A sleep spindle intelligent identification method, system, device, medium and product

Abstract

The present application discloses a method, system, device, medium, and product for intelligently identifying sleep spindles, which relates to the field of EEG characteristic wave recognition. The method comprises: obtaining a sleep EEG signal time series; performing bandpass filtering on the sleep EEG signal time series to obtain a filtered EEG signal time series; dividing the sleep EEG signal time series and the filtered EEG signal time series into overlapping window signals of different lengths to obtain a time segment data set; using a spindle wave recognition model based on the time segment data set to determine whether spindle waves exist in the sleep EEG signal time series and the start time and duration of the spindle waves; wherein the spindle wave recognition model is constructed based on a convolutional neural network and is pre-trained using a training sample set. The present application can detect spindle waves with different durations and amplitudes, thereby improving the recognition accuracy of sleep spindle waves.

Description

Technical Field

[0001] The present application relates to the field of EEG characteristic wave recognition, and in particular to a multi-window filtering coupled sleep spindle intelligent recognition method, system, equipment, medium and product. Background Art

[0002] Sleep spindles are bursts of 11-16 Hz sinusoidal cycles identifiable in sleep EEG. The American Academy of Sleep Medicine defines sleep spindles as "clearly discernible sinusoidal waves with a frequency of 11-16 Hz (usually 12-14 Hz) and a duration of at least 0.5 s, with the highest amplitude reached in the central leads." Sleep spindles play a vital role in sleep structure and are associated with various cognitive functions, such as memory consolidation and learning ability. In addition, abnormal spindle characteristics are associated with neurological and psychiatric disorders, including schizophrenia, depression, and neurodegenerative diseases. Therefore, accurate detection of spindles is of clinical significance.

[0003] The development of automated sleep spindle detection technologies has become an area of ​​growing interest due to the need for objective, valid, and reproducible assessments of sleep-related neural activity. Advances in machine learning, signal processing, and artificial intelligence have facilitated the emergence of new algorithms for highly accurate sleep spindle identification. Related research has employed various architectures, such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), and hybrid models that combine spatial and temporal features to improve detection accuracy. These models have demonstrated effectiveness in distinguishing spindle events from non-spindle electroencephalogram (EEG) activity. However, these models typically use fixed-length time windows to extract EEG signals. Spindle duration is uncertain and varies between individuals, and fixed time windows cannot accommodate this individual variability. Excessively long or short time windows can lead to inaccurate localization or incomplete spindle segmentation, resulting in missed or false detections. Summary of the Invention

[0004] The purpose of this application is to provide a method, system, device, medium and product for intelligent identification of sleep spindles, which can improve the identification accuracy of sleep spindles.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a method for intelligently identifying sleep spindles, comprising:

[0007] Obtain sleep EEG signal time series;

[0008] Performing bandpass filtering on the sleep EEG signal time series to obtain a filtered EEG signal time series;

[0009] Splitting the sleep EEG signal time series and the filtered EEG signal time series into overlapping window signals of different lengths to obtain time segment data sets;

[0010] According to the time segment data set, a spindle wave recognition model is used to determine whether spindle waves exist in the sleep EEG signal time series and the start time and duration of the spindle waves;

[0011] The spindle recognition model is constructed based on a convolutional neural network and is pre-trained using a training sample set; the training sample set includes multiple time segment sample sets and spindle labels for each time segment sample set; the spindle labels include whether spindles exist, the start time and duration of the spindles.

[0012] In a second aspect, the present application provides a sleep spindle intelligent recognition system, comprising:

[0013] Signal acquisition module, used to obtain sleep EEG signal time series;

[0014] a signal filtering module, configured to perform bandpass filtering on the sleep EEG signal time series to obtain a filtered EEG signal time series;

[0015] a signal segmentation module, configured to segment the sleep EEG signal time series and the filtered EEG signal time series into overlapping window signals of different lengths to obtain a time segment data set;

[0016] a signal recognition module, configured to determine, based on the time segment data set and using a spindle recognition model, whether spindles exist in the sleep EEG signal time series, as well as the start time and duration of the spindles;

[0017] The spindle recognition model is constructed based on a convolutional neural network and is pre-trained using a training sample set; the training sample set includes multiple time segment sample sets and spindle labels for each time segment sample set; the spindle labels include whether spindles exist, the start time and duration of the spindles.

[0018] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned sleep spindle intelligent identification method.

