Medical data feature extraction method and system based on machine learning

Abstract

According to the medical data feature extraction method and system based on machine learning disclosed by the invention, based on a machine learning algorithm, multiple pieces of medical data acquired from terminal monitoring equipment are subjected to machine learning analysis to obtain target medical data, and the target medical data are mapped into a high-dimensional space through a nonlinear mapping function; establishing a feature matrix, namely a kernel matrix, in a high-dimensional space, carrying out centralization processing on the kernel matrix, solving to obtain a feature value and a feature vector of the centralized kernel matrix, and according to the feature value and the feature vector of the centralized kernel matrix, calculating the feature value and the feature vector of the kernel matrix; and extracting a kernel principal component corresponding to each piece of target medical data of the original spatial data set from the high-dimensional space, and determining at least one piece of target medical data from the original spatial data set based on the extracted kernel principal component, and outputting the target medical data as standard monitoring data so as to perform medical data analysis by using the standard monitoring data to obtain an analysis result.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and specifically relates to a medical data feature extraction method and system based on machine learning. Background Art

[0002] At present, medical informationization and intelligence are increasingly becoming new development directions and trends in the medical field. Among them, the processing of medical data is the top priority of medical intelligence. Only by accurately obtaining the desired medical data from a variety of monitoring data can medical workers or computer programs make accurate classification, identification and diagnosis recommendations based on the medical data.

[0003] Most of the existing information feature extraction and processing for medical data are based on deep learning algorithms, such as "A medical data processing method, device, equipment and storage medium" with publication number CN111785340A, which generates medical item data based on the patient's medical data; extracts features from the medical data and medical item data to obtain a medical feature vector; and extracts features from the diagnostic data associated with the medical data to obtain a diagnostic feature vector; and determines the correlation between the medical feature vector and the diagnostic feature vector. However, this medical data processing method and feature extraction method are too dependent on the quality of the data, and the medical data in medical monitoring (such as medical monitoring data obtained by monitoring the blood oxygen, pulse rate, respiratory rate and end-tidal carbon dioxide of intubated patients) usually has large fuzziness and nonlinear characteristics, which will affect the classification and recognition efficiency and judgment accuracy.

[0004] Therefore, how to accurately extract nonlinear features when extracting medical data features has become a problem that needs to be solved urgently in this technical field. Summary of the invention

[0005] The purpose of the present invention is to provide a medical data feature extraction method and system based on machine learning to solve the above-mentioned problems existing in the prior art.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a method for extracting medical data features based on machine learning, comprising:

[0008] Acquire multiple medical data generated by the terminal monitoring device, and use a machine learning algorithm to parse each medical data to obtain n target medical data, so as to use the n target medical data to form an original spatial data set, wherein each target medical data has a different data type, and any target medical data is low-dimensional data;

[0009] A nonlinear mapping function is established, and the n target medical data in the original spatial data set are mapped into the high-dimensional space using the nonlinear mapping function to obtain the high-dimensional medical data corresponding to each target medical data;

[0010] Using n high-dimensional medical data, a kernel matrix is ​​constructed, and the kernel matrix is ​​centralized to obtain a centralized kernel matrix;

[0011] Calculate the eigenvalues ​​and eigenvectors of the centralized kernel matrix;

[0012] According to the eigenvalues ​​and eigenvectors of the centralized kernel matrix, the kernel principal components corresponding to the target medical data of the original spatial data set are extracted in the high-dimensional space, and based on the extracted kernel principal components, at least one target medical data is determined from the original spatial data set and output as standard monitoring data, so as to perform medical data analysis using the standard monitoring data and obtain analysis results.

