Myocardial infarction risk assessment method, system and device based on multi-modal interpretability

Abstract

The application discloses a myocardial infarction risk assessment method, system and device based on multi-modal interpretability. The method constructs a dynamic weighted fuzzy cognitive map model for myocardial infarction risk assessment. The aggregated feature vector is input into the dynamic weighted fuzzy cognitive map model, the activation value of each concept node is updated using the fine-tuned weight and activation function until the final activation value of each concept node is obtained after the preset condition is reached, and the final activation value of the myocardial infarction risk concept node is taken as the myocardial infarction risk assessment result. According to the myocardial infarction risk assessment result, the contribution degree of each physiological index in the aggregated feature vector is calculated. The maximum influence propagation path between each source node and the target node is found, and at least one maximum influence propagation path is obtained. Based on the contribution degree of each physiological index and the at least one maximum influence propagation path, a myocardial infarction risk assessment report with interpretability is constructed. The application realizes interpretable and high-precision model evaluation.

Description

Technical Field

[0001] This application relates to the field of medical and health technology, and in particular to a method, system and device for myocardial infarction risk assessment based on multimodal interpretability. Background Technology

[0002] Early and accurate diagnosis of cardiovascular diseases such as myocardial infarction is crucial for improving patient survival rates. With the development of information technology, intelligent risk assessment using multimodal data such as electrocardiograms (ECG), blood biomarkers, and medical imaging has become a research hotspot. Currently, intelligent diagnostic methods based on machine learning, especially deep learning, can automatically learn complex features from massive amounts of data, achieving significant progress in recognition accuracy. However, in the high-risk field of medicine, diagnostic decisions not only require accurate results but also transparent processes, clear logic, and understanding and trustworthiness from clinicians. The "black box" nature of existing data-driven models—the lack of transparency in their decision-making logic—constitutes a major obstacle to their widespread application in clinical practice.

[0003] While existing technologies have made some progress in diagnostic accuracy for myocardial infarction identification, they still generally face challenges such as high model complexity, large computational costs, and an inherent contradiction between model accuracy and interpretability. In particular, existing interpretability methods, as supplements to "black box" models, have not fundamentally solved the problem of opaque decision-making processes. This makes it difficult for the reliability and credibility of the results when the models process real medical data containing noise or missing information to meet the stringent requirements of clinical practice for decision transparency and traceability. Summary of the Invention

[0004] This application aims to propose a method, system, and device for myocardial infarction risk assessment based on multimodal interpretability, which can achieve interpretable and high-precision model assessment, thereby improving the reliability and credibility of myocardial infarction risk assessment results.

[0005] In a first aspect, embodiments of this application provide a method for myocardial infarction risk assessment based on multimodal interpretability, the method comprising:

[0006] Extract multimodal physiological indicators that include electrocardiogram clinical features, blood biomarker features, and radiomics features;

[0007] The features in the multimodal physiological index features are concatenated to obtain a concatenated feature vector;

[0008] Based on the spliced ​​feature vector, the attention score of each feature in the multimodal physiological indicator features is calculated, and based on the attention score, the weighted sum of each feature in the multimodal physiological indicator features is performed to obtain the aggregated feature vector.

[0009] A dynamic weighted fuzzy cognitive graph model for myocardial infarction risk assessment is constructed. The dynamic weighted fuzzy cognitive graph model includes a weight adjustment mechanism, multiple concept nodes, weighted edges, and an activation function. The weight adjustment mechanism is used to fine-tune the basic weights based on the context vector to obtain the fine-tuned weights. The weighted edges are edges connecting two concept nodes and are used to characterize the influence strength of one concept node on another concept node.

[0010] The aggregated feature vector is input into the dynamic weighted fuzzy cognitive graph model, and the activation value of each concept node is updated using the fine-tuned weights and activation function until the preset conditions are met to obtain the final activation value of each concept node. The final activation value of the myocardial infarction risk concept node is used as the myocardial infarction risk assessment result.

