An electroencephalogram signal quality scoring method, system and storage medium

Abstract

This application discloses a method, system, and storage medium for scoring the quality of electroencephalogram (EEG) signals. The EEG signal quality scoring method includes: acquiring raw multi-channel EEG signals corresponding to different locations in the user's brain within a preset time period; preprocessing the raw multi-channel EEG signals to divide them into multiple EEG signal segments of fixed duration; extracting multiple-dimensional features to characterize the signal quality for each EEG signal segment; inputting the extracted multiple-dimensional features corresponding to each EEG signal segment into a pre-trained EEG signal quality scoring model, which then outputs a quality score. The EEG signal quality scoring method of this application has many advantages, including reducing manual intervention, using multi-dimensional feature fusion to reflect signal quality, automatically providing a quality score for the EEG signal, and facilitating rapid judgment by technical personnel.

Description

Technical Field

[0001] This application relates to the field of electroencephalogram (EEG) signal processing technology, and more specifically, to an EEG signal quality scoring method, system, and storage medium. Background Technology

[0002] Electroencephalography (EEG) is a non-invasive method for recording brain electrical activity, widely used in clinical diagnosis, brain-computer interfaces, sleep monitoring, and other fields. With the increasing prevalence of portable EEG devices, signal acquisition quality has become a key factor affecting the accuracy of subsequent analysis and diagnosis. Currently, EEG signal quality assessment largely relies on manual visual examination or simple judgment based on a single indicator (such as impedance value or amplitude), which has the following shortcomings: First, manual assessment is inefficient, highly subjective, and difficult to standardize; second, a single indicator cannot comprehensively reflect signal quality, for example, it cannot distinguish between physiological artifacts (such as blinking, electromyography, etc.) and non-physiological noise (such as electrode detachment, power line interference, etc.); and third, the differences in signal quality between different brain regions and different frequency bands are difficult to comprehensively assess. Therefore, there is an urgent need for a method that can automatically, multidimensionally, and precisely assess EEG signal quality. Summary of the Invention

[0003] One objective of this application is to provide a method, system, and storage medium for scoring electroencephalogram (EEG) signals, which can at least solve the aforementioned technical problems in the prior art. To achieve the above objective, this application provides the following technical solutions.

[0004] A method for scoring the quality of electroencephalogram (EEG) signals according to a first aspect of this application includes: acquiring raw multi-channel EEG signals corresponding to different locations in a user's brain within a preset time period; preprocessing the raw multi-channel EEG signals to divide the preprocessed multi-channel EEG signals into multiple EEG signal segments of fixed duration; wherein each EEG signal segment includes EEG signals acquired from multiple channels at the same time period; extracting multiple dimensional features for characterizing the signal quality of each EEG signal segment; wherein the multiple dimensional features include at least one of time-domain features, frequency-domain features, artifact features, and channel consistency features; inputting the extracted multiple dimensional features corresponding to each EEG signal segment into a pre-trained EEG signal quality scoring model, and having the EEG signal quality scoring model output a quality score corresponding to each EEG signal segment; and generating a visualized EEG signal quality scoring table based on the quality score corresponding to each EEG signal segment.

[0005] Optionally, the preprocessing includes at least one of baseline drift removal processing, bandpass filtering processing, notch filtering processing, bad conductor detection and removal processing, and rereference processing.

[0006] Optionally, the fixed duration of the signal segment is 1 to 4 seconds, and adjacent signal segments may or may not overlap.

[0007] Optionally, the time-domain features include at least one of the amplitude range, variance, zero-crossing rate, kurtosis, and skewness extracted for each channel; the frequency-domain features include at least one of the power of each frequency band, power spectral entropy, and peak frequency extracted for each channel.

[0008] Optionally, artifact features include at least one of the following: correlation between frontal region channel and prefrontal electrooculography channel signals, high-frequency electromyographic noise energy of temporal region channel, power frequency energy ratio of each channel, amplitude abrupt change detection results of each channel, and signal flatness detection results of each channel; channel consistency features include at least one of the following: correlation between channels corresponding to adjacent brain locations, whole-brain spatial distribution entropy, and global field power.

[0009] Optionally, the EEG signal quality scoring model is a regression model trained based on a supervised learning algorithm, which includes one of random forest, XGBoost, LightGBM, or lightweight neural networks.

