A method for predicting the micro-mechanical state of a branch artery tissue and related equipment

Abstract

This application relates to the field of medical data processing and discloses a method and related equipment for predicting the micromechanical state of arterial tissue. The method acquires macroscopic clinical indicators of the object to be predicted and inputs these indicators into an arterial prediction model to obtain the predicted micromechanical state of the arterial tissue. The arterial prediction model is trained using multiple sets of original training data. During training, the macroscopic clinical indicators in each set of original training data are mapped to the micromechanical data of each region within the same set, generating multiple expanded training data sets to form an expanded training dataset. The model is then trained based on this expanded training dataset, thereby improving prediction stability.

Description

Technical Field

[0001] This application relates to the field of medical data processing, specifically to a method and related equipment for predicting the micromechanical state of branch artery tissue. Background Technology

[0002] The micromechanical state of branch arteries reflects the mechanical characteristics of the vessel wall and is of great significance for vascular condition analysis. The tunica media and adventitia of branch arteries differ in tissue composition and mechanical properties; therefore, analysis based solely on the overall vascular condition is insufficient to accurately reflect the actual differences between different layers.

[0003] In existing technologies, macroscopic clinical indicators such as aortic pulse wave velocity, body mass index, medical history indicators, and blood biochemical indicators are relatively easy to obtain. However, these macroscopic clinical indicators mainly reflect the overall or systemic condition and are difficult to directly and accurately characterize the micromechanical state of branch artery tissue, especially the tunica media and adventitia. On the other hand, although techniques such as atomic force microscopy and nanoindentation can accurately measure the micromechanical properties of isolated branch artery tissue, these techniques rely on isolated tissue samples and are difficult to directly apply to routine prediction scenarios.

[0004] In data analysis, existing methods often employ a single machine learning model to model macroscopic clinical indicators. However, the number of clinical samples with both macroscopic clinical indicators and micromechanical data of ex vivo arterial tissue is limited, resulting in a small sample size in the original training data. Under these circumstances, directly training a prediction model based on this limited amount of original training data can easily lead to problems such as insufficient model stability and large fluctuations in prediction results. Therefore, how to construct a usable training dataset using macroscopic clinical indicators and micromechanical data of ex vivo arterial tissue, given a limited number of original training samples, and achieve stable prediction of the micromechanical state of arterial tissue, has become an urgent technical problem to be solved. Summary of the Invention

[0005] The purpose of this application is to provide a method and related equipment for predicting the micromechanical state of branch artery tissue, so as to overcome the technical problems in the prior art where macroscopic clinical indicators are difficult to directly and accurately characterize the micromechanical state of branch artery tissue, and the small number of original training samples leads to insufficient stability of the prediction model.

[0006] To achieve the above objectives, this application adopts the following technical solution: In a first aspect, this application provides a method for predicting the micromechanical state of arterial tissue, comprising: acquiring macroscopic clinical indicators of the object to be predicted; inputting the macroscopic clinical indicators into an arterial prediction model to obtain prediction results of the micromechanical state of arterial tissue; the arterial prediction model is trained by: acquiring multiple sets of original training data, each set of original training data including macroscopic clinical indicators of the same object and micromechanical data of isolated arterial tissue, the micromechanical data of isolated arterial tissue including micromechanical data of multiple regions on the media and / or adventitia of the blood vessel; corresponding the macroscopic clinical indicators in each set of original training data with the micromechanical data of each region in the same set to generate multiple expanded training data to form an expanded training dataset; and training the arterial prediction model based on the expanded training dataset.

[0007] In some embodiments, before inputting the macro-clinical indicators into the branch artery prediction model, the method further includes: performing unique heat encoding on the medical history indicators and / or body mass index in the macro-clinical indicators to obtain processed macro-clinical indicators; and inputting the processed macro-clinical indicators into the branch artery prediction model.

[0008] In some embodiments, the macroscopic clinical indicators include medical history indicators, body mass index, aortic pulse wave velocity, and blood biochemical indicators; the medical history indicators include smoking history, diabetes history, and / or hypertension history; the blood biochemical indicators include C-peptide level, insulin level, total cholesterol, low-density lipoprotein, high-density lipoprotein, and / or triglycerides.

[0009] In some implementations, the step of mapping the macroscopic clinical indicators in each set of original training data to the microscopic mechanical data of each region in the same set to generate multiple sets of expanded training data includes: for each object, selecting several regions from the vascular media and / or adventitia of the isolated branch artery tissue of the object, and obtaining the microscopic mechanical data of each selected region; using the microscopic mechanical data of each selected region as independent region-level samples; and mapping the macroscopic clinical indicators of the object to each independent region-level sample to form corresponding expanded training data.

[0010] In some embodiments, the micromechanical data of the isolated branch artery tissue is elastic modulus data; the elastic modulus data is obtained by measuring the vascular media and / or adventitia of the isolated branch artery tissue slices using atomic force microscopy nanoindentation technology.