[0011] Based on the myocardial infarction risk assessment results, the contribution of each physiological indicator in the aggregated feature vector is calculated;

[0012] Based on the preset concept node sensitivity coefficient, path context decay factor, the multiple concept nodes and the weighted edges, find the maximum influence propagation path between each source node and the target node, and obtain at least one maximum influence propagation path;

[0013] Based on the contribution of each physiological indicator and the at least one maximum impact propagation path, construct an interpretable myocardial infarction risk assessment report.

[0014] Compared with the prior art, the first aspect of this application has the following beneficial effects:

[0015] This method extracts multimodal physiological indicator features, including ECG clinical features, blood biomarker features, and radiomics features. The features from these multimodal physiological indicator features are concatenated to obtain a concatenated feature vector. Based on this concatenated feature vector, attention scores are calculated for each feature in the multimodal physiological indicator features. Then, based on these attention scores, a weighted sum of the features is obtained to obtain an aggregated feature vector. A dynamically weighted fuzzy cognitive graph model for myocardial infarction risk assessment is constructed. This model includes a weight adjustment mechanism, multiple concept nodes, weighted edges, and an activation function. The weight adjustment mechanism fine-tunes the basic weights based on the context vector to obtain the adjusted weights. Weighted edges connect two concept nodes and represent a concept node. The influence strength of a point on another concept node is determined. The aggregated feature vector is input into a dynamically weighted fuzzy cognitive graph model, and the activation value of each concept node is updated using fine-tuned weights and activation functions until a preset condition is met to obtain the final activation value of each concept node. The final activation value of the myocardial infarction risk concept node is used as the myocardial infarction risk assessment result. Based on the myocardial infarction risk assessment result, the contribution of each physiological indicator in the aggregated feature vector is calculated. Based on the preset concept node sensitivity coefficient, path context decay factor, multiple concept nodes and weighted edges, the maximum influence propagation path between each source node and the target node is found, and at least one maximum influence propagation path is obtained. Based on the contribution of each physiological indicator and at least one maximum influence propagation path, an interpretable myocardial infarction risk assessment report is constructed. Therefore, by comprehensively considering the characteristics of multimodal physiological indicators, the accuracy of myocardial infarction risk assessment results can be improved. By inputting the aggregated feature vector into the dynamic weighted fuzzy cognitive graph model, and updating the activation value of each concept node using fine-tuned weights and activation functions, the myocardial infarction risk assessment results can be obtained. The dynamic weighted fuzzy cognitive graph model directly constructs model parameters through expert knowledge rather than data-driven training, effectively reducing the computational cost of the model. Finally, by calculating the contribution of each physiological indicator and the maximum impact propagation path, an interpretable myocardial infarction risk assessment report can be constructed, making the model interpretable and thus improving the reliability and credibility of the myocardial infarction risk assessment results.

[0016] In some embodiments, the extraction of multimodal physiological indicators comprising electrocardiogram clinical features, blood biomarker features, and radiomics features includes:

[0017] Baseline correction is performed on the electrocardiogram (ECG) signals obtained from the acquired ECG to obtain the corrected ECG signals.

[0018] The corrected electrocardiogram signal is subjected to harmonic wavelet packet decomposition to obtain a wavelet packet decomposition tree;

[0019] Based on the node coefficients in the wavelet packet decomposition tree, the time-frequency entropy of each node in the wavelet packet decomposition tree is calculated, and the optimal wavelet basis is obtained by minimizing the time-frequency entropy.

[0020] Extract ECG clinical features including ST segment shift and T wave symmetry from the optimal wavelet basis;

[0021] The extracted blood biomarkers were standardized to obtain their characteristics;

[0022] Radiomic features containing texture features are extracted from medical images to obtain image radiomic features;

[0023] The electrocardiogram clinical features, the blood biomarker features, and the radiomics features are constructed into a multimodal physiological indicator feature.

