram (EEG) signal data, the electrocardiogram (ECG) signal data and the electromyogram (EMG) signal data respectively to obtain preprocessed EEG signal data, preprocessed ECG signal data and preprocessed EMG signal data; wherein, the preprocessing includes: channel filtering, power frequency noise removal, baseline offset and high frequency noise removal, and rereference.
[0046] Furthermore, such as Figure 2 As shown, the suspicious segment detection module 102 includes: a segmentation unit, used to segment the EEG signal data after power frequency noise removal using a sliding rectangular window of fixed width to obtain multiple EEG signal data segments of fixed length; and a suspicious segment detection unit, used to acquire the signal amplitude of each EEG signal data segment and determine whether the absolute value of the signal amplitude is within a preset amplitude range. When the absolute value of the signal amplitude is determined to be within the preset amplitude range, the line length of the EEG signal data segment is calculated. When the line length reaches the line length reference value, the EEG signal data segment is regarded as a suspicious epileptic EEG signal data segment.
[0047] The calculation of the line length of the EEG signal data segment includes:
[0048]
[0049] Among them, L l N represents the line length. ll S(t) represents the window width used in the online length calculation, S(t) represents the signal of the EEG signal data segment, t represents the time sampling point, and abs[] is the function for calculating the absolute value.
[0050] Furthermore, such as Figure 2 As shown, the suspicious segment feature extraction module 103 includes: an acquisition unit, configured to acquire time period information of the suspicious epileptic EEG signal data segment, and acquire an ECG signal data segment corresponding to the time period from the preprocessed ECG signal data according to the time period information, and acquire an EMG signal data segment corresponding to the time period from the preprocessed EMG signal data according to the time period information; and a suspicious segment feature extraction unit, configured to obtain EEG suspicious segment features by performing feature extraction on the suspicious epileptic EEG signal data segment; obtain ECG suspicious segment features by performing feature extraction on the ECG signal data segment; and obtain EMG suspicious segment features by performing feature extraction on the EMG signal data segment.
[0051] The features of the suspected EEG segments include the temporal correlation coefficient and the temporal rate of change; the features of the suspected ECG segments include the heart rate value and whether a heart rate burst occurs; and the features of the suspected EMG segments include whether an EMG burst occurs and whether an EMG burst delay occurs.
[0052] Furthermore, such as Figure 2As shown, the epilepsy detection module 104 includes: a calculation unit, used to calculate the total feature score of the suspected EEG segment feature, suspected ECG segment feature, and suspected EMG segment feature according to a preset feature scoring mechanism table; and an epilepsy detection unit, used to detect the suspected epilepsy EEG signal data segment as a real epileptic seizure segment when the total feature score is greater than a preset feature threshold, and to detect the suspected epilepsy EEG signal data segment as a fake epileptic seizure segment when the total feature score is not greater than the preset feature threshold.
[0053] This application provides a storage medium storing a program, which, when executed by a processor, performs the following steps:
[0054] Step S1: Real-time acquisition of EEG, ECG, and EMG signal data from epilepsy subjects;
[0055] Step S2: Divide the EEG signal data into multiple EEG signal data segments according to time, and obtain suspected epileptic EEG signal data segments by detecting each EEG signal data segment separately;
[0056] Step S3: Based on the suspected epilepsy EEG signal data segment, obtain the corresponding ECG signal data segment and EMG signal data segment, and extract features from the suspected epilepsy EEG signal data segment, ECG signal data segment and EMG signal data segment respectively to obtain the features of the suspected EEG segment, the features of the suspected ECG segment and the features of the suspected EMG segment.
[0057] Step S4: Using the suspected EEG segment features, suspected ECG segment features, and suspected EMG segment features, detect whether the suspected epileptic EEG signal data segment is a real epileptic seizure segment.
