Multi-modal AI diagnosis and risk grading system, method and device suitable for vascular diseases, medium and program product

Abstract

The invention provides a multi-modal AI diagnosis and risk grading system, method and device suitable for vascular diseases, a medium and a program product, and provides an innovative method for cardiovascular diagnosis of old people by combining hemodynamics and structural indexes and utilizing a multi-modal deep learning framework. According to the method, the diagnosis accuracy is improved, and the prediction capability is enhanced through operable risk scoring. The method is unique in that double diagnosis of functions and structures is integrated, and compared with a traditional method focusing on single diagnosis, the method provides more comprehensive cardiovascular health assessment. Besides, the system has high applicability and expandability, can be integrated with clinical and wearable equipment, and realizes real-time monitoring and risk assessment. In the future, the system can further enlarge the diagnosis range by incorporating data such as biomarkers and the like, and shows good future adaptability.

Description

Technical Field

[0001] The present application relates to the fields of AI analysis and smart medical technology, and in particular to multimodal AI diagnosis and risk grading systems, methods, devices, media, and program products applicable to vascular diseases. Background Art

[0002] Vascular health issues are particularly prominent in the elderly, with common conditions including hypertension, vascular aging, and atherosclerosis. These issues not only affect vascular function but also lead to significant structural changes, such as thickening of the vessel wall, decreased elasticity, and narrowing of the lumen. These changes are often accompanied by complex physiological and pathological processes, making vascular health complex and variable.

[0003] However, existing diagnostic methods often separate functional hemodynamic testing from structural vascular assessment. Functional testing focuses primarily on dynamic changes in blood flow velocity, pressure, and other factors, while structural assessment focuses on the morphology and structure of the vascular wall. This separate diagnostic approach may lead to an insufficient understanding of the full picture of the disease, resulting in one-sided diagnostic results and difficulty in accurately revealing the underlying mechanisms of the disease. Therefore, in order to more comprehensively understand and diagnose vascular problems in elderly patients, it is necessary to develop comprehensive diagnostic techniques that can simultaneously assess vascular function and structure, so as to more accurately assess the disease status and formulate more effective treatment plans. Summary of the Invention

[0004] In view of the shortcomings of the prior art described above, the purpose of this application is to provide a multimodal AI diagnosis and risk grading system, method, device, medium and program product suitable for vascular diseases, which is used to solve the technical problem that the prior art separates functional hemodynamic detection and structural vascular assessment, resulting in inaccurate diagnosis.

[0005] To achieve the above-mentioned objectives and other related objectives, the first aspect of the present application provides a multimodal AI diagnosis and risk grading system for vascular diseases, including: a preprocessing and feature extraction module, which is used to preprocess and extract features of the collected hemodynamic data, imaging data and time series functional data using corresponding AI models respectively; a data fusion module, connected to the preprocessing and feature extraction module, which is used to use graph neural networks and hierarchical attention mechanisms to integrate and analyze information from different data modalities from the preprocessing and feature extraction module to generate a graph structure of each data modality and their correlations; a risk grading module, connected to the data fusion module, which is used to dynamically adjust the weights in the graph structure based on hemodynamic data, imaging data and time series functional data, and according to patient characteristics, to generate a vascular disease risk score to assess the risk of vascular diseases in the elderly.

[0006] In some embodiments of the first aspect of the present application, the preprocessing and feature extraction module uses corresponding AI models to preprocess and extract features of the collected hemodynamic data, imaging data and time series functional data, including: using a convolutional neural network to preprocess and extract features of hemodynamic data; using an image segmentation model to preprocess and extract features of imaging data; and using a Transformer model to preprocess and extract features of time series functional data.

[0007] In some embodiments of the first aspect of the present application, the method of using a convolutional neural network to preprocess and extract features of hemodynamic data includes: normalizing and filtering the hemodynamic data, and using a one-dimensional convolutional neural network to extract frequency domain features related to hemodynamics; the method of using an image segmentation model to preprocess and extract features of image data includes: using a U-Net model to segment vascular structures in MRI images or ultrasound images, and using a two-dimensional convolutional neural network to extract structural features from the segmented images that are influential in predicting vascular aging and degenerative changes; the method of using a Transformer model to preprocess and extract features of time series functional data includes: using a Transformer model to analyze and implement sequential functional data to capture the gradual changes in functional indicators of heart rate variability and blood flow stability.