[0019] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the above-mentioned sleep spindle intelligent identification method when executed by a processor.

[0020] In a fifth aspect, the present application provides a computer program product, including a computer program, which implements the above-mentioned sleep spindle intelligent identification method when executed by a processor.

[0021] According to the specific embodiments provided in this application, this application has the following technical effects:

[0022] The present application provides a method, system, device, medium and product for intelligent identification of sleep spindles. By performing bandpass filtering on the time series of sleep EEG signals, interference from muscle movement, eye movement and other non-neural activities is reduced, and the accuracy of spindle identification is improved. By dividing the time series of sleep EEG signals and the time series of filtered EEG signals into overlapping window signals of different lengths, the spindle identification model can capture short-term and long-term features. The spindle identification model can detect spindles with different durations and amplitudes, thereby improving the identification accuracy of sleep spindles. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 This is a diagram showing an application environment of a sleep spindle intelligent identification method in one embodiment of the present application;

[0025] Figure 2 A flowchart of a method for intelligently identifying sleep spindles provided in one embodiment of the present application;

[0026] Figure 3 A block diagram of a sleep spindle intelligent identification method provided in one embodiment of the present application;

[0027] Figure 4 This is a schematic diagram of the processing process of the spindle wave recognition model in one embodiment of the present application;

[0028] Figure 5 A schematic diagram of the functional modules of a sleep spindle intelligent recognition system provided in one embodiment of the present application;

[0029] Figure 6 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0031] The identification methods of spindles can be divided into the following categories:

[0032] (1) Template matching method: Based on the characteristics of spindles, the specific features of spindles are extracted from the EEG signal through signal technology, including time threshold and frequency threshold analysis.

[0033] (2) Traditional machine learning methods: extract time threshold, frequency domain, or nonlinear features from EEG signals, and learn spindle features from a large amount of labeled data by constructing feature vectors. Common algorithms include support vector machines, random forests, and K-nearest neighbors. Compared with template matching, these methods can better adapt to individual differences between signals, but the learned features are single and cannot consider the interactions between multiple windows.

[0034] (3) Deep learning-based methods: These methods do not require explicit feature extraction and can directly learn spindle features from the original signal. These methods include convolutional neural networks (for extracting spatial features) and recurrent neural networks (for extracting time series features), which can avoid feature engineering.

[0035] The purpose of this application is to study the integration of multiple EEG datasets, adopt advanced filtering techniques, and explore the effectiveness of multi-window architectures, in order to develop a more generalizable, robust, and accurate automatic spindle identification method.

[0036] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0037] The sleep spindle intelligent recognition method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be set up separately, integrated on the server 104, or placed on the cloud or other servers. The terminal 102 can send the sleep EEG signal time series to be identified to the server 104. After the server 104 receives the sleep EEG signal time series to be identified, it performs band-pass filtering, time series segmentation and identification in sequence to determine whether there are spindles in the sleep EEG signal time series and the start time and duration of the spindles. The server 104 can feedback the identification results of the spindles to the terminal 102. In addition, in some embodiments, the sleep spindle intelligent identification method can also be implemented by the server 104 or the terminal 102 alone.

[0038] Terminal 102 may include, but is not limited to, various desktop computers, laptops, smartphones, tablet computers, IoT devices, and portable wearable devices. IoT devices may include smart speakers, smart TVs, smart air conditioners, and smart car devices. Portable wearable devices may include smart watches, smart bracelets, and head-mounted devices. Server 104 may be implemented as a standalone server or a server cluster consisting of multiple servers, or may be a cloud server.

[0039] In an exemplary embodiment, Figure 2 As shown, a sleep spindle intelligent recognition method is provided, which is executed by a computer device, specifically a computer device such as a terminal or a server, or a terminal and a server. In the embodiment of the present application, the method is applied to Figure 1 The server 104 in the example is used for explanation, including the following steps 201 to 204.

[0040] Step 201: Acquire a time series of sleep EEG signals. Specifically, a polysomnography device is used to acquire the time series of sleep EEG signals, and the monitoring, acquisition, and storage are performed according to the standards of the American Academy of Sleep Medicine.

[0041] Step 202: performing bandpass filtering on the sleep EEG signal time series to obtain a filtered EEG signal time series.

