Multimodal hemodynamic analysis method and device applied to sepsis

Abstract

The application relates to the technical field of data processing, and discloses a multi-modal hemodynamic analysis method and device applied to sepsis, which comprises the following steps: monitoring multi-modal data of a plurality of sepsis patients; constructing an adjacency matrix of each view according to sepsis data in each view contained in the multi-modal data of each sepsis patient; performing embedding processing on the adjacency matrices of all views based on a dynamic fusion algorithm of a preset multi-view spectral clustering model to obtain a multi-view embedding matrix; and performing clustering analysis on all sepsis patients based on a preset clustering algorithm according to all multi-view spectral embedding matrices to obtain a clustering result, wherein all sepsis patients in each sepsis patient group contained in the clustering result correspond to a hemodynamic characteristic. It can be seen that the application can improve the classification accuracy of sepsis patients based on diversified monitoring data, and is beneficial to improving the analysis accuracy of the hemodynamic characteristics of each type of patient.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, and in particular to a multi-modal hemodynamic analysis method and device applied to sepsis. BACKGROUND

[0002] Sepsis can be caused by infection in any part of the body, such as the lungs, abdomen, urinary system, etc. Symptoms of sepsis can include fever, chills, rapid breathing, increased heart rate, decreased blood pressure, altered consciousness, etc., that is, the symptoms of sepsis are complex and varied.

[0003] However, in the management of sepsis in the prior art, hemodynamic analysis is usually simply performed by monitoring the function and state of the left ventricle of the patient through critical echocardiography (CCE), the monitored data is relatively single and incomplete, which can easily lead to low classification accuracy of sepsis patients, low analysis accuracy of hemodynamic characteristics of each type of patient, and thus affect the subsequent treatment effect. Therefore, it is particularly important to propose a new technical solution for hemodynamic data analysis of sepsis. SUMMARY

[0004] The present application provides a multi-modal hemodynamic analysis method and device applied to sepsis, which can improve the classification accuracy of sepsis patients based on diversified monitoring data, and is beneficial to improve the analysis accuracy of hemodynamic characteristics of each type of patient.

[0005] To solve the above technical problems, the present application discloses a multi-modal hemodynamic analysis method applied to sepsis in the first aspect, the method comprising:

[0006] Monitoring multi-modal data of each of a plurality of sepsis patients, the multi-modal data comprising two or more of medical index data, diagnostic text data and image data;

[0007] According to the sepsis data in each of the views contained in the multi-modal data of each of the sepsis patients, an adjacency matrix of each of the views is constructed;

[0008] Based on a dynamic fusion algorithm of a preset multi-view spectral clustering model, the adjacency matrix of all the views is embedded to obtain a multi-view embedding matrix;

[0009] Based on a preset clustering algorithm, all the sepsis patients are analyzed according to all the multi-view spectral embedding matrices to obtain a clustering result, the clustering result comprising a plurality of sepsis patient groups, all the sepsis patients in each of the sepsis patient groups corresponding to a hemodynamic characteristic; the hemodynamic characteristic being used as a basis for generating a treatment strategy for all the sepsis patients in the corresponding sepsis patient group.

[0010] As an optional implementation, in the first aspect of the present application, the sepsis data in each of the views included in the multi-modal data of each of the sepsis patients is used to construct an adjacency matrix of each of the views, including:

[0011] According to the preset deep neural network model, the sepsis data in each of the views included in the multi-modal data of each of the sepsis patients is used for feature extraction, and feature data of each of the sepsis patients in each of the views is obtained;

[0012] Based on the preset nearest neighbor analysis algorithm, the feature data of each of the sepsis patients in each of the views is used to construct a sample set corresponding to each of the sepsis patients, each sample in the sample set corresponds to a sample type, and the sample type includes a positive sample type or a negative sample type;

[0013] According to the feature data of each of the sepsis patients in each of the views, the mapping function included in the target network is optimized, and the target network is a twin network with the same network branch;

[0014] According to the optimized mapping function and the feature data of each of the sepsis patients in each of the views, the adjacency matrix of each of the views is calculated.

[0015] As an optional implementation, in the first aspect of the present application, the sample set corresponding to each of the sepsis patients is constructed based on the feature data of each of the sepsis patients in each of the views according to the preset nearest neighbor analysis algorithm, including:

[0016] Based on the preset distance metric algorithm, the similarity between each of the sepsis patients is calculated according to the feature data of each of the sepsis patients in each of the views;

[0017] Based on the preset nearest neighbor analysis algorithm, the neighbor set corresponding to each of the sepsis patients is constructed according to the similarity between each of the sepsis patients;

[0018] According to the determined sample type of each of the sepsis patients, the sample set corresponding to each of the sepsis patients is determined.

[0019] As an optional implementation, in the first aspect of the present application, the sample set corresponding to each of the sepsis patients is determined according to the determined sample type of each of the sepsis patients, including:

[0020] According to the determined sample type of each sepsis patient, determining samples of the same sample type as the sepsis patient from the neighbor set corresponding to each sepsis patient as the positive sample set corresponding to the sepsis patient;

[0021] According to the sample type of each sepsis patient, determining samples that are different from the sample type of the sepsis patient from the neighbor set corresponding to each sepsis patient as a negative sample set corresponding to the sepsis patient;

[0022] A positive sample set corresponding to each sepsis patient and a negative sample set corresponding to the sepsis patient are determined as the sample set corresponding to the sepsis patient.

[0023] As an optional embodiment, in the first aspect of the present invention, optimizing the mapping function included in the pre-constructed target network based on the feature data of the sample set corresponding to each sepsis patient in each view includes:

[0024] For each of the sepsis patients, the sample set corresponding to the sepsis patient is divided according to the sample types of all samples contained in the sample set to obtain multiple sample groups, each of the sample groups having a corresponding discriminant label, the discriminant label including a label indicating that the corresponding sample groups are mutually positive samples or a label indicating that the corresponding sample groups are mutually negative samples;

[0025] Inputting feature data of each sample group in each view into a pre-built target network, and outputting a feature vector of the sample group through all network branches included in the target network;

[0026] Calculating the contrast loss of the target network according to the feature vectors of all the sample groups, and optimizing the contrast loss of the target network to obtain an optimized contrast loss;

[0027] The mapping function included in the target network is optimized according to the optimized contrast loss.

[0028] As an optional embodiment, in the first aspect of the present invention, the multi-view spectral clustering model includes multiple deep neural network layers, and each of the deep neural network layers includes one of a local learning layer, a global learning layer, and an orthogonal constraint layer;

[0029] Furthermore, the dynamic fusion algorithm based on the preset multi-view spectral clustering model embeds the adjacency matrices of all the views to obtain a multi-view embedding matrix, including:

[0030] based on the local learning layer, processing the sepsis data of each of the sepsis patients in each of the views and the adjacency matrix of each of the views to obtain single-view feature information of each of the views;

[0031] based on the global learning layer, processing the single-view feature information of all the views to obtain a spectral embedding matrix of each of the views;

[0032] based on the orthogonal constraint layer, orthogonalizing the spectral embedding matrix of all the views to obtain a multi-view spectral embedding matrix.

[0033] As an optional implementation, in the first aspect of the present application, the method further comprises:

[0034] In the process of learning by using the local learning layer and the global learning layer, according to the obtained learning data of the local learning layer, a local learning loss of the local learning layer is calculated, and according to the obtained learning data of the global learning layer, a global learning loss of the global learning layer is calculated;

[0035] According to the local learning loss and the global learning loss, a total loss of the multi-view spectral clustering model is calculated;

[0036] According to the total loss, a performance indicator of the multi-view spectral clustering model is evaluated, and it is determined whether the performance indicator reaches a preset standard performance indicator;

[0037] When it is determined that the performance indicator does not reach the standard performance indicator, the total loss is optimized to obtain an optimized total loss, until the performance indicator of the multi-view spectral clustering model reaches the preset standard performance indicator;

[0038] When it is determined that the performance indicator reaches the standard performance indicator, it is determined that the learning of the local learning layer and the global learning layer is completed.