[0011] In some implementations, before training the branch artery prediction model based on the expanded training dataset, the process includes: performing principal component analysis on the original training data and the expanded training dataset to obtain the distribution results of the original training data and the expanded training dataset in the principal component space; comparing the overall distribution trends of the original training data and the expanded training dataset in the principal component space based on the distribution results; and, if the overall distribution trends of the original training data and the expanded training dataset in the principal component space are consistent, determining that the expanded training dataset does not introduce significant structural bias, and training the branch artery prediction model based on the expanded training dataset.

[0012] In some implementations, the branch artery prediction model is an ensemble learning model, which includes a logistic regression model, a random forest model, and a multilayer perceptron model. Training the branch artery prediction model based on the expanded training dataset includes: training the logistic regression model, the random forest model, and the multilayer perceptron model respectively based on the expanded training dataset; during prediction, the outputs of the logistic regression model, the random forest model, and the multilayer perceptron model are integrated using a hard voting mechanism to obtain the predicted result of the branch artery tissue micromechanical state.

[0013] Secondly, this application provides a branch artery tissue micromechanical state prediction system, comprising: an acquisition module for acquiring macroscopic clinical indicators of the object to be predicted; and a prediction module for inputting the macroscopic clinical indicators into a branch artery prediction model to obtain the prediction result of the branch artery tissue micromechanical state. The branch artery prediction model is trained in the following manner: acquiring multiple sets of original training data, each set of original training data including macroscopic clinical indicators of the same object and micromechanical data of isolated branch artery tissue; the micromechanical data of isolated branch artery tissue including micromechanical data of multiple regions on the vascular media and / or vascular adventitia; corresponding the macroscopic clinical indicators in each set of original training data with the micromechanical data of each region in the same set to generate multiple expanded training data, forming an expanded training dataset; and training the branch artery prediction model based on the expanded training dataset.

[0014] Thirdly, this application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method for predicting the micromechanical state of branch artery tissue as described above.

[0015] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method for predicting the micromechanical state of branch artery tissue as described above.

[0016] Compared with the prior art, this application has the following beneficial technical effects: Firstly, this application provides a method for predicting the micromechanical state of branch artery tissue. This method obtains macroscopic clinical indicators of the object to be predicted and inputs these indicators into a branch artery prediction model to obtain the predicted micromechanical state of the branch artery tissue. Simultaneously, the branch artery prediction model is not directly trained based on a limited number of original training data sets. Instead, it acquires multiple sets of original training data sets and maps the macroscopic clinical indicators in each set to the micromechanical data of each region within the same set, generating multiple expanded training data sets to form an expanded training dataset. The branch artery prediction model is then trained based on this expanded training dataset. This method is beneficial for expanding training data when the number of original training samples is small and improves the stability of the predicted micromechanical state of branch artery tissue.

[0017] Secondly, this application provides a micromechanical state prediction system for branch artery tissue. The system acquires macroscopic clinical indicators of the object to be predicted via an acquisition module, and inputs these macroscopic clinical indicators into a branch artery prediction model via a prediction module to obtain the predicted micromechanical state of the branch artery tissue. The branch artery prediction model acquires multiple sets of original training data, and maps the macroscopic clinical indicators in each set of original training data to the micromechanical data of each region within the same set, generating multiple expanded training data sets to form an expanded training dataset. The system is then trained based on this expanded training dataset. This allows the system to predict the micromechanical state of branch artery tissue based on macroscopic clinical indicators, and helps mitigate the impact of a small number of original training samples on model stability.

[0018] Thirdly, the computer device provided in this application executes a computer program stored in a memory through a processor to implement the steps of the above-mentioned method for predicting the micromechanical state of branch artery tissue, thereby enabling the processing of macroscopic clinical indicators of the object to be predicted, the invocation of the branch artery prediction model, and the output of the prediction results of the micromechanical state of branch artery tissue on the computer device.

[0019] Fourthly, this application provides a computer-readable storage medium in which a computer program is stored, so that when the computer program is executed by a processor, it implements the steps of the above-mentioned method for predicting the micromechanical state of branch artery tissue, thereby facilitating the deployment and operation of the above-mentioned method for predicting the micromechanical state of branch artery tissue on different computer devices. Attached Figure Description

[0020] Figure 1 This is a flowchart of a method for predicting the micromechanical state of arterial tissue in an embodiment of this application.

[0021] Figure 2 This is a flowchart of the branch artery prediction model training process in the embodiments of this application.

[0022] Figure 3 This is a schematic diagram of a micromechanical state prediction system for branch artery tissue in an embodiment of this application.

[0023] Figure 4 This is a schematic diagram showing the relationship between aortic pulse wave velocity and branch artery microelastic modulus in an embodiment of this application.

[0024] Figure 5 This is a graph showing the principal component analysis verification results in the embodiments of this application.

[0025] Figure 6 The figures show a comparison of the effects of different machine learning algorithms in the embodiments of this application. In the figure, (a) shows the effect on the outer membrane and (b) shows the effect on the middle membrane.

[0026] Figure 7 The figures show a comparison of the effects of the integrated learning model on the adventitia and media of blood vessels in the embodiments of this application. In the figure, (a) shows the effect on the adventitia and (b) shows the effect on the media.