[0013] In a possible design, multiple medical data generated by a terminal monitoring device are obtained, and a machine learning algorithm is used to parse each medical data to obtain n target medical data, so as to use the n target medical data to form an original spatial data set, including:

[0014] Acquire historical medical data of the terminal monitoring device, and classify the historical medical data according to preset data types to obtain multiple classification sets, wherein each classification set corresponds to a data type;

[0015] Dividing the historical medical data in each of the classification sets into an initial training data set and a test data set according to a preset ratio;

[0016] Setting a label for each historical medical data in each initial training data set to obtain each training data set, wherein the label of any historical medical data is used to characterize the data type of the any medical data;

[0017] Taking each historical medical data in each training data set as input and the data type of each historical medical data as output, a machine learning model is trained, and during the training process, the output of the machine learning model is used to adjust the parameters of the machine learning model, so as to obtain a pre-medical data classification model after the training is completed;

[0018] Use each test data set to test the pre-medical data classification model, obtain the test results, and determine whether the test results meet the preset conditions;

[0019] If not, the preliminary medical data classification model is optimized until the test results meet the preset conditions, and the final medical data classification model is obtained;

[0020] The multiple medical data are input into a medical data classification model to obtain a data type of each medical data, and based on the data type of each medical data, n target medical data of different data types are screened out from the multiple medical data.

[0021] In one possible design, a nonlinear mapping function is established, including:

[0022] The natural logarithm function is used as the nonlinear mapping function;

[0023] Accordingly, a nonlinear mapping function is used to map the n target medical data in the original spatial data set into a high-dimensional space to obtain high-dimensional medical data corresponding to each target medical data, including:

[0024] Each target medical data in the original space data set is mapped to the high-dimensional space through the natural logarithm function, where the high-dimensional medical data corresponding to the i-th target medical data in the original space data set is:

[0025]

[0026] In the above formula (1), Represents the i-th target medical data x i Corresponding to high-dimensional medical data, ln() represents the natural logarithm function, where i = 1, 2, ..., n.

[0027] In a possible design, n high-dimensional medical data are used to construct a kernel matrix, and the kernel matrix is ​​centralized to obtain a centralized kernel matrix, including:

[0028] For any high-dimensional medical data, calculate the dot product between the any high-dimensional medical data and n high-dimensional medical data to obtain n dot products, and after all high-dimensional medical data are polled, obtain n×n dot products;

[0029] Using n×n dot products, a kernel matrix is ​​constructed, where the size of the kernel matrix is ​​also n×n;

[0030] The kernel matrix is ​​centralized column by column to obtain a centralized kernel matrix.

[0031] In one possible design, the eigenvalues ​​and eigenvectors of the centered kernel matrix are calculated, including:

[0032] Calculate the covariance matrix of the centered kernel matrix;

[0033] Obtain the eigenvalues ​​and eigenvectors of the covariance matrix, so as to use the eigenvalues ​​of the covariance matrix as the eigenvalues ​​of the centralized kernel matrix;

[0034] According to the eigenvector sum of the covariance matrix and the centralized kernel matrix, the eigenvector corresponding to each eigenvalue of each centralized kernel matrix is ​​calculated.

[0035] In one possible design, the covariance matrix of the centered kernel matrix is ​​calculated;

[0036] According to the following formula (2), the covariance matrix is ​​calculated

[0037]

[0038] Among them, C H The covariance matrix used to represent the centered kernel matrix K', represents the vector composed of the elements in the i-th row of the centralized kernel matrix, and T represents the transpose operation;

[0039] Correspondingly, according to the eigenvectors of the covariance matrix and various high-dimensional medical data, the eigenvectors corresponding to the eigenvalues ​​of the centralized kernel matrix are calculated, including:

[0040] According to the following formula (3), the eigenvector corresponding to each eigenvalue is calculated;

[0041]

[0042] ν represents the eigenvector of the covariance matrix, and a represents the eigenvector of the centered kernel matrix.

[0043] In a possible design, according to the eigenvalues ​​and eigenvectors of the centralized kernel matrix, the kernel principal components corresponding to the target medical data of the original spatial data set are extracted in the high-dimensional space, and based on the extracted kernel principal components, at least one target medical data is determined from the original spatial data set and output as standard monitoring data, so as to perform medical data analysis using the standard monitoring data to obtain analysis results, including:

[0044] The eigenvectors corresponding to the eigenvalues ​​of the centralized kernel matrix are used as kernel principal elements;

[0045] Based on the variance cumulative contribution rate method, the eigenvectors corresponding to multiple eigenvalues ​​whose contribution rates are higher than the preset threshold are selected from the kernel principal components as the final kernel principal components;

[0046] Based on the nonlinear mapping function, the final kernel principal component is inversely operated to obtain the target medical data corresponding to the final kernel principal component in the original spatial data set, so that the target medical data corresponding to the final kernel principal component in the original spatial data set can be output as standard monitoring data.