[0058] Furthermore, when the program is executed by the processor, after real-time acquisition of EEG signal data, ECG signal data, and EMG signal data of the epileptic subject, it further includes: preprocessing the EEG signal data, ECG signal data, and EMG signal data respectively to obtain preprocessed EEG signal data, preprocessed ECG signal data, and preprocessed EMG signal data; wherein, the preprocessing includes: channel filtering, power frequency noise removal, baseline offset and high frequency noise removal, and rereference.
[0059] Furthermore, when the program is executed by the processor, it divides the EEG signal data into multiple EEG signal data segments according to time, and obtains suspected epilepsy EEG signal data segments by detecting each EEG signal data segment separately. This includes: segmenting the EEG signal data after power frequency noise removal using a sliding rectangular window of fixed width to obtain multiple EEG signal data segments of fixed length; acquiring the signal amplitude of each EEG signal data segment and determining whether the absolute value of the signal amplitude is within a preset amplitude range; when the absolute value of the signal amplitude is determined to be within the preset amplitude range, calculating the line length of the EEG signal data segment; and when the line length reaches a line length reference value, then the EEG signal data segment is identified as a suspected epilepsy EEG signal data segment.
[0060] This invention fully considers the different manifestations of different seizure types, and combines the characteristics of electrocardiogram (ECG) and electromyography (EMG) signals with simultaneous analysis of EEG signals, such as... Figure 3 As shown, it specifically includes the following:
[0061] 1. Real-time acquisition and preprocessing of EEG, ECG, and EMG signals; data preprocessing includes downsampling, rereference, notch filtering to remove power frequency, and bandpass filtering, specifically including:
[0062] EEG, ECG, and EMG signals are acquired using a clinically standard amplifier with a sampling rate of 200Hz or higher and 16 or more EEG channels. To improve computational efficiency, the sampling rate can be resampled to 200Hz.
[0063] Remove unnecessary channels. EEG signals may record channel information that is not needed later, such as bilateral mastoid points. These can be removed and do not need to be included in subsequent analysis.
[0064] Power frequency noise removal. Power frequency noise is caused by the power system; the AC power frequency is 50Hz. Power frequency noise can interfere with electrical and electronic equipment, causing abnormal equipment operation and also suppressing the effective information of the EEG signal. A notch filter with a frequency of 50Hz and / or multiples of 50Hz is designed to eliminate power frequency noise interference, obtaining the signal after power frequency noise removal.
[0065] Filtering and denoising. To remove baseline drift and high-frequency noise, a 1.6-70Hz bandpass filter was designed for EEG and EMG signals; a 5.3-40Hz bandpass filter was designed for ECG signals.
[0066] Rereference. Rereference methods include monopolar rereference, bipolar rereference, Laplacian rereference, and common average rereference. For EEG signals, a common average rereference is used, where the mean of all data from the entire brain is used as the reference data. Each value of the EEG signal is subtracted from the reference data to obtain the rereferenced EEG signal S(t).
[0067] 2. Suspicious Segment Extraction: Detection is performed in 20-second windows, filtering EEG segments with high-frequency or low-frequency signals standing out against the background as suspicious segments. The real-time detection step size is set to 1 second.
[0068] During EEG signal acquisition, electrodes may detach, appearing on the EEG as amplitudes much larger than normal signals. Normal EEG signals may also contain interfering electromyographic (EMG) signals, appearing as densely jittering waveforms on the EEG. These signals are considered artifacts in the normal EEG signal. Furthermore, EEG signals during epileptic seizures often stand out against the background, making them difficult to distinguish from artifact fragments. Therefore, it's possible to first determine whether a segment is suspicious based on whether it stands out against the background activity, and then score the suspicious segments based on the unique characteristics of the seizure phase, thereby achieving the purpose of seizure detection.
[0069] To provide a comparative effect, 20 minutes of EEG data without seizures were used as a baseline for setting the threshold.
[0070] Suspicious segments must meet both of the following criteria simultaneously.