[0027] Figure 8 The diagram shows the SHAP analysis in the embodiments of this application. In the diagram, (a) shows the effect on the outer membrane and (b) shows the effect on the middle membrane. Detailed Implementation

[0028] In the analysis of the micromechanical state of branch artery tissue, macroscopic clinical indicators such as aortic pulse wave velocity, body mass index, medical history indicators, and blood biochemical indicators are relatively easy to obtain. However, these macroscopic clinical indicators mainly reflect the overall or systemic state and are difficult to directly and accurately characterize the micromechanical state of branch artery tissue, especially the tunica media and adventitia. On the other hand, although techniques such as atomic force microscopy and nanoindentation can accurately measure the micromechanical properties of isolated branch artery tissue, the sources of such micromechanical data are limited, resulting in a small number of original training samples. When training prediction models directly based on the original training samples, the model is prone to insufficient stability.

[0029] Based on the above background, this application proposes a method and related equipment for predicting the micromechanical state of branch arteries. This method obtains macroscopic clinical indicators of the object to be predicted and inputs these indicators into a branch artery prediction model to obtain the prediction results of the branch artery tissue micromechanical state. Simultaneously, by acquiring multiple sets of original training data and mapping the macroscopic clinical indicators in each set of original training data to the micromechanical data of each region within the same set, multiple expanded training data are generated, forming an expanded training dataset. The branch artery prediction model is then trained based on this expanded training dataset. This approach is beneficial for expanding training data when the number of original training samples is small and improves the stability of the predicted branch artery tissue micromechanical state.

[0030] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0031] Example 1 Reference Figure 1 As shown, this embodiment provides a method for predicting the micromechanical state of arterial tissue, including: acquiring macroscopic clinical indicators of the object to be predicted; inputting the macroscopic clinical indicators into an arterial prediction model to obtain the predicted result of the micromechanical state of arterial tissue. The arterial prediction model uses... Figure 2 The training process shown is used to obtain the predicted micromechanical state of the branch artery tissue. The predicted micromechanical state of the branch artery tissue is used to characterize the micromechanical state of the object to be predicted.

[0032] In this embodiment, the macroscopic clinical indicators are data available for the object to be predicted in a clinical setting. By acquiring the macroscopic clinical indicators of the object to be predicted, input information can be provided for subsequent prediction of the micromechanical state of the branch arteries. As input to the branch artery prediction model, the macroscopic clinical indicators carry the possible correlation information between the overall state of the object to be predicted and the micromechanical state of the branch arteries, thereby enabling the branch artery prediction model to output the corresponding prediction result of the micromechanical state of the branch arteries based on the macroscopic clinical indicators.

[0033] After obtaining the macroscopic clinical indicators of the object to be predicted, these indicators are input into the branch artery prediction model. Based on the mapping relationship established during the training phase, the branch artery prediction model processes the input macroscopic clinical indicators and outputs the predicted results of the micromechanical state of the branch artery tissue. In this way, it is not necessary to obtain ex vivo branch artery tissue samples of the object to be predicted during the prediction phase; the corresponding predicted results of the micromechanical state of the branch artery tissue can be obtained based on the macroscopic clinical indicators of the object to be predicted, thereby achieving the prediction of the micromechanical state of the branch artery tissue.

[0034] In this embodiment, the branch artery prediction model is not directly preset, but is trained through a training process. (Refer to...) Figure 2 As shown, the training process of the branch artery prediction model includes: acquiring multiple sets of original training data, each set of original training data including macroscopic clinical indicators and micromechanical data of isolated branch artery tissue for the same subject; the micromechanical data of isolated branch artery tissue includes micromechanical data of multiple regions on the vascular media and / or adventitia; mapping the macroscopic clinical indicators in each set of original training data to the micromechanical data of each region in the same set to generate multiple expanded training data, forming an expanded training dataset; and training the branch artery prediction model based on the expanded training dataset.

[0035] The original training data is constructed on an object-by-object basis. Each set of original training data includes macroscopic clinical indicators of the same object and microscopic biomechanical data of the isolated arterial tissue of that object. This ensures that the macroscopic clinical indicators and microscopic biomechanical data of the isolated arterial tissue in the same set of original training data originate from the same object, thus providing a foundation for subsequently establishing the correspondence between the training data.

[0036] Furthermore, the micromechanical data of the ex vivo branch artery tissue does not correspond to a single location, but includes micromechanical data from multiple regions on the vascular media and / or adventitia. Since micromechanical data from multiple regions can be obtained from the ex vivo branch artery tissue of the same subject, multiple training datasets can be constructed based on the original training data, focusing on macroscopic clinical indicators of the same subject and the micromechanical data from multiple regions, thereby increasing the amount of training data. In this way, the micromechanical data of the ex vivo branch artery tissue serves two purposes: firstly, to characterize the micromechanical state of the branch artery tissue in different regions, and secondly, to provide a data foundation for the subsequent formation of an expanded training dataset.