[0008] In some embodiments of the first aspect of the present application, the graph neural network is used to construct a graph structure; the nodes in the graph structure represent different data modalities, including hemodynamic parameters, medical effects and functional parameters; the edges in the graph structure represent the correlation between different parameters; and the graph neural network dynamically assigns weights through an attention layer.

[0009] In some embodiments of the first aspect of the present application, the hierarchical attention mechanism includes: allocating importance weights within a modality, and / or allocating importance weights across modalities.

[0010] In some embodiments of the first aspect of the present application, the scoring components of the risk grading module include: a core hemodynamic score, a structural integrity score, and a comprehensive risk score; the core hemodynamic score evaluates the functional status of blood vessels based on hemodynamic data; the structural integrity score evaluates the physical status of blood vessels based on the structural parameters of the blood vessels; and the comprehensive risk score is a weighted sum of the core hemodynamic score and the structural integrity score.

[0011] To achieve the above-mentioned objectives and other related objectives, the second aspect of the present application provides a multimodal AI diagnosis and risk grading method suitable for vascular diseases, including: using corresponding AI models to preprocess and extract features of the collected hemodynamic data, imaging data and time series functional data; using graph neural networks and hierarchical attention mechanisms to integrate and analyze the information collected from different data modalities to generate and output diagnostic results for vascular diseases; based on hemodynamic data, imaging data and time series functional data, and dynamically adjusting weights according to patient characteristics, a vascular disease risk score is generated to assess the risk of vascular diseases in the elderly.

[0012] To achieve the above-mentioned objectives and other related objectives, the third aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the multimodal AI diagnosis and risk grading method for vascular diseases.

[0013] To achieve the above-mentioned objectives and other related objectives, the fourth aspect of the present application provides a computer program product, which includes computer program code. When the computer program code is run on a computer, the computer implements the multimodal AI diagnosis and risk grading method suitable for vascular diseases.

[0014] To achieve the above-mentioned objectives and other related objectives, the fifth aspect of the present application provides a computer device comprising a memory, a processor, and a computer program stored on the memory; the processor executes the computer program to implement the multimodal AI diagnosis and risk grading method suitable for vascular diseases.

[0015] As described above, the multimodal AI diagnosis and risk grading system, method, device, medium and program product for vascular diseases of the present application have the following beneficial effects: The present invention provides an innovative method for cardiovascular diagnosis in the elderly by combining hemodynamic and structural indicators and utilizing a multimodal deep learning framework. This method not only improves the accuracy of diagnosis, but also enhances predictive ability through actionable risk scoring. The uniqueness of the present invention lies in the integration of dual diagnosis of function and structure. Compared with traditional methods that focus on a single diagnosis, the present invention provides a more comprehensive cardiovascular health assessment. In addition, the system is highly applicable and scalable, and can be integrated with clinical and wearable devices to achieve real-time monitoring and risk assessment. In the future, the system can further expand the scope of diagnosis by incorporating data such as biomarkers, showing good future adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Shown is a structural diagram of a multimodal AI diagnosis and risk grading system for vascular diseases in one embodiment of the present application.

[0017] Figure 2 Shown is a flowchart of a multimodal AI diagnosis and risk grading method for vascular diseases in one embodiment of the present application.

[0018] Figure 3 Shown is a structural diagram of a computer device in one embodiment of the present application. DETAILED DESCRIPTION

[0019] The following describes the embodiments of the present application through specific examples. Those skilled in the art can easily understand the other advantages and effects of the present application from the content disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other unless they conflict.

[0020] To address the issue mentioned in the background technology above, namely that vascular disease in elderly patients is caused by both structural and functional changes, and current diagnostic methods fail to effectively link these two factors, this paper proposes an innovative diagnostic system and method that integrates multimodal data, including hemodynamic and vascular structural parameters, and uses deep learning and neural network models to provide patients with accurate and actionable risk levels. By integrating core hemodynamic data with vascular structural indicators, this system can more comprehensively assess disease risk.