[0024] In some implementations, updating the activation value of each concept node using fine-tuned weights and activation functions includes:

[0025] ;

[0026] in, Representing concept nodes In the The new activation value after the next iteration Indicates the first In this iteration, for concept nodes All other concept nodes that are affected activation value, Representing concept nodes Pointing to concept nodes The fine-tuned weights, Representing concept nodes The initial activation value, This represents the activation function. This represents the total number of concept nodes.

[0027] In some implementations, calculating the contribution of each physiological indicator in the aggregated feature vector based on the myocardial infarction risk assessment results includes:

[0028] Each physiological indicator in the aggregated feature vector is replaced with a clinical baseline value to construct a counterfactual input vector corresponding to each physiological indicator;

[0029] The counterfactual input vector is input into the dynamic weighted fuzzy cognitive graph model to obtain the counterfactual risk value corresponding to each physiological indicator;

[0030] The difference between the counterfactual risk value and the myocardial infarction risk assessment result is calculated to obtain the contribution of each physiological indicator in the aggregated feature vector.

[0031] In some implementations, finding the maximum influence propagation path from each source node to the target node based on preset concept node sensitivity coefficients, path context decay factors, the plurality of concept nodes, and the weighted edges, to obtain at least one maximum influence propagation path, includes:

[0032] Construct at least one path between each source node and the target node;

[0033] If each path contains multiple concept nodes and weighted edges, the influence flow between two adjacent concept nodes is calculated based on the preset concept node sensitivity coefficient, the final activation value of the sub-source node in two adjacent concept nodes, and the fine-tuned weight of the weighted edge between two adjacent concept nodes.

[0034] Based on the path context decay factor, the influence flow between all two adjacent concept nodes in each path is weighted and summed to obtain the total influence of each path;

[0035] Based on the total influence, find the maximum influence propagation path between each source node and the target node, and obtain at least one maximum influence propagation path.

[0036] In some implementations, constructing an interpretable myocardial infarction risk assessment report based on the contribution of each physiological indicator and the at least one most influential propagation path includes:

[0037] By sorting the contribution of each physiological indicator and filtering out multiple key physiological indicators, a quantitative attribution summary containing the indicator name corresponding to each key physiological indicator, the specific value derived from the aggregated feature vector, and the contribution is obtained.

[0038] Match the maximum impact propagation path corresponding to each of the key physiological indicators, and obtain the core logic path containing the name and final activation value of each concept node in the maximum impact propagation path;

[0039] The final activation value of each concept node and the maximum impact propagation path corresponding to each key physiological indicator are visually encoded to obtain a visualized decision map.

[0040] By integrating the quantitative attribution summary, the core logical path, and the visualized decision graph, an interpretable myocardial infarction risk assessment report is constructed.

[0041] In some implementations, after constructing an interpretable myocardial infarction risk assessment report, the method further includes:

[0042] The basic weights are constructed as a prior probability distribution of the weights;

[0043] Based on the prior probability distribution of the weights, the posterior probability distribution of the weights is obtained using Bayes' theorem.

[0044] The mean of the weight distribution obtained from the posterior probability distribution of the weights is used as the new weights in the dynamically weighted fuzzy cognitive graph model to update the model.

[0045] Secondly, embodiments of this application also provide a myocardial infarction risk assessment system based on multimodal interpretability, the system comprising:

[0046] The data extraction unit is used to extract multimodal physiological indicators that include electrocardiogram clinical features, blood biomarker features, and radiomics features.

[0047] The feature splicing unit is used to splice the features in the multimodal physiological index features to obtain a spliced ​​feature vector;

[0048] The feature aggregation unit is used to calculate the attention score of each feature in the multimodal physiological index features according to the spliced ​​feature vector, and to perform a weighted summation of each feature in the multimodal physiological index features based on the attention score to obtain the aggregated feature vector.