[0037] When creating the expanded training dataset, the macroscopic clinical indicators in each set of original training data are mapped to the microscopic biomechanical data of each region within the same set, generating multiple expanded training datasets. In other words, for the same object, instead of retaining only one object-level original training dataset, a mapping is established between the object's macroscopic clinical indicators and the microscopic biomechanical data of multiple regions on the ex vivo branch artery tissue of that object, allowing multiple expanded training datasets to be generated for the same object. In this way, multiple sets of original training data can be used to form the expanded training dataset, thereby increasing the amount of training data even with a small number of original training samples, providing more training data support for the subsequent training of the branch artery prediction model.

[0038] After obtaining the expanded training dataset, the branch artery prediction model is trained based on it. During training, the branch artery prediction model learns the correspondence between macroscopic clinical indicators and the micromechanical state of branch artery tissue based on the expanded training dataset. After training, the branch artery prediction model can be used to process the macroscopic clinical indicators of the object to be predicted and output the prediction results of the micromechanical state of branch artery tissue. In this way, the correspondence established in the training phase can be transferred to the prediction phase, thus forming a complete technical chain for the branch artery tissue micromechanical state prediction method.

[0039] In summary, the branch artery tissue micromechanical state prediction method provided in this embodiment obtains macroscopic clinical indicators of the object to be predicted and inputs these indicators into the trained branch artery prediction model to obtain the prediction result of the branch artery tissue micromechanical state. Simultaneously, during the training phase of the branch artery prediction model, multiple expanded training data are generated by corresponding the macroscopic clinical indicators in each group of original training data with the micromechanical data of each region in the same group, forming an expanded training dataset. The branch artery prediction model is then trained based on the expanded training dataset. This method is beneficial for expanding the training data when the number of original training samples is small and improves the stability of the branch artery tissue micromechanical state prediction.

[0040] Example 2 Based on Example 1, this example describes the processing method of macroscopic clinical indicators, the expansion method of the original training data, the acquisition method of micromechanical data of isolated branch artery tissue, the verification method of the expanded training dataset, and the specific implementation method of the branch artery prediction model.

[0041] In this embodiment, macroscopic clinical indicators include at least medical history indicators and body mass index (BMI). In some embodiments, macroscopic clinical indicators also include aortic pulse wave velocity (APV) and blood biochemical indicators. Medical history indicators may include smoking history, diabetes history, and / or hypertension history. Blood biochemical indicators may include C-peptide levels, insulin levels, total cholesterol, low-density lipoprotein (LDL), high-density lipoprotein (HDL), and / or triglycerides. Among the aforementioned macroscopic clinical indicators, medical history indicators and BMI classifications are typically categorical or hierarchical information, while aortic pulse wave velocity and blood biochemical indicators are typically numerical information. By simultaneously introducing medical history indicators, BMI, aortic pulse wave velocity, and blood biochemical indicators, the branch artery prediction model can obtain information related to the micromechanical state of branch artery tissue from multiple dimensions.

[0042] To facilitate the input of macro-clinical indicators into the branch artery prediction model, medical history indicators and body mass index (BMI) can be categorized and encoded using one-hot encoding before input. Specifically, smoking history, diabetes history, hypertension history, and BMI can be converted into multiple binary features to reduce the potential impact of illogical ordering when directly inputting categorical information into the model. After one-hot encoding, the processed macro-clinical indicators are then used as input to the branch artery prediction model along with numerical indicators such as aortic pulse wave velocity and blood biochemical parameters. This approach allows different types of macro-clinical indicators to be input into the branch artery prediction model in a form suitable for model processing, thereby improving the model's efficiency in utilizing macro-clinical indicators.

[0043] In this embodiment, the process of expanding the original training data can further include the following: For each object, select several regions from the vascular media and / or adventitia of the isolated branch artery tissue of the object, and obtain the micromechanical data of each selected region; use the micromechanical data of each selected region as an independent region-level sample; map the macroscopic clinical indicators of the object to each independent region-level sample to form corresponding expanded training data. That is, a set of macroscopic clinical indicators of the same object no longer corresponds to only one object-level data point of the object, but corresponds to the micromechanical data of multiple regions on the vascular media and / or adventitia of the object, thereby generating multiple expanded training data points. In this way, an expanded training dataset can be formed from a limited number of original training data points.

[0044] In some embodiments, when selecting regions from isolated arterial tissue, multiple regions can be selected from the adventitia and multiple regions from the media, respectively, so that the micromechanical data of both the adventitia and media can be included in the augmented training dataset. In some more specific embodiments, six regions can be selected from the adventitia and six regions from the media of each object. By treating the micromechanical data of each selected region as independent region-level samples, multiple augmented training datasets can be generated for the same object, thereby mitigating the impact of the small number of original training samples.

[0045] In this embodiment, the micromechanical data of the isolated branch artery tissue is elastic modulus data. This elastic modulus data can be obtained by measuring the media and / or adventitia of isolated branch artery tissue slices using atomic force microscopy nanoindentation technology. Specifically, isolated branch artery tissue slices can be obtained first, and then measurement areas can be selected on the media and / or adventitia. Atomic force microscopy nanoindentation technology can then be used to measure each measurement area to obtain the elastic modulus data for the corresponding region. This elastic modulus data is used to characterize the micromechanical properties of local areas of the isolated branch artery tissue and serves as an important component in forming expanded training data.