[0021] Before further explaining the present invention in detail, the nouns and terms involved in the embodiments of the present invention are explained. The nouns and terms involved in the embodiments of the present invention are subject to the following interpretations:

[0022] (a) Convolutional Neural Network (CNN), a deep learning model primarily used to process data with a grid-like topology, such as images. CNNs extract features from input data through convolutional layers, which use filters (or convolution kernels) to slide over the input data, calculating the weighted sum of local regions to generate feature maps. These feature maps capture the local features of the input data and are reduced in dimension through pooling layers, reducing the number of parameters and computational effort. CNNs typically contain multiple convolutional and pooling layers, which may be followed by fully connected layers for classification or regression tasks.

[0023] (b) The Transformer model, a deep learning architecture based on the self-attention mechanism, was originally designed for processing sequential data, particularly in the field of natural language processing (NLP). It abandons the traditional recurrent neural network structure and, through the self-attention mechanism, can process all positions in a sequence in parallel, capturing long-range dependencies within the sequence. The Transformer model consists of an encoder and a decoder. The encoder converts the input sequence into a continuous representation, while the decoder generates a target sequence based on the encoder's output and previous outputs.

[0024] (c) Graph Neural Network (GNN), a deep learning model for processing graph-structured data. Graph-structured data consists of nodes and edges. Nodes can contain features, and edges represent relationships between nodes. GNNs update the representation of each node by aggregating information from neighboring nodes. This process captures the local neighborhood structure and features of the node. GNNs can be used for a variety of graph-related tasks, including node classification, graph classification, and link prediction.

[0025] (d) U-Net model, a convolutional neural network architecture used for image segmentation tasks, particularly in the field of medical image segmentation. The U-Net design consists of a contraction path (encoder) and a symmetrical expansion path (decoder). The contraction path gradually reduces the spatial dimension of the image while increasing the number of feature channels to capture the image content; the expansion path gradually restores the spatial dimension and number of feature channels of the image for precise positioning. By adding the feature map of the contraction path (skip connections) to the expansion path, U-Net enables the model to better preserve the detailed information of the image, making it suitable for tasks requiring fine segmentation.

[0026] (e) The graph attention layer is a component in graph neural networks that allows the model to assign different weights to different neighbors when aggregating neighbor node information. This mechanism enables the model to pay more attention to those neighbor nodes that are more important. In the graph attention layer, the representation of each node is calculated using an attention mechanism that takes into account the node's own characteristics and the characteristics of its neighbors, as well as the interactions between them. This enables the model to capture complex relationships between nodes and improve performance on tasks with graph-structured data.

[0027] Figure 1A schematic diagram of the structure of a multimodal AI diagnosis and risk grading system for vascular diseases in an embodiment of the present invention is shown. The system provided by the present invention aims to use a multimodal AI framework, combined with convolutional neural networks (CNN), transformer models and graph neural networks (GNN), to integrate hemodynamic parameters with structural vascular indicators to provide real-time and accurate risk grading for elderly patients.

[0028] In an embodiment of the present invention, a multimodal AI diagnosis and risk grading system for vascular diseases includes: a preprocessing and feature extraction module 101 , a data fusion module 102 , a risk grading module 103 , and an explainability module 104 .

[0029] The preprocessing and feature extraction module 101 is used to preprocess and extract features from the collected hemodynamic data, imaging data, and time series functional data using corresponding AI models.

[0030] In an embodiment of the present application, the preprocessing and feature extraction module 101 uses corresponding AI models to preprocess the collected hemodynamic data, image data and time series functional data in the following ways: using a convolutional neural network to preprocess and extract features of hemodynamic data; using an image segmentation model to preprocess and extract features of image data; and using a Transformer model to preprocess and extract features of time series functional data.

[0031] In some examples, the method of using a convolutional neural network to preprocess and extract features from hemodynamic data includes: normalizing and filtering the hemodynamic data, and using a one-dimensional convolutional neural network to extract frequency domain features related to hemodynamics.