[0039] The second aspect of the present application discloses a multi-modal hemodynamic analysis device applied to sepsis, the device comprising:

[0040] a monitoring module for monitoring multi-modal data of each of a plurality of sepsis patients, the multi-modal data comprising two or more of medical indicator data, diagnostic text data and image data;

[0041] a construction module for constructing an adjacency matrix of each of the views according to the sepsis data in each of the views contained in the multi-modal data of each of the sepsis patients;

[0042] a processing module, configured to perform embedding processing on the adjacency matrices of all the views based on a preset dynamic fusion algorithm of a multi-view spectral clustering model, to obtain multi-view embedding matrices;

[0043] a clustering module, configured to perform clustering analysis on all the sepsis patients based on the preset clustering algorithm and the multi-view spectral embedding matrices, to obtain clustering results, the clustering results including a plurality of sepsis patient groups, and all the sepsis patients in each of the sepsis patient groups corresponding to a hemodynamic characteristic; and the hemodynamic characteristic serving as a basis for generating a treatment strategy for all the sepsis patients in the corresponding sepsis patient group.

[0044] As an optional implementation form, in the second aspect, the constructing module constructs the adjacency matrix of each of the views in the manner that:

[0045] extracts features from the sepsis data in each of the views included in the multi-modal data of each of the sepsis patients based on a preset deep neural network model, to obtain feature data of each of the sepsis patients in each of the views;

[0046] constructs a sample set corresponding to each of the sepsis patients based on a preset nearest neighbor analysis algorithm and the feature data of each of the sepsis patients in each of the views, each sample in the sample set corresponding to a sample type, the sample type including a positive sample type or a negative sample type;

[0047] optimizes a mapping function included in a target network based on the feature data of the sample set corresponding to each of the sepsis patients in each of the views, the target network being a twin network with the same network branch;

[0048] calculates the adjacency matrix of each of the views based on the optimized mapping function and the feature data of the sample set corresponding to each of the sepsis patients in each of the views.

[0049] As an optional implementation form, in the second aspect, the constructing module constructs the sample set corresponding to each of the sepsis patients in the manner that:

[0050] calculates the similarity between each of the sepsis patients based on a preset distance measurement algorithm and the feature data of each of the sepsis patients in each of the views;

[0051] constructing, according to the similarity between all the sepsis patients, a neighbor set corresponding to each of the sepsis patients based on a preset nearest neighbor analysis algorithm;

[0052] determining, according to the sample type of each of the sepsis patients, a sample set corresponding to each of the sepsis patients.

[0053] As an optional implementation, in the second aspect of the present application, the manner in which the constructing module determines the sample set corresponding to each of the sepsis patients according to the sample type of each of the sepsis patients specifically includes:

[0054] determining, according to the sample type of each of the sepsis patients, a sample in the neighbor set corresponding to the sepsis patient that is of the same sample type as the sample type of the sepsis patient as a positive sample set corresponding to the sepsis patient;

[0055] determining, according to the sample type of each of the sepsis patients, a sample in the neighbor set corresponding to the sepsis patient that is not of the same sample type as the sample type of the sepsis patient as a negative sample set corresponding to the sepsis patient;

[0056] determining the positive sample set corresponding to each of the sepsis patients and the negative sample set corresponding to the sepsis patient as the sample set corresponding to the sepsis patient.

[0057] As an optional implementation, in the second aspect of the present application, the manner in which the constructing module optimizes the mapping function contained in the target network according to the feature data of each of the sample sets in each of the views specifically includes:

[0058] dividing, for each of the sepsis patients, the sample set corresponding to the sepsis patient according to the sample type of all the samples contained in the sample set, to obtain a plurality of sample groups, each of the sample groups having a corresponding discrimination label, the discrimination label including a label indicating that the corresponding sample groups are positive samples or a label indicating that the corresponding sample groups are negative samples;

[0059] inputting the feature data of each of the sample groups in each of the views into the target network, and outputting a feature vector of the sample group through all the network branches contained in the target network;

[0060] calculating a contrastive loss of the target network according to the feature vectors of all the sample groups, and optimizing the contrastive loss of the target network to obtain an optimized contrastive loss;

[0061] optimizing the mapping function contained in the target network according to the optimized contrastive loss.

[0062] As an optional implementation, in the second aspect of the present application, the multi-view spectral clustering model comprises a plurality of deep neural network layers, and each of the deep neural network layers comprises one of a local learning layer, a global learning layer and an orthogonal constraint layer;

[0063] Furthermore, the processing module performs embedding processing on the adjacency matrices of all the views based on a preset dynamic fusion algorithm of the multi-view spectral clustering model to obtain a multi-view embedding matrix, and the manner specifically comprises:

[0064] Based on the local learning layer, the sepsis data of each of the sepsis patients in each of the views and the adjacency matrix of each of the views are processed to obtain single-view feature information of each of the views;

[0065] Based on the global learning layer, the single-view feature information of all the views is processed to obtain a spectral embedding matrix of each of the views;

[0066] Based on the orthogonal constraint layer, the spectral embedding matrices of all the views are processed to obtain a multi-view spectral embedding matrix.

[0067] As an optional implementation, in the second aspect of the present application, the device further comprises:

[0068] The computing module is configured to, in the process of learning by using the local learning layer and the global learning layer, calculate a local learning loss of the local learning layer according to the learning data of the local learning layer, and calculate a global learning loss of the global learning layer according to the learning data of the global learning layer;

[0069] The computing module is further configured to calculate a total loss of the multi-view spectral clustering model according to the local learning loss and the global learning loss;

[0070] The evaluation module is configured to evaluate a performance index of the multi-view spectral clustering model according to the total loss;

[0071] The judging module is configured to judge whether the performance index reaches a preset standard performance index;

[0072] The optimization module is configured to, when the judging module judges that the performance index does not reach the standard performance index, optimize the total loss to obtain an optimized total loss until the performance index of the multi-view spectral clustering model reaches the preset standard performance index;

[0073] The determining module is configured to, when the judging module judges that the performance index reaches the standard performance index, determine that the learning of the local learning layer and the global learning layer is completed.

[0074] The third aspect of the present application discloses another multi-modal hemodynamic analysis device for sepsis, the device comprising:

[0075] a memory storing executable program codes;

[0076] a processor coupled with the memory;

[0077] The processor invokes the executable program codes stored in the memory to execute the multi-modal hemodynamic analysis method for sepsis disclosed in the first aspect of the present application.

[0078] The fourth aspect of the present application discloses a computer storage medium storing computer instructions, which are invoked to execute the multi-modal hemodynamic analysis method for sepsis disclosed in the first aspect of the present application.

