A method, device, equipment and storage medium for preoperative assessment of cardiovascular and cerebrovascular risks in non-cardiac surgery

Abstract

The present application relates to a method, apparatus, device and storage medium for preoperative assessment of cardiovascular and cerebrovascular risks in non-cardiac surgery. The method includes: extracting features from the target patient's multimodal preoperative data to obtain multimodal data features; using a preoperative feature fusion module based on cross-modal global cross-attention to perform cross-modal interactive processing and global modeling and fusion on the multimodal data features to obtain cardiovascular and cerebrovascular risk features; inputting the cardiovascular and cerebrovascular risk features into a cardiovascular and cerebrovascular risk assessment module based on multidimensional reconstruction loss, the cardiovascular and cerebrovascular risk assessment module is guided by the comprehensive quantitative assessment of the risk to be predicted, and outputs a preoperative assessment result of cardiovascular and cerebrovascular risk based on multidimensional reconstruction loss. The present application improves the efficiency and accuracy of preoperative assessment of cardiovascular and cerebrovascular risks in non-cardiac surgery, optimizes the extraction and fusion process of multimodal data features, and enhances the model's preoperative assessment capability and generalization capability for cardiovascular and cerebrovascular risks.

Description

Technical Field

[0001] The present application belongs to the field of medical artificial intelligence models and medical health technology, and in particular relates to a method, apparatus, equipment and storage medium for preoperative assessment of cardiovascular and cerebrovascular risks in non-cardiac surgery. Background Art

[0002] Traditional preoperative cardiovascular risk assessment methods for non-cardiac surgery primarily involve collecting patient information through history-taking, physical examinations, and laboratory tests, and then integrating clinical experience into risk assessment. However, these methods often focus solely on physiological indicators or medical history, neglecting other important factors such as psychological indicators and social factors. Furthermore, traditional preoperative cardiovascular risk diagnosis relies heavily on the physician's clinical experience and subjective judgment when faced with complex and information-rich preoperative modal features. This preoperative data is not only diverse but also highly heterogeneous and heterogeneous, making comprehensive and accurate analysis difficult through a single processing method. Consequently, physicians must expend considerable time and effort on manual analysis and judgment, significantly reducing diagnostic and assessment efficiency and potentially leading to misjudgments or missed diagnosis due to human error, impacting surgical planning and patient outcomes. Furthermore, because preoperative data typically include a variety of different types, including high-definition medical images, gene sequencing information, and electronic medical records, the increasing volume and complexity of these data have led to limitations in traditional feature modeling and analysis methods for global feature extraction and inter-feature correlation analysis.

[0003] With the continuous advancement of medical technology and growing clinical needs, medical artificial intelligence models are playing an increasingly important role in medical research and clinical practice. Multimodal medical models can efficiently integrate and fuse diverse medical data resources, including high-definition medical images, in-depth genetic sequencing information, and detailed electronic medical records. These models provide solid and comprehensive data support for accurate disease diagnosis, scientific risk prediction, and personalized treatment plan design. They also strongly support precision medicine and personalized intervention for cardiovascular and cerebrovascular diseases. However, due to technical limitations and disease progression, some preoperative modality data are not fully incorporated into preoperative risk assessment. This can negatively impact the accuracy and reliability of multimodal medical models, potentially leading to biases or deficiencies in the model's prediction of cerebrovascular risk. Furthermore, the high heterogeneity and heterogeneity of collected multimodal data make it difficult to directly characterize pathological conditions. Therefore, optimizing the use of existing preoperative features and indicators without sacrificing model performance has become a key research topic in preoperative cerebrovascular risk assessment. Summary of the Invention

[0004] The present application provides a method, apparatus, device and storage medium for preoperative assessment of cardiovascular and cerebrovascular risks in non-cardiac surgery, aiming to solve at least one of the above-mentioned technical problems in the prior art to a certain extent.

[0005] In order to solve the above problems, this application provides the following technical solutions:

[0006] A method for preoperative assessment of cardiovascular and cerebrovascular risk in non-cardiac surgery, comprising:

[0007] The preoperative knowledge graph enhancement module based on the pre-trained large model is used to link the target patient's electronic medical record document to the knowledge graph, construct the target patient's preoperative knowledge graph, and use the pre-trained large model to perform risk prediction on the pre-operative knowledge graph to obtain a comprehensive quantitative assessment of the risks to be predicted;

[0008] A preoperative feature global interaction module based on feature aggregation graph convolution, a preoperative image feature extraction module based on multi-level low-rank matrix decomposition, and a preoperative time series data feature extraction module based on parallel time-frequency coupling are respectively used to extract features from the target patient's multimodal preoperative data to obtain multimodal data features; wherein, the multimodal preoperative data includes electronic health record data, preoperative medical image data, and preoperative time series data, and the multimodal data features include numerical and type-type preoperative electronic health data features, preoperative image data features, and preoperative time series data features;

[0009] It is beneficial for the preoperative feature fusion module based on cross-modal global cross attention to perform cross-modal interactive processing, global modeling and fusion of the multimodal data features to obtain a complete cardiovascular and cerebrovascular risk feature;

[0010] The cardiovascular and cerebrovascular risk characteristics are input into a cardiovascular and cerebrovascular risk assessment module based on multidimensional reconstruction loss. The cardiovascular and cerebrovascular risk assessment module is guided by the comprehensive quantitative assessment of the risk to be predicted and outputs a preoperative assessment result of the cardiovascular and cerebrovascular risk of non-cardiac surgery for the target patient based on multidimensional reconstruction loss.

[0011] The technical solution adopted by the embodiment of the present application also includes: the preoperative knowledge graph enhancement module based on the pre-trained large model links the target patient's electronic medical record document to the knowledge graph, constructs the target patient's preoperative knowledge graph, and uses the pre-trained large model to perform risk prediction on the preoperative knowledge graph to obtain a comprehensive quantitative assessment of the risk to be predicted, specifically:

[0012] Using a question-guided large language model to extract key information from the electronic medical record document, perform entity recognition on the key information, link each identified entity to a corresponding node in the knowledge graph, and adjust the relevance weight of the corresponding node pair based on entity type and contribution to obtain a preoperative knowledge graph for the target patient;

[0013] The collaborative risk prediction module based on the pre-trained large model is conducive to converting the preoperative knowledge graph into a semi-structured knowledge graph, and inputting the semi-structured knowledge graph into the pre-trained large model, and using the pre-trained large model to quantitatively evaluate the correlation between the risk to be predicted and various related dimensions, so as to obtain a comprehensive quantitative evaluation of the risk to be predicted.