[0046] Before training the branch artery prediction model based on the expanded training dataset, principal component analysis can be performed on both the original and expanded training datasets. Specifically, the distribution results of the original and expanded training datasets in the principal component space can be obtained separately, and the overall distribution trends of the original and expanded training datasets in the principal component space can be compared based on these distribution results. (Refer to...) Figure 5 As shown, in some embodiments, if the overall distribution trends of the original training data and the expanded training dataset are consistent in the principal component space, it can be determined that the expanded training dataset does not introduce significant structural bias, and the branch artery prediction model can be trained based on the expanded training dataset. Principal component analysis can verify the usability of the expanded training dataset, thereby providing support for the subsequent training of the branch artery prediction model.

[0047] In this embodiment, the branch artery prediction model can be an ensemble learning model. The ensemble learning model includes a logistic regression model, a random forest model, and a multilayer perceptron model. When training the branch artery prediction model based on the expanded training dataset, the logistic regression model, random forest model, and multilayer perceptron model can be trained separately on the expanded training dataset. The logistic regression model can be used to model the relationship between input features and classification results, the random forest model can be used to handle complex nonlinear relationships, and the multilayer perceptron model can be used to learn more complex feature patterns. By training multiple models separately, different models can learn the relationship between macroscopic clinical indicators and the micromechanical state of branch artery tissue from different perspectives.

[0048] During prediction, a hard voting mechanism is used to integrate the outputs of the logistic regression model, random forest model, and multilayer perceptron model to obtain the predicted micromechanical state of the branch artery tissue. Specifically, the prediction results output by the logistic regression model, random forest model, and multilayer perceptron model are summarized, and the prediction result with the most occurrences is used as the final prediction result of the micromechanical state of the branch artery tissue. By adopting a hard voting mechanism, the output results of multiple models can be integrated during the prediction stage, thereby improving the stability of the predicted micromechanical state of the branch artery tissue.

[0049] In summary, this embodiment clarifies the specific implementation process of the branch artery tissue micromechanical state prediction method by explaining the processing methods of macroscopic clinical indicators, the expansion methods of the original training data, the acquisition methods of micromechanical data of isolated branch artery tissue, the verification methods of the expanded training dataset, and the specific implementation methods of the branch artery prediction model. By performing one-hot encoding on medical history indicators and body mass index, macroscopic clinical indicators can be made more suitable for input into the branch artery prediction model. By mapping the macroscopic clinical indicators of the same object to the micromechanical data of multiple regions of that object, an expanded training dataset can be formed. Principal component analysis can be used to determine whether the expanded training dataset introduces significant structural bias. By employing an ensemble learning model including logistic regression, random forest, and multilayer perceptron models, and using a hard voting mechanism for integration during prediction, the prediction results of the branch artery tissue micromechanical state can be obtained.

[0050] Example 3 Based on Examples 1 and 2, this example, combined with experimental analysis results, explains the relationship between macroscopic clinical indicators and micromechanical data of branch artery tissue in the prediction method of branch artery tissue micromechanical state, the availability of the expanded training dataset, and the prediction effect of the branch artery prediction model.

[0051] Reference Figure 4 As shown, Figure 4 The relationship between aortic pulse wave velocity and branch artery microelastic modulus is shown, where, Figure 4 The x-axis represents the aortic pulse wave velocity, and the y-axis represents the microelastic modulus of the branch arteries. Aortic pulse wave velocity is one of the macroscopic clinical indicators, while the microelastic modulus of the branch arteries is one of the micromechanical data points for isolated branch artery tissue. Figure 4 It can be shown that there is a certain correlation between aortic pulse wave velocity and the microelastic modulus of branch arteries, but the two are not a simple, complete correspondence. In other words, it is difficult to accurately characterize the micromechanical state of branch artery tissue based solely on aortic pulse wave velocity, a single macroscopic clinical indicator. Therefore, inputting macroscopic clinical indicators such as medical history, body mass index, aortic pulse wave velocity, and blood biochemical indicators into the branch artery prediction model can help improve the ability to characterize the micromechanical state of branch artery tissue.

[0052] Reference Figure 5 As shown, Figure 5 The results of principal component analysis are shown. Figure 5 This is used to characterize the distribution of the original training data and the expanded training dataset in the principal component space. By performing principal component analysis on the original training data and the expanded training dataset separately, the distribution results of both in the principal component space can be obtained. In some embodiments, the overall distribution trends of the original training data and the expanded training dataset in the principal component space are compared based on these distribution results; if the overall distribution trends of the original training data and the expanded training dataset in the principal component space are consistent, it can be determined that the expanded training dataset has not introduced significant structural bias, and the branch artery prediction model can be trained based on the expanded training dataset. Therefore, Figure 5 The results shown are used to verify the availability of the expanded training dataset, thereby providing support for the training of subsequent branch artery prediction models.

