Spike Detection System, Epilepsy Detection Device

Abstract

The present invention belongs to the technical field of physiological electrical signal detection, and particularly relates to a spike detection system and an epilepsy detection device. This spike detection system first uses a spike detection unit to detect signal data to obtain a spike detection result, then filters out spikes from the spike detection result according to a spike screening unit and displays the spike waveforms in a waveform display area, and finally adjusts the preset value of the sensitivity to change the displayed spike waveforms. It can not only better handle the detection errors caused by the morphological differences of spikes, but also combine the spike morphology with the screening threshold to meet the general requirements, making the entire system highly adaptable.

Description

Technical Field

[0001] The present invention belongs to the technical field of physiological electrical signal detection, and in particular relates to a spike wave detection system and an epilepsy detection device. Background Art

[0002] Epilepsy spike detection is crucial in clinical diagnosis because spikes are the hallmark electrophysiological features of epileptic activity. With the advancement of artificial intelligence and signal processing technology, researchers have developed some automatic spike detection methods to reduce the workload of clinicians. These methods include algorithms based on machine learning and deep learning, which can automatically extract features from EEG and detect spikes. However, due to the significant differences in spike morphology between individuals, these methods have limited versatility among different patients and are difficult to adapt to the individual characteristics of each patient. For example, patent number CN114869301A discloses a method and device for detecting epileptic discharges, which mentions the comparison of at least one of the following waveform features with a waveform template to identify whether there is an epileptic discharge in the signal sub-segment: the amplitude difference of the rising edge of the spike or sharp wave in the signal sub-segment; the duration of the rising edge of the spike or sharp wave in the signal sub-segment; the amplitude difference of the falling edge of the spike or sharp wave in the signal sub-segment; the duration of the falling edge of the spike or sharp wave in the signal sub-segment; the amplitude difference of the rising edge of the slow wave following the falling edge of the spike or sharp wave in the signal sub-segment; the standard deviation of the amplitude before the rising edge of the spike or sharp wave in the signal sub-segment. This technical solution detects spikes based on the time-frequency similarity or frequency division energy of a fixed template. Since the spike wave morphology will change under different epileptic seizure states, the fixed template will have a template range that cannot cover more types of spike waves. If there is a difference between the spike wave waveform and the fixed template, the detection system may have a large number of missed detections or false detections. Summary of the invention

[0003] The present invention provides a technical field of physiological electrical signal detection, and specifically relates to a spike wave detection system and an epilepsy detection device, wherein a screening threshold is adjusted according to the spike wave recognition result to solve the problem of missed detection or false detection of the device.

[0004] In order to solve the above technical problems, the present invention provides a spike wave detection system, comprising: a processor and a data calculation unit respectively connected to the processor, a human-computer interaction machine, a spike wave detection unit, and a spike wave screening unit; wherein the spike wave detection unit detects signal data to obtain a spike wave detection result; the spike wave screening unit screens spike waves from the spike wave detection result to obtain a spike wave screening result; a sensitivity option and a waveform display area are provided on the operation interface of the human-computer interaction machine; the waveform display area is used to display the spike wave waveform in the spike wave screening result; the sensitivity option is used to input a preset value of sensitivity to adjust the displayed spike wave waveform.

[0005] Further, the spike screening unit includes: a first screening module that sets the sensitivity to the quantile corresponding to the amplitude difference threshold of the spike to screen the spike, including: in the spike detection result, using the data calculation unit to calculate the amplitude difference from the peak to the end point of each spike and its mean and standard deviation; sorting the amplitude differences of each spike according to size; using the data calculation unit to calculate the proportion value of the amplitude difference of each spike in the sorted position to obtain the quantile corresponding to each amplitude difference; the processor determines whether the standard deviation of the amplitude difference is greater than the quantile at the corresponding position; if so, taking the amplitude difference at the sensitivity corresponding position as the amplitude difference threshold; if not, taking the difference between the mean of the amplitude difference and N times the standard deviation as the amplitude difference threshold.