[0014] The technical solution adopted in the embodiment of the present application also includes: the preoperative feature global interaction module based on feature aggregation graph convolution, the preoperative image feature extraction module based on multi-level low-rank matrix decomposition, and the preoperative time series data feature extraction module based on parallel time-frequency coupling respectively extract features from the multimodal preoperative data of the target patient to obtain multimodal data features, including:

[0015] The preoperative feature global interaction module based on feature aggregation graph convolution is conducive to standardizing and preprocessing the electronic health record data, and performing feature aggregation graph representation on the standardized preprocessed electronic health record data to obtain numerical and species-type preoperative electronic health data features; wherein, the feature extraction process of the preoperative feature global interaction module based on feature aggregation graph convolution includes: normalizing the numerical data in the electronic health record data to a normal distribution, and performing one-hot encoding on the species-type data, expressing both the numerical data and the species-type data in numerical form, and combining the numerical forms to generate a data vector x; upgrading the data vector x to a square feature matrix H 0, and construct an initial adjacency matrix in a graph convolution manner A 0 and a degree matrix D 0, so that the numerical data and the type data form a mutually related graph representation, and obtain the final numerical and type preoperative electronic health data features through multi-layer graph convolution.

[0016] The technical solution adopted by the embodiment of the present application also includes: the preoperative feature global interaction module based on feature aggregation graph convolution, the preoperative image feature extraction module based on multi-level low-rank matrix decomposition, and the preoperative time series data feature extraction module based on parallel time-frequency coupling respectively extract features from the multimodal preoperative data of the target patient to obtain multimodal data features, and also includes:

[0017] The preoperative image feature extraction module based on multi-level low-rank matrix decomposition is conducive to performing feature extraction on the preoperative medical image data to obtain preoperative image data features; wherein, the feature extraction process of the preoperative image feature extraction module based on multi-level low-rank matrix decomposition includes: performing preliminary convolution feature extraction on the preoperative medical image data, and then using low-rank constrained multi-level matrix decomposition technology to decompose the preoperative medical image data step by step into a plurality of low-rank represented matrix products to form a new feature space, and projecting the preoperative medical image data into the new feature space, and reconstructing the preoperative medical image data using the projection results and spatial information to obtain the final preoperative image data features.

[0018] The technical solution adopted by the embodiment of the present application also includes: the preoperative feature global interaction module based on feature aggregation graph convolution, the preoperative image feature extraction module based on multi-level low-rank matrix decomposition, and the preoperative time series data feature extraction module based on parallel time-frequency coupling respectively extract features from the multimodal preoperative data of the target patient to obtain multimodal data features, and also includes:

[0019] The preoperative time series data feature extraction module based on parallel time-frequency coupling is used to perform multi-dimensional representation of the preoperative time series data in the time domain and frequency domain, and extracts preoperative time series data features with time-frequency consistency through periodic analysis and main frequency analysis; wherein, the preoperative time series data feature extraction module based on parallel time-frequency coupling includes a time series analysis submodule and a frequency domain analysis submodule, and the time series analysis submodule captures the long-term trend and change law of the preoperative time series data through a polynomial matrix to obtain time domain features; the frequency domain analysis submodule uses frequency domain decomposition to convert the preoperative time series data to the frequency domain, obtains the sequence data with the largest spectrum amplitude through sampling, and converts it back to the time domain to obtain frequency features; finally, the time domain features and frequency domain features are respectively passed through a linear layer and then spliced ​​to obtain the final preoperative time series data features.

[0020] The technical solution adopted by the embodiment of the present application also includes: the preoperative feature fusion module based on cross-modal global cross attention performs cross-modal interactive processing and global modeling and fusion on the multimodal data features to obtain a complete cardiovascular and cerebrovascular risk feature, specifically:

[0021] Taking the electronic health data features, preoperative image data features or preoperative time series data features as the baseline, cross-attention calculations are performed with other modal data features respectively, so that the multimodal data features obtain global interaction and similarity modeling, and then an independent self-attention calculation is performed on the electronic health data features, preoperative image data features and preoperative time series data features respectively, and a complete cardiovascular and cerebrovascular risk feature is obtained through feature alignment.

[0022] The technical solution adopted by the embodiment of the present application further includes: inputting the cardiovascular and cerebrovascular risk characteristics into a cardiovascular and cerebrovascular risk assessment module based on multidimensional reconstruction loss, and the cardiovascular and cerebrovascular risk assessment module based on multidimensional reconstruction loss outputting a preoperative cardiovascular and cerebrovascular risk assessment result for non-cardiac surgery of the target patient guided by the comprehensive quantitative assessment of the risk to be predicted, specifically:

[0023] The cardiovascular risk assessment module based on multi-dimensional reconstruction loss aggregates cardiovascular risk features through a multi-layer perceptron and connects them with an activation function, and outputs a cardiovascular risk prediction probability under the guidance of a comprehensive quantitative assessment of the risk to be predicted; wherein, the multi-dimensional reconstruction loss of the cardiovascular risk assessment module based on multi-dimensional reconstruction loss includes classification cross entropy loss, low-rank matrix decomposition loss, and time-frequency domain coupling similarity loss, the classification cross entropy loss is used to measure the difference between the cardiovascular risk prediction probability output by the model and the true expert prior knowledge, the low-rank matrix decomposition loss is used to ensure that the model can extract the supplementary information implicit in the preoperative image data features, and the time-frequency domain coupling similarity loss is used to constrain and optimize the model during the back-propagation process.