[0032] Normalization is a processing method that eliminates the influence of different dimensions and magnitudes, allowing data to be compared and analyzed on a unified scale. This process ensures that each feature contributes evenly to the model, which helps with subsequent machine learning or statistical analysis. Through normalization, hemodynamic data from different sources and different magnitudes can be compared and processed under the same analytical framework, enhancing the flexibility and universality of data processing. Filtering refers to the removal of noise and outliers in the data to improve data quality. In hemodynamic data, filtering can help remove instantaneous fluctuations or equipment errors in physiological measurements, thereby obtaining more accurate hemodynamic parameters.

[0033] For example, hemodynamic data such as cardiac output and arterial resistance can be processed by Z-score normalization. Z-score normalization converts the data into a distribution with a mean of 0 and a standard deviation of 1. The formula is:

[0034]

[0035] Here, x is the original data, μ is the mean, and σ is the standard deviation. This processing helps with subsequent machine learning or statistical analysis because it ensures that each feature contributes evenly to the model.

[0036] One-dimensional convolutional neural networks (1D-CNN) are a deep learning technique used in hemodynamic analysis to extract local features from time series data, such as pressure and flow changes within the cardiac cycle. Because hemodynamic data contains not only time-domain information but also rich frequency-domain information, 1D-CNN can effectively extract frequency-domain features from time-domain signals through convolution operations. These features reflect the dynamic characteristics of the cardiovascular system, such as heart rate variability and the effects of breathing on the cardiovascular system, and are important for the diagnosis and monitoring of cardiovascular disease. Therefore, the frequency-domain features extracted by 1D-CNN can capture subtle changes in hemodynamic data that are difficult to detect using time-domain analysis, thereby improving feature recognition and diagnostic accuracy.

[0037] In some examples, the method of using an image segmentation model to preprocess and extract features from imaging data includes: using a U-Net model to segment vascular structures in MRI images or ultrasound images, and using a two-dimensional convolutional neural network to extract structural features from the segmented images that are influential in predicting vascular aging and degenerative changes.

[0038] U-Net is a popular deep learning model, particularly suitable for medical image segmentation tasks. It has a symmetrical U-shaped structure, consisting of a contraction path (downsampling) and an expansion path (upsampling). This structure enables the network to capture contextual information at a deep level while retaining more positional information through skip connections, which is particularly important for accurate image segmentation.

[0039] The first half of the U-Net architecture consists of convolutional layers (for feature extraction), and the second half consists of upsampling layers (for restoring the spatial resolution of the original input). This symmetrical design allows the model to capture global contextual information while accurately performing pixel-level classification. The skip connections in U-Net merge low-level feature maps with high-level feature maps, which helps preserve detailed image information and combine it with abstract semantic information. This combination is crucial for accurate image segmentation, especially in medical images, where both details and overall structure need to be considered. In addition, U-Net has a relatively large model capacity and can learn complex feature representations, which is very beneficial for the recognition of various complex lesions in medical images. Medical images are usually expensive to annotate and the datasets are relatively small. The structure of U-Net allows it to be trained on limited data and achieve good performance because it is parameter-efficient and can effectively avoid overfitting.

[0040] Two-dimensional convolutional neural networks (2D-CNNs) are effective in image processing for learning spatial features and local patterns in images. They can extract features such as texture, shape, and edges, facilitating more accurate classification and fault identification. In extracting vascular structural features, 2D-CNNs use convolution kernels to operate and aggregate features, extracting local features from images to identify structural features such as vessel wall thickness and plaque formation. By learning spatial features and local patterns in images, 2D-CNNs can identify vascular structural features such as wall thickness and plaque formation that are important for predicting vascular aging and degenerative changes.

[0041] It should be noted here that by combining the U-Net network and the 2D-CNN network, information from different modalities can be integrated to more comprehensively describe the characteristics of the data, accurately segment and extract vascular structure features, and provide a powerful tool for the diagnosis and treatment of vascular diseases.

[0042] In some examples, the method of using the Transformer model to preprocess and extract features from time series functional data includes: using the Transformer model to analyze the sequential functional data to capture the gradual changes in functional indicators of heart rate variability and blood flow stability.