[0053] To quantify the performance of different machine learning algorithms on tasks related to the adventitia and media of blood vessels, refer to Table 1, which shows the AUC results of different machine learning algorithms on tasks related to the adventitia and media of blood vessels.

[0054] Table 1. AUC Results of Machine Learning Algorithms

[0055] In the task corresponding to the adventitia, the AUC of the logistic regression model was 0.79, the AUC of the random forest model was 0.87, the AUC of the multilayer perceptron model was 0.90, and the AUC of the ensemble learning model was 0.88. In the task corresponding to the media, the AUCs of the logistic regression model and the random forest model were 0.72, the AUC of the multilayer perceptron model was 0.73, and the AUC of the ensemble learning model was 0.72. Table 1 illustrates the differences in performance between different machine learning algorithms in the adventitia and media correspondence tasks from a numerical perspective.

[0056] Reference Figure 6 As shown, Figure 6 The comparison of the performance of different machine learning algorithms is shown. Figure 6 (a) shows a comparison of the performance of different machine learning algorithms for tasks related to the adventitia of blood vessels. Figure 6 (b) shows a comparison of the performance of different machine learning algorithms for the task corresponding to the vascular media. Figure 6 The machine learning algorithms compared in the text include logistic regression, random forest, multilayer perceptron, and ensemble learning models. Figure 6 This indicates that different machine learning algorithms perform differently on tasks related to the adventitia and media of blood vessels, and different models have different adaptability to predicting the micromechanical state of branch arteries at different layers. Figure 6 This is used to illustrate the predictive performance of logistic regression, random forest, and multilayer perceptron models as base learners, and the performance of ensemble learning models in predicting the micromechanical state of branch arteries.

[0057] Reference Figure 7 As shown, Figure 7 The comparison of the performance of the ensemble learning model on the adventitia and media of blood vessels is shown. Figure 7 (a) shows the performance of the ensemble learning model in the adventitia correspondence task. Figure 7 (b) in the figure shows the performance of the ensemble learning model in the vascular membrane correspondence task. Figure 7 This can include multiple evaluation parameters of the ensemble learning model for the corresponding task, used to characterize the predictive performance of the ensemble learning model at different levels of the corresponding task. Figure 7 It can be shown that the ensemble learning model performs differently in the tasks corresponding to the adventitia and media of blood vessels, and the prediction results for tasks corresponding to different layers can be analyzed separately. Therefore, the branch artery prediction model in this application can not only be used to obtain the prediction results of the micromechanical state of branch artery tissue, but also to characterize the prediction performance in the tasks corresponding to the adventitia and media of blood vessels, respectively.

[0058] In some implementations, the ensemble learning model includes logistic regression, random forest, and multilayer perceptron models, and the outputs of these models are integrated during prediction using a hard voting mechanism. Since logistic regression, random forest, and multilayer perceptron models have different characteristics in processing input features, integrating the outputs of each model through a hard voting mechanism during the prediction phase can synthesize the results of different models. Figure 6 and Figure 7 The results shown allow for a comparative analysis of the performance of the ensemble learning model and individual models on tasks at different levels, thereby illustrating the application of the ensemble learning model in predicting the micromechanical state of arterial tissue.

[0059] Reference Figure 8 As shown, Figure 8 The SHAP analysis diagram is shown. Among them, Figure 8 (a) shows the SHAP analysis results for the task corresponding to the adventitia of blood vessels. Figure 8 (b) shows the SHAP analysis results for the corresponding task of the vascular media. Figure 8 This is used to characterize the contribution of each input feature to the output of the branch artery prediction model. In some implementations, input features include medical history indicators, body mass index, aortic pulse wave velocity, and blood biochemical indicators. Figure 8 The contributions of each input feature in the adventitia and medial vascular correspondence tasks can be analyzed to illustrate the differences in the roles of different input features in tasks corresponding to different vascular layers. Therefore, Figure 8 It can be used to demonstrate the impact of various input features on the prediction results when using macroscopic clinical indicators to predict the micromechanical state of branch artery tissue in a branch artery prediction model.

[0060] In summary, this embodiment combines Figures 4 to 8 This paper explains some of the validation results of the method for predicting the micromechanical state of arterial tissue. Figure 4 This is used to illustrate the relationship between aortic pulse wave velocity and branch artery microelastic modulus in macroscopic clinical indicators. Figure 5 This is used to illustrate the distribution of the expanded training dataset and the original training data in the principal component space; Figure 6 and Figure 7 This is used to illustrate the effectiveness of different machine learning algorithms and ensemble learning models in corresponding tasks on the adventitia and media of blood vessels. Figure 8 This is used to illustrate the contribution of each input feature to the output of the branch artery prediction model. The above figures further illustrate part of the implementation process and analysis results of the branch artery tissue micromechanical state prediction method in this application.

[0061] Example 4 Based on Examples 1 to 3, and referring to Figure 3 As shown, this embodiment provides a micromechanical state prediction system for branch artery tissue, including an acquisition module and a prediction module.