[0024] Another technical solution adopted in the embodiment of the present application is: a preoperative cardiovascular and cerebrovascular risk assessment device for non-cardiac surgery, comprising:

[0025] Preoperative knowledge graph enhancement module: used to link the target patient's electronic medical record documents to the knowledge graph, construct the target patient's preoperative knowledge graph, and use the pre-trained large model to perform risk prediction on the preoperative knowledge graph to obtain a comprehensive quantitative assessment of the risks to be predicted;

[0026] Multimodal feature extraction module: used to respectively facilitate a preoperative feature global interaction module based on feature aggregation graph convolution, a preoperative image feature extraction module based on multi-level low-rank matrix decomposition, and a preoperative time series data feature extraction module based on parallel time-frequency coupling to extract features from the target patient's multimodal preoperative data to obtain multimodal data features; wherein the multimodal preoperative data includes electronic health record data, preoperative medical image data, and preoperative time series data, and the multimodal data features include numerical and type-type preoperative electronic health data features, preoperative image data features, and preoperative time series data features;

[0027] Preoperative feature fusion module: used to use cross-modal global cross-attention to perform cross-modal interactive processing, global modeling and fusion of the multimodal data features to obtain a complete cardiovascular and cerebrovascular risk feature;

[0028] Cardiovascular risk assessment module: used to receive input cardiovascular risk characteristics, and guided by the comprehensive quantitative assessment of the risk to be predicted, use multidimensional reconstruction loss to output the preoperative assessment results of cardiovascular risk for non-cardiac surgery of the target patient.

[0029] Another technical solution adopted by the embodiment of the present application is: a device, the device comprising a processor and a memory coupled to the processor, wherein:

[0030] The memory stores program instructions for implementing the method for preoperative assessment of cardiovascular and cerebrovascular risks of non-cardiac surgery;

[0031] The processor is configured to execute the program instructions stored in the memory to control a preoperative cardiovascular and cerebrovascular risk assessment method for non-cardiac surgery.

[0032] Another technical solution adopted in the embodiment of the present application is: a storage medium storing program instructions executable by a processor, wherein the program instructions are used to execute the preoperative assessment method for cardiovascular and cerebrovascular risks of non-cardiac surgery.

[0033] Compared with the prior art, the beneficial effects of the embodiments of the present application are as follows: the preoperative assessment method, device, equipment and storage medium of the cardiovascular and cerebrovascular risk of non-cardiac surgery in the embodiments of the present application use a parallel multimodal feature extraction framework as the network architecture, and introduce pre-trained large models, knowledge graphs, graph convolutional neural networks, low-rank matrix decomposition and other methods to perform multi-dimensional processing of multimodal preoperative data of non-cardiac surgery to obtain multimodal data features, which is conducive to cross-modal global cross-attention to perform cross-modal interactive processing and global modeling and fusion of multimodal data features to obtain complete cardiovascular and cerebrovascular risk features, and the cardiovascular and cerebrovascular risk assessment module based on multi-dimensional reconstruction loss uses prior knowledge as guidance to perform preoperative assessment of cardiovascular and cerebrovascular risk of non-cardiac surgery, which not only improves the efficiency and accuracy of preoperative assessment of cardiovascular and cerebrovascular risk of non-cardiac surgery, but also optimizes the extraction and fusion process of multimodal data features, improves the model's preoperative assessment ability and generalization ability of cardiovascular and cerebrovascular risk, and has good interpretability. In addition, the embodiments of the present application also help optimize the allocation of medical resources, improve medical efficiency, reduce medical costs, and provide important references for clinical research such as exploring the causes and mechanisms of cardiovascular and cerebrovascular events. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 This is a flow chart of a method for preoperative assessment of cardiovascular and cerebrovascular risks in non-cardiac surgery according to the first embodiment of the present application;

[0035] Figure 2 This is a flow chart of a method for preoperative assessment of cardiovascular and cerebrovascular risks in non-cardiac surgery according to a second embodiment of the present application;

[0036] Figure 3 Schematic diagram of the knowledge graph structure of the electronic medical record document constructed by the preoperative knowledge graph enhancement module based on the pre-trained large model in an embodiment of the present application;

[0037] Figure 4 Schematic diagram of the cross-modal global cross-attention architecture of the preoperative feature fusion module based on cross-modal global cross-attention in an embodiment of the present application;

[0038] Figure 5 This is a schematic structural diagram of a preoperative cardiovascular and cerebrovascular risk assessment device for non-cardiac surgery according to an embodiment of the present application;

[0039] Figure 6 This is a schematic diagram of the device structure of an embodiment of the present application;

[0040] Figure 7 A schematic diagram of the structure of the storage medium of an embodiment of the present application. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0042] The terms "first," "second," and "third" in this application are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features identified. Therefore, features identified as "first," "second," or "third" may explicitly or implicitly include at least one of such features. In the description of this application, "plurality" means at least two, for example, two, three, etc., unless otherwise specifically defined. All directional designations in the embodiments of this application (such as up, down, left, right, front, back, etc.) are intended only to illustrate the relative positional relationships and movement of components in a specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional designations will also change accordingly. Furthermore, the terms "including," "having," and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or apparatus comprising a series of steps or elements is not limited to the listed steps or elements but may optionally include steps or elements not listed, or may optionally include other steps or elements inherent to such process, method, product, or apparatus.

[0043] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0044] Specifically, see Figure 1 , is a flow chart of the method for preoperative assessment of cardiovascular and cerebrovascular risk of non-cardiac surgery according to the first embodiment of the present application. The method for preoperative assessment of cardiovascular and cerebrovascular risk of non-cardiac surgery according to the first embodiment of the present application comprises the following steps:

[0045] S100: The preoperative knowledge graph enhancement module based on the pre-trained large model links the target patient's electronic medical record document to the knowledge graph, constructs the target patient's preoperative knowledge graph, and uses the pre-trained large model to perform risk prediction on the pre-operative knowledge graph to obtain a comprehensive quantitative assessment of the risks to be predicted;

[0046] S110: extracting features from the target patient's multimodal preoperative data using a preoperative feature global interaction module based on feature aggregation graph convolution, a preoperative image feature extraction module based on multi-level low-rank matrix decomposition, and a preoperative time series data feature extraction module based on parallel time-frequency coupling, respectively, to obtain multimodal data features; wherein the multimodal preoperative data includes electronic health record data, preoperative medical image data, and preoperative time series data, and the extracted multimodal data features include numerical and type-type preoperative electronic health data features, preoperative image data features, and preoperative time series data features;

[0047] S120: A preoperative feature fusion module based on cross-modal global cross-attention facilitates cross-modal interactive processing, global modeling, and fusion of multimodal data features to obtain a complete cardiovascular and cerebrovascular risk profile;

[0048] S130: The cardiovascular and cerebrovascular risk characteristics are input into a cardiovascular and cerebrovascular risk assessment module based on multidimensional reconstruction loss. The cardiovascular and cerebrovascular risk assessment module is guided by the comprehensive quantitative assessment of the risk to be predicted and outputs a preoperative assessment result of the cardiovascular and cerebrovascular risk of non-cardiac surgery for the target patient based on the multidimensional reconstruction loss.