[0006] Further, when the sensitivity is set in the range of [0%, 20%), N takes 0.5; when the sensitivity is set in the range of [20%, 80%), N takes 1; when the sensitivity is set in the range of [80%, 100%], N takes 2.

[0007] Further, the spike screening unit further includes: a second screening module that sets the sensitivity to the quantile corresponding to the slope threshold of the spike to screen the spike, including: in the spike detection result, using the data calculation unit to calculate the slope of each spike; sorting the slopes of each spike according to size; using the data calculation unit to calculate the proportion value of the slope of each spike in the sorted position as the sensitivity.

[0008] Further, the spike detection unit includes: a data slicing module that performs sliding window slicing processing on the signal data to obtain data segments and their corresponding start times; a spike detection model, whose input end is the data segment and the output end is the spike detection result of each data segment, and the output end result being 1 indicates that there is a spike signal, and the output end being 0 indicates that there is no spike signal.

[0009] Further, it further includes: a data correction unit; the data correction unit is used to delete the resting data and high-amplitude artifact data in the spike detection result; marking the data segments with a standard deviation greater than the set threshold as spike signals; marking the data segments with a standard deviation less than the set threshold as normal signals, that is, resting data; calculating the difference between the maximum value and the minimum value in each data segment and its mean and standard deviation, and recording the data segments with a difference greater than the mean and standard deviation as artifact signals, that is, high-amplitude artifact data.

[0010] Further, it further includes: a peak positioning unit, which marks the peak positions of spike signals based on the spike detection results, that is, calculates the difference between the maximum value and the minimum value in the spike signal using a data calculation unit, sorts the differences from large to small, and takes the difference at a preset position multiplied by a peak threshold coefficient as the peak threshold; searches for the peak of the spike signal based on the peak threshold and the half-wave width threshold, and takes the sum of the moment corresponding to the peak and the starting moment as the peak position of the spike; merges the peaks with a time difference less than a set time threshold into one peak, and calculates the starting position and the ending position of each peak.

[0011] Further, it further includes: a multi-data integration unit; the multi-data integration unit includes: a preprocessing module, which preprocesses the signal data to obtain first data; a data reconstruction module, which performs independent component decomposition and reconstruction on the signal data to obtain second data; an integration module, which integrates the spike detection results after the spike detection unit detects the first data and the second data respectively.

[0012] Further, the integration of the spike detection results includes: statistically counting the channels in the first data with the number of detected spikes not less than a set value as merged channels, and the channels less than the set value as other channels; in the merged channels, taking the union of the detection results corresponding to the first data and the second data, and merging the spikes with a spike position interval less than a threshold as the same spike in the union; in the other channels, retaining the detection results corresponding to the first data.

[0013] In a second aspect, the present invention provides an epilepsy detection device, including: the spike detection system described above; a data display area is further provided on the operation interface of the human-computer interaction machine to display the quantitative statistical results of epileptic waveforms; the quantitative statistical results of the epileptic waveforms include: the number and peak size of the spike peaks of the EEG signals in the spike screening results.

[0014] The beneficial effect of the present invention is that the spike detection system of the present invention first uses a spike detection unit to detect signal data to obtain spike detection results, then uses a spike screening unit to screen spikes from the spike detection results, and displays the spike waveforms in the spike screening results in a waveform display area. Finally, setting a preset value of sensitivity to adjust the spike screening results can not only better handle the detection errors caused by the morphological differences of spikes, but also combine the spike detection results with the screening threshold to meet the generality requirements, making the entire system have high adaptability.

[0015] Other features and advantages of the present invention will be described in the following specification, and, in part, will be obvious from the specification, or will be understood by implementing the present invention. To make the above objects, features, and advantages of the present invention more obvious and understandable, the following specific embodiments are given, and in conjunction with the accompanying drawings, the detailed description is as follows. Description of the Drawings

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0017] Figure 1 It is the workflow diagram of the spike detection system using two types of data for detection.