[0071] (1) Check whether the absolute value amplitude of the signal after removing power frequency noise is within the range of 300-800 micro-amplitude;
[0072] (2) Perform line length transformation on the preprocessed signal segment to obtain the line length of the signal segment, and detect the line length L of the signal segment. l Has the line length baseline value been reached? eeg If this condition is met, the segment is considered suspicious. Among them, ll eeg The selectable range is 2-8 times the median of the 20-minute comparison signal after line length transformation;
[0073] The formula for line length transformation is as follows:
[0074]
[0075] Where S(t) represents the signal S(t) after removing power frequency noise, t represents the time sampling point, and N llL represents the window width used in the online length calculation, abs[] is the function for calculating the absolute value, and L... l Indicates the line length.
[0076] 3. Feature Extraction: For suspicious segments, calculate the corresponding heart rate and its rate of change, detect electromyographic signals within the corresponding time period, and determine if electromyographic bursts are present. For suspicious segments, a temporal correlation matrix also needs to be calculated.
[0077] 31. Calculation of heart rate and its rate of change
[0078] For comparison, 20 minutes of ECG data without seizures were used as a baseline for threshold setting. In the preprocessing stage, the ECG signal was resampled to 200Hz, and power frequency components were removed, with a filtering range of 5.3-40Hz. Data was processed in blocks, with an average segment length of 150 seconds. A corresponding threshold was set for each segment.
[0079] thresh = 0.8 * mean * std
[0080] Where mean and std represent the mean and variance of the 20-minute ECG signal used for comparison after line length transformation, respectively.
[0081] The system detects peak points in consecutive data points exceeding a threshold and calculates heart rate values using two adjacent peak points. The rate of change of heart rate is calculated by the difference between the preceding and following heart rate values. Outliers in both heart rate and rate of change are removed, and the data is smoothed. The smoothed heart rate is then used as a feature for burst detection. The classic CUSUM algorithm is employed for heart rate burst detection.
[0082] 32. Electromyographic signal processing
[0083] In the preprocessing stage, the electromyography (EMG) signal is resampled to 200Hz and the power frequency component is removed, with a filtering range of 1.6-70Hz. The processed EMG signal is then smoothed, and the RMS envelope is taken from the smoothed data. The classic CUSUM algorithm is used to detect EMG signal bursts from the enveloped EMG values, and it is determined whether there is a delay between the EMG burst time and the EEG signal abrupt change time.
[0084] 33. Calculate the time-domain correlation matrix
[0085] To prevent eye movement interference, 17 leads (excluding FP1 (left frontal pole) and FP2 (right frontal pole) were selected from the EEG data. After smoothing the data (17 leads), the dataset was divided into 10-second segments, and the time-domain correlation coefficient was calculated. The correlation coefficient was calculated by taking the average of the upper triangular values after matrix autocorrelation as the correlation coefficient for that moment. The rate of change of the time-domain correlation coefficient was then calculated. The threshold for the time-domain correlation matrix was set to five times the median time-domain correlation coefficient of the 20-minute seizure-free segments.
[0086] The formula for calculating the correlation coefficient matrix is: Where C ij These are elements in the voltage matrix of the EEG data. The final correlation coefficient calculation formula is: The correlation coefficient matrix has an n*n dimension.
[0087] 4. Determine whether a real attack occurred by scoring suspicious segments.
[0088] Heart rate, heart rate bursts, electromyographic bursts, electromyographic burst delay, and temporal correlation matrix were used as features to identify suspicious segments. By assigning different weights to each feature value, suspicious segments were scored, and segments with a score higher than 8 points were identified as true seizures.
[0089] 41. Scoring of ECG and EMG signal data
[0090] Add 4 points if a heart rate burst is detected, add 2 points if an electromyography (EMG) burst is detected, add 1 point if an EMG burst is detected with a delay, and add 1 point if the heart rate value is greater than 120 and accounts for more than 10% of the total.