[0049] See also Figure 2 , is a flow chart of a method for preoperative assessment of cardiovascular and cerebrovascular risk of non-cardiac surgery according to the second embodiment of the present application. The method for preoperative assessment of cardiovascular and cerebrovascular risk of non-cardiac surgery according to the second embodiment of the present application comprises the following steps:

[0050] S200: Acquire multimodal preoperative data of the target patient, including electronic medical record documents, electronic health record data, preoperative medical image data, and preoperative time series data;

[0051] In this step, the electronic medical record documents include but are not limited to data such as the target patient's symptoms, medical history, medication details, and previous medical records. The electronic health record data include but are not limited to data such as the target patient's age, BMI (Body Mass Index), gender, blood type, and medical history. The preoperative medical image data include but are not limited to data such as the target patient's X-rays. The preoperative time series data include but are not limited to data such as the target patient's electrocardiograms.

[0052] S210: The preoperative knowledge graph enhancement module based on the pre-trained large model extracts key information from the electronic medical record documents and links the key information to the knowledge graph to construct the preoperative knowledge graph of the target patient;

[0053] In this step, the construction process of the preoperative knowledge graph is as follows: first, a series of questions are used to guide the large language model (LLMs) to extract key information from the electronic medical record document, and then the key information is identified as an entity, and each identified entity is linked to the corresponding node in the knowledge graph (KGs), and the corresponding node pair relevance weight is adjusted according to the entity type and contribution, so as to obtain the global relevance representation of the electronic medical record document, thereby constructing the preoperative knowledge graph of the target patient, which provides strong support for the subsequent cardiovascular and cerebrovascular risk assessment. Figure 3 As shown, this is a schematic diagram of the knowledge graph structure of the electronic medical record document constructed by the preoperative knowledge graph enhancement module based on the pre-trained large model in an embodiment of the present application.

[0054] It can be understood that before using the pre-trained large model, the embodiment of the present application is conducive to constructing a pre-operative knowledge graph suitable for the input of the pre-trained large model based on the pre-operative knowledge graph enhancement module of the pre-trained large model, which can better adapt to the input requirements of the pre-trained large model and improve the efficiency and accuracy of model evaluation.

[0055] S220: Facilitating the collaborative risk prediction module based on the pre-trained large model to deeply reconstruct the pre-operative knowledge graph, converting it into a semi-structured knowledge graph, and inputting the semi-structured knowledge graph into the pre-trained large model. The pre-trained large model is used to quantitatively evaluate the correlation between the risk to be predicted and various related dimensions, thereby obtaining a comprehensive quantitative evaluation of the risk to be predicted;

[0056] In this step, the pre-operative knowledge graph is deeply reconstructed through a collaborative risk prediction module based on the pre-trained large model. First, it is converted into a more flexible and easy-to-process semi-structured knowledge graph to better adapt to the input requirements of the pre-trained large model. Then, using prompt templates, the pre-trained large model is guided to conduct a refined quantitative assessment of the correlation between the predicted risk and various relevant dimensions. This results in a comprehensive quantitative assessment of the predicted risk, laying a solid foundation for subsequent cardiovascular and cerebrovascular risk assessment and improving the accuracy of the model assessment.

[0057] S230: Facilitating a preoperative feature global interaction module based on feature aggregation graph convolution to perform standardized preprocessing on the electronic health record data, and performing feature aggregation graph representation on the standardized preprocessed electronic health record data to obtain numerical and typological preoperative electronic health data features;

[0058] In this step, the feature extraction process of the preoperative feature global interaction module based on feature aggregation graph convolution specifically includes: first, the numerical data such as age and BMI in the electronic health record data are normalized, and then the species data such as gender and blood type are one-hot encoded, and the numerical data and species data are expressed in numerical form and combined to generate a data vector x; the data vector x is dimensioned into a square feature matrix H 0, and construct an initial adjacency matrix in a graph convolution manner A 0 and a degree matrix D 0, so that the numerical data and the type of data form a mutually related graph representation, and the final numerical and type of preoperative electronic health data features are obtained through multi-layer graph convolution. The specific processing process can be expressed as the following formula:

[0059] (1)

[0060] in H i For the i Layer feature matrix, A i For the i layer adjacency matrix, D i For the i Degree matrix, U i , V i , W i For the i The three weight matrices of the layer, is the sigmoid activation function, H n is the feature matrix of the last layer, is the feature vector of the final output of the module.

[0061] It can be understood that the embodiment of the present application uses the idea of ​​graph convolution to construct the initialization adjacency matrix and degree matrix, so that different data types form interrelated graph representations, and obtain the final numerical and species-type preoperative electronic health data features through multi-layer graph convolution, which can fully explore the implicit feature correlations in electronic health record data, realize global correlation modeling of electronic health record data, provide richer and more reliable feature representations for subsequent feature fusion, and improve the accuracy of preoperative risk assessment.

[0062] S240: facilitating a preoperative image feature extraction module based on multi-level low-rank matrix decomposition to extract features from the preoperative medical image data to obtain preoperative image data features;

[0063] In this step, the process of extracting features from preoperative image data specifically includes: first, performing preliminary convolution feature extraction on the preoperative medical image data, then using a low-rank constrained multi-level matrix decomposition technique to decompose the preoperative medical image data step by step into multiple low-rank matrix products to form a new feature space, and then projecting the preoperative medical image data into the new feature space. The preoperative medical image data is reconstructed using the projection results and spatial information to obtain the final preoperative image data features. The low-rank constraint formula is expressed as follows:

[0064] (2)

[0065] in, X Input image feature matrix to the module, M i For the i The feature matrix after layer decomposition, Z i For the i The weight matrix of the layer matrix decomposition, X l For the l The feature matrix decomposed by the layer (that is, the last layer) is represents the matrix infinity norm, Representation matrix F norm.

[0066] It can be understood that the embodiment of the present application adds low-rank constraints to the preoperative image feature extraction module based on multi-level low-rank matrix decomposition, which can efficiently extract the spatial semantic features of preoperative medical image data, effectively improve the feature extraction efficiency of preoperative medical image data, and ensure the stability and convergence of the model during the training process, providing support for subsequent multimodal feature fusion and preoperative risk assessment.