[0010] Optionally, the multiple dimensional features corresponding to each extracted EEG signal segment are input into a pre-trained EEG signal quality scoring model, and the EEG signal quality scoring model outputs a quality score corresponding to each EEG signal segment. This includes: sequentially extracting multiple dimensional features from each channel of the EEG signal in each EEG signal segment; inputting the multiple dimensional features corresponding to each channel of the extracted EEG signal in each EEG signal segment into a pre-trained EEG signal quality scoring model corresponding to each channel; and having each EEG signal quality scoring model output a quality score corresponding to each channel of the EEG signal in each EEG signal segment.

[0011] A brainwave signal quality scoring system according to a second aspect of this application includes: a signal acquisition module for acquiring raw multi-channel brainwave signals corresponding to different locations in the user's brain within a preset time period; a signal processing module for preprocessing the raw multi-channel brainwave signals; a signal segmentation module for dividing the preprocessed multi-channel brainwave signals into multiple brainwave signal segments of fixed duration; wherein each brainwave signal segment includes brainwave signals acquired from multiple channels at the same time period; a feature extraction module for extracting multiple dimensional features for characterizing signal quality for each brainwave signal segment; wherein the multiple dimensional features include at least one of time-domain features, frequency-domain features, artifact features, and channel consistency features; and a quality scoring module for inputting the extracted multiple dimensional features corresponding to each brainwave signal segment into a pre-trained brainwave signal quality scoring model, and outputting a quality score corresponding to each brainwave signal segment from the brainwave signal quality scoring model.

[0012] According to a third aspect of this application, a computer-readable storage medium has a computer program stored thereon that, when executed by a processor, implements the electroencephalogram (EEG) signal quality scoring method of the first aspect.

[0013] The EEG signal quality scoring method of the present invention has many advantages, such as realizing automatic scoring of signal quality, reducing manual intervention and improving efficiency; using multi-dimensional feature fusion to comprehensively reflect signal quality and distinguish different types of artifacts and noise; and being able to score each channel independently, finely locating problem channels or brain regions, and facilitating rapid judgment by technicians.

[0014] Other features and advantages of this application will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0015] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments of the present application and, together with their description, serve to explain the principles of the present application.

[0016] Figure 1 This is a flowchart illustrating an electroencephalogram (EEG) signal quality scoring method according to one embodiment; Figure 2 This is a hardware configuration structure diagram of an electronic device for implementing one embodiment; Figure 3 This is a schematic diagram of an EEG signal quality scoring system that can be used to implement one embodiment; Figure 4 This is a schematic diagram of the International 10-20 system, which can be used to implement another embodiment. Detailed Implementation

[0017] Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the present application.

[0018] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the scope of this application and its application or use.

[0019] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0020] In all the examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0021] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0022] Implementation environment and hardware configuration:

[0023] Figure 2 This is a hardware configuration diagram of an electronic device 1000 that can be applied to the EEG signal quality scoring method in the embodiments of the present invention.

[0024] like Figure 2 As shown, the electronic device 1000 may include a processor 1100, a memory 1200, an interface device 1300, a display device 1400, an input device 1500, etc. The processor 1100 executes computer programs, which may employ instruction sets based on architectures such as x86, Arm, RISC, MIPS, and SSE. The memory 1200 may include, for example, ROM (Read-Only Memory), RAM (Random Access Memory), or non-volatile memory such as a hard disk. The interface device 1300 is a physical interface, such as a USB interface or a headphone jack. The display device 1400 may be a screen, which may be a touch screen. The input device 1500 may include a keyboard, a mouse, etc., or may include a touch device.

[0025] In this embodiment, the memory 1200 of the electronic device 1000 is used to store a computer program that controls the processor 1100 to operate in order to implement the EEG signal quality scoring method according to any embodiment. Those skilled in the art can design the computer program based on the scheme disclosed in this specification. How the computer program controls the processor 1100 to operate is well known in the art and will not be described in detail here.

[0026] Those skilled in the art should understand that, although in Figure 1 The present invention illustrates a plurality of devices of an electronic device 1000. The electronic device 1000 of the present invention may involve only some of the devices, or may include other devices, and is not limited herein.

[0027] Method Implementation Examples: The method for scoring the quality of electroencephalogram (EEG) signals according to embodiments of this application is described in detail below with reference to the accompanying drawings.