[0049] The model building unit is used to build a dynamic weighted fuzzy cognitive graph model for myocardial infarction risk assessment. The dynamic weighted fuzzy cognitive graph model includes a weight adjustment mechanism, multiple concept nodes, weighted edges, and an activation function. The weight adjustment mechanism is used to fine-tune the basic weights based on the context vector to obtain the fine-tuned weights. The weighted edges are edges connecting two concept nodes and are used to characterize the influence strength of one concept node on another concept node.

[0050] The risk assessment unit is used to input the aggregated feature vector into the dynamic weighted fuzzy cognitive graph model, update the activation value of each concept node with fine-tuned weights and activation functions, and obtain the final activation value of each concept node after reaching the preset conditions, and use the final activation value of the myocardial infarction risk concept node as the myocardial infarction risk assessment result.

[0051] The contribution calculation unit is used to calculate the contribution of each physiological indicator in the aggregated feature vector based on the myocardial infarction risk assessment results.

[0052] The path finding unit is used to find the maximum influence propagation path between each source node and the target node based on the preset concept node sensitivity coefficient, path context decay factor, the multiple concept nodes and the weighted edge, so as to obtain at least one maximum influence propagation path.

[0053] The assessment report construction unit is used to construct an interpretable myocardial infarction risk assessment report based on the contribution of each physiological indicator and the at least one maximum impact propagation path.

[0054] Thirdly, embodiments of this application also provide an electronic device, including at least one control processor and a memory for communicatively connecting to the at least one control processor; the memory stores instructions executable by the at least one control processor, the instructions being executed by the at least one control processor to enable the at least one control processor to perform a myocardial infarction risk assessment method based on multimodal interpretability as described above.

[0055] Fourthly, embodiments of this application also provide a computer-readable storage medium storing computer-executable instructions for causing a computer to execute a myocardial infarction risk assessment method based on multimodal interpretability as described above.

[0056] It is understood that the beneficial effects of the second to fourth aspects compared with the related technologies are the same as the beneficial effects of the first aspect compared with the related technologies. Please refer to the relevant description in the first aspect above, which will not be repeated here. Attached Figure Description

[0057] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0058] Figure 1 This is a flowchart illustrating an embodiment of the myocardial infarction risk assessment method based on multimodal interpretability provided in this application;

[0059] Figure 2 This is a schematic diagram of the structure of an embodiment of the myocardial infarction risk assessment system based on multimodal interpretability provided in this application;

[0060] Figure 3 This is a schematic diagram of the structure of an embodiment of the electronic device provided in this application. Detailed Implementation

[0061] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0062] In the description of this application, the use of terms such as "first," "second," etc., is for the purpose of distinguishing technical features only and should not be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated or the order of the technical features indicated.

[0063] In the description of this application, it should be understood that the orientation descriptions, such as up, down, etc., are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this application.

[0064] In the description of this application, it should be noted that, unless otherwise explicitly defined, terms such as "setup," "installation," and "connection" should be interpreted broadly, and those skilled in the art can reasonably determine the specific meaning of the above terms in this application in conjunction with the specific content of the technical solution.

[0065] While existing technologies have made some progress in diagnostic accuracy for myocardial infarction identification, they still generally face challenges such as high model complexity, large computational costs, and an inherent contradiction between model accuracy and interpretability. In particular, existing interpretability methods, as supplements to "black box" models, have not fundamentally solved the problem of opaque decision-making processes. This makes it difficult for the reliability and credibility of the results when the models process real medical data containing noise or missing information to meet the stringent requirements of clinical practice for decision transparency and traceability.

[0066] To address the issue of low reliability and credibility of the results in the existing technologies, this application proposes a method, system, and device for myocardial infarction risk assessment based on multimodal interpretability.