[0067] S250: using a preoperative time series data feature extraction module based on parallel time-frequency coupling to perform multi-dimensional representation of the preoperative time series data in the time domain and frequency domain, and extracting preoperative time series data features with time-frequency consistency through periodicity analysis and dominant frequency analysis;

[0068] In this step, the preoperative time series data feature extraction module based on parallel time-frequency coupling not only fully considers the periodic characteristics contained in the preoperative time series data, but also deeply explores and reveals the potential frequency domain correlation characteristics behind the preoperative time series data. It reflects the changing laws and trends of medical data through periodic characteristics, and reflects the dynamic characteristics and internal mechanisms of preoperative time series data through frequency domain correlation characteristics. It can enhance the feature extraction efficiency of preoperative time series data and provide a more solid and reliable data foundation for preoperative risk prediction.

[0069] Furthermore, the preoperative time series data feature extraction module based on parallel time-frequency coupling uses a dual-domain coupled parallel structure to process time domain information and frequency domain information in parallel, achieving a comprehensive and in-depth analysis of preoperative time series data, greatly improving the accuracy and efficiency of feature extraction. Specifically, the preoperative time series data feature extraction module based on parallel time-frequency coupling includes a time series analysis submodule and a frequency domain analysis submodule, wherein the time series analysis submodule uses a polynomial matrix to model the inherent overall trend of the preoperative time series data, captures the long-term trend and change law of the preoperative time series data, and obtains time domain features; the frequency domain analysis submodule uses frequency domain decomposition to convert the preoperative time series data to the frequency domain, obtains the sequence data with the largest spectrum amplitude through sampling, and converts it back to the time domain to obtain frequency domain features; finally, the obtained time domain features and frequency domain features are respectively passed through a linear layer and then spliced ​​to obtain the final preoperative time series data features. Specifically, the feature extraction process of the preoperative time series data feature extraction module based on parallel time-frequency coupling can be expressed as the following formula:

[0070]

[0071] (3)

[0072] in i Representative i A parallel encoder, Q Represents a polynomial matrix, argTopK represents the top K maximum values K i (1) arrive K i (K) , Linear represents the linear layer, concat represents concatenation and alignment in dimension, FFT represents fast Fourier transform, and IFFT represents inverse fast Fourier transform. Ai (k) Representative i The kth frequency component is obtained after FFT of the input of the encoder, M is the aggregated time domain feature vector, and S is the aggregated frequency domain feature vector.

[0073] S260: A preoperative feature fusion module based on cross-modal global cross-attention facilitates cross-modal interactive processing, global modeling, and fusion of numerical and typological preoperative electronic health data features, preoperative image data features, and preoperative time series data features to obtain a complete cardiovascular and cerebrovascular risk profile.

[0074] In this step, in order to ensure that the multimodal data features such as numerical and species-type preoperative electronic health data features, preoperative image data features, and preoperative time series data features extracted in parallel can maintain a high degree of consistency in semantic representation and can be integrated into independent and complete cardiovascular risk features, this application introduces a preoperative feature fusion module based on cross-modal global cross-attention for cross-modal interactive processing and global modeling and fusion, and uses a multi-dimensional complementary strategy to model the similarity and consistency between different modal features, so that each modal data feature can fully consider the information of other modal data features to strengthen different modalities. The representation commonality between data features allows for information exchange and integration on a global scale, resulting in more representative and discriminative cardiovascular and cerebrovascular risk features. This enables the model to strengthen the semantic consistency of multimodal data features, helping to overcome the fusion difficulties brought about by the high heterogeneity of preoperative data. It not only deepens the model's ability to understand complex multimodal preoperative data, but also significantly improves the model's interpretability and generalization capabilities, further improving the accuracy and reliability of preoperative assessments of cardiovascular and cerebrovascular risks in non-cardiac surgery, and providing clinicians with more scientific and powerful decision-making support. Specifically, Figure 4 As shown, this is an overall schematic diagram of the cross-modal global cross-attention architecture of the preoperative feature fusion module based on cross-modal global cross-attention in an embodiment of the present application. Specifically, taking the electronic health data feature X, the preoperative image data feature Z or the preoperative time series data feature Y as the baseline, cross-attention calculations are performed with the other modal data features respectively, so that the data features of the three modalities are fully globally interacted and similarity modeled, and then an independent self-attention calculation is performed on the electronic health data feature, the preoperative image data feature and the preoperative time series data feature respectively, and a complete cardiovascular and cerebrovascular risk feature is obtained by feature alignment. The cross-attention and self-attention calculation formulas are as follows:

[0075] (4)

[0076] in q represents the query vector, k represents the key vector,d k represent k Dimensions, v represents the value vector and softmax represents the activation function.

[0077] S270: Inputting the cardiovascular and cerebrovascular risk characteristics into a cardiovascular and cerebrovascular risk assessment module based on multidimensional reconstruction loss. The cardiovascular and cerebrovascular risk assessment module based on multidimensional reconstruction loss outputs a preoperative cardiovascular and cerebrovascular risk assessment result for a non-cardiac surgery of the target patient, guided by a comprehensive quantitative assessment of the risk to be predicted.

[0078] In this step, the multi-dimensional reconstruction loss-based cardiovascular risk assessment module utilizes a perceptron structure containing multiple layers of neurons. It receives cardiovascular risk features output by the preoperative feature fusion module based on cross-modal global cross-attention, and uses the comprehensive quantitative assessment generated by the collaborative risk prediction module based on the pre-trained large model to perform cardiovascular risk assessment. Specifically, the multi-dimensional reconstruction loss-based cardiovascular risk assessment module aggregates cardiovascular risk features through a multi-layer perceptron and connects them with an activation function. Guided by the comprehensive quantitative assessment of the risk to be predicted, it outputs a predicted cardiovascular risk probability and a probability value between 0 and 1, representing the risk of a cardiovascular event. Among them, the multidimensional reconstruction loss of the cardiovascular risk assessment module based on multidimensional reconstruction loss consists of three parts: classification cross entropy loss, low-rank matrix decomposition loss, and time-frequency domain coupling similarity loss. The classification cross entropy loss is used to measure the difference between the cardiovascular risk prediction probability output by the model and the real expert prior knowledge. The low-rank matrix decomposition loss is used to ensure that the model can efficiently extract implicit supplementary information when processing preoperative image data features, improve model stability and accelerate model convergence; the time-frequency domain coupling similarity loss is used to further constrain and optimize the model during the backpropagation process to ensure the stability and consistency of the model during the reverse derivation process. Through the organic combination of these three, not only is the prior knowledge of the pre-trained large model fully utilized to guide model training, but the model is also guided to continuously converge during the training process through multi-dimensional loss optimization, ultimately achieving an accurate assessment of cardiovascular risk, and further improving the model's preoperative assessment ability and generalization ability for cardiovascular risk. Specifically, the multidimensional reconstruction loss calculation formula is as follows:

[0079] (5)

[0080] in is the time-frequency domain coupling similarity loss.