[0028] In one embodiment of this application, such as Figure 1As shown, the EEG signal quality scoring method according to an embodiment of this application includes: acquiring raw multi-channel EEG signals corresponding to different locations in the user's brain within a preset time period; preprocessing the raw multi-channel EEG signals to divide the preprocessed multi-channel EEG signals into multiple EEG signal segments of fixed duration; wherein each EEG signal segment includes EEG signals acquired from multiple channels at the same time period; extracting multiple dimensional features for characterizing signal quality for each EEG signal segment; wherein the multiple dimensional features include at least one of time-domain features, frequency-domain features, artifact features, and channel consistency features; inputting the extracted multiple dimensional features corresponding to each EEG signal segment into a pre-trained EEG signal quality scoring model, and outputting a quality score corresponding to each EEG signal segment from the EEG signal quality scoring model.

[0029] Specifically, in this embodiment, a 32-lead EEG acquisition device is first used as follows: Figure 4 The international 10-20 system, as shown, places multiple electrodes to collect EEG signals from different locations in the user's brain. Specifically, it collects EEG signals from the user in a resting state for 5 minutes, obtaining the user's raw, multi-channel EEG signals. The international 10-20 system is a standard electrode placement method established by the International Federation of Electroencephalography (EEG) Societies in 1958. It is currently the most widely used electrode placement system for EEG recordings worldwide.

[0030] Secondly, the acquired raw multi-channel EEG signals are preprocessed. Specifically, high-pass filtering is applied to remove baseline drift, followed by band-pass filtering to retain effective frequency bands, and then notch filtering to remove power frequency interference, ultimately yielding the preprocessed multi-channel EEG signals. The continuous multi-channel EEG signals are then divided into multiple fixed-duration EEG signal segments. In other words, each EEG signal segment can contain EEG signals from 32 channels within the same time period.

[0031] Next, multiple dimensional features are extracted from each EEG signal segment. These multiple dimensional features can include at least one of the following: temporal features, frequency domain features, artifact features, and channel consistency features. Specifically, temporal features can be used to calculate the amplitude range, variance, and zero-crossing rate of each channel. Frequency domain features can be used to calculate the δ, θ, α, β, and γ band powers of each channel, as well as the power spectral entropy. Artifact features can be used to calculate the correlation between frontal channels and prefrontal ocular channels (e.g., to detect blinking), to calculate high-frequency electromyographic energy in temporal channels (e.g., to detect electromyographic interference), and to calculate the power frequency energy ratio of each channel (e.g., to detect electrode detachment). Channel consistency features can be used to calculate the correlation coefficient between adjacent channels and to calculate the whole-brain spatial distribution entropy.

[0032] Finally, the multi-dimensional features corresponding to each extracted EEG signal segment are input into a pre-trained EEG signal quality scoring model, and then the EEG signal quality scoring model outputs the quality score corresponding to each EEG signal segment. The specific process will be detailed below.

[0033] In another embodiment of this application, the preprocessing includes at least one of baseline drift removal processing, bandpass filtering processing, notch filtering processing, bad conductor detection and removal processing, and rereference processing.

[0034] Specifically, preprocessing of the raw multichannel EEG signals can include at least one of the following: baseline drift removal, bandpass filtering, notch filtering, bad conductor detection and removal, and rereference processing. Baseline drift removal can employ a 0.5Hz high-pass filter to remove slow baseline drift caused by respiration, sweating, etc. Bandpass filtering can use a 0.5-45Hz bandpass filter to retain the effective frequency band of the EEG and filter out high-frequency noise and extremely low-frequency interference. Notch filtering can attenuate 50Hz power frequency interference using a notch filter. Bad conductor detection and removal involves calculating the variance of each channel; if the variance is close to 0 or remains saturated for an extended period, it is marked as a bad conductor and excluded from subsequent analysis. Rereference processing involves subtracting the average value of all channels from the signal of all channels to eliminate common interference from the reference electrodes. These steps are understood by those skilled in the art and will not be elaborated further.

[0035] In another embodiment of this application, the fixed duration of the signal segment is 1 to 4 seconds, and adjacent signal segments may or may not overlap.