[0018] Figure 2 It is the schematic diagram of data sliding window slicing.

[0019] Figure 3 It is the iterative curve graph of the accuracy rate of the spike detection model after training and verification.

[0020] Figure 4 It is the iterative curve graph of the loss of the spike detection model after training and verification.

[0021] Figure 5 It is the confusion matrix of the detection results.

[0022] Figure 6 It is the schematic flow diagram of two types of spike screening modules.

[0023] Figure 7 It is the schematic diagram of the amplitude difference in the waveform graph.

[0024] Figure 8 It is the schematic diagram of the selection of the trial channel.

[0025] Figure 9 It is the schematic diagram of the steepness in the waveform graph.

[0026] Figure 10 It is the schematic diagram of taking the quantile corresponding to the spike slope as the sensitivity in the trial channel.

[0027] Figure 11 It is the schematic flow diagram of obtaining the waveform parameters of the spike.

[0028] Figure 12 It is the schematic diagram of the division of the first half wave and the second half wave in the waveform graph.

[0029] Figure 13 It is the schematic diagram of the calculation process of the termination point of the second half wave.

[0030] Figure 14 It is the schematic diagram of the calculation process of the starting point of the first half wave.

[0031] Figure 15 It is the detection result of the C4 channel of epilepsy patients using a non-overlapping sliding window of 1s window.

[0032] Figure 16 It is the detection result of the C4 channel of epilepsy patients with an 80% overlapping sliding window of 1 s window.

[0033] Figure 17 It is the detection result of setting 0% sensitivity for multiple channels of epilepsy patients.

[0034] Figure 18 It is the detection result of setting 25% sensitivity for multiple channels of epilepsy patients. Detailed implementation manners

[0035] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Apparently, the described embodiments are some but not all of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0036] See Figures 1-14 , this embodiment provides a spike detection system, including: a processor and a data calculation unit, a human-machine interaction machine, a spike detection unit, and a spike screening unit respectively connected to the processor; wherein the spike detection unit detects signal data to obtain a spike detection result; the spike screening unit screens spikes from the spike detection result to obtain a spike screening result; a sensitivity option and a waveform display area are provided on the operation interface of the human-machine interaction machine; the waveform display area is used to display the spike waveforms in the spike screening result; the sensitivity option is used to input a preset value of sensitivity to adjust the displayed spike waveforms. The spike detection unit includes but is not limited to: functional modules such as a data slicing module and a spike detection model. Specifically, the data slicing module is used to perform sliding window slicing processing on the data, the spike detection model is used to detect whether spike signals are contained in each data slice and mark the spike slices, the data correction unit is used to eliminate mis-identified spike slices, and the peak positioning unit calculates the waveform parameters and spike positions of the spikes based on the spike slices, and finally is used to display the spike waveforms in the waveform display area or calculate the screening threshold (i.e., sensitivity) of the spikes. Generally, corresponding functions can be selected on the operation interface of the human-machine interaction machine. When the preset value of sensitivity (i.e., the screening threshold of the spikes) is set, the data passes through the spike detection unit (data slicing module, spike detection model, data correction unit, peak positioning unit) and the spike screening unit in sequence, and finally the spike waveforms in the spike screening result are displayed in the waveform display area (such as Figure 17 or Figure 18 the waveforms highlighted in Figure 17If there are many obvious false detections of medium-spikes, increase the preset value of the sensitivity; conversely, if there are many missed detections of spikes, decrease the preset value of the sensitivity.