[0081] Based on the above, the preoperative assessment method of cardiovascular and cerebrovascular risk of non-cardiac surgery in the embodiment of the present application uses a parallel multimodal feature extraction framework as the network architecture, and introduces pre-trained large models, knowledge graphs, graph convolutional neural networks, low-rank matrix decomposition and other methods to perform multi-dimensional processing on the multimodal preoperative data of non-cardiac surgery to obtain multimodal data features, and facilitate cross-modal global cross-attention to perform cross-modal interactive processing and global modeling and fusion of multimodal data features to obtain complete cardiovascular and cerebrovascular risk features, and a cardiovascular and cerebrovascular risk assessment module based on multi-dimensional reconstruction loss uses prior knowledge as a guide to perform preoperative assessment of cardiovascular and cerebrovascular risk of non-cardiac surgery. The embodiment of the present application integrates deep learning with prior knowledge of cardiovascular and cerebrovascular risk in the medical field, and constructs a stable and reliable cardiovascular and cerebrovascular risk feature through multiple carefully designed network modules, which not only improves the efficiency and accuracy of preoperative assessment of cardiovascular and cerebrovascular risk of non-cardiac surgery, but also optimizes the extraction and fusion process of multimodal data features, improves the model's preoperative assessment ability and generalization ability of cardiovascular and cerebrovascular risk, and has good interpretability. In addition, the embodiments of the present application also help optimize the allocation of medical resources, improve medical efficiency, reduce medical costs, and provide important references for clinical research such as exploring the causes and mechanisms of cardiovascular and cerebrovascular events.

[0082] See also Figure 5 , is a schematic diagram of the structure of a preoperative cardiovascular and cerebrovascular risk assessment device for non-cardiac surgery according to an embodiment of the present application. The preoperative cardiovascular and cerebrovascular risk assessment method device 40 for non-cardiac surgery according to an embodiment of the present application comprises:

[0083] Preoperative knowledge graph enhancement module 41: used to link the target patient's electronic medical record document to the knowledge graph, construct the target patient's preoperative knowledge graph, and use the pre-trained large model to perform risk prediction on the preoperative knowledge graph to obtain a comprehensive quantitative assessment of the risks to be predicted;

[0084] Multimodal feature extraction module 42: used to respectively facilitate a preoperative feature global interaction module based on feature aggregation graph convolution, a preoperative image feature extraction module based on multi-level low-rank matrix decomposition, and a preoperative time series data feature extraction module based on parallel time-frequency coupling to extract features from the target patient's multimodal preoperative data to obtain multimodal data features; wherein the multimodal preoperative data includes electronic health record data, preoperative medical image data, and preoperative time series data, and the multimodal data features include numerical and type-type preoperative electronic health data features, preoperative image data features, and preoperative time series data features;

[0085] Preoperative feature fusion module 43: used to perform cross-modal interactive processing, global modeling and fusion of the multimodal data features using cross-modal global cross attention to obtain a complete cardiovascular and cerebrovascular risk feature;

[0086] The cardiovascular risk assessment module 44 is used to receive the input cardiovascular risk characteristics, and use the comprehensive quantitative assessment of the risk to be predicted as a guide to output the preoperative cardiovascular risk assessment results of the target patient's non-cardiac surgery using multidimensional reconstruction loss.

[0087] It should be noted that the information interaction, execution process, etc. between the above-mentioned devices / units are based on the same concept as the method embodiment of this application. Their specific functions and technical effects can be found in the method embodiment section and will not be repeated here.

[0088] The device provided in the embodiment of the present application can be applied in the aforementioned method embodiment. For details, please refer to the description of the aforementioned method embodiment, which will not be repeated here.

[0089] See also Figure 6 , is a schematic diagram of the device structure of an embodiment of the present application. The device 50 includes:

[0090] A memory 51 storing executable program instructions;

[0091] a processor 52 connected to the memory 51;

[0092] The processor 52 is used to call the executable program instructions stored in the memory 51 and perform the following steps: facilitate the preoperative knowledge graph enhancement module based on the pre-trained large model to link the target patient's electronic medical record document to the knowledge graph, construct the target patient's preoperative knowledge graph, and use the pre-trained large model to predict the risk of the preoperative knowledge graph to obtain a comprehensive quantitative assessment of the risk to be predicted; facilitate the preoperative feature global interaction module based on feature aggregation graph convolution, the preoperative image feature extraction module based on multi-level low-rank matrix decomposition, and the preoperative time series data feature extraction module based on parallel time-frequency coupling to extract features from the target patient's multimodal preoperative data to obtain multimodal data features; wherein, the multimodal preoperative data packet The multimodal data include electronic health record data, preoperative medical image data and preoperative time series data, and the multimodal data features include numerical and type preoperative electronic health data features, preoperative image data features and preoperative time series data features; it is conducive to the preoperative feature fusion module based on cross-modal global cross-attention to perform cross-modal interactive processing and global modeling and fusion of the multimodal data features to obtain complete cardiovascular and cerebrovascular risk features; the cardiovascular and cerebrovascular risk features are input into the cardiovascular and cerebrovascular risk assessment module based on multidimensional reconstruction loss, and the cardiovascular and cerebrovascular risk assessment module is guided by the comprehensive quantitative assessment of the risk to be predicted, and outputs the preoperative assessment results of the cardiovascular and cerebrovascular risk of non-cardiac surgery for the target patient based on multidimensional reconstruction loss.