[0079] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0080] In the embodiment of the present application, the multi-modal data of each sepsis patient in a plurality of sepsis patients is monitored, and the multi-modal data includes two or more of medical index data, diagnostic text data and image data; an adjacency matrix of each view is constructed according to the sepsis data in each view of a plurality of views included in the multi-modal data of each sepsis patient; embedding processing is performed on the adjacency matrices of all views based on a dynamic fusion algorithm of a preset multi-view spectral clustering model to obtain a multi-view embedding matrix; clustering analysis is performed on all sepsis patients based on a preset clustering algorithm according to all multi-view spectral embedding matrices to obtain a clustering result, and the clustering result includes a plurality of sepsis patient groups, and all sepsis patients in each sepsis patient group correspond to a hemodynamic characteristic; and the hemodynamic characteristic is used as a basis for generating a treatment strategy for all sepsis patients in the corresponding sepsis patient group. It can be seen that the present application can monitor the multi-modal data of each sepsis patient in a plurality of sepsis patients, such as two or more of medical index data, diagnostic text data and image data, and construct an adjacency matrix of each view according to the sepsis data in each view of a plurality of views included in the multi-modal data of each sepsis patient, which can improve the construction accuracy and efficiency of the adjacency matrix of each view. Then, embedding processing is performed on the adjacency matrices of all views based on a dynamic fusion algorithm of a preset multi-view spectral clustering model to obtain a multi-view embedding matrix, which can improve the accuracy and reliability of the multi-view embedding matrix obtained by processing. Then, clustering analysis is performed on all sepsis patients based on a preset clustering algorithm according to all multi-view spectral embedding matrices to obtain a clustering result, so that all sepsis patients in each sepsis patient group included in the clustering result correspond to a hemodynamic characteristic, which can improve the classification accuracy of sepsis patients based on multi-modal data, is conducive to improving the analysis accuracy of the hemodynamic characteristics of each type of sepsis patient, thereby being conducive to improving the generation accuracy of the treatment strategy of each type of patient based on the accurately analyzed hemodynamic characteristics, and further being conducive to achieving good treatment effect. BRIEF DESCRIPTION OF DRAWINGS

[0081] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0082] Figure 1 is a flowchart of a multi-modal hemodynamic analysis method for sepsis disclosed by the embodiment of the present application;

[0083] Figure 2is another flowchart of a method for multi-modal hemodynamic analysis applied to sepsis disclosed by the embodiment of the present application;

[0084] Figure 3 is a structural diagram of a device for multi-modal hemodynamic analysis applied to sepsis disclosed by the embodiment of the present application;

[0085] Figure 4 is another structural diagram of a device for multi-modal hemodynamic analysis applied to sepsis disclosed by the embodiment of the present application;

[0086] Figure 5 is still another structural diagram of a device for multi-modal hemodynamic analysis applied to sepsis disclosed by the embodiment of the present application. DETAILED DESCRIPTION

[0087] In order to enable persons skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by persons skilled in the art without creative labor fall within the scope of protection of the present application.

[0088] The terms "first", "second", and the like in the specification and claims of the present application and the above-mentioned drawings are used to distinguish different objects, and are not used to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product, or end including a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to the process, method, product, or end.

[0089] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The appearance of the phrase in various places in the specification does not necessarily all refer to the same embodiment, nor does it necessarily refer to a separate or alternative embodiment. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0090] The application discloses a multi-modal hemodynamic analysis method and device applied to sepsis, can monitor two or more than two multi-modal data of each sepsis patient in multiple sepsis patients, such as medical index data, diagnostic text data and image data, and constructs the adjacency matrix of each view according to the sepsis data in each view in multiple views contained in the multi-modal data of each sepsis patient, can improve the construction accuracy and efficiency of the adjacency matrix of each view, then based on the dynamic fusion algorithm of the preset multi-view spectral clustering model, the adjacency matrix of all views is embedded, and a multi-view embedding matrix is obtained, which can improve the accuracy and reliability of the obtained multi-view embedding matrix; and based on the preset clustering algorithm, all sepsis patients are clustered and analyzed according to all multi-view spectral embedding matrices, and a clustering result is obtained, so that all sepsis patients in each sepsis patient group contained in the clustering result correspond to a hemodynamic characteristic, the classification accuracy of sepsis patients can be improved based on multi-modal data, which is beneficial to improve the analysis accuracy of the hemodynamic characteristics of each type of sepsis patient, thereby being beneficial to improve the generation accuracy of the treatment strategy of each type of patient based on the accurately analyzed hemodynamic characteristics, and further being beneficial to achieve good treatment effect. The following will be described in detail.

[0091] Embodiment one

[0092] Please refer to Figure 1 , Figure 1 is a flowchart of a multi-modal hemodynamic analysis method applied to sepsis disclosed by the embodiment of the application. Wherein, Figure 1 The multi-modal hemodynamic analysis method applied to sepsis described can be applied to a multi-modal hemodynamic analysis device applied to sepsis, wherein the device can include an analysis device or an analysis server, wherein the analysis server can include a cloud server or a local server, and the embodiment of the application is not limited. As shown in Figure 1 The multi-modal hemodynamic analysis method applied to sepsis can include the following operations:

[0093] 101, monitor the multi-modal data of each sepsis patient in multiple sepsis patients.

[0094] In the embodiment of the application, optionally, the multi-modal data can include two or more than two of medical index data, diagnostic text data and image data.

[0095] The medical index data refers to numerical data obtained by detecting physiological, biochemical, immune and other aspects of the sepsis patient, such as heart rate, blood pressure, white blood cell count, blood lactic acid level and the like. These indexes can reflect the hemodynamic state and organ function of the sepsis patient, and are important basis for evaluating the severity of the disease and the treatment effect. The diagnostic text data refers to the textual data recorded by the doctor during the clinical diagnosis and treatment plan formulation of the sepsis patient, such as medical records, prescriptions, examination reports and the like. These texts can reflect the etiology, symptoms, complications, drug reactions and other information of the sepsis patient, and are important basis for understanding the individual differences of the patient and formulating individualized treatment plan. The image data refers to image or video data obtained by imaging examination of the sepsis patient, such as X-ray film, ultrasound, CT scan and the like. These data can reflect the anatomical structure and functional changes of the sepsis patient, and are important basis for observing the organ damage degree and treatment effect of the patient. In this way, through the multi-modal data, more comprehensive and comprehensive information can be provided, which reflects the condition and treatment effect of the sepsis patient from different angles and levels, which helps to improve the accuracy and robustness of clustering; and the shortcomings and deficiencies of single modal data can be made up, for example, the medical index may have measurement error or lack of sensitivity, the diagnostic text may have subjective bias or unclear expression, the image data may have noise or low resolution. Through the fusion of multi-modal data, the complementary and synergistic effect between different modalities can be utilized to improve the quality and reliability of the data.

[0096] 102. Construct an adjacency matrix of each view according to the sepsis data in each view of the multi-view contained in the multi-modal data of each sepsis patient.

[0097] 103. Perform embedding processing on the adjacency matrices of all views based on a preset dynamic fusion algorithm of multi-view spectral clustering model, to obtain a multi-view embedding matrix.

[0098] In the embodiment of the application, the multi-view spectral clustering model comprises a plurality of deep neural network layers, and each deep neural network layer comprises one of a local learning layer, a global learning layer and an orthogonal constraint layer.

[0099] Specifically, first, the data of each view is processed by the local learning layer to learn the unique embedding of each view and retain the local invariance of each view; second, the unique embedding of all views is processed by the global learning layer to learn the shared embedding between the multi-views and mine the global consistency and complementarity between the multi-views; and finally, the output of the global learning layer is processed by the orthogonal constraint layer to ensure the orthogonality of the multi-view spectral embedding matrix, to obtain a consistent spectral embedding. The mathematical expression can be Y N×c = [Y 1 ,Y 2 ,...,YV ]; wherein, is the data of the vth view in the Nth sample, is a mapping function of the multi-view spectral clustering model, Y N×c is a consistency spectral embedding matrix obtained after mapping by the multi-view spectral clustering model, each row of which represents a data sample (e.g., the Nth row represents the Nth data sample), and each column of which represents a class cluster (e.g., the cth column represents the cth class cluster, c = 1, 2,..., v). This matrix can be used for clustering data, that is, each sample will be assigned to a class cluster, which is the column where the maximum value in the matrix is located. The generalization ability and adaptability of the multi-view spectral clustering model can be enhanced based on the adjacency matrix of multi-modal data. Through the fusion of multi-modal data, the multi-view spectral clustering model can be more easily adapted to different scenarios and environments, improving the portability and universality of the multi-view spectral clustering model.