[0062] The acquisition module is used to acquire macroscopic clinical indicators of the object to be predicted. These macroscopic clinical indicators can be relevant data acquired from the object in a clinical setting, used as input for subsequent branch artery prediction models. The acquisition module can obtain these macroscopic clinical indicators from hospital information systems, laboratory information systems, electronic medical record systems, or other medical data sources, or from local databases, terminal input interfaces, or other data interfaces.

[0063] The prediction module is used to input the macroscopic clinical indicators into the branch artery prediction model to obtain the prediction results of the micromechanical state of the branch artery tissue. The branch artery prediction model is used to process the input macroscopic clinical indicators and output the corresponding prediction results of the micromechanical state of the branch artery tissue. By setting the prediction module, the micromechanical state of the branch artery tissue can be predicted directly based on the branch artery prediction model after obtaining the macroscopic clinical indicators of the object to be predicted.

[0064] In this embodiment, the branch artery prediction model is trained as follows: multiple sets of original training data are acquired, each set including macroscopic clinical indicators and micromechanical data of isolated branch artery tissue for the same object; the micromechanical data of isolated branch artery tissue includes micromechanical data of multiple regions on the vascular media and / or adventitia; the macroscopic clinical indicators in each set of original training data are mapped to the micromechanical data of each region in the same set to generate multiple expanded training data, forming an expanded training dataset; the branch artery prediction model is trained based on the expanded training dataset. Therefore, the branch artery prediction model called by the prediction module is not statically preset, but trained based on the expanded training dataset, enabling the prediction module to predict the micromechanical state of the branch artery tissue of the object to be predicted based on the correspondence between macroscopic clinical indicators and the micromechanical state of branch artery tissue learned during the training phase.

[0065] In some implementations, the macro-clinical indicators acquired by the acquisition module include at least medical history indicators and body mass index (BMI). Further, the macro-clinical indicators may also include aortic pulse wave velocity and blood biochemical indicators. Correspondingly, before the prediction module inputs the macro-clinical indicators into the branch artery prediction model, the medical history indicators and BMI can be categorized and subjected to one-hot encoding to obtain processed macro-clinical indicators, which are then input into the branch artery prediction model. In this way, different types of macro-clinical indicators acquired by the acquisition module can be input into the branch artery prediction model invoked by the prediction module in a form suitable for model processing.

[0066] In some implementations, the branch artery prediction model can be an ensemble learning model. The ensemble learning model may include a logistic regression model, a random forest model, and a multilayer perceptron model. When performing predictions, the prediction module can integrate the outputs of the logistic regression model, the random forest model, and the multilayer perceptron model through a hard voting mechanism to obtain the predicted results of the branch artery tissue micromechanical state. In this way, the prediction module can complete the prediction process by combining the outputs of different models.

[0067] The branch artery tissue micromechanical state prediction system provided in this embodiment acquires macroscopic clinical indicators of the object to be predicted through an acquisition module, and inputs the macroscopic clinical indicators into the branch artery prediction model through a prediction module to obtain the prediction results of the branch artery tissue micromechanical state. At the same time, the branch artery prediction model is obtained by expanding the original training data at the regional level and training it based on the expanded training dataset. This enables the system to predict the micromechanical state of branch artery tissue based on macroscopic clinical indicators, and helps to alleviate the impact of the small number of original training samples on the stability of the model.

[0068] For details regarding the training method of the branch artery prediction model, the specific content of the macroscopic clinical indicators, the expansion method of the original training data, the acquisition method of the micromechanical data of the ex vivo branch artery tissue, the verification method of the expanded training dataset, and the specific implementation method of the ensemble learning model, please refer to the relevant descriptions in Examples 1 to 3. The corresponding technical content is also applicable in this example and will not be repeated here.

[0069] This application also provides a computer device in its specific embodiments. The computer device includes a processor and a memory. The memory stores a computer program. When the processor executes the computer program, it implements the steps of the branch artery tissue micromechanical state prediction method described in any of the foregoing embodiments. Specifically, the processor can perform the following steps: acquiring macroscopic clinical indicators of the object to be predicted; inputting the macroscopic clinical indicators into the branch artery prediction model to obtain the branch artery tissue micromechanical state prediction result; wherein, the branch artery prediction model acquires multiple sets of original training data, and corresponds the macroscopic clinical indicators in each set of original training data with the micromechanical data of each region in the same set to generate multiple expanded training data, forming an expanded training dataset, and then trains based on the expanded training dataset.

[0070] This application also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the steps of the branch artery tissue micromechanical state prediction method described in any of the foregoing embodiments. Specifically, when the computer program is executed by a processor, it can perform the following steps: acquiring macroscopic clinical indicators of the object to be predicted; inputting the macroscopic clinical indicators into a branch artery prediction model to obtain the branch artery tissue micromechanical state prediction result; wherein, the branch artery prediction model acquires multiple sets of original training data, and corresponds the macroscopic clinical indicators in each set of original training data with the micromechanical data of each region in the same set to generate multiple expanded training data, forming an expanded training dataset, and then trains based on the expanded training dataset.