[0043] The Transformer model is a deep learning model based on the self-attention mechanism. Its core strength lies in its ability to handle long-range dependencies and parallelize processing of long sequences of data. The self-attention layer in the Transformer model enables the model to capture the dependencies between any two time points in the sequence, regardless of their distance in the sequence. This mechanism allows the model to process sequence data in parallel, thereby improving processing efficiency. It is worth noting that self-attention allows each time step to attend to information from all other time steps, extracting global temporal information in parallel, a capability lacking in RNNs (serial global) and CNNs (parallel local). The Transformer's self-attention mechanism allows it to simultaneously process all positions in the sequence, capturing long-range dependencies and providing a more accurate understanding of the time series. Therefore, using the Transformer model to analyze time series functional data can effectively capture gradual changes in functional indicators such as heart rate variability and blood flow stability. Its advantages lie in its powerful ability to capture global information, parallel processing capabilities, the ability to handle long-term dependencies, and the model's flexibility and generalization.

[0044] In an embodiment of the present application, the data fusion module 102 is connected to the preprocessing and feature extraction module 101, and is used to use a graph neural network and a hierarchical attention mechanism to integrate and analyze information from different data modalities from the preprocessing and feature extraction module 101 to generate a graph structure of each data modality and the correlation between them.

[0045] Graph neural networks (GNNs) are used to construct graph structures, where nodes represent different data modalities, including hemodynamic parameters, medical effects, and functional parameters, unifying these diverse data sources within a unified framework. Edges within the graph structure represent correlations between different parameters. For example, physiological parameters such as pulse wave velocity (PWV) and arterial resistance are linked to reflect their relationships. GNNs dynamically assign weights through a graph attention network (GAT). This allows the model to adjust the correlations between modalities based on individual patient characteristics, making the model more flexible and adaptable. The graph attention layer introduces an attention mechanism to achieve weighted aggregation of node features. In GAT, each node learns the weights of its neighboring nodes, allowing it to dynamically adjust the contribution of neighboring nodes to its own features. This mechanism makes GAT not only robust to noisy nodes but also imparts a degree of interpretability to the model.

[0046] Hierarchical attention machines are a method that applies attention at different levels. They can assign importance weights both intramodally (within the same data type) and cross-modally (between different data types). Intramodal attention assigns importance weights within each data type, highlighting key features, such as changes in PWV, and enhancing the model's ability to capture key information. Cross-modal attention applies attention layers across different modalities, determining the relevance of each feature and dynamically adjusting the weights based on the clinical context to achieve effective fusion of multimodal data. This mechanism enables the model to focus more on key features and dynamically adjust the relevance weights of different features based on the clinical context.

[0047] It's important to note that in the medical field, single-modality data often fails to provide comprehensive disease information. Integrating multimodal data can provide a more comprehensive representation of the disease. Complex interrelationships exist between different data modalities, and GNNs can effectively capture these relationships, improving the model's predictive accuracy. Differences in individual patient characteristics require flexible model adjustments, and the attention mechanism provides this flexibility.

[0048] The risk grading module 103 is used to dynamically adjust the weights in the graph structure based on hemodynamic data, imaging data, and time series functional data according to patient characteristics, thereby generating a vascular disease risk score to assess the risk of elderly vascular diseases.

[0049] The scoring components of the risk stratification module 103 include: a core hemodynamic score (CHS), a structural integrity score (SIS), and a comprehensive risk score (CRS). The core hemodynamic score assesses vascular functional status based on key hemodynamic parameters such as cardiac output (CO), stroke volume (SV), and arterial resistance (AR). The structural integrity score assesses the physical status of blood vessels based on structural parameters of the blood vessels, such as pulse wave velocity (PWV), vascular stiffness (VS), and arterial wall thickness (AWT). The comprehensive risk score is a weighted sum of the CHS and SIS, where the weights α and β are dynamically adjusted based on patient characteristics to reflect the individualized risk level.

[0050] CHS=w1·CO+w2·SV+w3·AR; formula (2)

[0051] SIS=w4·PWV+w5·VS+w6·AWT; formula (3)

[0052] CRS=α·CHS+β·SIS; formula (4)

[0053] Where CHS denotes core hemodynamic score, CO denotes cardiac output, SV denotes stroke volume, and AR denotes arterial resistance. SIS denotes structural integrity score, PWV denotes pulse wave velocity, SV denotes vascular stiffness, and AWT denotes arterial wall thickness. CRS denotes comprehensive risk score. w1, w2, w3, w4, w5, w6, α, and β denote weights.