[0036] Specifically, the fixed duration of an EEG signal segment can be set between 1 and 4 seconds. In this embodiment, taking a fixed duration of 2 seconds as an example, a 2-second EEG signal segment can capture the transient characteristics of EEG without being too short and causing insufficient frequency domain resolution. Furthermore, adjacent EEG signal segments in this embodiment can be set to be non-overlapping or partially overlapping. When adjacent EEG signal segments are set to be non-overlapping, EEG signal segment 1 is from 0 to 2 seconds, EEG signal segment 2 is from 2 to 4 seconds, EEG signal segment 3 is from 4 to 6 seconds, and so on. This method of segmenting EEG signal segments is suitable for scoring large amounts of EEG signal data.

[0037] In some other embodiments, adjacent signal segments can also be divided into overlapping segments. Specifically, when adjacent EEG signal segments are set to overlap, EEG signal segment 1 is from second 0 to second 2, EEG signal segment 2 is from second 1 to third 3, EEG signal segment 3 is from second 2 to fourth 4, and so on, with an overlap rate of 50%. The advantage of this method is that it avoids abrupt changes in scoring caused by segment boundary truncation, resulting in a smoother scoring curve and easier trend observation.

[0038] In some embodiments of this application, the time-domain features include at least one of amplitude range, variance, zero-crossing rate, kurtosis, and skewness extracted for each channel; the frequency-domain features include at least one of power of each frequency band, power spectral entropy, and peak frequency extracted for each channel.

[0039] Specifically, extracting multiple dimensions of features to characterize the signal quality for each EEG signal segment can be achieved by extracting time-domain and frequency-domain features for each channel signal in each EEG signal segment. More specifically, the time-domain features extracted in this embodiment can include one or more values ​​of amplitude range, variance, zero-crossing rate, kurtosis, and skewness extracted for each channel in each EEG signal segment.

[0040] The amplitude range reflects the dynamic range of the signal, subtracting the minimum value from the maximum value within a segment. Normal EEG amplitude ranges are generally from tens to hundreds of microvolts; an excessively large range may indicate noise, while an excessively small range may indicate electrode detachment. Variance reflects the energy level of the signal within a segment. A small variance suggests a potentially flat signal. Zero-crossing rate is the number of times the signal crosses zero per unit time. Normal EEG zero-crossing rates have a certain range; an excessively high rate may indicate high-frequency noise, while an excessively low rate may indicate low-frequency drift. Kurtosis measures the steepness of the waveform; blink artifacts typically exhibit kurtosis. Skewness measures waveform asymmetry; certain artifacts can cause abnormal skewness. These parameters are self-evident to those skilled in the art and will not be elaborated upon further.

[0041] Furthermore, the frequency domain features extracted in this embodiment may include one or more values ​​of each frequency band power, power spectral entropy, and peak frequency extracted for each channel in each EEG signal segment. The power of each frequency band can be the absolute and relative power of the five frequency bands (δ, θ, α, β, and γ waves) calculated using Fast Fourier Transform, i.e., the power of each frequency band divided by the total power. The power spectral entropy can be calculated based on the proportion of power in each frequency band; a higher entropy value indicates more chaotic frequency components and a greater likelihood of signal contamination by noise. The peak frequency can be the frequency point with the highest energy in the power spectrum, used to determine the presence of abnormal rhythms. This will be understood by those skilled in the art and will not be elaborated further here.

[0042] In one embodiment of this application, artifact features include at least one of the following: correlation between frontal region channel and prefrontal electrooculography channel signals, high-frequency electromyographic noise energy of temporal region channel, power frequency energy ratio of each channel, amplitude abrupt change detection results of each channel, and signal flatness detection results of each channel; channel consistency features include at least one of the following: correlation between channels corresponding to adjacent brain locations, whole-brain spatial distribution entropy, and global field power.

[0043] Specifically, extracting multiple dimensional features to characterize the signal quality for each EEG signal segment may further include extracting artifact features and channel consistency features for each channel signal in each EEG signal segment.

[0044] In detail, the artifact features extracted in this embodiment may include one or more of the following values: the correlation between the frontal channel and the frontoocardial channel extracted for each channel in each EEG signal segment; the high-frequency electromyographic noise energy of the temporal channel; the power frequency energy ratio of each channel; the amplitude abrupt change detection of each channel; and the signal flatness detection of each channel.