[0037] In some embodiments, the spike detection system can complete the online detection work and data calculation work with the aid of a computer and its corresponding software and hardware. Its working method can be stored in a memory or a readable storage medium to form a computer program or a computer program product. When the computer program is executed by a processor, the steps of the working method are implemented. See Figure 1 , and its working method includes the following steps:

[0038] Step S1, sliding window slicing processing. Use the data slicing module in the spike detection unit to perform sliding window slicing processing on the online multi-channel EEG data, and output data segments and the start time of each slice through an operation interface or a display interface. As Figure 2 shown, to ensure that the spike can be covered by at least one complete window, overlapping sliding window slicing processing is performed on the data. Assume that the duration of a single-channel data is L seconds, the data is sliced with a sliding window, the window length is T, and the step length is ΔT = 200ms; then the start time of the i-th slice can also be determined . For example, perform multiple 1s data truncation processes with random non-repeating starting points on each 1.5s spike data to ensure that the spike slices contain at least one complete spike. Similarly, perform multiple 1s data truncation processes with random non-repeating starting points on the background EEG data.

[0039] Step S2, data normalization processing. Perform zscore normalization processing on each 1s slice of data.

[0040] Step S3, spike detection. Input the data slices into the trained spike detection model, and output the spike detection results of each slice. Calculate the standard deviation of the slice data, mark the slices with a standard deviation less than the preset value th1 as normal EEG and update the corresponding labels; calculate the difference between the maximum value and the minimum value of each slice data channel by channel, and calculate the mean and standard deviation of all maximum and minimum value differences of each channel. Mark the slices in each channel that are greater than the mean plus the standard deviation of their respective channels as artifact slices and update the corresponding labels. That is, a label of 1 indicates the presence of a spike signal, and a label of 0 indicates the absence of a spike signal.

[0041] Optionally, the spike detection model is, for example but not limited to, the EEGNet deep EEG model, with the convolution kernel set to k = (1, 32), the feature map F = 16, using the Adam optimizer, the learning rate lr = 0.01, using offline data or historical data to detect signal data, and selecting the training group model with the smallest validation loss as the final fixed model to train the spike detection model. For example, Figure 3The iterative curve of the verification accuracy of the spike detection model after training is shown. It can be seen that the accuracy of model training and verification gradually increases with the progress of training, indicating that the judgment of spikes is accurate in actual use. Figure 4 The iterative curve of the verification loss of the spike detection model after training is shown. It can be seen that the loss of model training and verification gradually decreases with the progress of training, indicating that the training of the model is stable and reliable. Of course, in order to ensure the robustness of the model, the offline data or historical data can be marked with the positions of spikes in some epilepsy data by a semi-automatic spike expert system model recognized by clinicians. Taking the spike peak as the center, 1.5s of data is intercepted as the training data of spike types, and another 1.5s of EEG data without spikes of the same quantity is taken as the background EEG; additional artifact data can also be added as the background. It is found in practice that after the spike detection model is trained and verified, compared with the spike detection model that has not participated in training, the test effect on spike and background EEG data has been greatly improved, such as Figure 5 shown, and its test accuracy can reach more than 90%, with true positives and true negatives being 99% and 89% respectively.

[0042] Step S4, data correction. The slices marked with 1 in the spike detection results (i.e., spike slices) are input into the data correction unit to eliminate misidentified spike slices; the correction module eliminates the data of the two misidentified spike types of resting data and high-amplitude artifact data. Specifically, the data segments with a standard deviation greater than the set threshold are marked as spike signals; the data segments with a standard deviation less than the set threshold are marked as normal signals, i.e., resting data; calculate the difference between the maximum and minimum values, their mean, and standard deviation in each data segment, and mark the data segments with a difference greater than the mean and standard deviation as artifact signals, i.e., high-amplitude artifact data.