[0054] It is worth noting that this method combines hemodynamic and vascular structural parameters to comprehensively assess vascular health from both functional and structural dimensions. By dynamically adjusting weights, it can more accurately reflect individual differences and provide a more accurate basis for risk assessment of vascular disease in the elderly. More specifically, by combining hemodynamic and structural parameters, a more comprehensive assessment of vascular health is provided. By dynamically adjusting weights, the scoring system can provide personalized risk assessments based on the patient's specific circumstances. Accurate risk assessments help identify high-risk individuals early, allowing for timely intervention and management.

[0055] Finally, the dividing lines between low, medium, and high risks are determined based on the CRS value, and the machine learning algorithm is used to dynamically adjust the risk levels. For example, the risk levels are as follows: (1) High risk: CRS>0.8; (2) Medium risk: 0.5≤CRS≤0.8; (3) Low risk: CRS<0.5.

[0056] In an embodiment of the present application, the explainability module 104 provides an easy-to-understand visualization of patient risk score influencing factors and key risk factors through Shapley value attribution, attention heat map, and feature attribution.

[0057] Shapley value attribution is a game-theory-based method used to determine the contribution of each feature to a model's predictions. In medical risk assessment, Shapley values can help physicians understand which characteristics (such as age and blood pressure) have the greatest impact on a patient's vascular disease risk score. Attention heatmaps and feature attribution aim to generate visual heatmaps that intuitively demonstrate which features contribute most to the risk score. This visualization approach helps physicians quickly identify and locate key risk factors, leading to more accurate clinical decisions.

[0058] It should be understood that this approach is intended to improve model interpretability, meaning that physicians can understand how the model arrives at its risk assessment. This is particularly important for medical decision-making, as physicians need to clearly understand the factors that contribute to a high-risk score so they can explain it to their patients and take appropriate preventive or treatment measures. For example, through explanatory components, the model's decision-making process becomes more transparent, helping to build trust between physicians and patients. Physicians can use these explanatory tools to better understand the logic behind risk scores and make more targeted treatment decisions. These tools can help physicians explain to patients why they are at a specific risk level, enhancing their understanding and acceptance of treatment options.

[0059] It should be understood that the division of modules in the embodiments of the present application is illustrative and is merely a logical functional division. In actual implementation, other division methods may be used. Furthermore, the functional modules in the various embodiments of the present application may be integrated into a single processor, or may exist physically as separate modules, or two or more modules may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or software functional modules.

[0060] It should also be understood that in the embodiments of the present application, words such as "first" and "second" are used to distinguish between identical or similar items with substantially the same functions and effects. For example, the first XX and the second XX are merely to distinguish between different XXs and do not limit their order of precedence. Those skilled in the art will understand that words such as "first" and "second" do not limit the quantity and execution order, and words such as "first" and "second" do not necessarily limit them to be different. In the embodiments of the present application, words such as "exemplary" or "for example" represent examples, illustrations, or descriptions. Any embodiment or design described in this application as "exemplary" or "for example" should not be interpreted as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way.

[0061] In the embodiments of the present application, "at least one" refers to one or more, and "more" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: the existence of A alone, the existence of A and B at the same time, and the existence of B alone, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can represent: a, b, c, ab, ac, bc or abc, where a, b, c can be single or multiple.

[0062] Figure 2 : This is a flow chart of a multimodal AI diagnosis and risk grading method for vascular disease provided in an embodiment of the present application. The method includes the following:

[0063] Step S21: Use the corresponding AI models to preprocess and extract features from the collected hemodynamic data, imaging data, and time series functional data.

[0064] Step S22: Utilize graph neural networks and hierarchical attention mechanisms to integrate and analyze the collected information of different data modalities to generate and output diagnostic results for vascular diseases.

[0065] Step S23: Based on the hemodynamic data, imaging data, and time series functional data, and dynamically adjusting the weights according to patient characteristics, a vascular disease risk score is generated to assess the risk of vascular disease in the elderly.

[0066] It should be understood that the specific process of executing each method step has been described in detail in the above system embodiment, and for the sake of brevity, it will not be repeated here.