[0045] Furthermore, the correlation between the frontal region channel and the prefrontal electrooculography channel can be used to select, for example... Figure 4The signals from the frontal electrodes (e.g., Fp1, Fp2) and dedicated electrooculography (EOG) channels (e.g., VEOG) are used to calculate the Pearson correlation coefficient. If the absolute value of the correlation coefficient is greater than 0.6, it indicates that the segment may be contaminated by blinking or eye movements. The high-frequency electromyographic noise energy of the temporal channel can be calculated by selecting temporal electrodes (e.g., T3, T4) and calculating the root mean square (RMS) or mean absolute value (MAV) in the 35-45Hz frequency band. Simply put, if this value exceeds a preset threshold (e.g., 30μV), it indicates the presence of electromyographic interference. The power frequency energy ratio of each channel can be calculated by determining the proportion of energy near 50Hz (or 60Hz) in the total energy of that channel. If the proportion exceeds 30%, it indicates that the channel may have poor contact and severe power frequency interference. Amplitude mutation detection for each channel can be performed on the number of abrupt changes in the difference between adjacent sampling points in the detection signal that exceed the normal range (e.g., greater than 100 μV). Too many abrupt changes indicate the presence of sudden artifacts. Signal flatness detection for each channel can be performed on cases where the amplitude change of the detection signal is less than 1 μV over a continuous period (e.g., 0.5 seconds). If this occurs, it indicates that the front electrode may have detached. This is understandable to those skilled in the art and will not be elaborated further. Furthermore, the dedicated electrooculogram (VEOG) channel mentioned above is used to record potential changes caused by vertical eye movements (e.g., blinking, vertical gaze), and is a crucial artifact monitoring and correction channel in EEG and event-related potential experiments. Obtaining a vertical electrooculogram (VEOG) requires bipolar leads: a pair of electrodes are placed above the orbit (e.g., 1 cm above Fp1) and below the orbit (midpoint of the infraorbital margin), respectively, recording the corneal-retinal potential difference to reflect vertical eye movements. VEOG (Vertical Electrooculography) is defined by the potential difference between two electrodes. These two electrodes are placed above and below the eye socket, respectively, forming a bipolar lead that records the potential changes at two points in the dipole electric field between the cornea and retina during vertical eye movements. When the eye moves upward, the upper electrode presents a positive potential relative to the lower electrode, and vice versa. The difference between the two is the VEOG signal. This is something that those skilled in the art will understand, and will not be elaborated further here.

[0046] In detail, the channel consistency features extracted in this embodiment may include one or more values ​​of adjacent channel correlation, whole brain spatial distribution entropy, and global field power extracted for each channel in each EEG signal segment.

[0047] Furthermore, the correlation between adjacent channels can be specifically calculated by taking the EEG signal of each channel in the same EEG signal segment and calculating its average correlation coefficient with its four neighboring channels (e.g., F3 with Fz, F4, C3, etc.). Simply put, if the average correlation coefficient of a particular channel is below 0.3, it indicates that the channel may be abnormal. The whole-brain spatial distribution entropy can be specifically calculated based on the power values ​​of all channels, showing the entropy value of their spatial distribution. An excessively high entropy value suggests that the power distribution is too dispersed, indicating the presence of global noise. An excessively low entropy value suggests that the power is too concentrated, indicating a single-channel abnormality. The global field power can be specifically calculated as the square root of the sum of the squares of the instantaneous values ​​of all channel signals, used to reflect the overall EEG intensity. If the global field power suddenly drops significantly, it indicates a device malfunction or a change in the user's state of consciousness. This is understandable to those skilled in the art and will not be elaborated further here.

[0048] In some embodiments of this application, the EEG signal quality scoring model is a regression model trained based on a supervised learning algorithm, which includes one of random forest, XGBoost, LightGBM, or lightweight neural networks.

[0049] Specifically, in this embodiment, a supervised learning algorithm can be used to train the regression model. The supervised learning algorithm used in this embodiment can be one of Random Forest, XGBoost, LightGBM, or a lightweight neural network. The following explanation uses the XGBoost regressor selected in this embodiment as an example. In the training data preparation stage, 10,000 2-second EEG signal segments are collected, covering various quality conditions, such as clean, slightly noisy, severely artifacts, and electrode detachment. Further, multiple experts can independently score each EEG signal segment (0-100 points), and the average score (i.e., quality score) is taken as the standard label for each EEG signal segment, resulting in 10,000 2-second EEG signal segments with independent labels. Then, multi-dimensional features, including time-domain features, frequency-domain features, artifact features, and channel consistency features, are extracted from each channel of the EEG signal in each segment. These features are then concatenated into a feature vector corresponding to each EEG signal segment.