[0067] Reference Figure 1 This application provides a flowchart illustrating a multimodal interpretability-based myocardial infarction risk assessment method. This method is applied to an electronic device, which may be a server or a mobile terminal, etc. Figure 1 As shown, this myocardial infarction risk assessment method based on multimodal interpretability may include the following steps:

[0068] Step S101: Extract multimodal physiological indicators including electrocardiogram clinical features, blood biomarker features, and radiomics features;

[0069] Step S102: Concatenate the features in the multimodal physiological index features to obtain the concatenated feature vector;

[0070] Step S103: Based on the spliced ​​feature vector, calculate the attention score of each feature in the multimodal physiological index features, and based on the attention score, perform a weighted summation of each feature in the multimodal physiological index features to obtain the aggregated feature vector.

[0071] Step S104: Construct a dynamic weighted fuzzy cognitive graph model for myocardial infarction risk assessment. The dynamic weighted fuzzy cognitive graph model includes a weight adjustment mechanism, multiple concept nodes, weighted edges, and an activation function. The weight adjustment mechanism is used to fine-tune the basic weights based on the context vector to obtain the fine-tuned weights. The weighted edges are the edges connecting two concept nodes and are used to characterize the influence strength of one concept node on another concept node.

[0072] Step S105: Input the aggregated feature vector into the dynamic weighted fuzzy cognitive graph model, update the activation value of each concept node using the fine-tuned weights and activation function, until the preset conditions are met to obtain the final activation value of each concept node, and use the final activation value of the myocardial infarction risk concept node as the myocardial infarction risk assessment result.

[0073] Step S106: Based on the myocardial infarction risk assessment results, calculate the contribution of each physiological indicator in the aggregated feature vector;

[0074] Step S107: Based on the preset concept node sensitivity coefficient, path context decay factor, multiple concept nodes and weighted edges, find the maximum influence propagation path between each source node and the target node, and obtain at least one maximum influence propagation path;

[0075] Step S108: Based on the contribution of each physiological indicator and at least one maximum impact propagation path, construct an interpretable myocardial infarction risk assessment report.

[0076] In this embodiment, multimodal physiological indicator features, including electrocardiogram clinical features, blood biomarker features, and radiomics features, are extracted. The features in the multimodal physiological indicator features are concatenated to obtain a concatenated feature vector. Based on the concatenated feature vector, the attention score of each feature in the multimodal physiological indicator features is calculated, and based on the attention score, the features in the multimodal physiological indicator features are weighted and summed to obtain an aggregated feature vector. A dynamic weighted fuzzy cognitive graph model for myocardial infarction risk assessment is constructed. The dynamic weighted fuzzy cognitive graph model includes a weight adjustment mechanism, multiple concept nodes, weighted edges, and an activation function. The weight adjustment mechanism is used to fine-tune the basic weights based on the context vector to obtain the fine-tuned weights. The weighted edges are edges connecting two concept nodes and are used to represent a concept. The influence strength of a node on another concept node is determined. The aggregated feature vector is input into a dynamically weighted fuzzy cognitive graph model. The activation value of each concept node is updated using fine-tuned weights and activation functions until a preset condition is met, at which point the final activation value of each concept node is obtained. The final activation value of the myocardial infarction risk concept node is used as the myocardial infarction risk assessment result. Based on the myocardial infarction risk assessment result, the contribution of each physiological indicator in the aggregated feature vector is calculated. Based on the preset concept node sensitivity coefficient, path context decay factor, multiple concept nodes, and weighted edges, the maximum influence propagation path between each source node and the target node is found, resulting in at least one maximum influence propagation path. Based on the contribution of each physiological indicator and at least one maximum influence propagation path, an interpretable myocardial infarction risk assessment report is constructed. Therefore, by comprehensively considering the characteristics of multimodal physiological indicators, the accuracy of myocardial infarction risk assessment results can be improved. By inputting the aggregated feature vector into the dynamic weighted fuzzy cognitive graph model, and updating the activation value of each concept node using fine-tuned weights and activation functions, the myocardial infarction risk assessment results can be obtained. The dynamic weighted fuzzy cognitive graph model directly constructs model parameters through expert knowledge rather than data-driven training, effectively reducing the computational cost of the model. Finally, by calculating the contribution of each physiological indicator and the maximum impact propagation path, an interpretable myocardial infarction risk assessment report can be constructed, making the model interpretable and thus improving the reliability and credibility of the myocardial infarction risk assessment results.

