Non-uniform dynamic brain network node determination method and system
By dividing the EEG time series into several sub-sequences and constructing a dynamic brain network with mutual information entropy matrix, the shortcomings of the dynamic determination method of network nodes in the existing technology are solved, and the dynamic brain network construction of multiple non-uniform nodes is realized, improving the The sparseness and small world attributes of the network are suitable for feature extraction and classification analysis of EEG time series.
Patent Information
- Application Number
- CN202411404444.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-10
- Publication Date
- 2024-11-08
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The lack of a number of sub-sequences in the prior art to dynamic determination methods for network nodes to construct dynamic brain networks for nodes, resulting in unreasonable sparseness and small-world attributes of the network.
By dividing the EEG time series into several sub-sequences, the expected value of the mutual information entropy matrix is calculated and sparse, a dynamic brain network of the mutual information entropy matrix is constructed, and the number of network nodes is updated according to the sparseness and small world attribute indicators are constructed to build a target dynamic Brain network.
Automatic division of network nodes is realized, and a dynamic brain network with multiple non-uniform nodes is constructed, which improves the sparseness of the network and the rationality of small-world attributes, and is suitable for feature extraction and classification analysis of EEG time series.
Smart Images

Figure CN118917347A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of brain-like intelligence technology, and in particular, relates to a method and system for determining nodes in a non-uniform dynamic brain network. Background Art
[0002] The time series time division methods mainly include fixed (uniform) sampling point division, Jaccard coefficient division of edge sets before and after network changes, and division of estimated community changes between two adjacent time slices. Although there are time series time division methods, they are all single sampling point divisions. There is no method for selecting the number of subsequences composed of several sampling points, or a method for determining the network nodes of a dynamic brain network constructed with several subsequences as nodes. The network constructed based on sparsity and "small world" properties has higher credibility and more reasonable sparsity. Summary of the invention
[0003] The present invention provides a method and system for determining nodes in a non-uniform dynamic brain network, which are used to solve the current situation that there is no method for dynamically determining network nodes that uses several subsequences as nodes to construct a dynamic brain network.
[0004] In a first aspect, the present invention provides a method for determining nodes in a non-uniform dynamic brain network, comprising: Set the number of network nodes of the initial network to be equal to the number of EEG channels C, that is, the number of subsequences C; Divide the EEG time series into a number of EEG time subsequences, and select the first C EEG time subsequences that are equal to the number of network nodes; According to the mutual information entropy between each time subsequence, a mutual information entropy matrix M is created, and an expected value E of the mutual information entropy matrix M is calculated, and the mutual information entropy matrix M is sparsely processed according to the expected value E; A dynamic brain network with mutual information entropy matrix M is constructed with EEG time subsequences as network nodes and mutual information entropy as edge weights, and the sparsity of the dynamic brain network is calculated ; Taking the number of nodes in the dynamic brain network as the number of nodes, setting the connection probability p between nodes, constructing an ER random network, and calculating the small-world property index of the dynamic brain network based on the ER random network ; According to the sparsity and the small-world property index The dynamic brain network is updated to obtain a target dynamic brain network.
[0005] In a second aspect, the present invention provides a non-uniform dynamic brain network node determination system, comprising: A setting module is configured to set the number of network nodes of the initial network to be equal to the number of EEG channels C, that is, the number of subsequences C; a partitioning module configured to partition the EEG time series into a plurality of EEG time subsequences, and select the first C EEG time subsequences having the same number as the number of network nodes; A first calculation module is configured to create a mutual information entropy matrix M according to the mutual information entropy between each time subsequence, calculate an expected value E of the mutual information entropy matrix M, and perform sparseness on the mutual information entropy matrix M according to the expected value E; The second computing module is configured to construct a dynamic brain network with the mutual information entropy matrix M using the EEG time subsequences as network nodes and the mutual information entropy as edge weights, and calculate the sparsity of the dynamic brain network ; A construction module is configured to take the number of nodes in the dynamic brain network as the number of nodes, set the connection probability p between nodes, construct an ER random network, and calculate the small-world attribute index of the dynamic brain network based on the ER random network ; An update module is configured to update the and the small-world property index The dynamic brain network is updated to obtain a target dynamic brain network.
[0006] According to a third aspect, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the steps of the method for determining a non-uniform dynamic brain network node according to any embodiment of the present invention.
[0007] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor executes the steps of the method for determining nodes of a non-uniform dynamic brain network according to any embodiment of the present invention.