[0100] 104. Based on the preset clustering algorithm, all sepsis patients are clustered and analyzed according to all multi-view spectral embedding matrices to obtain clustering results.

[0101] In the embodiments of the present application, the clustering results can include multiple sepsis patient groups, and all sepsis patients in each sepsis patient group correspond to a hemodynamic characteristic; the hemodynamic characteristic is used as a basis for generating a treatment strategy for all sepsis patients in the corresponding sepsis patient group.

[0102] It can be seen that the implementation Figure 1 The multi-modal hemodynamic analysis method for sepsis described above can monitor the multi-modal data of each sepsis patient in multiple sepsis patients, such as two or more of medical index data, diagnostic text data, and image data, and construct an adjacency matrix for each view according to the sepsis data in each view included in the multi-modal data of each sepsis patient, which can improve the construction accuracy and efficiency of the adjacency matrix of each view. Subsequently, based on the dynamic fusion algorithm of the preset multi-view spectral clustering model, the adjacency matrices of all views are embedded to obtain a multi-view embedding matrix, which can improve the accuracy and reliability of the multi-view embedding matrix obtained by processing. Then, based on the preset clustering algorithm, all sepsis patients are clustered and analyzed according to all multi-view spectral embedding matrices to obtain clustering results, so that all sepsis patients in each sepsis patient group included in the clustering results correspond to a hemodynamic characteristic. The classification accuracy of sepsis patients can be improved based on multi-modal data, which is beneficial to improving the analysis accuracy of the hemodynamic characteristics of each type of sepsis patient, thereby improving the generation accuracy of the treatment strategy for each type of patient based on the accurately analyzed hemodynamic characteristics, and further achieving good treatment effect.

[0103] In an optional embodiment, the dynamic fusion algorithm based on the preset multi-view spectral clustering model of step 103, the embedding processing is performed on the adjacency matrix of all views to obtain a multi-view embedding matrix, which can include:

[0104] Based on the local learning layer, the sepsis data of each sepsis patient in each view and the adjacency matrix of each view are processed to obtain single-view feature information of each view;

[0105] Based on the global learning layer, the single-view feature information of all views is processed to obtain a spectral embedding matrix of each view;

[0106] Based on the orthogonal constraint layer, the spectral embedding matrix of all views is orthogonalized to obtain a multi-view spectral embedding matrix.

[0107] In the embodiment of the application, the local learning layer can adapt to different forms of data and learn the unique embedding of each view to extract single-view unique features. Specifically, the local learning layer uses a fully connected layer and a ReLU function to perform nonlinear transformation on the original data of each view to obtain new single-view feature representation.

[0108] In the global learning layer, the parameter sharing and feature transfer method can be used to learn the consistency features and complementary features of multiple views to obtain a consistent spectral embedding that integrates multi-view information.

[0109] Specifically, the parameter sharing method includes cascading the output of the local learning layer of all views, and then using a fully connected layer for learning to embed the features of multiple views. The feature transfer method includes transferring different view information through nonlinear mapping and fusing it with the local information of the current view to achieve feature transfer between different views and form a new view.

[0110] In the orthogonal constraint layer, Cholesky decomposition can be used to orthogonalize the spectral embedding of each view, so that the product of the spectral embedding matrix of each view and its transpose is equal to the unit matrix. The feature embedding output by the model satisfies the orthogonality condition, thereby avoiding invalid solutions.

[0111] It can be seen that the optional embodiment can process the sepsis data of each view and the adjacency matrix of each view of each sepsis patient based on the local learning layer to obtain the single-view feature information of each view, can improve the extraction accuracy and efficiency of the single-view feature of each view, and process the single-view feature information of all views based on the global learning layer to obtain the spectral embedding matrix of each view, can improve the accuracy and reliability of the spectral embedding matrix of each view, and then orthogonalize the spectral embedding matrix of all views based on the orthogonal constraint layer to obtain the multi-view spectral embedding matrix, can improve the accuracy and reliability of the multi-view spectral embedding matrix.

[0112] In the optional embodiment, as an optional implementation, the method can further include:

[0113] In the learning process using the local learning layer and the global learning layer, the local learning loss of the local learning layer is calculated according to the obtained learning data of the local learning layer, and the global learning loss of the global learning layer is calculated according to the obtained learning data of the global learning layer;

[0114] The overall loss of the multi-view spectral clustering model is calculated according to the local learning loss and the global learning loss;

[0115] The performance index of the multi-view spectral clustering model is evaluated according to the overall loss, and it is determined whether the performance index reaches a preset standard performance index;

[0116] When it is determined that the performance index does not reach the standard performance index, the overall loss is optimized to obtain an optimized overall loss until the performance index of the multi-view spectral clustering model reaches the preset standard performance index.

[0117] When it is determined that the performance index reaches the standard performance index, it is determined that the learning of the local learning layer and the global learning layer is completed.

[0118] In the embodiment of the application, the weight of the loss in each view can be adjusted in the local learning layer, and the weight of the loss between multiple views can be adjusted in the global learning layer, so as to optimize the overall loss.

[0119] It can be seen that in the process of learning by using the local learning layer and the global learning layer, the optional embodiment can calculate the local learning loss of the local learning layer according to the obtained learning data of the local learning layer, calculate the global learning loss of the global learning layer according to the obtained learning data of the global learning layer, and calculate the overall loss of the multi-view spectral clustering model according to the local learning loss and the global learning loss, thereby improving the calculation accuracy and reliability of the overall loss of the model. Then, the performance index of the multi-view spectral clustering model is evaluated according to the overall loss, the evaluation accuracy and efficiency of the performance index of the model are improved, it is judged whether the performance index reaches the preset standard performance index, when it is judged that the performance index does not reach the standard performance index, the overall loss is optimized to obtain the optimized overall loss, until the performance index of the multi-view spectral clustering model reaches the preset standard performance index, the optimization timeliness and accuracy of the overall loss can be improved, so that the accurate standard of the performance index of the model can be realized, and when it is judged that the performance index reaches the standard performance index, it is determined that the learning of the local learning layer and the global learning layer is completed. The training accuracy of the local learning layer and the global learning layer included in the model can be improved through the performance index evaluated based on the overall loss.

[0120] Embodiment two

[0121] Please refer to Figure 2 , Figure 2 is a flowchart of a multi-modal hemodynamic analysis method for sepsis disclosed by the embodiment of the application. Wherein, Figure 2 The multi-modal hemodynamic analysis method for sepsis described can be applied to a multi-modal hemodynamic analysis device for sepsis, wherein the device can include an analysis equipment or an analysis server, wherein the analysis server can include a cloud server or a local server, and the embodiments of the application are not limited. As shown in Figure 2 The multi-modal hemodynamic analysis method for sepsis can include the following operations:

[0122] 201, monitoring multi-modal data of each sepsis patient in a plurality of sepsis patients.

[0123] 202, according to a preset deep neural network model, performing feature extraction on sepsis data in each view in a plurality of views included in the multi-modal data of each sepsis patient, to obtain feature data of each sepsis patient in each view.