[0077] The aforementioned construction of a dynamically weighted fuzzy cognitive graph model for myocardial infarction risk assessment can be achieved by using each component of the aggregated feature vector as a concept node in the dynamically weighted fuzzy cognitive graph. This graph also includes concept nodes such as "ST segment abnormality," "myocardial enzyme level," "degree of coronary artery stenosis," and ultimately, "myocardial infarction risk." The edges connecting two concept nodes are called weighted edges. An activation function updates the activation values ​​of each concept node. A weight adjustment mechanism is introduced during model inference to fine-tune the basic weights of the dynamically weighted fuzzy cognitive graph itself based on the context vector. The final activation value of the "myocardial infarction risk" concept node is used as the myocardial infarction risk assessment result. Therefore, the dynamically weighted fuzzy cognitive graph constitutes the model described in this embodiment.

[0078] In some implementations, multimodal physiological indicators comprising electrocardiogram clinical features, blood biomarker features, and radiomics features are extracted, including:

[0079] Baseline correction is performed on the electrocardiogram (ECG) signals obtained from the acquired ECG to obtain the corrected ECG signals.

[0080] Harmonic wavelet packet decomposition was performed on the corrected electrocardiogram signal to obtain the wavelet packet decomposition tree;

[0081] Based on the node coefficients in the wavelet packet decomposition tree, calculate the time-frequency entropy of each node in the wavelet packet decomposition tree, and obtain the optimal wavelet basis by minimizing the time-frequency entropy;

[0082] Extract ECG clinical features including ST segment deviation and T wave symmetry from the optimal wavelet basis;

[0083] The extracted blood biomarkers were standardized to obtain their characteristics;

[0084] Radiomic features containing texture features are extracted from medical images to obtain image radiomic features;

[0085] The clinical features of electrocardiograms, blood biomarkers, and radiomics were combined to construct a multimodal physiological indicator feature.

[0086] In this embodiment, baseline correction can eliminate ECG signal baseline drift and low-frequency noise, ensuring that subsequent features reflect the true physiological state. Through harmonic wavelet packet decomposition and time-frequency entropy guidance, the optimal decomposition basis that best characterizes the signal structure can be adaptively selected, thereby capturing transient change information related to myocardial infarction more precisely and extracting accurate ECG clinical features, blood biomarker features, and radiomics features. This lays a good data foundation for subsequent dynamic weighted fuzzy cognitive graph model for myocardial infarction risk assessment and can improve the accuracy of myocardial infarction risk assessment results.

[0087] In some implementations, the activation value of each concept node is updated using fine-tuned weights and activation functions, including:

[0088] ;

[0089] in, Representing concept nodes In the The new activation value after the next iteration Indicates the first In this iteration, for concept nodes All other concept nodes that are affected activation value, Representing concept nodes Pointing to concept nodes The fine-tuned weights, Representing concept nodes The initial activation value, This represents the activation function. This represents the total number of concept nodes.

[0090] In this embodiment, by considering the influence of all other concept nodes on the current concept node, the dynamically weighted fuzzy cognitive graph model can obtain more accurate myocardial infarction risk assessment results.

[0091] In some implementations, based on the myocardial infarction risk assessment results, the contribution of each physiological indicator in the aggregated feature vector is calculated, including:

[0092] Each physiological indicator in the aggregated feature vector is replaced with a clinical baseline value to construct a counterfactual input vector corresponding to each physiological indicator;

[0093] Input the counterfactual input vector into the dynamic weighted fuzzy cognitive graph model to obtain the counterfactual risk value corresponding to each physiological indicator;

[0094] The contribution of each physiological indicator in the aggregated feature vector is obtained by calculating the difference between the counterfactual risk value and the myocardial infarction risk assessment result.