[0093] The processor 52 may also be referred to as a CPU (Central Processing Unit). The processor 52 may be an integrated circuit chip with signal processing capabilities. The processor 52 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component. The general-purpose processor may be a microprocessor or any conventional processor.

[0094] See also Figure 7, which is a structural diagram of the storage medium of the embodiment of the present application. The storage medium of the embodiment of the present application stores program instructions 61 that can implement the following steps: facilitating the preoperative knowledge graph enhancement module based on the pre-trained large model to link the target patient's electronic medical record document to the knowledge graph, constructing the target patient's preoperative knowledge graph, and using the pre-trained large model to perform risk prediction on the preoperative knowledge graph to obtain a comprehensive quantitative assessment of the risk to be predicted; facilitating the preoperative feature global interaction module based on feature aggregation graph convolution, the preoperative image feature extraction module based on multi-level low-rank matrix decomposition, and the preoperative time series data feature extraction module based on parallel time-frequency coupling to extract features from the target patient's multimodal preoperative data to obtain multimodal data features; wherein, the multimodal preoperative data includes Electronic health record data, preoperative medical image data, and preoperative time series data, wherein the multimodal data features include numerical and typological preoperative electronic health data features, preoperative image data features, and preoperative time series data features; a preoperative feature fusion module based on cross-modal global cross-attention is used to perform cross-modal interactive processing, global modeling, and fusion on the multimodal data features to obtain a complete cardiovascular and cerebrovascular risk feature; the cardiovascular and cerebrovascular risk feature is input into a cardiovascular and cerebrovascular risk assessment module based on multidimensional reconstruction loss, the cardiovascular and cerebrovascular risk assessment module is guided by the comprehensive quantitative assessment of the risk to be predicted, and outputs a preoperative cardiovascular and cerebrovascular risk assessment result for the target patient's non-cardiac surgery based on multidimensional reconstruction loss. The program instructions 61 can be stored in the above-mentioned storage medium in the form of a software product, including a number of instructions for causing a device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the method of each embodiment of the present application. The aforementioned storage media include: USB flash drives, mobile hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks or optical disks, and other media that can store program instructions, or terminal devices such as computers, servers, mobile phones, and tablets. The server can be a standalone server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0095] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0096] In addition, each functional unit in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional unit. The above is only an implementation method of the present application and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the content of the description and drawings of this application, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for preoperative assessment of cardiovascular and cerebrovascular risk in non-cardiac surgery, characterized in that: include: The preoperative knowledge graph enhancement module based on the pre-trained large model is used to link the target patient's electronic medical record document to the knowledge graph, construct the target patient's preoperative knowledge graph, and use the pre-trained large model to perform risk prediction on the pre-operative knowledge graph to obtain a comprehensive quantitative assessment of the risks to be predicted; A preoperative feature global interaction module based on feature aggregation graph convolution, a preoperative image feature extraction module based on multi-level low-rank matrix decomposition, and a preoperative time series data feature extraction module based on parallel time-frequency coupling are respectively used to extract features from the target patient's multimodal preoperative data to obtain multimodal data features; wherein, the multimodal preoperative data includes electronic health record data, preoperative medical image data, and preoperative time series data, and the multimodal data features include numerical and type-type preoperative electronic health data features, preoperative image data features, and preoperative time series data features; It is beneficial for the preoperative feature fusion module based on cross-modal global cross attention to perform cross-modal interactive processing, global modeling and fusion of the multimodal data features to obtain a complete cardiovascular and cerebrovascular risk feature; Inputting the cardiovascular and cerebrovascular risk characteristics into a cardiovascular and cerebrovascular risk assessment module based on multidimensional reconstruction loss, the cardiovascular and cerebrovascular risk assessment module being guided by the comprehensive quantitative assessment of the risk to be predicted and outputting a preoperative cardiovascular and cerebrovascular risk assessment result for non-cardiac surgery of the target patient based on the multidimensional reconstruction loss; The preoperative knowledge graph enhancement module based on the pre-trained large model links the target patient's electronic medical record document to the knowledge graph, constructs the target patient's preoperative knowledge graph, and uses the pre-trained large model to perform risk prediction on the preoperative knowledge graph to obtain a comprehensive quantitative assessment of the risk to be predicted, specifically: Using a question-guided large language model to extract key information from the electronic medical record document, perform entity recognition on the key information, link each identified entity to a corresponding node in the knowledge graph, and adjust the relevance weight of the corresponding node pair based on entity type and contribution to obtain a preoperative knowledge graph for the target patient; The collaborative risk prediction module based on the pre-trained large model is conducive to converting the preoperative knowledge graph into a semi-structured knowledge graph, and inputting the semi-structured knowledge graph into the pre-trained large model, and using the pre-trained large model to quantitatively evaluate the correlation between the risk to be predicted and various related dimensions, so as to obtain a comprehensive quantitative evaluation of the risk to be predicted.

2. The method for preoperative assessment of cardiovascular and cerebrovascular risk in non-cardiac surgery according to claim 1, characterized in that: The preoperative feature global interaction module based on feature aggregation graph convolution, the preoperative image feature extraction module based on multi-level low-rank matrix decomposition, and the preoperative time series data feature extraction module based on parallel time-frequency coupling are respectively used to extract features from the multimodal preoperative data of the target patient to obtain multimodal data features, including: The preoperative feature global interaction module based on feature aggregation graph convolution is conducive to standardizing and preprocessing the electronic health record data, and performing feature aggregation graph representation on the standardized preprocessed electronic health record data to obtain numerical and species-type preoperative electronic health data features; wherein, the feature extraction process of the preoperative feature global interaction module based on feature aggregation graph convolution includes: normalizing the numerical data in the electronic health record data to a normal distribution, and performing one-hot encoding on the species-type data, expressing both the numerical data and the species-type data in numerical form, and combining the numerical forms to generate a data vector x; upgrading the data vector x to a square feature matrix H 0, and construct an initial adjacency matrix in a graph convolution manner A 0 and a degree matrix D 0, so that the numerical data and the type data form a mutually related graph representation, and obtain the final numerical and type preoperative electronic health data features through multi-layer graph convolution.