[0047] In a second aspect, the present invention provides a medical data feature extraction system based on machine learning, comprising:

[0048] A data acquisition unit, used to acquire multiple medical data generated by the terminal monitoring device, and analyze the medical data using a machine learning algorithm;

[0049] A mapping establishment unit, used to establish a nonlinear mapping function, and use the nonlinear mapping function to map n target medical data in the original space data set to the high-dimensional space to obtain high-dimensional medical data corresponding to each target medical data, and also used to use the n high-dimensional medical data to construct a kernel matrix;

[0050] A mean processing unit, used for performing centralization processing on the kernel matrix to obtain a centralized kernel matrix;

[0051] Matrix operation unit, used to calculate the eigenvalues ​​and eigenvectors of the centralized kernel matrix;

[0052] The data screening unit is used to extract the core principal components corresponding to each target medical data of the original spatial data set in the high-dimensional space according to the eigenvalues ​​and eigenvectors of the centralized kernel matrix, and based on the extracted core principal components, determine at least one target medical data from the original spatial data set and output it as standard monitoring data.

[0053] In a third aspect, the present invention provides an electronic device comprising a memory, a processor and a transceiver which are communicatively connected in sequence, wherein the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the medical data feature extraction method based on machine learning as described in the first aspect or any possible design of the first aspect.

[0054] In a fourth aspect, the present invention provides a computer program product, comprising a computer program or instructions, which, when executed on a computer, cause the computer to execute the medical data feature extraction method based on machine learning as described in the first aspect or any possible design of the first aspect.

[0055] Beneficial effects: The medical data feature extraction method and system based on machine learning disclosed in the present invention, based on a machine learning algorithm, performs machine learning analysis on multiple medical data obtained from a terminal monitoring device to obtain target medical data, maps the data to a high-dimensional space through a nonlinear mapping function, and establishes a feature matrix in the high-dimensional space, that is, a kernel matrix. After the kernel matrix is ​​centralized, the eigenvalues ​​and eigenvectors of the centralized kernel matrix are obtained. According to the eigenvalues ​​and eigenvectors of the centralized kernel matrix, the kernel principal components corresponding to each target medical data of the original spatial data set are extracted in the high-dimensional space, and based on the extracted kernel principal components, at least one target medical data is determined from the original spatial data set and output as standard monitoring data, so that the standard monitoring data can be used to perform medical data analysis to obtain analysis results. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a flowchart of a medical data feature extraction method based on machine learning;

[0057] Figure 2 This is the functional structure diagram of the medical data feature extraction system based on machine learning. DETAILED DESCRIPTION

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the present invention will be briefly introduced below in combination with the drawings and the description of the embodiments or the prior art. Obviously, the following description of the structure of the drawings is only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. It should be noted that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation of the present invention.

[0059] It should be understood that although the terms first, second, etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are only used to distinguish one unit from another unit. For example, a first unit can be referred to as a second unit, and similarly, a second unit can be referred to as a first unit without departing from the scope of the exemplary embodiments of the present invention.

[0060] It should be understood that the term "and / or" that may appear in this article is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, A and / or B can represent three situations: A exists alone, B exists alone, and A and B exist at the same time. The term " / and" that may appear in this article describes another type of association object relationship, indicating that two relationships may exist. For example, A / and B can represent two situations: A exists alone, and A and B exist alone. In addition, the character " / " that may appear in this article generally indicates that the previous and next associated objects are in an "or" relationship.

[0061] Embodiment 1:

[0062] like Figure 1 As shown, this embodiment provides a method for extracting medical data features based on machine learning, which specifically includes:

[0063] S100: Acquire multiple medical data generated by the terminal monitoring device, and use a machine learning algorithm to parse each medical data to obtain n target medical data, so as to use the n target medical data to form an original spatial data set, wherein each target medical data has a different data type, and any target medical data is low-dimensional data;

[0064] Among them, n target medical data are defined as x, and the original spatial data set is defined as X, where X = {x i |i=1,2,…,n;x i ∈R}, i is the index number of the target medical data x, and R is a real number set.