[0124] In the embodiment of the present application, specifically, the multi-modal data of each sepsis patient contains sepsis data in each view in multiple views, and the sepsis data is subjected to feature extraction according to a preset deep neural network model corresponding to the data type of the sepsis data, to obtain feature data in the view. The data type of each sepsis data can include one of a medical index type, a diagnosis text data type and an image data type.

[0125] 203. Based on a preset nearest neighbor analysis algorithm, a sample set corresponding to each sepsis patient is constructed according to the feature data of all sepsis patients in each view.

[0126] In the embodiment of the present application, each sample in the sample set corresponds to a sample type, and the sample type includes a positive sample type or a negative sample type. The sample set corresponding to each sepsis patient can include a positive sample set corresponding to each sepsis patient and a negative sample set corresponding to each sepsis patient. All samples in the positive sample set are positive samples, and all samples in the negative sample set are negative samples.

[0127] 204. The mapping function contained in the target network is optimized according to the feature data of each sepsis patient in each view.

[0128] In the embodiment of the present application, the target network is a twin network with the same network branch. Specifically, the mapping function in the target network is optimized by neighbor learning through a contrastive loss.

[0129] 205. The adjacency matrix of each view is calculated according to the optimized mapping function and the feature data of each sepsis patient in each view.

[0130] In the embodiment of the present application, specifically, a Gaussian kernel function is used to calculate the final adjacency matrix, wherein δ is a width parameter in the Gaussian kernel function, and δ is usually greater than 0. The Gaussian kernel function is used to map data to a high-dimensional feature space, and then calculate the distance between samples. The calculation formula of the adjacency matrix is as follows:

[0131] Neighbors

[0132] wherein, represents the feature data of the i-th sepsis patient in the v-th view, represents the mapping function in the twin network, which is used to embed sample data into a latent space; the calculated adjacency matrix can be represented as [S 1 ,S 2 ,...,S v], for spectrum embedding learning based on knowledge transfer.

[0133] 206, based on the dynamic fusion algorithm of the preset multi-view spectral clustering model, the adjacency matrix of all views is embedded to obtain a multi-view embedding matrix.

[0134] 207, based on the preset clustering algorithm, all sepsis patients are clustered according to all multi-view spectral embedding matrices to obtain a clustering result.

[0135] In the embodiment of the application, for other descriptions of steps 201, 206 and 207, please refer to the detailed description of steps 101, 103 and 104 in embodiment one, and the embodiment of the application will not be repeated here.

[0136] It can be seen that the implementation Figure 2The multi-modal hemodynamic analysis method applied to sepsis can monitor two or more kinds of multi-modal data of each sepsis patient in a plurality of sepsis patients, such as medical index data, diagnostic text data and image data, and construct an adjacency matrix of each view according to the sepsis data in each view included in the multi-modal data of each sepsis patient, which can improve the construction accuracy and efficiency of the adjacency matrix of each view. Subsequently, based on the dynamic fusion algorithm of the preset multi-view spectral clustering model, the adjacency matrices of all views are embedded to obtain a multi-view embedding matrix, which can improve the accuracy and reliability of the obtained multi-view embedding matrix. Then, based on the preset clustering algorithm, the clustering analysis is performed on all sepsis patients according to all multi-view spectral embedding matrices to obtain a clustering result, so that all sepsis patients in each sepsis patient group included in the clustering result correspond to a hemodynamic characteristic. The classification accuracy of sepsis patients can be improved based on multi-modal data, which is beneficial to improve the analysis accuracy of the hemodynamic characteristics of each type of sepsis patient, thereby improving the generation accuracy of the treatment strategy of each type of patient based on the accurately analyzed hemodynamic characteristics, and further improving the good treatment effect. In addition, the feature data of each sepsis patient in each view can be obtained by extracting the sepsis data in each view included in the multi-modal data of each sepsis patient according to the preset deep neural network model, which improves the extraction accuracy and efficiency of the feature data of each view. Based on the preset nearest neighbor analysis algorithm, a sample set corresponding to each sepsis patient is constructed according to the feature data of each sepsis patient in each view, which improves the construction accuracy and efficiency of the sample set corresponding to each sepsis patient. The mapping function included in the target network is optimized according to the feature data of each sepsis patient in each view in the sample set corresponding to each sepsis patient, which can accurately optimize the mapping function in the target network by taking the feature data as the input of the target network. Then, the adjacency matrix of each view is calculated according to the optimized mapping function and the feature data of each sepsis patient in each view in the sample set corresponding to each sepsis patient, which improves the calculation accuracy and reliability of the adjacency matrix of each view.

[0137] In an optional embodiment, the step 203 of constructing a sample set corresponding to each sepsis patient according to the feature data of all sepsis patients in each view based on the preset nearest neighbor analysis algorithm can include:

[0138] calculating the similarity between each sepsis patient according to the feature data of all sepsis patients in each view based on the preset distance measurement algorithm;

[0139] construct a neighbor set corresponding to each sepsis patient according to the similarity between all sepsis patients based on a preset nearest neighbor analysis algorithm;

[0140] According to the sample type of each sepsis patient determined, a sample set corresponding to each sepsis patient is determined.

[0141] In the embodiment of the present application, the distance measurement algorithm can include the Euclidean distance algorithm, or can include the cosine similarity algorithm, or can include any other algorithm that can calculate the distance / similarity between each sepsis patient. Through the nearest neighbor analysis algorithm, the k nearest neighbors of each sepsis patient can be found as the neighbor set corresponding to the sepsis patient, which is not limited in the embodiment of the present application.

[0142] It can be seen that the optional embodiment can calculate the similarity between each sepsis patient based on the feature data of all sepsis patients in each view according to the preset distance measurement algorithm, improve the calculation accuracy of the similarity between each sepsis patient, then construct a neighbor set corresponding to each sepsis patient according to the similarity between all sepsis patients based on a preset nearest neighbor analysis algorithm, which can improve the construction accuracy and reliability of the neighbor set corresponding to each sepsis patient, and then determine the sample set corresponding to each sepsis patient according to the sample type of each sepsis patient determined, which improves the determination accuracy of the sample set used as the input data sample of the twin network, thereby facilitating the optimization accuracy of the mapping function of the twin network subsequently.

[0143] In the optional embodiment, as an optional implementation, according to the sample type of each sepsis patient determined, the sample set corresponding to each sepsis patient can include:

[0144] According to the sample type of each sepsis patient determined, the sample with the same sample type as the sepsis patient is determined from the neighbor set corresponding to the sepsis patient as the positive sample set corresponding to the sepsis patient;

[0145] According to the sample type of each sepsis patient, the sample with a sample type different from that of the sepsis patient is determined from the neighbor set corresponding to the sepsis patient as the negative sample set corresponding to the sepsis patient;

[0146] The positive sample set corresponding to each sepsis patient and the negative sample set corresponding to the sepsis patient are determined as the sample set corresponding to the sepsis patient.

[0147] In the embodiments of the present application, specifically, if most of the neighbors of a sample are positive samples, the sample is also determined as a positive sample; if most of the neighbors of a sample are negative samples, the sample is determined as a negative sample; for a sample of uncertain type, a probabilistic division strategy can be used, such as determining the probability of belonging to a positive / negative class according to the proportion of positive / negative neighbors.

[0148] It can be seen that the optional implementation can determine, according to the sample type of each sepsis patient, the samples of the same sample type as the sample type of the sepsis patient from the corresponding neighbor set of the sepsis patient as the corresponding positive sample set of the sepsis patient, thereby improving the determination accuracy of the corresponding positive sample set of each sepsis patient, and determine, according to the sample type of each sepsis patient, the samples of different sample types from the corresponding neighbor set of the sepsis patient as the corresponding negative sample set of the sepsis patient, thereby improving the determination accuracy of the corresponding negative sample set of each sepsis patient, and then determine the corresponding positive sample set of each sepsis patient and the corresponding negative sample set of the sepsis patient as the corresponding sample set of the sepsis patient, thereby improving the determination accuracy and reliability of the corresponding sample set of each sepsis patient.