[0095] In this embodiment, by introducing counterfactual thinking, the marginal effect of each physiological indicator is systematically evaluated, thereby enabling a quantitative assessment of the contribution of each physiological indicator to the risk outcome and the identification of key risk drivers.

[0096] In some implementations, based on preset concept node sensitivity coefficients, path context decay factors, multiple concept nodes, and weighted edges, the maximum influence propagation path between each source node and the target node is found, resulting in at least one maximum influence propagation path, including:

[0097] Construct at least one path between each source node and the target node;

[0098] If each path contains multiple concept nodes and weighted edges, the influence flow between two adjacent concept nodes is calculated based on the preset concept node sensitivity coefficient, the final activation value of the child source node in two adjacent concept nodes, and the fine-tuned weight of the weighted edge between two adjacent concept nodes.

[0099] Based on the path context decay factor, the influence flow between all two adjacent concept nodes in each path is weighted and summed to obtain the total influence of each path;

[0100] Based on the total influence, find the path with the maximum influence propagation between each source node and the target node, and obtain at least one path with the maximum influence propagation.

[0101] In this embodiment, by deeply analyzing the internal reasoning process of the dynamic weighted fuzzy cognitive graph model, the medical logic chain of important reasons is clearly revealed, that is, why the indicators are important is revealed; by introducing the sensitivity coefficient of concept nodes, the complex and nonlinear causal effects in medicine can be simulated, and the contribution of activation values ​​to the influence flow can be adjusted; by introducing the path context decay factor, the contribution of the single-step influence between two adjacent concept nodes to the total influence can be correspondingly reduced, which is more in line with the loss characteristics of information transmission in medical logic.

[0102] In some implementations, an interpretable myocardial infarction risk assessment report is constructed based on the contribution of each physiological indicator and at least one maximum-impact propagation pathway, including:

[0103] By ranking the contribution of each physiological indicator and selecting multiple key physiological indicators, a quantitative attribution summary containing the indicator name, the specific value derived from the aggregated feature vector, and the contribution of each key physiological indicator is obtained.

[0104] Match the maximum impact propagation path corresponding to each key physiological indicator, and obtain the core logical path containing the name and final activation value of each concept node in the maximum impact propagation path;

[0105] Visually encode the final activation value of each concept node and the maximum impact propagation path corresponding to each key physiological indicator to obtain a visualized decision map;

[0106] By integrating quantitative attribution summaries, core logical paths, and visualized decision maps, an interpretable myocardial infarction risk assessment report is constructed.

[0107] In this embodiment, contribution analysis and path activation information are integrated, and key indicators are weighted and ranked based on medical semantics to generate a comprehensive and interpretable report that is well-organized and conforms to clinical logic, thereby enhancing the system's ability to assist medical staff in making decisions.

[0108] In some implementations, after constructing an interpretable myocardial infarction risk assessment report, the method further includes:

[0109] The basic weights are constructed as a prior probability distribution of the weights;

[0110] Based on the prior probability distribution of the weights, the posterior probability distribution of the weights is obtained using Bayes' theorem.

[0111] The mean of the weight distribution obtained from the posterior probability distribution of the weights is used as the new weights in the dynamically weighted fuzzy cognitive graph model to update the model.

[0112] In this embodiment, the mean of the weight distribution obtained from the posterior probability distribution of the weights is used as the new weights in the dynamically weighted fuzzy cognitive graph model to update the model. This allows the dynamically weighted fuzzy cognitive graph model to have an uncertainty measure of its own knowledge. A single expert feedback will not drastically change the dynamically weighted fuzzy cognitive graph model, but will be smoothly integrated with the original knowledge. Furthermore, each update has a clear record of the Bayesian inference process, which has data traceability.