[0050] The model training for the EEG signal quality scoring model in this embodiment mainly includes: dividing a dataset of 10,000 EEG signal segments into a training set (80%), a validation set (10%), and a test set (10%) according to a preset ratio. Specifically, the feature vector corresponding to each EEG signal segment is used as input, and the quality score in the standard label corresponding to each EEG signal segment is used as output to train the XGBoost regressor. Based on this, using the XGBoost regressor, the model is trained using labeled EEG signal segments in the training set. Then, the parameters are adjusted on the validation set to prevent overfitting, and finally, the performance is evaluated on the test set. The resulting trained EEG signal quality scoring model can then be obtained for subsequent use. This is a concept that those skilled in the art should understand, and will not be elaborated further here.

[0051] Based on the above, in the subsequent use of the EEG signal quality scoring model, for each newly acquired EEG signal segment, all dimensions of features are extracted from each channel of the EEG signal in each segment. These features are then concatenated into a feature vector corresponding to each EEG signal segment, and then input into the trained EEG signal quality scoring model. The EEG signal quality scoring model can directly output the quality score corresponding to each EEG signal segment.

[0052] In some other embodiments of this application, multiple dimensional features corresponding to each extracted EEG signal segment are input into a pre-trained EEG signal quality scoring model, and the EEG signal quality scoring model outputs a quality score corresponding to each EEG signal segment. This includes: sequentially extracting multiple dimensional features from each channel of the EEG signal in each EEG signal segment; inputting the multiple dimensional features corresponding to each channel of the extracted EEG signal in each EEG signal segment into a pre-trained EEG signal quality scoring model corresponding to each channel; and having each EEG signal quality scoring model output a quality score corresponding to each channel of the EEG signal in each EEG signal segment.

[0053] Specifically, this embodiment, based on embodiment 6, further implements independent scoring of the EEG signal for each channel. Specifically, in some other embodiments of this application, supervised learning algorithms can also be used to train the regression model; in this embodiment, the supervised learning algorithm for training the regression model can be the XGBoost model. The following explanation uses the XGBoost regressor as an example. During the training data preparation phase, 10,000 2-second EEG signal segments are collected. Each EEG signal segment includes EEG signals corresponding to 32 channels. These EEG signal segments cover various quality conditions, such as clean, slightly noisy, severely artifactive, and electrode detachment. Furthermore, it is important to note that 32 XGBoost regressors are also required, each corresponding to one channel. If the collected EEG signal has 64 channels, then 64 XGBoost regressors are needed, one-to-one, and so on.

[0054] Furthermore, multiple experts can independently score the EEG signals corresponding to each channel in each EEG signal segment (0-100 points). Simply put, each EEG signal segment can include EEG signals corresponding to 32 channels, and 10,000 EEG signal segments can include 3,200 EEG signals, each corresponding to a channel and lasting 2 seconds. Finally, the average of multiple scores from multiple experts for the same EEG signal (i.e., the quality score) is taken as the standard label corresponding to each channel's EEG signal in each EEG signal segment, resulting in 320,000 independently labeled 2-second EEG signal segments corresponding to each channel.

[0055] Next, time-domain features, frequency-domain features, and artifact features are extracted from the EEG signal corresponding to each channel in each EEG signal segment, but channel consistency features are not included. These features are then concatenated to form a feature vector corresponding to the EEG signal of each channel in each EEG signal segment.

[0056] The model training for the EEG signal quality scoring model in this application embodiment mainly includes: firstly, dividing the 320,000 2-second segments of EEG signals corresponding to each channel according to different channel numbers to obtain a dataset including 10,000 2-second EEG signals corresponding to the first channel, 10,000 2-second EEG signals corresponding to the second channel, and so on up to 10,000 2-second EEG signals corresponding to the thirty-second channel.

[0057] Secondly, taking the first channel as an example, the dataset containing 10,000 EEG signal segments corresponding to the first channel is divided into a training set (80%), a validation set (10%), and a test set (10%) according to a preset ratio.

[0058] Then, the feature vector of each EEG signal corresponding to the first channel in the training set is used as input, and the quality score in the standard label corresponding to each EEG signal of the first channel is used as output, so as to train an XGBoost regressor corresponding to the first channel.