[0043] Step S5, peak localization. Specifically, see Figure 11 , calculate the difference between the maximum and minimum values of the spike signals (i.e., spike slices marked with 1) for each channel one by one. After sorting the differences from large to small, take the difference at the percentile X1 or X2 (for example, in the position of 20% - 30%) and multiply it by the peak threshold coefficient as the peak threshold of this channel; use the calculated peak threshold and the set half-wave width threshold to search for the peak positions of the spike slices of each channel. Add the corresponding slice start time to each slice peak position to obtain the true time of each peak; merge the peaks with close true times in each channel into one peak. By close, it means the time difference between two peaks is less than the tolerance time limit th2; calculate the start and end positions of each peak for each channel one by one; calculate the peak width of each peak: the width from the start to the end point, and exclude the peaks not within the peak width threshold range [w1 w2]. Finally, based on the peak quantity and peaks of the spikes in each channel, a quantitative statistical result of the epileptic waveform is given. For example, see Figure 12, Given the peak point position of a certain peak, since the spike width is approximately 40 ms - 70 ms, 100 ms is taken forward from the peak point as the first half of the wave for searching the starting point, and 200 ms is taken backward from the peak point as the second half of the wave for searching the ending point; 200 ms is taken backward from the peak point as the second half of the wave Taking the search ending point as an example, the peak position is denoted as , and the position 200 ms backward is denoted as . See Figure 13 . The ending point of the second half of the wave is obtained through the following steps: ① Perform the first-order difference on the second half of the wave ; ② Judge the sign sign(x) of , and the result is denoted as ; ③ Perform the first-order difference on ; ④ Find the positions of all points greater than zero in ; ⑤ Calculate the proportion r of "1" from to in . If r > 0.5, it indicates that this is the ending point. If r < 0.5, continue to judge the proportion r of "1" in the next position to until . If none of them are satisfied, the ending point is . See . The calculation of the starting point of the first half of the wave is similar to that of the ending point of the second half of the wave. It only needs to be calculated after flipping the first half of the wave symmetrically with the peak as the center. Finally, the waveform parameters of the spike are obtained through the peak positioning unit for calculating the screening threshold of the spike . See Figure 14 . Step S6, screening spikes. See

[0044] . The spike screening unit calculates the screening threshold of the spike according to the waveform parameters of the spike. Specifically, the spike screening unit includes: a first screening module that sets the sensitivity to the quantile corresponding to the amplitude difference threshold of the spike to screen the spike; a second screening module that sets the sensitivity to the quantile corresponding to the slope threshold of the spike to screen the spike. Obviously, the two screening modules can be used separately or simultaneously. In actual operation, generally, the first screening module is used for primary screening first, and then the second screening module is used for secondary screening of the spikes after primary screening, and finally, it is used as the spike screening result. However, the order of the two can also be reversed Figure 6

[0045] ​Optionally, in the first screening module, setting the sensitivity to the quantile corresponding to the amplitude difference threshold of the spike wave includes: in the spike wave detection result, obtaining the amplitude difference from the peak value to the end point of the spike wave, its mean value, and standard deviation; sorting the amplitude differences of each spike wave in ascending order; using the data calculation unit to calculate the proportion value of the amplitude difference of each spike wave in the sorted position to obtain the quantile corresponding to each amplitude difference; the processor determines whether the standard deviation of the amplitude difference is greater than the quantile at the corresponding position; if so, taking the amplitude difference at the corresponding position of the sensitivity as the amplitude difference threshold; if not, taking the difference between the mean value of the amplitude difference and N times the standard deviation as the amplitude difference threshold. Specifically, see Figure 7 , setting the relative amplitude to be defined as: recording the peak position obtained in step S5 as , the peak amplitude is recorded as , the end point position , the end point amplitude is ; the amplitude difference ; calculating the amplitude difference from the peak value to the end point amplitude for each peak in each channel, denoted as the amplitude difference; re-sorting the amplitude differences of all peaks in each channel, and taking the amplitude difference at the X1 percentile as the amplitude difference P1 of the channel. Calculate the mean value M and standard deviation S of the amplitude differences of all peaks in each channel. If the standard deviation S of a channel is greater than P1, then select P1 as the amplitude difference threshold th3 of the channel; otherwise, select the difference between the standard deviation and the mean value as the amplitude difference threshold th3; judge channel by channel, and remove the peaks with amplitude differences less than the amplitude difference threshold th3. The remaining peaks are used as the peaks of the second-screened spike waves. Among them, the percentile X1 corresponding to the amplitude difference threshold of the spike wave is an adjustable parameter exposed to the user, and the user can adjust it according to their own needs. If the amplitude differences of the spike waves are sorted from small to large, the larger the value of X1, the higher the specificity of spike wave detection, and the smaller the value, the higher the sensitivity of spike wave detection. Generally, in order to overcome the problem of data distribution differences, if the amplitude differences of the spike waves are very scattered, it means that the data fluctuates violently and there are extreme values or outliers. It is suitable to use the amplitude difference at the quantile corresponding to the preset value of the sensitivity as the amplitude difference threshold to remove outliers; if the amplitude differences of the spike waves are relatively concentrated, it means that the data is relatively stable and the values are relatively stable. The difference between the mean value of the amplitude difference at the quantile corresponding to the preset value of the sensitivity and N times the standard deviation can be used as the amplitude difference threshold to remove individual outliers. Preferably, when the sensitivity is set in the range of [0%, 20%), in order to obtain higher detection sensitivity, N is taken as 0.5; when the sensitivity is set in the range of [20%, 80%), in order to balance sensitivity and specificity, N is taken as 1; when the sensitivity is set in the range of [80%, 100%], in order to obtain higher specificity, N is taken as 2.