[0042] In an exemplary embodiment, the sleep EEG signal time series is first digitized, and the baseline and noise waves are removed to obtain a preprocessed EEG signal time series, and then a Butterworth filter is used to filter the preprocessed EEG signal time series to obtain a filtered EEG signal time series.

[0043] The extremely low-frequency baseline, high-frequency, and low-frequency noise waves in the sleep EEG signal time series were removed. The sampling frequency was 256 Hz to complete the signal preprocessing. A Butterworth filter was used to perform frequency calibration at 11 Hz to 16 Hz to verify the validity of the data in the EEG signal time series. The calculation formula is:

[0044]

[0045] Where x[t] is the time series of the preprocessed EEG signal, y[t] is the impulse response of the bandpass filter, N is the order of the Butterworth filter, and b k is the filter coefficient calculated based on the order of the Butterworth filter, a m is the filter coefficient calculated based on the cutoff frequency.

[0046] Step 203 : Segment the sleep EEG signal time series and the filtered EEG signal time series into overlapping window signals of different lengths to obtain time segment data sets.

[0047] In an exemplary embodiment, the time segment dataset includes a first dataset and a second dataset. The first dataset includes a plurality of signal segments of different time lengths of the filtered EEG signal time series. The second dataset includes a plurality of signal segments of different time lengths of the sleep EEG signal time series.

[0048] Specifically, spindle waves generally exist in EEG signals for no more than 2 seconds, so this application divides the entire sleep EEG signal time series and the entire filtered EEG signal time series into segments of different time lengths of 2 seconds, 4 seconds, and 6 seconds, respectively.

[0049] Step 204: Based on the time segment data set, a spindle wave recognition model is used to determine whether spindle waves exist in the sleep EEG signal time series and the start time and duration of the spindle waves. If spindle waves do not exist, the start time and duration are both 0.

[0050] The spindle recognition model is constructed based on a convolutional neural network and pre-trained using a training sample set. The training sample set includes multiple time segment sample sets and spindle labels for each time segment sample set. The spindle labels include the presence of spindles, the spindle start time, and the spindle duration.

[0051] In an exemplary embodiment, Figure 3As shown, the spindle wave recognition model includes a first convolutional neural network, a second convolutional neural network, and a fully connected neural network. The time segment sample set is divided into a first sample set and a second sample set. The first sample set includes multiple signal segments of different time lengths of the filtered EEG signal time series, and the second sample set includes multiple signal segments of different time lengths of the sleep EEG signal time series. A first convolutional neural network is constructed based on the first sample set, and a second convolutional neural network is constructed based on the second sample set. The features extracted by the first and second convolutional neural networks are aggregated through the fully connected neural network, and further iteration is performed to output the final recognition result.

[0052] This application constructs a training sample set by extracting multiple time segments containing spindles and multiple time segments not containing spindles. The training sample set is divided into a training set, a validation set, and a test set. The training set is used to train the spindle recognition model, and the predicted value of the presence of spindles is obtained. The predicted value is compared with the spindle labels manually annotated by experts, and the loss function is calculated. The optimal spindle recognition model parameters are obtained through iterative training.

[0053] The loss function used in the spindle wave recognition model training is the focus loss function to address the problem of class imbalance in target detection. The formula of the focus loss function is:

[0054] FL(p t )=-α t (1-p t ) γ log(p t );

[0055] Among them, FL(p t ) is the loss function value, p t is the recognition result output by the spindle wave recognition model, α t and γ are factors, which is a dynamically scaled cross entropy loss that gradually decreases to zero as the confidence of the spindle recognition model in the correct category increases. The focus loss function adds a regulation term to the cross entropy loss in order to focus on samples that are difficult to classify during training.

[0056] The spindle wave recognition model was optimized using the Adam optimizer, which is particularly suitable for processing non-stationary data sets (such as EEG signals). The learning rate was initially set to 0.001 and gradually decreased as the spindle wave recognition model converged, reaching a minimum of 1×10 -8 .

[0057] Step 204 specifically includes the following steps 301 to 303 .

[0058] Step 301: Use the first convolutional neural network to perform feature extraction on each signal segment in the first data set to obtain a first feature set.

[0059] Step 302: Use the second convolutional neural network to perform feature extraction on each signal segment in the second data set to obtain a second feature set.

[0060] Step 303: After concatenating the first feature set and the second feature set, the fully connected neural network is used to determine whether spindles exist in the sleep EEG signal time series and the start time and duration of the spindles.