[0067] Figure 3 : is a schematic block diagram of a computer device provided in an embodiment of the present application. Figure 3 As shown, the computer device includes: at least one processor 301, memory 302, at least one network interface 303 and a user interface 305. The various components in the device are coupled together via a bus system 304. It is understood that the bus system 304 is used to achieve connection and communication between these components. In addition to including a data bus, the bus system 304 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, the bus system 304 is not described in detail. Figure 3 Various buses are labeled as bus systems.

[0068] The user interface 305 may include a display, a keyboard, a mouse, a trackball, a click gun, keys, buttons, a touch pad or a touch screen.

[0069] It will be appreciated that the memory 302 may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM) or a programmable read-only memory (PROM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM) and synchronous static random access memory (SSRAM). The memory described in the embodiments of the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0070] The memory 302 in the embodiment of the present invention is used to store various categories of data to support the operation of the electronic terminal 300. Examples of such data include: any executable program for operating on the electronic terminal 300, such as an operating system 3021 and an application 3022; the operating system 3021 includes various system programs, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks. The application 3022 can include various applications, such as a media player (Media Player), a browser (Browser), etc., for implementing various application services. The implementation of the XX method provided in the embodiment of the present invention can be included in the application 3022.

[0071] The methods disclosed in the above embodiments of the present invention can be applied to or implemented by processor 301. Processor 301 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by hardware integrated logic circuits in processor 301 or by software instructions. The above processor 301 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 301 can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor 301 can be a microprocessor or any conventional processor. The steps of the accessory optimization method provided in the embodiments of the present invention can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium located in a memory. The processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method.

[0072] In an exemplary embodiment, the electronic terminal 300 may be configured to execute the aforementioned method using one or more application specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), or complex programmable logic devices (CPLDs).

[0073] According to the method provided in the embodiments of the present application, the present application also provides a computer program product, which includes: computer program code, which, when running on a computer, enables the computer to execute the multimodal AI diagnosis and risk grading method for vascular diseases in the above embodiment.

[0074] According to the method provided in the embodiments of the present application, the present application also provides a computer-readable storage medium, which stores program code. When the program code runs on a computer, the computer executes the above-mentioned multimodal AI diagnosis and risk grading method applicable to vascular diseases.

[0075] As used in this specification, the terms "component," "module," "system," and the like are used to represent computer-related entities, hardware, firmware, a combination of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable file, an execution thread, a program, and / or a computer. By way of illustration, both an application running on a computing device and a computing device can be a component. One or more components can reside in a process and / or an execution thread, and a component can be located on a computer and / or distributed between two or more computers. In addition, these components can be executed from various computer-readable media having various data structures stored thereon. Components can communicate, for example, via local and / or remote processes based on signals having one or more data packets (e.g., data from two components interacting with another component on a local system, a distributed system, and / or a network, such as the Internet interacting with other systems via signals).

[0076] Those skilled in the art will appreciate that the various illustrative logical blocks and steps described in conjunction with the embodiments disclosed herein can be implemented using electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0077] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0078] 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 device embodiments described above are merely schematic. 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 through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0079] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0080] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0081] In the above embodiments, the functions of each functional unit can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions (programs). When the computer program instructions (program) are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated therein. Available media may be magnetic media (e.g., floppy disks, hard disks, tapes), optical media (e.g., high-density digital video discs (DVDs), or semiconductor media (e.g., solid state disks (SSDs)).

[0082] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, and other media that can store program codes.

[0083] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

[0084] In summary, the present application provides a multimodal AI diagnosis and risk grading system, method, device, medium and program product suitable for vascular diseases. The present invention provides an innovative method for cardiovascular diagnosis in the elderly by combining hemodynamic and structural indicators and utilizing a multimodal deep learning framework. This method not only improves the accuracy of diagnosis, but also enhances predictive ability through actionable risk scoring. The uniqueness of the present invention lies in the integration of dual diagnosis of function and structure. Compared with the traditional method that focuses on a single diagnosis, the present invention provides a more comprehensive cardiovascular health assessment. In addition, the system is highly applicable and scalable, and can be integrated with clinical and wearable devices to achieve real-time monitoring and risk assessment. In the future, the system can further expand the scope of diagnosis by incorporating data such as biomarkers, showing good future adaptability. Therefore, the present application effectively overcomes the various shortcomings of the existing technology and has a high industrial utilization value.