[0046] Optionally, in the second screening module, setting the sensitivity to the quantile corresponding to the spike slope includes: in the spike detection result, using the data calculation unit to calculate the slope of each spike; sorting the slopes of each spike according to the magnitude; using the data calculation unit to calculate the proportional value of the slope of each spike in the sorting position to obtain the slope quantile of the spike, which is used as the sensitivity. Taking the spike slope at the quantile corresponding to the preset value of the sensitivity as the slope threshold to screen the spikes. Specifically, see Figure 8 , counting the number of remaining candidate peaks in each channel, and selecting the top two channels with the largest number as the trial channels. If there is only one channel, only select this channel as the trial channel. See Figure 9 and Figure 10 , setting the steepness defined as: the peak position is denoted as , the peak amplitude is denoted as , the termination point position , the termination point amplitude is ; there is , ; the steepness calculation formula is ; calculating the ratio of the difference between the peak value and the termination point amplitude to the width between the peak and the termination point for each peak in each channel, denoted as the steepness (i.e., the slope of the spike); re-sorting the steepness of all peaks in the trial channels according to the magnitude, taking the steepness at the X2 percentile as the preset value of the sensitivity, and removing the peaks in all channels with a steepness less than the sensitivity threshold. The remaining peaks are the finally identified spike peaks; the X2 percentile of the steepness is an adjustable parameter exposed to the user, and the user can adjust it according to their own needs. If the spike slopes are sorted from small to large, the larger the value of X2, the higher the specificity of spike detection, and the smaller the value, the higher the sensitivity of spike detection.

[0047] In addition, in order to avoid missing detections in a single recognition result, independent component analysis and reconstruction can also be performed on online multi-channel EEG data. The original data and the reconstructed data are respectively subjected to spike detection, and finally the spike detection results are integrated by a multi-data integration unit to complement each other. The multi-data integration unit includes: a preprocessing module that preprocesses signal data to obtain first data; a data reconstruction module that performs independent component analysis and reconstruction on the signal data to obtain second data; and an integration module that integrates the spike detection results after the spike detection unit detects the first data and the second data respectively. Specifically, the independent component analysis and reconstruction include: performing ICA decomposition (independent component analysis) on the preprocessed EEG data to obtain different source signals; detecting spikes in each source signal based on a deep spike detection method and counting the number of spikes in each source signal; retaining the source signals with the top 30% of the spike counts, and reconstructing these source signals into EEG signals. The integration module counts the channels with the top 30% of the detected spike counts in the first data, directly takes the union of the detection results of these corresponding channels in the first data and the second data, and merges the spikes with a spike position interval less than a threshold (usually 10 ms) in the union as the same spikes, and retains the detection results of the first data for other channels. Of course, the multi-data integration unit can be selected according to requirements and may not be selected when processing a single type of data.