3. The method for preoperative assessment of cardiovascular and cerebrovascular risk in non-cardiac surgery according to claim 2, characterized in that: The method further comprises: extracting features from the multimodal preoperative data of the target patient by a preoperative feature global interaction module based on feature aggregation graph convolution, a preoperative image feature extraction module based on multi-level low-rank matrix decomposition, and a preoperative time series data feature extraction module based on parallel time-frequency coupling to obtain multimodal data features; and The preoperative image feature extraction module based on multi-level low-rank matrix decomposition is conducive to performing feature extraction on the preoperative medical image data to obtain preoperative image data features; wherein, the feature extraction process of the preoperative image feature extraction module based on multi-level low-rank matrix decomposition includes: performing preliminary convolution feature extraction on the preoperative medical image data, and then using low-rank constrained multi-level matrix decomposition technology to decompose the preoperative medical image data step by step into a plurality of low-rank represented matrix products to form a new feature space, and projecting the preoperative medical image data into the new feature space, and reconstructing the preoperative medical image data using the projection results and spatial information to obtain the final preoperative image data features.

4. The method for preoperative assessment of cardiovascular and cerebrovascular risk in non-cardiac surgery according to claim 3, characterized in that: The method further comprises: extracting features from the multimodal preoperative data of the target patient by a preoperative feature global interaction module based on feature aggregation graph convolution, a preoperative image feature extraction module based on multi-level low-rank matrix decomposition, and a preoperative time series data feature extraction module based on parallel time-frequency coupling to obtain multimodal data features; and The preoperative time series data feature extraction module based on parallel time-frequency coupling is used to perform multi-dimensional representation of the preoperative time series data in the time domain and frequency domain, and extracts preoperative time series data features with time-frequency consistency through periodic analysis and main frequency analysis; wherein, the preoperative time series data feature extraction module based on parallel time-frequency coupling includes a time series analysis submodule and a frequency domain analysis submodule, and the time series analysis submodule captures the long-term trend and change law of the preoperative time series data through a polynomial matrix to obtain time domain features; the frequency domain analysis submodule uses frequency domain decomposition to convert the preoperative time series data to the frequency domain, obtains the sequence data with the largest spectrum amplitude through sampling, and converts it back to the time domain to obtain frequency features; finally, the time domain features and frequency domain features are respectively passed through a linear layer and then spliced ​​to obtain the final preoperative time series data features.

5. The method for preoperative assessment of cardiovascular and cerebrovascular risk in non-cardiac surgery according to any one of claims 1 to 4, characterized in that: The preoperative feature fusion module based on cross-modal global cross-attention performs cross-modal interactive processing, global modeling, and fusion on the multimodal data features to obtain a complete cardiovascular and cerebrovascular risk feature, specifically: Taking the electronic health data features, preoperative image data features or preoperative time series data features as the baseline, cross-attention calculations are performed with other modal data features respectively, so that the multimodal data features obtain global interaction and similarity modeling, and then an independent self-attention calculation is performed on the electronic health data features, preoperative image data features and preoperative time series data features respectively, and a complete cardiovascular and cerebrovascular risk feature is obtained through feature alignment.

6. The method for preoperative assessment of cardiovascular and cerebrovascular risk in non-cardiac surgery according to claim 5, characterized in that: The cardiovascular and cerebrovascular risk characteristics are input into a cardiovascular and cerebrovascular risk assessment module based on multidimensional reconstruction loss. The cardiovascular and cerebrovascular risk assessment module based on multidimensional reconstruction loss outputs a preoperative cardiovascular and cerebrovascular risk assessment result for a target patient undergoing non-cardiac surgery, guided by a comprehensive quantitative assessment of the risk to be predicted, specifically: The cardiovascular risk assessment module based on multi-dimensional reconstruction loss aggregates cardiovascular risk features through a multi-layer perceptron and connects them with an activation function, and outputs a cardiovascular risk prediction probability under the guidance of a comprehensive quantitative assessment of the risk to be predicted; wherein, the multi-dimensional reconstruction loss of the cardiovascular risk assessment module based on multi-dimensional reconstruction loss includes classification cross entropy loss, low-rank matrix decomposition loss, and time-frequency domain coupling similarity loss, the classification cross entropy loss is used to measure the difference between the cardiovascular risk prediction probability output by the model and the true expert prior knowledge, the low-rank matrix decomposition loss is used to ensure that the model can extract the supplementary information implicit in the preoperative image data features, and the time-frequency domain coupling similarity loss is used to constrain and optimize the model during the back-propagation process.

7. A preoperative assessment device for cardiovascular and cerebrovascular risk in non-cardiac surgery, characterized in that: include: Preoperative knowledge graph enhancement module: used to link the target patient's electronic medical record documents to the knowledge graph, construct the target patient's preoperative knowledge graph, and use the pre-trained large model to perform risk prediction on the preoperative knowledge graph to obtain a comprehensive quantitative assessment of the risks to be predicted; Multimodal feature extraction module: used to respectively facilitate a preoperative feature global interaction module based on feature aggregation graph convolution, a preoperative image feature extraction module based on multi-level low-rank matrix decomposition, and a preoperative time series data feature extraction module based on parallel time-frequency coupling to extract features from the target patient's multimodal preoperative data to obtain multimodal data features; wherein the multimodal preoperative data includes electronic health record data, preoperative medical image data, and preoperative time series data, and the multimodal data features include numerical and type-type preoperative electronic health data features, preoperative image data features, and preoperative time series data features; Preoperative feature fusion module: used to use cross-modal global cross-attention to perform cross-modal interactive processing, global modeling and fusion of the multimodal data features to obtain a complete cardiovascular and cerebrovascular risk feature; Cardiovascular risk assessment module: used to receive input cardiovascular risk characteristics, and guided by the comprehensive quantitative assessment of the risk to be predicted, use multidimensional reconstruction loss to output the preoperative assessment results of cardiovascular risk for non-cardiac surgery of the target patient.

8. An electronic device, characterized in that: The device includes a processor and a memory coupled to the processor, wherein: The memory stores program instructions for implementing the method for preoperative assessment of cardiovascular and cerebrovascular risks in non-cardiac surgery according to any one of claims 1 to 6; The processor is configured to execute the program instructions stored in the memory to control a preoperative cardiovascular and cerebrovascular risk assessment method for non-cardiac surgery.

9. A storage medium, characterized in that: Program instructions executable by a processor are stored, and the program instructions are used to execute the preoperative assessment method for cardiovascular and cerebrovascular risks of non-cardiac surgery as described in any one of claims 1 to 6.

Images