[0149] In another optional embodiment, the step 204 of optimizing the mapping function contained in the target network according to the feature data of each sepsis patient in each view can include:

[0150] For each sepsis patient, the sample set is divided according to the sample types of all samples contained in the sample set of the sepsis patient, to obtain a plurality of sample groups;

[0151] The feature data of each sample group in each view is input into the pre-constructed target network, and the feature vector of the sample group is output through all network branches contained in the target network;

[0152] The contrast loss of the target network is calculated according to the feature vectors of all sample groups, and the contrast loss of the target network is optimized to obtain an optimized contrast loss;

[0153] The mapping function contained in the target network is optimized according to the optimized contrast loss.

[0154] In the embodiments of the present application, specifically, the contrast loss can be minimized to optimize the above contrast loss. Each sample group has a corresponding discrimination label, and the discrimination label includes a label that the corresponding sample groups are positive samples or a label that the corresponding sample groups are negative samples. The expression of the contrast loss can be:

[0155]

[0156] wherein P e {0, 1} is a discriminant label, that is, P = 1 represents and are positive samples, and P = 0 represents and are negative samples; and γ represents the distance from the boundary (usually defined as 1).

[0157] It can be seen that the optional embodiment can divide the sample set of each sepsis patient according to the sample types of all samples contained in the sample set corresponding to the sepsis patient, obtain a plurality of sample groups, improve the grouping accuracy of the sample set corresponding to each sepsis patient, input the feature data of each sample group in each view into the target network constructed in advance, and output the feature vector of the sample group through all network branches contained in the target network, thereby improving the output accuracy of the sample feature vector based on the network branch, then calculating the contrast loss of the target network according to the feature vectors of all sample groups, improving the calculation accuracy and reliability of the contrast loss of the target network, and optimizing the contrast loss of the target network to obtain an optimized contrast loss, improving the optimization accuracy and efficiency of the contrast loss, and optimizing the mapping function contained in the target network according to the optimized contrast loss, which is conducive to precise optimization of the mapping function by optimizing the contrast loss, thereby improving the adjacency matrix construction accuracy of each subsequent view based on the accurately optimized mapping function.

[0158] Embodiment three

[0159] Please refer to Figure 3 , Figure 3 is a structural schematic diagram of a multi-modal hemodynamic analysis device for sepsis disclosed by the embodiment of the present application. Wherein, Figure 3 The multi-modal hemodynamic analysis device for sepsis described can include an analysis device or an analysis server, wherein the analysis server can include a cloud server or a local server, and the embodiment of the present application does not limit it. As Figure 3 shown, the multi-modal hemodynamic analysis device for sepsis can include:

[0160] The monitoring module 301 is used for monitoring the multi-modal data of each sepsis patient in a plurality of sepsis patients, and the multi-modal data includes two or more of medical index data, diagnostic text data and image data.

[0161] The construction module 302 is used for constructing the adjacency matrix of each view according to the sepsis data in each view contained in the multi-modal data of each sepsis patient.

[0162] The processing module 303 is configured to embed the adjacency matrices of all views based on a dynamic fusion algorithm of a preset multi-view spectral clustering model to obtain a multi-view embedding matrix;

[0163] The clustering module 304 is configured to perform cluster analysis on all sepsis patients based on a preset clustering algorithm and all multi-view spectrum embedding matrices to obtain clustering results. The clustering results include multiple sepsis patient groups, and all sepsis patients in each sepsis patient group correspond to a hemodynamic characteristic. The hemodynamic characteristic is used as a basis for generating a treatment strategy for all sepsis patients in the corresponding sepsis patient group.

[0164] It can be seen that implementation Figure 3 The multimodal hemodynamic analysis device for sepsis described herein can monitor the multimodal data of each septic patient in a plurality of septic patients, such as two or more of medical indicator data, diagnostic text data, and imaging data data, and construct an adjacency matrix of each view based on the sepsis data in each view in a plurality of views contained in the multimodal data of each septic patient, thereby improving the accuracy and efficiency of constructing the adjacency matrix of each view, and then embedding the adjacency matrices of all views based on the dynamic fusion algorithm of the preset multi-view spectral clustering model to obtain a multi-view embedding matrix, thereby improving the processing accuracy and efficiency. The accuracy and reliability of the multi-view embedding matrix are obtained; then, based on the preset clustering algorithm, all sepsis patients are clustered according to all multi-view spectrum embedding matrices to obtain clustering results, so that all sepsis patients in each sepsis patient group included in the clustering results correspond to a hemodynamic feature, which can improve the classification accuracy of sepsis patients based on multimodal data, and is conducive to improving the analysis accuracy of the hemodynamic features of each type of sepsis patient, thereby facilitating the improvement of the accuracy of generating treatment strategies for each type of patient based on the accurately analyzed hemodynamic features, and thus facilitating the achievement of good treatment effects.

[0165] In an optional embodiment, the construction module 302 constructs the adjacency matrix of each view according to the sepsis data in each of the multiple views included in the multimodal data of each sepsis patient, specifically including:

[0166] According to a preset deep neural network model, feature extraction is performed on the sepsis data in each of the multiple views contained in the multimodal data of each sepsis patient to obtain feature data of each sepsis patient in each view;

[0167] Based on the preset nearest neighbor analysis algorithm, a sample set corresponding to each sepsis patient is constructed according to the characteristic data of all sepsis patients in each view. Each sample in the sample set corresponds to a sample type, which includes a positive sample type or a negative sample type.

[0168] According to the feature data of the sample set corresponding to each sepsis patient in each view, the mapping function contained in the pre-constructed target network is optimized, the target network being a twin network with the same network branch;

[0169] According to the optimized mapping function and the feature data of the sample set corresponding to each sepsis patient in each view, the adjacency matrix of each view is calculated.

[0170] It can be seen that the optional embodiment can extract features of sepsis data in each view in multiple views contained in multi-modal data of each sepsis patient according to the preset deep neural network model, obtain feature data of each sepsis patient in each view, improve the extraction accuracy and efficiency of feature data of each view, and based on the preset nearest neighbor analysis algorithm, construct a sample set corresponding to each sepsis patient according to the feature data of all sepsis patients in each view, improve the construction accuracy and efficiency of the sample set corresponding to each sepsis patient, and according to the feature data of the sample set corresponding to each sepsis patient in each view, the mapping function contained in the pre-constructed target network is optimized, which can accurately optimize the mapping function in the target network by taking the output feature obtained by taking the feature data as the input of the target network, and then according to the optimized mapping function and the feature data of the sample set corresponding to each sepsis patient in each view, the adjacency matrix of each view is calculated, which improves the calculation accuracy and reliability of the adjacency matrix of each view.

[0171] In this optional embodiment, as an optional implementation, the construction module 302 can construct a sample set corresponding to each sepsis patient based on the preset nearest neighbor analysis algorithm according to the feature data of all sepsis patients in each view in the following manner:

[0172] According to the feature data of all sepsis patients in each view, the similarity between each sepsis patient is calculated based on a preset distance metric algorithm;

[0173] According to the similarity between all sepsis patients, a neighbor set corresponding to each sepsis patient is constructed based on a preset nearest neighbor analysis algorithm;

[0174] According to the determined sample type of each sepsis patient, the sample set corresponding to each sepsis patient is determined.