[0048] In some embodiments, there is provided an epilepsy detection device, including: the spike detection system; a data display area is further provided on the operation interface of the human-computer interaction machine to display the quantitative statistical results of epilepsy waveforms; the quantitative statistical results of the epilepsy waveforms include: the number and peak size of spikes in the EEG signal in the spike screening results.

[0049] In some embodiments, when the function or model of the spike detection system is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or a part of this 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 for causing a computer device or its processor (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, external hard drives, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs and other various media that can store program codes.

[0050] Test Example 1.

[0051] Detect the electroencephalogram (EEG) data and compare the performance indicators under different test conditions. Data type: EEG signals of the C4 channel of epilepsy patients from 50 to 80 s; Performance indicators: correct recognition rate and false detection rate of sharp waves.

[0052] Test condition 1: Detection results of non-overlapping sliding window with a 1 s window, from Figure 15 It can be seen that the continuous line is the waveform of the EEG data, and the intermittent protruding lines are the detection results. Correct detection rate of sharp waves: 22 / 29 ≈ 75.8%; False recognition rate: 0.

[0053] Test condition 2: Detection results of overlapping 80% sliding window with a 1 s window, from Figure 16 It can be seen that the continuous line is the waveform of the EEG data, and the intermittent protruding lines are the detection results. Correct detection rate of sharp waves: 29 / 29 = 100%; False recognition rate: 0.

[0054] From this, it can be found that overlapping sliding windows can reduce the missed detection of sharp waves.

[0055] Test example 2.

[0056] Detect the EEG data and compare the performance indicators under different test conditions. Data type: EEG signals of multiple channels of epilepsy patients from 800 to 820 s; Performance indicators: correct recognition rate and false detection rate of sharp waves.

[0057] Test condition 3: Figure 17 Shows the detection results at 0% sensitivity (i.e., no sensitivity, setting the screening threshold as the amplitude difference at the 0% position and the steepness at the 0% position of the peak passing through the sharp wave in sequence) (i.e., Figure 17 the waveforms highlighted in). Correct detection rate of sharp waves: 16 / 20 ≈ 80%; False recognition rate: relatively high.

[0058] Test condition 4: Figure 18 Shows the detection results with the peak of the sharp wave set at 25% sensitivity (i.e., setting the screening threshold as the amplitude difference at the 25th percentile and the steepness at the 25th percentile in sequence) (i.e., Figure 18 the waveforms highlighted in). Correct detection rate of sharp waves: 16 / 20 ≈ 80%; False recognition rate: 0.

[0059] From the comparison results, it can be seen that the sharp wave screening unit can effectively reduce false positives caused by false detections and improve the accuracy by adjusting the sensitivity. Of course, users can adjust the values of percentile X1 or percentile X2 after selecting different sensitivity preset values according to their needs, so as to adjust the sensitivity of the device and balance the problems of false detection and missed detection.

[0060] Inspired by the above-described ideal embodiments of the present invention, through the above description, relevant staff can completely make various changes and modifications without departing from the technical idea of the present invention. That is, the technical scope of the present invention is not limited to the content in the specification.