[0065] In step S100, the following steps may also be included:

[0066] S1001: Acquire historical medical data of a terminal monitoring device, and classify the historical medical data according to preset data types to obtain multiple classification sets, wherein each classification set corresponds to a data type;

[0067] The specific medical data classification categories include the blood oxygen, pulse rate, respiratory rate and end-tidal carbon dioxide of the intubated patient sent by the terminal monitoring device, which are classified and processed by the data acquisition unit according to different classification categories. The terminal monitoring device here can be a monitoring terminal instrument such as SpO2 / EtCO2 monitor.

[0068] S1002: Divide the historical medical data in each classification category into an initial training data set and a test data set according to a preset ratio;

[0069] S1003: Setting a label for each historical medical data in each initial training data set to obtain each training data set, wherein the label of any historical medical data is used to characterize the classification type of the any medical data;

[0070] The setting of the training data set here can provide training samples for the machine learning algorithm to make the subsequent machine learning testing process smoother.

[0071] S1004: using each historical medical data in each training data set as input and the data type of each historical medical data as output to train a machine learning model, and in the training process, using the output of the machine learning model to adjust the parameters of the machine learning model, so as to obtain a pre-medical data classification model after the training is completed;

[0072] S1005: Using each test data set to test the pre-medical data classification model, obtain the test result, and determine whether the test result meets the preset conditions;

[0073] S1006: If not, the preliminary medical data classification model is optimized until the test result meets the preset conditions, and the final medical data classification model is obtained;

[0074] S1007: Input the plurality of medical data into a medical data classification model to obtain a data type of each medical data, and based on the data type of each medical data, screen out n target medical data of different data types from the plurality of medical data.

[0075] S200: establishing a nonlinear mapping function, and using the nonlinear mapping function to map n target medical data in the original spatial data set to a high-dimensional space, so as to obtain high-dimensional medical data corresponding to each target medical data;

[0076] Specifically, these may include:

[0077] The natural logarithm function is used as the nonlinear mapping function;

[0078] Accordingly, a nonlinear mapping function is used to map the n target medical data in the original spatial data set into a high-dimensional space to obtain high-dimensional medical data corresponding to each target medical data, including:

[0079] Each target medical data in the original space data set is mapped to the high-dimensional space through the natural logarithm function, where the high-dimensional medical data corresponding to the i-th target medical data in the original space data set is:

[0080]

[0081] The above formula (1) Represents the high-dimensional medical data x corresponding to the i-th target medical data i Corresponding to high-dimensional medical data, ln() represents the natural logarithm function, where i = 1, 2, ..., n.

[0082] S300: constructing a kernel matrix using n high-dimensional medical data, and centralizing the kernel matrix to obtain a centralized kernel matrix;

[0083] In step S300, the following steps may also be included:

[0084] S3001: For any high-dimensional medical data, calculate the dot product between the any high-dimensional medical data and n high-dimensional medical data to obtain n dot products, and after all high-dimensional medical data are polled, obtain n×n dot products;

[0085] S3002: construct a kernel matrix using n×n dot products, where the size of the kernel matrix is ​​also n×n;

[0086] S3003: performing centering processing on the kernel matrix by column to obtain a centralized kernel matrix.

[0087] Among them, centering by column is essentially to perform zero mean processing on the kernel matrix, which can be completed by subtracting the mean of each column from the mean of the column; centering by column can ensure that the mean of each feature is zero, so that the machine learning algorithm will not give higher weights to some features because of their large means, thereby improving the accuracy of weight distribution of the main features. And processing by column can perform more accurate and specific analysis on different features. The kernel matrix is ​​represented by K, and the centralized kernel matrix is ​​represented by K'.