[0008] The non-uniform dynamic brain network node determination method and system of the present application divides the continuous time series into multiple subsequences, selects several subsequences to construct the brain network, and realizes automatic division of the number of network nodes. In this way, one EEG time series can construct multiple dynamic brain networks with non-uniform node numbers. BRIEF DESCRIPTION OF THE DRAWINGS
[0009] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.
[0010] Figure 1 A flowchart of a method for determining nodes in a non-uniform dynamic brain network provided by one embodiment of the present invention; Figure 2 A structural block diagram of a non-uniform dynamic brain network node determination system provided by one embodiment of the present invention; Figure 3 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0011] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0012] See also Figure 1 , which shows a flow chart of a method for determining nodes in a non-uniform dynamic brain network of the present application.
[0013] like Figure 1 As shown, the method for determining nodes in a non-uniform dynamic brain network specifically includes the following steps: Step S101, setting the number of network nodes of the initial network to be equal to the number of EEG channels C, that is, the number of subsequences C.
[0014] Step S102, dividing the EEG time series into a number of EEG time sub-sequences, and selecting the first C EEG time sub-sequences having the same number as the number of network nodes.
[0015] In this step, the EEG time series is divided into a number of EEG time subsequences based on the sampling frequency, and the first C EEG time subsequences having the same number as the network nodes are selected.
[0016] For example, assuming that the number of EEG time series channels C = 30, the sampling frequency fs = 1000 Hz / s, the sampling time T = 300 s, a total of N = 300,000 data samples are sampled, and according to the sampling frequency fs, N is divided into 300 time subsequences N = {N1, N2, …, N300}.
[0017] Step S103: create a mutual information entropy matrix M according to the mutual information entropy between each time subsequence, calculate an expected value E of the mutual information entropy matrix M, and perform sparseness on the mutual information entropy matrix M according to the expected value E.
[0018] In this step, the mutual information entropy I between each time subsequence is calculated, and the expression is: , In the formula, For network nodes and network nodes The mutual information entropy between is the joint probability when N1=x, N2=y, is the probability when N1=x, The probability when N2=y; According to the mutual information entropy between each time subsequence, a mutual information entropy matrix M is created. The expression of the mutual information entropy matrix M is: , In the formula, is the time subsequence and subsequences The mutual information entropy between ,i,j∈C.
[0019] The expected value E of the mutual information entropy matrix M is calculated, the values in the mutual information entropy matrix M that are greater than the expected value E are retained, and the rest are deleted. The retained values are the network edge weights, and a dynamic brain network is constructed.
[0020] Step S104, constructing a dynamic brain network with mutual information entropy matrix M using EEG time subsequences as network nodes and mutual information entropy as edge weights, and calculating the sparsity of the dynamic brain network .
[0021] In this step, the sparsity of the dynamic brain network is calculated The expression is: , where e is the number of edges in the dynamic brain network and C is the number of network nodes.
[0022] Step S105, taking the number of nodes in the dynamic brain network as the number of nodes, setting the connection probability p between nodes, constructing an ER random network, and calculating the small-world attribute index of the dynamic brain network based on the ER random network .
[0023] In this step, the number of network nodes C of the ER random network is initialized, and C nodes are isolated nodes; two unconnected nodes are randomly selected from the number of network nodes C; a random number r between 0 and 1 is generated; if r is greater than or equal to the probability p, the two nodes are connected until all nodes are traversed and the ER random network is constructed.
[0024] Furthermore, the small-world property index of the dynamic brain network is calculated The expression is: , , , In the formula, is the path length, is the clustering coefficient, is the clustering coefficient of the dynamic brain network, is the clustering coefficient of the random network, is the shortest path length of the dynamic brain network, is the shortest path length in a random network.
[0025] Step S106: according to the sparsity and the small-world property index The dynamic brain network is updated to obtain a target dynamic brain network.
[0026] In this step, it is determined whether the sparsity is within the preset range and the small-world attribute index Is it greater than a preset threshold, where the preset range is the interval range from the first preset threshold to the second preset threshold; if the sparsity is within the preset range and the small world attribute index If the sparsity is less than the first preset threshold, or the small-world attribute index is If the sparsity is not greater than a preset threshold, the number of network nodes is increased and the dynamic brain network is reconstructed; if the sparsity is greater than a second preset threshold, the number of network nodes is reduced and the dynamic brain network is reconstructed.