Claims

1. A spike detection system, characterized in that, Comprising: A processor and a data calculation unit, a human-computer interaction unit, a spike detection unit, and a spike screening unit respectively connected to the processor; wherein The spike detection unit detects signal data to obtain a spike detection result; The spike screening unit screens spikes from the spike detection result to obtain a spike screening result; A sensitivity option and a waveform display area are provided on the operation interface of the human-computer interaction unit; The waveform display area is used to display the spike waveforms in the spike screening result; The sensitivity option is used to input a preset value of sensitivity to adjust the displayed spike waveforms; The spike screening unit includes: A first screening module that sets the sensitivity to the quantile corresponding to the amplitude difference threshold of the spike to screen the spike, including: In the spike detection result, the data calculation unit calculates the amplitude difference from the peak to the end point of each spike and its mean and standard deviation; Sort the amplitude differences of each spike according to their magnitudes; The data calculation unit calculates the proportion value of the amplitude difference of each spike in the sorted position to obtain the quantile corresponding to each amplitude difference; The processor determines whether the standard deviation of the amplitude difference is greater than the quantile at the corresponding position; If so, the amplitude difference at the corresponding position of the sensitivity is used as the amplitude difference threshold; If not, the difference between the mean of the amplitude difference and N times the standard deviation is used as the amplitude difference threshold.

2. The spike detection system according to claim 1, wherein When the sensitivity is set within the range of [0%, 20%), N takes 0.5; When the sensitivity is set within the range of [20%, 80%), N takes 1; When the sensitivity is set within the range of [80%, 100%], N takes 2.

3. The spike detection system according to claim 1, wherein The spike screening unit further includes: A second screening module that sets the sensitivity to the quantile corresponding to the slope threshold of the spike to screen the spike, including: In the spike detection result, the data calculation unit calculates the slope of each spike; Sort the slopes of each spike according to their magnitudes; The data calculation unit calculates the proportion value of the slope of each spike in the sorted position as the sensitivity.

4. The spike detection system according to claim 1, wherein The spike detection unit includes: A data slicing module that performs sliding window slicing processing on the signal data to obtain data segments and their corresponding starting times; A spike detection model, whose input end is the data segment and the output end is the spike detection result of each data segment. The output result of 1 indicates the presence of a spike signal, and the output of 0 indicates the absence of a spike signal.

5. The spike detection system according to claim 1, wherein, It further includes: A data correction unit; The data correction unit is used to delete resting data and high-amplitude artifact data from the spike detection result; Mark the data segments with a standard deviation greater than the set threshold as spike signals; Mark the data segments with a standard deviation less than the set threshold as normal signals, i.e., resting data; Calculate the difference between the maximum value and the minimum value in each data segment and its mean and standard deviation. Mark the data segments with a difference greater than the mean and standard deviation as artifact signals, i.e., high-amplitude artifact data.

6. The spike detection system according to claim 1 or 5, characterized in that, It further includes: A peak localization unit that marks the peak positions of the spike signals based on the spike detection result, that is Use the data calculation unit to calculate the difference between the maximum value and the minimum value in the spike signal, sort the differences from large to small, and multiply the difference at the preset position by the peak threshold coefficient as the peak threshold; Based on the peak threshold and the half-wave width threshold, search for the peak of the spike signal, and take the sum of the moment corresponding to the peak and the starting moment as the peak position of the spike; Merge the peaks with a time difference less than the set time threshold into one peak, and calculate the starting position and the ending position of each peak.

7. The spike detection system according to claim 1, wherein It further includes: A multi-data integration unit; The multi-data integration unit includes: A preprocessing module that preprocesses the signal data to obtain the first data; A data reconstruction module that performs independent component decomposition and reconstruction on the signal data to obtain the second data; An integration module that integrates the spike detection results after the spike detection unit detects the first data and the second data respectively.

8. The spike detection system according to claim 7, The integration of the spike detection results includes: Statistically, the channels in the first data with the number of detected spikes not less than the set value are used as the merged channels, and the channels less than the set value are used as other channels; In the merged channels, take the union of the detection results corresponding to the first data and the second data, and merge the spikes with a spike position interval less than the threshold as the same spike in the union; In other channels, retain the detection results corresponding to the first data.

9. An epilepsy detection device, characterized in that, It includes: The spike detection system according to any one of claims 1-8; A data display area is further set on the operation interface of the human-computer interaction machine to display the quantitative statistical results of the epileptic waveforms; The quantitative statistical results of the epileptic waveforms include: the number and peak size of the spikes in the electroencephalogram signal in the spike screening results.

Images