[0061] In this application, each signal segment is processed by CNN, and the CNN architecture uses a 5×5 convolution kernel with a step size of 2. The spindle wave recognition model also includes a padding operation to ensure that the output dimension matches the input dimension, thereby achieving consistent feature extraction in each layer. After the signal segment is convolved in the convolutional neural network, the ReLU activation function is applied to introduce nonlinearity, enhancing the spindle wave recognition model's ability to model complex patterns. ReLU continues to be the main activation function of these layers to maintain nonlinearity. Subsequently, maximum pooling is performed to reduce the spatial dimension and retain only the most significant features, further simplifying the calculation. After convolution and maximum pooling, the output results are flattened and passed to a fully connected neural network with 4 hidden layers. The first input layer of the fully connected neural network has 256 nodes to adapt to the complexity and diversity of the spindle wave pattern. The subsequent hidden layers gradually reduce the number of nodes (128 nodes, 64 nodes, 32 nodes), and gradually refine the feature representation. The final output layer contains two nodes, corresponding to the start time and duration of the spindle wave respectively. The 32 nodes are connected to the two nodes by multiplying each node by the weight, and further through the activation function operation, an output value is obtained, which is the start time. The other output is the duration after the same operation, but with a different weight.

[0062] like Figure 4 As shown in the figure, assuming that the input sleep EEG signal time series has a duration of 8 hours and a frequency of 64 Hz, there are a total of 28,800 seconds × 64 data points. Then, a 2-second signal segment, a 4-second signal segment, and a 6-second signal segment are selected respectively. After the signal segment passes through a convolution layer with a convolution kernel of 5×5, a step size of 2, and a padding size of 1, and then passes through a convolution layer with a convolution kernel of 3×3, a step size of 1, and a padding size of 0, a vector with 256 nodes is obtained by splicing and input into the fully connected neural network. After passing through the hidden layer of the fully connected neural network, the number of nodes is reduced to 128, 64, and 32 respectively, and finally the start time and duration are output through two nodes.

[0063] This application uses bandpass filtering to screen candidate bands. Filtering improves signal stability in noisy EEG environments, reduces interference from muscle movement, eye movements, and other non-neural activity, and improves the accuracy of spindle identification. Using multiple time windows to segment the sleep EEG signal time series improves the accuracy of spindle capture. The duration of spindles is 0.5 seconds to 3 seconds. Setting a reasonable size for segmenting sleep data segments is very important for the accurate identification of spindles. Segmentation that is too short or too long may cause loss of spindle identification. This application uses multiple time windows to segment sleep EEG data. The spindle recognition model concatenates features from multiple windows before passing them to the fully connected layer. After the convolution branch processes the EEG signal, the extracted features are flattened and concatenated into a single feature vector. This feature vector integrates information at different time scales, enabling the spindle recognition model to fuse short-term and long-term features. By fusing information from different windows, the spindle recognition model can detect spindles with different durations and amplitudes, ensuring that spindles that may be missed by the single-window model are fully captured, better adapting to individual differences in spindle duration, and improving the accuracy of spindle recognition.

[0064] The sleep spindle intelligent recognition method provided by this application has a recall rate of 0.92, a precision rate of 0.88, and an F1 score of 0.90 in its own dataset. It has been verified in the DREAM and MASS datasets of public sleep EEG data. In the DREAM dataset, the recall rate of this application reached 0.91, the precision rate reached 0.89, and the F1 score reached 0.90. In the MASS dataset, the recall rate of this application reached 0.92, the precision rate reached 0.88, and the F1 score reached 0.90.

[0065] Based on the same inventive concept, embodiments of the present application also provide a sleep spindle intelligent identification system for implementing the aforementioned sleep spindle intelligent identification method. The solution provided by this system is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more sleep spindle intelligent identification system embodiments provided below can be found in the limitations of the sleep spindle intelligent identification method described above and will not be repeated here.

[0066] In an exemplary embodiment, Figure 5 As shown, a sleep spindle intelligent recognition system is provided, which includes: a signal acquisition module 501, a signal filtering module 502, a signal segmentation module 503 and a signal recognition module 504.

[0067] The signal acquisition module 501 is used to acquire a time series of sleep EEG signals.

[0068] The signal filtering module 502 is used to perform bandpass filtering on the sleep EEG signal time series to obtain a filtered EEG signal time series.