[0027] For example, if 5%<=ρ<=50%, and Greater than 1, the network is successfully built; if Less than 5% or If it is less than 1, increase the number of network nodes and rebuild the network; if If it is greater than 50%, reduce the number of network nodes and reconstruct the network; In other cases, the network construction is completed with the number of channels and p.
[0028] In summary, the method of the present application divides a continuous time series into multiple subsequences, selects several subsequences to construct a brain network, and realizes automatic division of the number of network nodes. Because one time series constructs multiple brain networks, and the number of nodes in the network is not equal, it has the characteristic of non-uniform number of nodes. Using the time-varying nature of the time series, the time series is divided from the time dimension, and the correlation between each subsequence is analyzed over time. One time series constructs multiple non-uniform brain networks, preparing for feature extraction, classification, and analysis of EEG time series.
[0029] In a specific embodiment, a private resting-state dataset is used as an example to construct a dynamic brain network and determine the number of nodes in the network.
[0030] Import channel 1 data, N=300000 sampling points, divide the EEG time series into 300 subsequences, each with 1000 sampling points.
[0031] Initialize the variable s=0 to increase or decrease the dynamic brain network; The number of nodes in the initial network is C = 30, which is the number of subsequences; Ni = pandas.cut(Ni,100); Nj = pandas.cut(Nj,100), where 0 <i,j<=C; Call metrics.normalized_mutual_info_score(Ni,Nj) to calculate the mutual information entropy I between nodes N1, N2, ..., NC in the network: ; According to the mutual information entropy between 30 time subsequences, create the matrix M: , Calculate the expected value E(M) = numpy.mean(M) of the matrix M, add edges to the values in the matrix M that are greater than the expected value E, and construct a dynamic brain network.
[0032] Calculate the dynamic brain network sparsity ρ=e / (C(C-1)*0.5); Calculate the clustering coefficient of dynamic brain networks; Calculate the shortest path length in dynamic brain networks; Call generate_random_network2(C,p=0.2) in networkx to generate an ER random network; Calculation indicators ; If 5%<= <=50% and Greater than 1, the network is successfully built; if Less than 5% or If it is less than 1, increase the number of network nodes C=C+1 and rebuild the network; if If it is greater than 50%, reduce the number of network nodes to C=C-1 and reconstruct the network; The other networks are constructed with the number of channels and p=0.2, and the construction is completed.
[0033] See also Figure 2 , which shows a structural block diagram of a non-uniform dynamic brain network node determination system of the present application.
[0034] like Figure 2 As shown, the non-uniform dynamic brain network node determination system 200 includes a setting module 210, a division module 220, a first calculation module 230, a second calculation module 240, a construction module 250 and an update module 260.
[0035] Among them, the setting module 210 is configured to set the number of network nodes of the initial network to be equal to the number of EEG channels C, that is, the number of subsequences C; the division module 220 is configured to divide the EEG time series into a number of EEG time subsequences, and select the first C EEG time subsequences with the same number of network nodes; the first calculation module 230 is configured to create a mutual information entropy matrix M according to the mutual information entropy between each time subsequence, and calculate the expected value E of the mutual information entropy matrix M, and sparse the mutual information entropy matrix M according to the expected value E; the second calculation module 240 is configured to construct a dynamic brain network of the mutual information entropy matrix M with the EEG time subsequences as network nodes and the mutual information entropy as edge weights, and calculate the sparsity of the dynamic brain network ; Construction module 250 is configured to use the number of nodes in the dynamic brain network as the number of nodes, set the connection probability p between nodes, construct an ER random network, and calculate the small-world attribute index of the dynamic brain network based on the ER random network ; Update module 260, configured to update the sparsity according to the and the small-world property index The dynamic brain network is updated to obtain a target dynamic brain network.
[0036] It should be understood that Figure 2 Modules and references documented in Figure 1 Therefore, the operations and features described above for the method and the corresponding technical effects are also applicable to Figure 2 The modules in it will not be described in detail here.