[0059] Based on this, the XGBoost regressor corresponding to the first channel is used to train the model using labeled EEG signals from the training set. Then, the parameters are adjusted on the validation set to prevent overfitting, and finally, the performance is evaluated on the test set. The resulting trained EEG signal quality scoring model can then be used for subsequent applications.

[0060] Similarly, the same method can be used to train, validate, and test the XGBoost regressors for each channel as described above, resulting in 32 EEG signal quality scoring models corresponding to each channel.

[0061] Based on the above, in the subsequent use of the EEG signal quality scoring model, for each channel of the newly acquired EEG signal corresponding to each EEG signal segment, time-domain features, frequency-domain features, and artifact features are extracted for each channel of the EEG signal in each EEG signal segment. These features of different dimensions are then concatenated to form a feature vector corresponding to each channel of the EEG signal in each EEG signal segment. This feature vector is then input into the corresponding channel's EEG signal quality scoring model. This EEG signal quality scoring model corresponding to the channel number can directly output the quality score for the EEG signal in that channel. For example, the EEG signal corresponding to the first channel in an EEG signal segment can be input into the EEG signal quality scoring model corresponding to the first channel, rather than the model corresponding to the second channel. Based on this, the EEG signal quality scoring model can output the quality score corresponding to the first channel of the EEG signal in this EEG signal segment. And so on, ultimately obtaining the quality scores corresponding to the EEG signals of each channel in each EEG signal segment.

[0062] In summary, the EEG signal quality scoring method of the present invention has many advantages, such as realizing automatic scoring of signal quality, reducing manual intervention, improving scoring efficiency, using multi-dimensional feature fusion to comprehensively reflect signal quality, distinguishing different types of artifacts and noise, being able to score each channel independently, finely locating problem channels or brain regions, and facilitating rapid judgment by technicians.

[0063] System Implementation Example: In this embodiment, an electroencephalogram (EEG) signal quality scoring system 2000 is also provided. For example... Figure 3As shown, the EEG signal quality scoring system 2000 includes a signal acquisition module 2100, a signal processing module 2200, a feature extraction module 2300, and a quality scoring module 2400. Wherein: The signal acquisition module 2100 acquires raw multi-channel EEG signals corresponding to different locations in the user's brain within a preset time period; the signal processing module 2200 preprocesses the raw multi-channel EEG signals; and the signal segmentation module divides the preprocessed multi-channel EEG signals into multiple EEG signal segments of fixed duration; wherein each EEG signal segment includes EEG signals acquired from multiple channels at the same time period. The feature extraction module 2300 extracts multiple dimensional features for characterizing signal quality for each EEG signal segment; wherein the multiple dimensional features include at least one of time domain features, frequency domain features, artifact features, and channel consistency features; the quality scoring module 2400 inputs the multiple dimensional features corresponding to each extracted EEG signal segment into a pre-trained EEG signal quality scoring model, and the EEG signal quality scoring model outputs a quality score corresponding to each EEG signal segment.

[0064] The EEG signal quality scoring system 2000 provided in the embodiments of this application has many advantages, such as realizing automatic scoring of signal quality, reducing manual intervention, improving scoring efficiency, adopting multi-dimensional feature fusion to comprehensively reflect signal quality, distinguishing different types of artifacts and noise, and being able to score each channel independently, finely locating problem channels or brain regions, and facilitating rapid judgment by technicians.

[0065] It should be noted that although several modules or units of the system for executing actions are mentioned in the detailed description above, this division is not mandatory. In fact, according to the implementation method of this application, the characteristics and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the characteristics and functions of one module or unit described above can be further divided and embodied by multiple modules or units. Medium Examples

[0066] In this embodiment, a computer-readable storage medium is also provided, on which computer instructions are stored. When the computer instructions are executed by a processor, the electroencephalogram (EEG) signal quality scoring method as described in any embodiment of the present invention is implemented.

[0067] The present invention may be a system, method, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing processor 1100 to implement various aspects of the present invention.

[0068] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0069] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0070] The computer program instructions used to perform the operations of this invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing state information from the computer-readable program instructions. This electronic circuitry can execute the computer-readable program instructions to implement various aspects of the invention.