[0088] S400: Calculate the eigenvalues ​​and eigenvectors of the centralized kernel matrix;

[0089] In step S400, the following steps may also be included:

[0090] S4001: Calculate the covariance matrix of the centered kernel matrix;

[0091] According to the following formula (2), the covariance matrix is ​​calculated

[0092]

[0093] Among them, C H The covariance matrix used to represent the centered kernel matrix K', represents the vector composed of the elements of the ith row in the centered kernel matrix, and T represents the transposition operation S4002: obtaining the eigenvalues ​​and eigenvectors of the covariance matrix, so as to use the eigenvalues ​​of the covariance matrix as the eigenvalues ​​of the centered kernel matrix;

[0094] S4002: Obtain the eigenvalues ​​and eigenvectors of the covariance matrix, so as to use the eigenvalues ​​of the covariance matrix as the eigenvalues ​​of the centralized kernel matrix;

[0095] S4003: Calculate the eigenvectors corresponding to the eigenvalues ​​of each centered kernel matrix according to the eigenvectors of the covariance matrix and the centered kernel matrix.

[0096] According to the following formula (3), the eigenvector corresponding to each eigenvalue is calculated;

[0097]

[0098] ν represents the eigenvector of the covariance matrix, and a represents the eigenvector of the centered kernel matrix.

[0099] Specifically, by substituting the above formula (3) into the covariance matrix in formula (2), we can obtain the following formula (4):

[0100]

[0101] Among them, taking the original spatial data set in 2D space as an example, the corresponding kernel matrix is ​​in 3D space. The eigenvalues ​​of each covariance matrix and the eigenvectors of the covariance matrix corresponding to each covariance matrix are all subspaces mapped by nonlinear functions (in the following derivation, i, j and k are the index numbers of the target medical data x, i∈{1,2,…,n}, j∈{1,2,…,n}, k∈{1,2,…,n}), so formula (4) can be rewritten as the following formula (5):

[0102]

[0103] Solving formula (5) can obtain ν and λ;

[0104] Substituting formula ν and λ back into formula (1), we can obtain the following formula (6):

[0105]

[0106] According to the specific centralization process, the characteristic equation λna=K'a can be obtained, and the characteristic vector a of the centralized kernel matrix can be obtained accordingly.

[0107] Through the above process, the specified type of medical data completes the process of conversion from the original space (low-dimensional space) to the high-dimensional space. The nonlinear data can be converted into linear operations in the high-dimensional space for processing. The size of the eigenvalue λ obtained according to λna=K'a is the dimension of the corresponding eigenvector a.

[0108] S500: According to the eigenvalues ​​and eigenvectors of the centralized kernel matrix, kernel principal components corresponding to each target medical data of the original spatial data set are extracted in the high-dimensional space, and based on the extracted kernel principal components, at least one target medical data is determined from the original spatial data set and output as standard monitoring data, so as to perform medical data analysis using the standard monitoring data to obtain analysis results.

[0109] In step S500, the following steps may also be included:

[0110] S5001: taking the eigenvectors corresponding to the eigenvalues ​​of the centralized kernel matrix as the kernel principal element;

[0111] S5002: Based on the variance cumulative contribution rate method, eigenvectors corresponding to multiple eigenvalues ​​whose contribution rates are higher than a preset threshold are selected from the kernel principal components as the final kernel principal components;

[0112] S5003: Based on the nonlinear mapping function, the final kernel principal component is inversely operated to obtain the target medical data corresponding to the final kernel principal component in the original spatial data set, so that the target medical data corresponding to the final kernel principal component in the original spatial data set can be output as standard monitoring data.

[0113] Among them, the selected multiple kernel principal components are selected according to the variance cumulative contribution rate method, which is a common method of principal component analysis. Through this method, the eigenvectors with the highest cumulative contribution rate are selected as the kernel principal components. The eigenvector with the highest cumulative contribution rate is the first kernel principal component, the second is the second kernel principal component, and so on. Each kernel principal component is brought into the mapping function An inverse operation is performed to obtain the original image of each kernel principal element in the original space (corresponding to the specified type of medical data), thus completing the dimensionality reduction and extraction of feature data.