[0175] It can be seen that the optional implementation can calculate the similarity between each sepsis patient based on the preset distance metric algorithm according to the feature data of all sepsis patients in each view, improve the calculation accuracy of the similarity between each sepsis patient, then construct the neighbor set corresponding to each sepsis patient based on the preset nearest neighbor analysis algorithm according to the similarity between all sepsis patients, and improve the construction accuracy and reliability of the neighbor set corresponding to each sepsis patient. According to the determined sample type of each sepsis patient, the sample set corresponding to each sepsis patient is determined, the determination accuracy of the sample set used as the input data sample of the twin network is improved, thereby facilitating the optimization accuracy of the mapping function of the twin network.

[0176] In the optional implementation, optionally, the manner in which the construction module 302 determines the sample set corresponding to each sepsis patient according to the determined sample type of each sepsis patient specifically includes:

[0177] According to the sample type of each sepsis patient, samples with the same sample type as the sepsis patient are determined from the neighbor set corresponding to the sepsis patient as the positive sample set corresponding to the sepsis patient.

[0178] According to the sample type of each sepsis patient, samples with the same sample type as the sepsis patient are determined from the neighbor set corresponding to the sepsis patient as the positive sample set corresponding to the sepsis patient.

[0179] The positive sample set corresponding to each sepsis patient and the negative sample set corresponding to the sepsis patient are determined as the sample set corresponding to the sepsis patient.

[0180] It can be seen that the optional implementation can also determine samples with the same sample type as the sepsis patient from the neighbor set corresponding to the sepsis patient as the positive sample set corresponding to the sepsis patient according to the determined sample type of each sepsis patient, improve the determination accuracy of the positive sample set corresponding to each sepsis patient, and determine samples with the same sample type as the sepsis patient from the neighbor set corresponding to the sepsis patient as the negative sample set corresponding to the sepsis patient according to the sample type of each sepsis patient, improve the determination accuracy of the negative sample set corresponding to each sepsis patient, and then determine the positive sample set corresponding to each sepsis patient and the negative sample set corresponding to the sepsis patient as the sample set corresponding to the sepsis patient, improve the determination accuracy and reliability of the sample set corresponding to each sepsis patient.

[0181] In this optional embodiment, as another optional implementation, the construction module 302 optimizes the mapping function included in the pre-constructed target network according to the feature data of the sample set corresponding to each sepsis patient in each view, and the specific method may include:

[0182] For each sepsis patient, the sample set is divided according to the sample types of all samples contained in the sample set corresponding to the sepsis patient to obtain multiple sample groups. Each sample group has a corresponding discriminant label, which includes a label indicating that the corresponding sample groups are mutually positive samples or a label that the corresponding sample groups are mutually negative samples;

[0183] Input the feature data of each sample group in each view into the pre-built target network, and output the feature vector of the sample group through all network branches contained in the target network;

[0184] According to the feature vectors of all sample groups, the contrast loss of the target network is calculated, and the contrast loss of the target network is optimized to obtain the optimized contrast loss;

[0185] According to the optimized contrast loss, the mapping function contained in the target network is optimized.

[0186] It can be seen that this optional implementation method can divide the sample set for each sepsis patient according to the sample types of all samples contained in the sample set corresponding to the sepsis patient to obtain multiple sample groups, thereby improving the grouping accuracy of the sample set corresponding to each sepsis patient, and inputting the feature data of each sample group in each view into the pre-constructed target network, and outputting the feature vector of the sample group through all network branches contained in the target network, thereby improving the output accuracy of the sample feature vector based on the network branch, and then calculating the contrast loss of the target network based on the feature vectors of all sample groups, thereby improving the calculation accuracy and reliability of the contrast loss of the target network, and optimizing the contrast loss of the target network to obtain the optimized contrast loss, thereby improving the optimization accuracy and efficiency of the contrast loss, and then optimizing the mapping function contained in the target network based on the optimized contrast loss, which is conducive to achieving accurate optimization of the mapping function by optimizing the contrast loss, thereby facilitating improving the accuracy of constructing the adjacency matrix of each subsequent view based on the accurately optimized mapping function.

[0187] In another optional embodiment, the multi-view spectral clustering model includes multiple deep neural network layers, and each deep neural network layer includes one of a local learning layer, a global learning layer, and an orthogonal constraint layer. Furthermore, the processing module 303 embeds the adjacency matrices of all views based on a preset dynamic fusion algorithm of the multi-view spectral clustering model. The method for obtaining the multi-view embedding matrix may specifically include:

[0188] Based on the local learning layer, the sepsis data of each sepsis patient in each view and the adjacency matrix of each view are processed to obtain the single-view feature information of each view;

[0189] Based on the global learning layer, the single-view feature information of all views is processed to obtain the spectral embedding matrix of each view;

[0190] Based on the orthogonal constraint layer, the spectral embedding matrices of all views are orthogonalized to obtain the multi-view spectral embedding matrix.

[0191] It can be seen that this optional embodiment can process the sepsis data of each septic patient in each view and the adjacency matrix of each view based on the local learning layer to obtain the single-view feature information of each view, which can improve the extraction accuracy and efficiency of the single-view features of each view, and based on the global learning layer, process the single-view feature information of all views to obtain the spectral embedding matrix of each view, which can improve the accuracy and reliability of the processed spectral embedding matrix of each view, and then based on the orthogonal constraint layer, orthogonalize the spectral embedding matrices of all views to obtain a multi-view spectral embedding matrix, which can improve the accuracy and reliability of the processed multi-view spectral embedding matrix.

[0192] In this optional embodiment, as an optional implementation, as Figure 4 As shown, Figure 4 is a schematic structural diagram of another multimodal hemodynamic analysis device for sepsis disclosed in an embodiment of the present invention, wherein the device may further include:

[0193] The calculation module 305 is used to calculate the local learning loss of the local learning layer based on the learning data of the local learning layer obtained during the learning process using the local learning layer and the global learning layer; and to calculate the global learning loss of the global learning layer based on the learning data of the global learning layer obtained.

[0194] The calculation module 305 is further configured to calculate the overall loss of the multi-view spectral clustering model based on the local learning loss and the global learning loss.

[0195] The evaluation module 306 is used to evaluate the performance index of the multi-view spectral clustering model according to the overall loss.

[0196] The judgment module 307 is used to judge whether the performance index reaches a preset standard performance index.

[0197] The optimization module 308 is used to optimize the overall loss when the judgment module 307 determines that the performance index does not reach the standard performance index, and obtain the optimized overall loss until the performance index of the multi-view spectrum clustering model reaches the preset standard performance index.

[0198] The determining module 309 is configured to determine that the local learning layer and the global learning layer are learned when the judging module 307 judges that the performance index reaches the standard performance index.

[0199] It can be seen that, in the process of learning by using the local learning layer and the global learning layer, the optional embodiment can calculate the local learning loss of the local learning layer according to the obtained learning data of the local learning layer, calculate the global learning loss of the global learning layer according to the obtained learning data of the global learning layer, and calculate the overall loss of the multi-view spectral clustering model according to the local learning loss and the global learning loss, thereby improving the calculation accuracy and reliability of the overall loss of the model. Then, the performance index of the multi-view spectral clustering model is evaluated according to the overall loss, thereby improving the evaluation accuracy and efficiency of the performance index of the model. It is judged whether the performance index reaches the preset standard performance index. When it is judged that the performance index does not reach the standard performance index, the overall loss is optimized to obtain the optimized overall loss, until the performance index of the multi-view spectral clustering model reaches the preset standard performance index. The timeliness and accuracy of the optimization of the overall loss can be improved, so that the performance index of the model can be accurately reached. When it is judged that the performance index reaches the standard performance index, it is determined that the local learning layer and the global learning layer are learned. The training accuracy of the local learning layer and the global learning layer included in the model can be improved by the performance index evaluated based on the overall loss.