[0071] Various aspects of the present invention are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0072] These computer-readable program instructions can be provided to the processor 1100 of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor 1100 of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0073] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0074] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction, which contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions. It will be known to those skilled in the art that implementation in hardware, implementation in software, and implementation using a combination of software and hardware are equivalent.

[0075] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, and are not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein. The scope of the invention is defined by the appended claims.

[0076] While specific embodiments of this application have been described in detail by way of examples, those skilled in the art should understand that the above examples are for illustrative purposes only and are not intended to limit the scope of this application. Those skilled in the art should understand that modifications can be made to the above embodiments without departing from the scope and spirit of this application. The scope of this application is defined by the appended claims.

Claims

1. A method for scoring the quality of electroencephalogram (EEG) signals, characterized in that, include: Collect raw multi-channel EEG signals corresponding to different locations in the user's brain within a preset time period; The original multi-channel EEG signal is preprocessed, and the preprocessed multi-channel EEG signal is divided into multiple EEG signal segments of fixed duration; wherein, each EEG signal segment includes EEG signals collected from multiple channels at the same time period; For each EEG signal segment, multiple dimensional features are extracted to characterize the signal quality; wherein, the multiple dimensional features include at least one of time-domain features, frequency-domain features, artifact features, and channel consistency features; The multiple dimensional features corresponding to each extracted EEG signal segment are input into a pre-trained EEG signal quality scoring model, which then outputs a quality score corresponding to each EEG signal segment.

2. The EEG signal quality scoring method according to claim 1, characterized in that, The preprocessing includes at least one of baseline drift removal processing, bandpass filtering processing, notch filtering processing, bad conductor detection and removal processing, and rereference processing.

3. The EEG signal quality scoring method according to claim 2, characterized in that, The fixed duration of the signal segment is 1 to 4 seconds, and adjacent signal segments may or may not overlap.

4. The EEG signal quality scoring method according to claim 3, characterized in that, The time-domain features include at least one of amplitude range, variance, zero-crossing rate, kurtosis, and skewness extracted for each channel; the frequency-domain features include at least one of power, power spectral entropy, and peak frequency extracted for each channel.

5. The EEG signal quality scoring method according to claim 4, characterized in that, The artifact features include at least one of the following: correlation between the frontal region channel and the frontal electrooculography channel signals, high-frequency electromyographic noise energy of the temporal region channel, power frequency energy ratio of each channel, amplitude abrupt change detection results of each channel, and signal flatness detection results of each channel. The channel consistency features include at least one of the following: correlation between channels corresponding to adjacent brain locations, whole-brain spatial distribution entropy, and global field power.

6. The electroencephalogram (EEG) signal quality scoring method according to claim 5, characterized in that, The EEG signal quality scoring model is a regression model trained based on a supervised learning algorithm, which includes one of random forest, XGBoost, LightGBM, or lightweight neural networks.

7. The EEG signal quality scoring method according to claim 6, characterized in that, The step of inputting the multiple dimensional features corresponding to each extracted EEG signal segment into a pre-trained EEG signal quality scoring model, and having the EEG signal quality scoring model output a quality score corresponding to each EEG signal segment, further includes: The multiple dimensional features are extracted sequentially for each channel of the EEG signal in each EEG signal segment. The multiple dimensional features corresponding to each channel of the extracted EEG signal segment are input into a pre-trained EEG signal quality scoring model corresponding to each channel. Each EEG signal quality scoring model outputs a quality score corresponding to each channel of the EEG signal in each EEG signal segment.

8. A brainwave signal quality scoring system, characterized in that, include: The signal acquisition module collects raw multi-channel EEG signals corresponding to different locations in the user's brain within a preset time period; The signal processing module preprocesses the raw multichannel EEG signals; The signal segmentation module divides the preprocessed multi-channel EEG signal into multiple EEG signal segments of fixed duration; wherein each EEG signal segment includes EEG signals collected from multiple channels at the same time period; The feature extraction module extracts multiple dimensional features for characterizing signal quality for each EEG signal segment; wherein the multiple dimensional features include at least one of time-domain features, frequency-domain features, artifact features, and channel consistency features; The quality scoring module inputs the multiple dimensional features corresponding to each extracted EEG signal segment into a pre-trained EEG signal quality scoring model, and the EEG signal quality scoring model outputs a quality score corresponding to each EEG signal segment.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the electroencephalogram (EEG) signal quality scoring method as described in any one of claims 1 to 7.

Images