[0114] Embodiment 2:

[0115] like Figure 2 As shown, this embodiment provides a medical data feature extraction system based on machine learning, which specifically includes:

[0116] A data acquisition unit, used to acquire multiple medical data generated by the terminal monitoring device and analyze each medical data using a machine learning algorithm;

[0117] Mapping establishment unit, used to establish nonlinear mapping functions And use the nonlinear mapping function The n specified types of medical data x in the original spatial dataset X i Map it into the high-dimensional space H to obtain the high-dimensional medical data corresponding to each medical data It is also used to map nonlinear functions Construct a kernel matrix K in high-dimensional space;

[0118] A mean processing unit, used for performing centralization processing on the kernel matrix K to obtain a centralized kernel matrix K';

[0119] Matrix operation unit, used to obtain the inner product K'(x i ,x j ), and is also used to obtain the eigenvalue λ and eigenvector a of the centralized kernel matrix K' according to the characteristic equation λna=Ka;

[0120] The data screening unit is used to extract the medical data x corresponding to the specified type in the high-dimensional space H according to the eigenvalue λ and eigenvector a of the centralized kernel matrix K'. i The multiple core components are obtained through inverse operation, and the medical data x corresponding to the specified types of the multiple core components in the original spatial data set X are obtained. i and output it as standard monitoring data to facilitate medical diagnosis recommendations.

[0121] Each unit works under the logic of the machine learning algorithm and realizes the extraction of specific medical features based on the mechanism of kernel principal component analysis, especially the accurate extraction of nonlinear data.

[0122] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for extracting medical data features based on machine learning, characterized in that: Acquire multiple medical data generated by the terminal monitoring device, and use a machine learning algorithm to parse each medical data to obtain n target medical data, so as to use the n target medical data to form an original spatial data set, wherein each target medical data has a different data type, and any target medical data is low-dimensional data; A nonlinear mapping function is established, and the n target medical data in the original spatial data set are mapped into the high-dimensional space using the nonlinear mapping function to obtain the high-dimensional medical data corresponding to each target medical data; Using n high-dimensional medical data, a kernel matrix is ​​constructed, and the kernel matrix is ​​centralized to obtain a centralized kernel matrix; Calculate the eigenvalues ​​and eigenvectors of the centralized kernel matrix; According to the eigenvalues ​​and eigenvectors of the centralized kernel matrix, the kernel principal components corresponding to the target medical data of the original spatial data set are extracted in the high-dimensional space, and based on the extracted kernel principal components, at least one target medical data is determined from the original spatial data set and output as standard monitoring data, so as to perform medical data analysis using the standard monitoring data and obtain analysis results.

2. The method for extracting medical data features based on machine learning according to claim 1, characterized in that: Obtain multiple medical data generated by the terminal monitoring device, and use machine learning algorithms to parse each medical data to obtain n target medical data, including: Acquire historical medical data of the terminal monitoring device, and classify the historical medical data according to preset data types to obtain multiple classification sets, wherein each classification set corresponds to a data type; Dividing the historical medical data in each of the classification sets into an initial training data set and a test data set according to a preset ratio; Setting a label for each historical medical data in each initial training data set to obtain each training data set, wherein the label of any historical medical data is used to characterize the data type of the any historical medical data; Taking each historical medical data in each training data set as input and the data type of each historical medical data as output, a machine learning model is trained, and during the training process, the output of the machine learning model is used to adjust the parameters of the machine learning model, so as to obtain a pre-medical data classification model after the training is completed; Use each test data set to test the pre-medical data classification model, obtain the test results, and determine whether the test results meet the preset conditions; If not, the preliminary medical data classification model is optimized until the test results meet the preset conditions, and the final medical data classification model is obtained; The multiple medical data are input into a medical data classification model to obtain a data type of each medical data, and based on the data type of each medical data, n target medical data of different data types are screened out from the multiple medical data.

3. The method for extracting medical data features based on machine learning according to claim 1, characterized in that: Establish nonlinear mapping functions, including: The natural logarithm function is used as the nonlinear mapping function; Accordingly, a nonlinear mapping function is used to map the n target medical data in the original spatial data set to a high-dimensional space to obtain high-dimensional medical data corresponding to each target medical data, which includes: Each target medical data in the original space data set is mapped to the high-dimensional space through the natural logarithm function, where the high-dimensional medical data corresponding to the i-th target medical data in the original space data set is: In formula (1), Represents the i-th target medical data x i Corresponding to high-dimensional medical data, ln() represents the natural logarithm function, where i = 1, 2, ..., n.