[0200] Embodiment four

[0201] Please refer to Figure 5 , Figure 5 is another structure schematic view of the multi-modal hemodynamic analysis device for sepsis disclosed in the embodiment of the application. As shown in Figure 5 , the multi-modal hemodynamic analysis device for sepsis can include:

[0202] a memory 401 in which executable program codes are stored;

[0203] a processor 402 coupled with the memory 401;

[0204] The processor 402 invokes the executable program codes stored in the memory 401 to execute the steps in the multi-modal hemodynamic analysis method for sepsis described in the embodiment one of the application or the embodiment two of the application.

[0205] Embodiment five

[0206] The embodiment of the present application discloses a computer storage medium, which stores computer instructions, and the computer instructions are used to execute the steps of the multi-modal hemodynamic analysis method applied to sepsis described in the embodiment one or the embodiment two of the present application when called.

[0207] Embodiment six

[0208] The embodiment of the present application discloses a computer program product, which comprises a non-transitory computer readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute the steps of the multi-modal hemodynamic analysis method applied to sepsis described in the embodiment one or the embodiment two.

[0209] The device embodiments described above are only schematic, wherein the modules described as separate components can or can not be physically separate, and the components displayed as modules can or can not be physical modules, that is, can be located in one place, or can be distributed on multiple network modules. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0210] Those skilled in the art can clearly understand the implementation of the various embodiments by means of software and necessary general hardware platforms through the above specific description of the embodiments, and of course, the embodiments can also be implemented by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part of the prior art that makes a contribution, can be embodied in the form of a software product. The computer software product can be stored in a computer readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disk storage, a magnetic disk storage, a magnetic tape storage, or any other computer readable medium that can be used to carry or store data.

[0211] Finally, it should be noted that: the multi-modal hemodynamic analysis method and device applied to sepsis disclosed by the embodiments of the present application are only the preferred embodiments of the present application, and are used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A multi-modal hemodynamic analysis method applied to sepsis, characterized in that, The method comprises: monitoring multi-modal data of each of a plurality of sepsis patients, the multi-modal data comprising two or more of medical indicator data, diagnostic text data, and image data; constructing an adjacency matrix for each of a plurality of views based on sepsis data included in the multi-modal data of each of the sepsis patients in each of the views; performing embedding processing on the adjacency matrices of all of the views based on a dynamic fusion algorithm of a preset multi-view spectral clustering model to obtain multi-view spectral embedding matrices; performing clustering analysis on all of the sepsis patients based on all of the multi-view spectral embedding matrices according to a preset clustering algorithm to obtain clustering results, the clustering results comprising a plurality of sepsis patient groups, all of the sepsis patients in each of the sepsis patient groups corresponding to a hemodynamic characteristic; and the hemodynamic characteristic being used as a basis for generating a treatment strategy for all of the sepsis patients in the corresponding sepsis patient group; wherein the constructing of the adjacency matrix for each of the views based on the sepsis data included in the multi-modal data of each of the sepsis patients in each of the views comprises: performing feature extraction on the sepsis data included in the multi-modal data of each of the sepsis patients in each of the views according to a preset deep neural network model to obtain feature data of each of the sepsis patients in each of the views; constructing a sample set corresponding to each of the sepsis patients based on the feature data of all of the sepsis patients in each of the views according to a preset nearest neighbor analysis algorithm, each sample in the sample set corresponding to a sample type, the sample type comprising a positive sample type or a negative sample type; optimizing a mapping function included in a target network having the same network branches according to the feature data of the sample set corresponding to each of the sepsis patients in each of the views; and calculating the adjacency matrix for each of the views according to the optimized mapping function and the feature data of the sample set corresponding to each of the sepsis patients in each of the views.

2. The multi-modal hemodynamic analysis method for use in sepsis according to claim 1, characterized in that, The constructing of the sample set corresponding to each of the sepsis patients based on the feature data of all of the sepsis patients in each of the views according to the preset nearest neighbor analysis algorithm comprises: calculating the similarity between each of the sepsis patients based on the feature data of all of the sepsis patients in each of the views according to a preset distance metric algorithm; constructing a neighbor set corresponding to each of the sepsis patients based on the similarity between all of the sepsis patients according to the preset nearest neighbor analysis algorithm; and determining the sample set corresponding to each of the sepsis patients according to the determined sample type of each of the sepsis patients.

3. The multi-modal hemodynamic analysis method for use in sepsis according to claim 2, characterized in that, The determining of the sample set corresponding to each of the sepsis patients according to the determined sample type of each of the sepsis patients comprises: determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient; determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient; determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient.

4. The multi-modal hemodynamic analysis method for sepsis according to claim 1, wherein, The method further comprises: determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient; determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient; determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient. The method further comprises:

5. The multi-modal hemodynamic analysis method applied to sepsis according to any one of claims 1-4, characterized in that, determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient; determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient; determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient. The method further comprises: determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient; 6. The multi-modal hemodynamic analysis method for sepsis according to claim 5, characterized in that, determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient; determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient. The method further comprises: determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient; determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient; determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient. The method further comprises: determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient; determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient; determine, according to the sample type of each sepsis patient, samples identical to the sample type of the sepsis patient from the neighbor set corresponding to the sepsis patient as a positive sample set corresponding to the sepsis patient. According to the total loss, a performance index of the multi-view spectral clustering model is evaluated; and it is determined whether the performance index reaches a preset standard performance index; When it is determined that the performance index does not reach the standard performance index, the total loss is optimized to obtain an optimized total loss until the performance index of the multi-view spectral clustering model reaches the preset standard performance index; When it is determined that the performance index reaches the standard performance index, it is determined that the local learning layer and the global learning layer are learned.

7. A multi-modal hemodynamic analysis device applied to sepsis, characterized in that, The device is used to perform the multi-modal hemodynamic analysis method for sepsis according to any one of claims 1-6; and the device comprises: A monitoring module is configured to monitor multi-modal data of each of a plurality of sepsis patients, the multi-modal data including two or more of medical index data, diagnostic text data, and image data; A construction module is configured to construct an adjacency matrix of each of a plurality of views according to sepsis data in each of the views included in the multi-modal data of each of the sepsis patients; A processing module is configured to perform embedding processing on the adjacency matrices of all the views based on a preset dynamic fusion algorithm of a multi-view spectral clustering model to obtain a multi-view spectral embedding matrix; A clustering module is configured to perform clustering analysis on all the sepsis patients based on a preset clustering algorithm according to all the multi-view spectral embedding matrices to obtain a clustering result, the clustering result including a plurality of sepsis patient groups, all the sepsis patients in each of the sepsis patient groups corresponding to a hemodynamic feature; and the hemodynamic feature is used as a basis for generating a treatment strategy for all the sepsis patients in the corresponding sepsis patient group.

8. A multi-modal hemodynamic analysis device applied to sepsis, characterized in that, The device comprises: a memory storing executable program codes; a processor coupled to the memory; the processor invokes the executable program codes stored in the memory to perform the multi-modal hemodynamic analysis method for sepsis according to any one of claims 1-6.

9. A computer storage medium, characterized in that The computer storage medium stores computer instructions, which are invoked to perform the multi-modal hemodynamic analysis method for sepsis according to any one of claims 1-6.

Images