[0091] 42. Scoring of the temporal correlation matrix of EEG signal data
[0092] If the proportion of the time-domain correlation coefficient exceeding the threshold exceeds 10%, add 2 points; if the time-domain correlation coefficient is less than 0.2, add 2 points; if the time-domain change rate exceeds 0.05, add 2 points. The characteristics corresponding to the actual seizure segments are as follows: Figure 4 As shown.
[0093] According to another aspect of this application, a device for detecting epileptic seizures is proposed, comprising: an acquisition unit configured to acquire electroencephalogram (EEG) signals of a subject across multiple lead channels, acquire electromyographic (EMG) signals of a subject across multiple channels, and acquire electrocardiogram (ECG) signals of a subject; and a detection unit configured to extract candidate signal fragments in real time from features of the EEG signals and the ECG and EMG signals across multiple dimensions; and to analyze correlation indices of the candidate signal fragments to determine whether the candidate signal fragments indicate an epileptic seizure.
[0094] In summary, the embodiments of the present invention are applied to real-time monitoring of epileptic seizures in PICU wards, allowing doctors to be notified promptly when a patient experiences an epileptic seizure. This approach includes acquiring electroencephalogram (EEG) signals from multiple lead channels, where the EEG signals have multiple signal segments; extracting candidate signal segments in real-time from multiple dimensions of features of the EEG and electrocardiogram / electromyographic (ECG / EMG) signals; and analyzing the correlation indices of the candidate signal segments to determine whether they indicate an epileptic seizure.
[0095] According to the solution provided in the embodiments of the present invention, it is applied to the real-time monitoring of epileptic seizures in the PICU ward, and the doctor can be notified in a timely manner when a patient has an epileptic seizure.
[0096] Although the present invention has been described in detail above, it is not limited thereto, and those skilled in the art can make various modifications based on the principles of the present invention. Therefore, all modifications made in accordance with the principles of the present invention should be understood to fall within the protection scope of the present invention.
Claims
1. A detection system for epileptic seizures in a PICU, characterized in that, include: The data acquisition module is used to collect EEG, ECG, and EMG signal data of epilepsy subjects in real time. The suspicious segment detection module is used to divide the EEG signal data into multiple EEG signal data segments according to time, and to obtain suspicious epileptic EEG signal data segments by detecting each EEG signal data segment separately. The suspicious segment feature extraction module is used to obtain electrocardiogram (ECG) and electromyogram (EMG) signal data segments corresponding to the suspicious epilepsy EEG signal data segments based on the suspicious epilepsy EEG signal data segments, and to obtain EEG suspicious segment features, ECG suspicious segment features, and EMG suspicious segment features by performing feature extraction on the suspicious epilepsy EEG signal data segments, ECG signal segments, and EMG signal segments respectively. The epilepsy detection module is used to detect whether the suspicious epileptic EEG signal data segment is a real epileptic seizure segment by utilizing the suspicious EEG segment features, suspicious ECG segment features, and suspicious EMG segment features. The PICU refers to the Pediatric Intensive Care Unit.
2. The system according to claim 1, characterized in that, Also includes: The data preprocessing module is used to preprocess the electroencephalogram (EEG) signal data, the electrocardiogram (ECG) signal data, and the electromyogram (EMG) signal data respectively to obtain preprocessed EEG signal data, preprocessed ECG signal data, and preprocessed EMG signal data. The preprocessing includes: channel filtering, power frequency noise removal, baseline offset and high frequency noise removal, and rereference.
3. The system according to claim 2, characterized in that, The suspicious fragment detection module includes: The segmentation unit is used to segment the EEG signal data after power frequency noise removal using a sliding rectangular window of fixed width to obtain multiple EEG signal data segments of fixed length. The suspicious segment detection unit is used to acquire the signal amplitude of each EEG signal data segment and determine whether the absolute value of the signal amplitude is within a preset amplitude range. When the absolute value of the signal amplitude is determined to be within the preset amplitude range, the line length of the EEG signal data segment is calculated. When the line length reaches the line length reference value, the EEG signal data segment is regarded as a suspicious epileptic EEG signal data segment.