[0069] The signal segmentation module 503 is used to segment the sleep EEG signal time series and the filtered EEG signal time series into overlapping window signals of different lengths to obtain time segment data sets.

[0070] The signal recognition module 504 is configured to determine whether spindles exist in the sleep EEG signal time series and the start time and duration of the spindles using a spindle recognition model based on the time segment data set.

[0071] The spindle recognition model is constructed based on a convolutional neural network and pre-trained using a training sample set. The training sample set includes multiple time segment sample sets and spindle labels for each time segment sample set. The spindle labels include the presence of spindles, the spindle start time, and the spindle duration.

[0072] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, I / O) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store sleep EEG signal time series, and the input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for intelligently identifying sleep spindles is implemented.

[0073] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0074] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0075] In an exemplary embodiment, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0076] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0077] In this application, all actions to obtain signals, information or data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0078] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0079] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0080] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0081] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A sleep spindle intelligent identification method, characterized in that: The sleep spindle intelligent identification method comprises: Obtain sleep EEG signal time series; Performing bandpass filtering on the sleep EEG signal time series to obtain a filtered EEG signal time series; Splitting the sleep EEG signal time series and the filtered EEG signal time series into overlapping window signals of different lengths to obtain time segment data sets; According to the time segment data set, a spindle wave recognition model is used to determine whether spindle waves exist in the sleep EEG signal time series and the start time and duration of the spindle waves; The spindle recognition model is constructed based on a convolutional neural network and is pre-trained using a training sample set; the training sample set includes multiple time segment sample sets and spindle labels for each time segment sample set; the spindle labels include whether spindles exist, the start time and duration of the spindles.

2. The sleep spindle intelligent identification method according to claim 1, characterized in that: Obtain sleep EEG signal time series, including: Polysomnography was used to collect the time series of sleep EEG signals.

3. The sleep spindle intelligent identification method according to claim 1, characterized in that: Performing bandpass filtering on the sleep EEG signal time series to obtain a filtered EEG signal time series specifically includes: digitizing the sleep EEG signal time series and removing the baseline and noise waves to obtain a preprocessed EEG signal time series; The preprocessed EEG signal time series is filtered using a Butterworth filter to obtain a filtered EEG signal time series.

4. The sleep spindle intelligent identification method according to claim 1, characterized in that: The time segment data set includes a first data set and a second data set; the first data set includes multiple signal segments of different time lengths of the filtered EEG signal time series; the second data set includes multiple signal segments of different time lengths of the sleep EEG signal time series.

5. The sleep spindle intelligent identification method according to claim 4, characterized in that: The spindle wave recognition model includes a first convolutional neural network, a second convolutional neural network and a fully connected neural network; Based on the time segment data set, a spindle wave recognition model is used to determine whether spindle waves exist in the sleep EEG signal time series and the start time and duration of the spindle waves, specifically including: Using the first convolutional neural network to perform feature extraction on each signal segment in the first data set to obtain a first feature set; Using the second convolutional neural network to perform feature extraction on each signal segment in the second data set to obtain a second feature set; After concatenating the first feature set and the second feature set, the fully connected neural network is used to determine whether spindles exist in the sleep EEG signal time series and the start time and duration of the spindles.

6. The sleep spindle intelligent identification method according to claim 1, characterized in that: The loss function during the training of the spindle wave recognition model is a focal loss function.

7. A sleep spindle intelligent recognition system, applied to the sleep spindle intelligent recognition method according to any one of claims 1 to 6, characterized in that: The sleep spindle intelligent recognition system includes: Signal acquisition module, used to obtain sleep EEG signal time series; a signal filtering module, configured to perform bandpass filtering on the sleep EEG signal time series to obtain a filtered EEG signal time series; a signal segmentation module, configured to segment the sleep EEG signal time series and the filtered EEG signal time series into overlapping window signals of different lengths to obtain a time segment data set; a signal recognition module, configured to determine, based on the time segment data set and using a spindle recognition model, whether spindles exist in the sleep EEG signal time series, as well as the start time and duration of the spindles; The spindle recognition model is constructed based on a convolutional neural network and is pre-trained using a training sample set; the training sample set includes multiple time segment sample sets and spindle labels for each time segment sample set; the spindle labels include whether spindles exist, the start time and duration of the spindles.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the sleep spindle intelligent identification method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the sleep spindle intelligent recognition method according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the sleep spindle intelligent recognition method according to any one of claims 1 to 6 is implemented.

Images