[0085] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Anyone skilled in the art may modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by one of ordinary skill in the art without departing from the spirit and technical concepts disclosed in this application shall be covered by the claims of this application.

Claims

1. A multimodal AI diagnosis and risk grading system for vascular diseases, characterized by: include: The preprocessing and feature extraction module is used to preprocess and extract features from the collected hemodynamic data, imaging data, and time series functional data using corresponding AI models; A data fusion module, connected to the preprocessing and feature extraction module, is used to integrate and analyze information from different data modalities from the preprocessing and feature extraction module using a graph neural network and a hierarchical attention mechanism to generate a graph structure of each data modality and their correlations; The risk grading module is connected to the data fusion module and is used to dynamically adjust the weights in the graph structure based on hemodynamic data, imaging data and time series functional data, and according to patient characteristics, to generate a vascular disease risk score to assess the risk of vascular disease in the elderly.

2. The multimodal AI diagnosis and risk grading system for vascular diseases according to claim 1, characterized in that: The preprocessing and feature extraction modules use corresponding AI models to preprocess and extract features from the collected hemodynamic data, imaging data, and time series functional data, respectively, including: Use convolutional neural networks to preprocess and extract features from hemodynamic data; use image segmentation models to preprocess and extract features from imaging data; and use Transformer models to preprocess and extract features from time series functional data.

3. The multimodal AI diagnosis and risk grading system for vascular disease according to claim 2, characterized in that: The method of using convolutional neural network to preprocess and extract features of hemodynamic data includes: normalizing and filtering the hemodynamic data, and using one-dimensional convolutional neural network to extract frequency domain features related to hemodynamics; Methods for preprocessing and extracting features from image data using image segmentation models include: using a U-Net model to segment vascular structures in MRI images or ultrasound images, and using a two-dimensional convolutional neural network to extract structural features from the segmented images that are influential in predicting vascular aging and degenerative changes; Methods for preprocessing and feature extraction of time series functional data using the Transformer model include: using the Transformer model to analyze and implement sequence functional data to capture gradual changes in functional indicators of heart rate variability and blood flow stability.

4. The multimodal AI diagnosis and risk grading system for vascular diseases according to claim 1, characterized in that: The graph neural network is used to construct a graph structure; the nodes in the graph structure represent different data modalities, including hemodynamic parameters, medical effects and functional parameters; the edges in the graph structure represent the correlation between different parameters; the graph neural network dynamically allocates weights through an attention layer.

5. The multimodal AI diagnosis and risk grading system for vascular diseases according to claim 1, characterized in that: The hierarchical attention mechanism includes: allocating importance weights within a modality, and / or allocating importance weights across modalities.

6. The multimodal AI diagnosis and risk grading system for vascular diseases according to claim 1, characterized in that: The scoring components of the risk grading module include: a core hemodynamic score, a structural integrity score, and a comprehensive risk score; the core hemodynamic score assesses the functional status of blood vessels based on hemodynamic data; the structural integrity score assesses the physical status of blood vessels based on the structural parameters of the blood vessels; and the comprehensive risk score is the weighted sum of the core hemodynamic score and the structural integrity score.

7. A multimodal AI diagnosis and risk grading method for vascular diseases, characterized by: include: Use corresponding AI models to preprocess and extract features from the collected hemodynamic data, imaging data, and time series functional data; Utilize graph neural networks and hierarchical attention mechanisms to integrate and analyze information from different data modalities to generate and output diagnostic results for vascular diseases; Based on hemodynamic data, imaging data and time series functional data, and dynamically adjusting weights according to patient characteristics, a vascular disease risk score is generated to assess the risk of vascular disease in the elderly.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the multimodal AI diagnosis and risk grading method for vascular diseases according to claim 7 is implemented.

9. A computer program product, characterized in that The computer program product includes computer program code, and when the computer program code is run on a computer, the computer implements the multimodal AI diagnosis and risk grading method for vascular diseases as claimed in claim 7.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory, wherein: The processor executes the computer program to implement the multimodal AI diagnosis and risk grading method for vascular diseases as described in claim 7.

Images