[0037] In some other embodiments, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor executes the method for determining nodes of a non-uniform dynamic brain network in any of the above method embodiments; As an implementation mode, the computer-readable storage medium of the present invention stores computer-executable instructions, and the computer-executable instructions are configured as follows: Set the number of network nodes of the initial network to be equal to the number of EEG channels C, that is, the number of subsequences C; Divide the EEG time series into a number of EEG time subsequences, and select the first C EEG time subsequences that are equal to the number of network nodes; According to the mutual information entropy between each time subsequence, a mutual information entropy matrix M is created, and an expected value E of the mutual information entropy matrix M is calculated, and the mutual information entropy matrix M is sparsely processed according to the expected value E; A dynamic brain network with mutual information entropy matrix M is constructed with EEG time subsequences as network nodes and mutual information entropy as edge weights, and the sparsity of the dynamic brain network is calculated ; Taking the number of nodes in the dynamic brain network as the number of nodes, setting the connection probability p between nodes, constructing an ER random network, and calculating the small-world property index of the dynamic brain network based on the ER random network ; According to the sparsity and the small-world property index The dynamic brain network is updated to obtain a target dynamic brain network.
[0038] The computer-readable storage medium may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the non-uniform dynamic brain network node determination system, etc. In addition, the computer-readable storage medium may include a high-speed random access memory, and may also include a memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the non-uniform dynamic brain network node determination system via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0039] Figure 3 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 3 As shown, the device includes: a processor 310 and a memory 320. The electronic device may also include: an input device 330 and an output device 340. The processor 310, the memory 320, the input device 330 and the output device 340 may be connected via a bus or other means. Figure 3 In the example, the connection through the bus is taken as an example. The memory 320 is the above-mentioned computer-readable storage medium. The processor 310 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 320, that is, the non-uniform dynamic brain network node determination method of the above-mentioned method embodiment is realized. The input device 330 can receive input digital or character information, and generate key signal input related to user settings and function control of the non-uniform dynamic brain network node determination system. The output device 340 may include a display device such as a display screen.
[0040] The electronic device can execute the method provided by the embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not described in detail in this embodiment, please refer to the method provided by the embodiment of the present invention.
[0041] As an implementation mode, the electronic device is applied to a non-uniform dynamic brain network node determination system and is used for a client, and includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: Set the number of network nodes of the initial network to be equal to the number of EEG channels C, that is, the number of subsequences C; Divide the EEG time series into a number of EEG time subsequences, and select the first C EEG time subsequences that are equal to the number of network nodes; According to the mutual information entropy between each time subsequence, a mutual information entropy matrix M is created, and an expected value E of the mutual information entropy matrix M is calculated, and the mutual information entropy matrix M is sparsely processed according to the expected value E; A dynamic brain network with mutual information entropy matrix M is constructed with EEG time subsequences as network nodes and mutual information entropy as edge weights, and the sparsity of the dynamic brain network is calculated ; Taking the number of nodes in the dynamic brain network as the number of nodes, setting the connection probability p between nodes, constructing an ER random network, and calculating the small-world property index of the dynamic brain network based on the ER random network ; According to the sparsity and the small-world property index The dynamic brain network is updated to obtain a target dynamic brain network.
[0042] Through the description of the above implementation modes, those skilled in the art can clearly understand that each implementation mode can be implemented by means of software plus a necessary general hardware platform, or of course by hardware. Based on such an understanding, the above technical solution can essentially or in other words be embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.
[0043] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for determining nodes in a non-uniform dynamic brain network, characterized in that: include: Set the number of network nodes of the initial network to be equal to the number of EEG channels C, that is, the number of subsequences C; Divide the EEG time series into a number of EEG time subsequences, and select the first C EEG time subsequences that are equal to the number of network nodes; According to the mutual information entropy between each time subsequence, a mutual information entropy matrix M is created, and an expected value E of the mutual information entropy matrix M is calculated, and the mutual information entropy matrix M is sparsely processed according to the expected value E; A dynamic brain network with mutual information entropy matrix M is constructed with EEG time subsequences as network nodes and mutual information entropy as edge weights, and the sparsity of the dynamic brain network is calculated ; Taking the number of nodes in the dynamic brain network as the number of nodes, setting the connection probability p between nodes, constructing an ER random network, and calculating the small-world property index of the dynamic brain network based on the ER random network ; According to the sparsity and the small-world property index The dynamic brain network is updated to obtain a target dynamic brain network.
2. A method for determining nodes in a non-uniform dynamic brain network according to claim 1, characterized in that: The step of dividing the EEG time series into a plurality of EEG time subsequences and selecting the first C EEG time subsequences having the same number as the number of network nodes comprises: The EEG time series is divided into a number of EEG time sub-sequences based on the sampling frequency, and the first C EEG time sub-sequences having the same number as the network nodes are selected.