[0071] The foregoing has shown and described the basic principles, main features, and advantages of this application. It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from the spirit or basic characteristics of this application. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of this application is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within this application. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0072] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can be appropriately combined to form other embodiments that can be understood by those skilled in the art. The above content is only for illustrating the technical concept of this application and should not be used to limit the scope of protection of this application. Any modifications made to the technical solutions based on the technical concept proposed in this application fall within the scope of protection of the claims of this application.

Claims

1. A method for predicting the micromechanical state of arterial tissue, characterized in that, include: Obtain macroscopic clinical indicators of the subjects to be predicted; The macroscopic clinical indicators are input into the branch artery prediction model to obtain the prediction results of the micromechanical state of the branch artery tissue; The branch artery prediction model was trained in the following way: Multiple sets of raw training data were acquired. Each set of raw training data included macroscopic clinical indicators and micromechanical data of isolated branch artery tissue for the same subject. The micromechanical data of the isolated branch artery tissue included micromechanical data of multiple regions on the vascular media and / or adventitia. The macroscopic clinical indicators in each group of original training data are matched with the microscopic mechanical data of each region in the same group to generate multiple expanded training data, forming an expanded training dataset. The branch artery prediction model is trained based on the expanded training dataset.

2. The method for predicting the micromechanical state of arterial tissue according to claim 1, characterized in that, The macro-clinical indicators include at least medical history indicators and body mass index; Inputting the macroscopic clinical indicators into the branch artery prediction model includes: The medical history indicators and the body mass index are categorized and subjected to unique thermal coding to obtain the processed macroscopic clinical indicators. The processed macroscopic clinical indicators are input into the branch artery prediction model.

3. The method for predicting the micromechanical state of arterial tissue according to claim 2, characterized in that, The macroscopic clinical indicators include medical history indicators, body mass index, aortic pulse wave velocity, and blood biochemistry indicators; The medical history indicators include smoking history, diabetes history and / or hypertension history; The blood biochemical indicators include C-peptide levels, insulin levels, total cholesterol, low-density lipoprotein, high-density lipoprotein, and / or triglycerides.

4. The method for predicting the micromechanical state of arterial tissue according to claim 1, characterized in that, The process involves mapping the macroscopic clinical indicators in each group of original training data to the microscopic mechanical data of each region within the same group, generating multiple supplementary training data sets, including: For each object, several regions are selected from the vascular media and / or adventitia of the isolated branch artery tissue of that object, and the micromechanical data of each selected region are obtained. The micromechanical data of each selected region are used as independent region-level samples; The macroscopic clinical indicators of the object are mapped one-to-one with each independent regional sample to form corresponding expanded training data.

5. The method for predicting the micromechanical state of arterial tissue according to claim 4, characterized in that, The micromechanical data of the isolated arterial tissue are elastic modulus data; The elastic modulus data were obtained by measuring the vascular media and / or adventitia of isolated arterial tissue sections using atomic force microscopy nanoindentation technology.

6. The method for predicting the micromechanical state of arterial tissue according to claim 1, characterized in that, Before training the branch artery prediction model based on the expanded training dataset, the following steps are also included: Principal component analysis is performed on the original training data and the expanded training dataset to obtain the distribution results of the original training data and the expanded training dataset in the principal component space, respectively. Based on the distribution results, compare the overall distribution trends of the original training data and the expanded training dataset in the principal component space; If the overall distribution trend of the original training data and the expanded training dataset is consistent in the principal component space, it is determined that the expanded training dataset does not introduce significant structural bias, and the branch artery prediction model is trained based on the expanded training dataset.

7. The method for predicting the micromechanical state of arterial tissue according to claim 1, characterized in that, The branch artery prediction model is an ensemble learning model, which includes a logistic regression model, a random forest model, and a multilayer perceptron model. The branch artery prediction model is trained based on the expanded training dataset, including: The logistic regression model, the random forest model, and the multilayer perceptron model are trained respectively based on the expanded training dataset. During prediction, the outputs of the logistic regression model, the random forest model, and the multilayer perceptron model are integrated through a hard voting mechanism to obtain the prediction results of the micromechanical state of the branch artery tissue.

8. A system for predicting the micromechanical state of arterial tissue, characterized in that, include: The acquisition module is used to acquire macroscopic clinical indicators of the object to be predicted. The prediction module is used to input the macroscopic clinical indicators into the branch artery prediction model to obtain the prediction results of the micromechanical state of the branch artery tissue. The branch artery prediction model was trained in the following way: Multiple sets of raw training data were acquired. Each set of raw training data included macroscopic clinical indicators and micromechanical data of isolated branch artery tissue for the same subject. The micromechanical data of the isolated branch artery tissue included micromechanical data of multiple regions on the vascular media and / or adventitia. The macroscopic clinical indicators in each group of original training data are matched with the microscopic mechanical data of each region in the same group to generate multiple expanded training data, forming an expanded training dataset. The branch artery prediction model is trained based on the expanded training dataset.

9. A computer device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method for predicting the micromechanical state of branch artery tissue as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method for predicting the micromechanical state of branch artery tissue as described in any one of claims 1 to 7.

Images