4. The method for extracting medical data features based on machine learning according to claim 1, characterized in that: Using n high-dimensional medical data, a kernel matrix is ​​constructed and then the kernel matrix is ​​centralized to obtain a centralized kernel matrix, including: For any high-dimensional medical data, calculate the dot product between the any high-dimensional medical data and n high-dimensional medical data to obtain n dot products, and after all high-dimensional medical data are polled, obtain n×n dot products; Using n×n dot products, a kernel matrix is ​​constructed, where the size of the kernel matrix is ​​also n×n; The kernel matrix is ​​centralized column by column to obtain a centralized kernel matrix.

5. The method for extracting medical data features based on machine learning according to claim 1, characterized in that: Calculate the eigenvalues ​​and eigenvectors of the centralized kernel matrix, including: Calculate the covariance matrix of the centered kernel matrix; Obtain the eigenvalues ​​and eigenvectors of the covariance matrix, so as to use the eigenvalues ​​of the covariance matrix as the eigenvalues ​​of the centralized kernel matrix; According to the eigenvectors of the covariance matrix and the centralized kernel matrix, the eigenvectors corresponding to the eigenvalues ​​of each centralized kernel matrix are calculated.

6. The method for extracting medical data features based on machine learning according to claim 5, characterized in that: Calculate the covariance matrix of the centered kernel matrix, including: According to the following formula (2), the covariance matrix is ​​calculated Among them, C H The covariance matrix used to represent the centered kernel matrix K', represents the vector composed of the elements in the i-th row of the centralized kernel matrix, and T represents the transpose operation; Correspondingly, according to the eigenvectors of the covariance matrix and various high-dimensional medical data, the eigenvectors corresponding to the eigenvalues ​​of the centralized kernel matrix are calculated, including: According to the following formula (3), the eigenvector corresponding to each eigenvalue is calculated; ν represents the eigenvector of the covariance matrix, and a represents the eigenvector of the centered kernel matrix.

7. The method for extracting medical data features based on machine learning according to claim 1, characterized in that: According to the eigenvalues ​​and eigenvectors of the centralized kernel matrix, kernel principal components corresponding to the target medical data of the original spatial data set are extracted in the high-dimensional space, and based on the extracted kernel principal components, at least one target medical data is determined from the original spatial data set and output as standard monitoring data, including: The eigenvectors corresponding to the eigenvalues ​​of the centralized kernel matrix are used as kernel principal elements; Based on the variance cumulative contribution rate method, the eigenvectors corresponding to multiple eigenvalues ​​whose contribution rates are higher than the preset threshold are selected from the kernel principal components as the final kernel principal components; Based on the nonlinear mapping function, the final kernel principal component is inversely operated to obtain the target medical data corresponding to the final kernel principal component in the original spatial data set, so that the target medical data corresponding to the final kernel principal component in the original spatial data set can be output as standard monitoring data.

8. A medical data feature extraction system based on machine learning, characterized in that: include: A data acquisition unit, used to acquire multiple medical data generated by the terminal monitoring device, and analyze the medical data using a machine learning algorithm; A mapping establishment unit, used to establish a nonlinear mapping function, and use the nonlinear mapping function to map n target medical data in the original space data set to the high-dimensional space to obtain high-dimensional medical data corresponding to each target medical data, and also used to use the n high-dimensional medical data to construct a kernel matrix; A mean processing unit, used for performing centralization processing on the kernel matrix to obtain a centralized kernel matrix; Matrix operation unit, used to calculate the eigenvalues ​​and eigenvectors of the centralized kernel matrix; The data screening unit is used to extract the core principal components corresponding to each target medical data of the original spatial data set in the high-dimensional space according to the eigenvalues ​​and eigenvectors of the centralized kernel matrix, and based on the extracted core principal components, determine at least one target medical data from the original spatial data set and output it as standard monitoring data.

9. An electronic device, characterized in that: It includes a memory, a processor and a transceiver which are communicatively connected in sequence, wherein the memory is used to store a computer program, the transceiver is used to send and receive messages, and the processor is used to read the computer program to execute the medical data feature extraction method based on machine learning as described in any one of claims 1 to 7.

10. A computer program product comprising a computer program or instructions, characterized in that When the computer program or the instruction is executed by a computer, the method for extracting medical data features based on machine learning as described in any one of claims 1 to 7 is implemented.

Images