3. A method for determining nodes in a non-uniform dynamic brain network according to claim 1, characterized in that: The creating a mutual information entropy matrix M according to the mutual information entropy between each time subsequence, calculating an expected value E of the mutual information entropy matrix M, and performing sparse processing on the mutual information entropy matrix M according to the expected value E comprises: Calculate the mutual information entropy I between time subsequences, the expression is: , In the formula, For network nodes and network nodes The mutual information entropy between is the joint probability when N1=x, N2=y, is the probability when N1=x, The probability when N2=y; According to the mutual information entropy between each time subsequence, a mutual information entropy matrix M is created. The expression of the mutual information entropy matrix M is: , In the formula, is the time subsequence and subsequences The mutual information entropy between ,i,j∈C; The expected value E of the mutual information entropy matrix M is calculated, the values in the mutual information entropy matrix M that are greater than the expected value E are retained, and the rest are deleted. The retained values are the network edge weights, and a dynamic brain network is constructed.
4. The method for determining nodes in a non-uniform dynamic brain network according to claim 1, characterized in that: Calculating the sparsity of the dynamic brain network The expression is: , Where e is the number of edges in the dynamic brain network, and C is the number of network nodes.
5. The method for determining nodes in a non-uniform dynamic brain network according to claim 1, characterized in that: The method of taking the number of nodes in the dynamic brain network as the number of nodes, setting the connection probability p between nodes, and constructing the ER random network includes: Initialize the number of network nodes C of the ER random network, and C nodes are isolated nodes; Randomly select two unconnected nodes from the number of network nodes C; Generate a random number r between 0 and 1; If r is greater than or equal to the probability p, the two nodes are connected until all nodes are traversed and the ER random network is constructed.
6. A method for determining nodes in a non-uniform dynamic brain network according to claim 1, characterized in that: Calculating the small-world property index of the dynamic brain network The expression is: , , , In the formula, is the path length, is the clustering coefficient, is the clustering coefficient of the dynamic brain network, is the clustering coefficient of the random network, is the shortest path length of the dynamic brain network, is the shortest path length in a random network.
7. The method for determining nodes in a non-uniform dynamic brain network according to claim 1, characterized in that: According to the sparsity and the small-world property index The dynamic brain network is updated to obtain a target dynamic brain network including: Determine whether the sparsity is within a preset range and whether the small-world attribute index Whether it is greater than a preset threshold, wherein the preset range is an interval range from a first preset threshold to a second preset threshold; If the sparsity is within a preset range and the small-world attribute index If the dynamic brain network is greater than a preset threshold, the dynamic brain network is directly used as the target dynamic brain network; If the sparsity is less than the first preset threshold, or the small world attribute index If it is not greater than the preset threshold, the number of network nodes is increased to reconstruct the dynamic brain network; If the sparsity is greater than a second preset threshold, the number of network nodes is reduced to reconstruct the dynamic brain network.
8. A non-uniform dynamic brain network node determination system, characterized in that: include: A setting module is configured to set the number of network nodes of the initial network to be equal to the number of EEG channels C, that is, the number of subsequences C; a partitioning module configured to partition the EEG time series into a plurality of EEG time subsequences, and select the first C EEG time subsequences having the same number as the number of network nodes; A first calculation module is configured to create a mutual information entropy matrix M according to the mutual information entropy between each time subsequence, calculate an expected value E of the mutual information entropy matrix M, and perform sparseness on the mutual information entropy matrix M according to the expected value E; The second computing module is configured to construct a dynamic brain network with the mutual information entropy matrix M using the EEG time subsequences as network nodes and the mutual information entropy as edge weights, and calculate the sparsity of the dynamic brain network ; A construction module is configured to take the number of nodes in the dynamic brain network as the number of nodes, set the connection probability p between nodes, construct an ER random network, and calculate the small-world attribute index of the dynamic brain network based on the ER random network ; An update module is configured to update the and the small-world property index The dynamic brain network is updated to obtain a target dynamic brain network.
9. An electronic device, characterized in that: include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method described in any one of claims 1 to 7.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.
Citation Information
Patent Citations
Complicated network construction and analysis method with microstate EEG (electroencephalogram) time domain features
CN109497996A
Dynamic brain network node consistent-behavior analysis method based on electroencephalogram signals
CN110338785A
Rest-state brain network construction method, device and equipment and computer storage medium
CN111753947A
Processing method and device for near-infrared brain function imaging data, and storage medium
CN112274144A
Brain function network construction method and device, electronic equipment and storage medium
CN113558641A