4. The system according to claim 3, characterized in that, The calculation of the line length of the EEG signal data segment includes: in, Indicates line length; This refers to the window width used in online length calculations. Signals representing segments of electroencephalogram (EEG) signal data. Indicates the time sampling point. This is for calculating the absolute value function.
5. The system according to claim 3, characterized in that, The suspicious fragment feature extraction module includes: The acquisition unit is used to acquire time period information of the suspected epileptic electroencephalogram (EEG) signal data segment, and to acquire an ECG signal data segment corresponding to the time period from the preprocessed ECG signal data according to the time period information, and to acquire an electromyography (EMG) signal data segment corresponding to the time period from the preprocessed EMG signal data according to the time period information. The suspicious segment feature extraction unit is used to obtain suspicious EEG segment features by extracting features from the suspicious epileptic EEG signal data segments; to obtain suspicious ECG segment features by extracting features from the ECG signal data segments; and to obtain suspicious ECG segment features by extracting features from the ECG signal data segments.
6. The system according to claim 5, characterized in that, The features of the suspicious EEG segments include the temporal correlation coefficient and the temporal rate of change; the features of the suspicious ECG segments include the heart rate value and whether a heart rate burst occurs; the features of the suspicious EMG segments include whether an EMG burst occurs and whether an EMG burst delay occurs.
7. The system according to claim 6, characterized in that, The epilepsy detection module includes: The calculation unit is used to calculate the total feature score of the suspected EEG segment features, suspected ECG segment features, and suspected EMG segment features based on a preset feature scoring mechanism table. An epilepsy detection unit is configured to detect the suspicious epilepsy EEG signal data segment as a real epileptic seizure segment when the total feature score is greater than a preset feature threshold, and to detect the suspicious epilepsy EEG signal data segment as a spurious epileptic seizure segment when the total feature score is not greater than the preset feature threshold.
8. A storage medium, characterized in that, The storage medium stores a detection program for PICU seizures, which, when executed by a processor, performs the following steps: Real-time acquisition of electroencephalogram (EEG), electrocardiogram (ECG), and electromyogram (EMG) data from subjects with epilepsy; The EEG signal data is divided into multiple EEG signal data segments according to time, and each EEG signal data segment is detected separately to obtain suspected epileptic EEG signal data segments. Based on the suspected epilepsy EEG signal data segment, obtain the corresponding ECG signal data segment and EMG signal data segment, and obtain the suspected EEG segment, ECG segment, and EMG segment features by performing feature extraction on the suspected epilepsy EEG signal data segment, ECG segment, and EMG segment respectively. Using the features of the suspected EEG segments, suspected ECG segments, and suspected EMG segments, it is determined whether the suspected epileptic EEG signal data segments are actual epileptic seizure segments. The PICU refers to the Pediatric Intensive Care Unit.
9. The storage medium according to claim 8, characterized in that, When the program is executed by the processor, it also includes: The electroencephalogram (EEG) signal data, the electrocardiogram (ECG) signal data, and the electromyogram (EMG) signal data are preprocessed to obtain preprocessed EEG signal data, preprocessed ECG signal data, and preprocessed EMG signal data. The preprocessing includes: channel filtering, power frequency noise removal, baseline offset and high frequency noise removal, and rereference.
10. The storage medium according to claim 9, characterized in that, When the program is executed by the processor, it divides the EEG signal data into multiple EEG signal data segments according to time, and obtains suspected epileptic EEG signal data segments by detecting each EEG signal data segment separately. The EEG signal data after power frequency noise removal is segmented using a sliding rectangular window of fixed width to obtain multiple EEG signal data segments of fixed length. The signal amplitude of each EEG signal data segment is acquired, and it is determined whether the absolute value of the signal amplitude is within a preset amplitude range. When it is determined that the absolute value of the signal amplitude is within the preset amplitude range, the line length of the EEG signal data segment is calculated. When the line length reaches the line length reference value, the EEG signal data segment is regarded as a suspected epileptic EEG signal data segment.