[0113] To facilitate understanding by those skilled in the art, a set of preferred embodiments is provided below:

[0114] Existing technologies include data-driven black-box evaluation models and post-analysis methods for the interpretability of black-box models. Specifically, data-driven black-box evaluation models are represented by deep learning algorithms such as deep neural networks (DNNs) and convolutional neural networks (CNNs). Their core is to learn the nonlinear mapping relationship between raw input data (such as ECG waveforms) and risk conclusions by training on large-scale labeled datasets. While such models can achieve high predictive accuracy, their decision-making process is deeply hidden within millions of model parameters, forming a "black box." Doctors cannot know which specific clinical indicators the model relies on or what medical logic chain leads to its final conclusion, making it difficult to fully rely on the model's output in critical clinical decisions.

[0115] Post-interpretability analysis methods for black-box models, such as gradient attribution analysis and hidden semantic analysis, attempt to explain the model's behavior after it has made predictions by calculating the contribution or sensitivity of input features to the output. Their limitation lies in the fact that these explanations are essentially external observations and approximate inferences of a known, opaque model, rather than a direct representation of the model's internal decision-making logic. They can reveal "what the model focuses on," but cannot clearly answer "why the model makes those decisions," and the computation of these methods usually incurs additional system overhead, increasing the overall complexity of the model.

[0116] The aforementioned prior art has the following disadvantages:

[0117] 1. The decision-making process lacks internal explainability and medical logic.

[0118] The core flaw of existing technologies lies in the fact that their data-driven nature dictates that their decision-making logic is statistical, rather than based on medical causal knowledge. Internally, the model is a complex mathematical function that lacks explicit medical concepts and reasoning rules for scrutiny, leading to a fundamental difference between its decision-making process and the thinking paradigm of clinicians, making it difficult to establish trust. Post-explanation techniques can only provide limited clues and cannot fundamentally make the model "transparent."

[0119] 2. It is highly dependent on large-scale data for training and has high computational costs.

[0120] Building deep learning models requires significant computational resources and time to train on large-scale, high-quality labeled medical data. Once trained, the model's knowledge is embedded in the network weights. When medical knowledge is updated or fine-tuning is needed for specific situations, costly data collection and model retraining are often required.

[0121] 3. Difficulty in updating and calibrating knowledge.

[0122] In data-driven models, medical knowledge is implicitly encoded and cannot be directly and precisely modified. If a certain judgment logic in the model is found to be inconsistent with the latest clinical guidelines, researchers cannot directly correct the model as they would modify rules, lacking a flexible and efficient model knowledge iteration and maintenance mechanism led by experts.

[0123] To address the shortcomings of existing technologies, this embodiment aims to provide a multimodal interpretability-based method for myocardial infarction risk assessment. By proposing an innovative, expert-knowledge-driven approach instead of a data-driven one, it fundamentally overcomes the inherent deficiencies of existing technologies in terms of interpretability, computational efficiency, and knowledge maintenance. This embodiment aims to provide an intelligent risk assessment tool that is both highly accurate and completely transparent, ensuring that physicians can understand and trust every decision made by the model, thereby significantly improving the clinical application value and reliability of intelligent auxiliary diagnostic systems for myocardial infarction. The main objective of this embodiment is:

[0124] 1. Provide an inherently transparent and medically knowledge-driven risk assessment framework.

[0125] This embodiment aims to construct an assessment engine centered on an expert-predefined Dynamic Weighted Fuzzy Cognitive Graph (DW-FCM). The concept nodes and weighted edges of this engine directly correspond to explicit medical concepts and causal relationships, making the entire risk reasoning process completely transparent. Its logic is consistent with clinical thinking, fundamentally solving the "black box" problem.

[0126] 2. Risk assessment with zero training and low computational overhead has been achieved.

[0127] This embodiment aims to completely eliminate the reliance on large-scale data and computationally intensive training processes by directly transforming medical expert knowledge into a model structure. This not only significantly reduces the cost of building and deploying the system model but also enables real-time evaluation in resource-constrained environments.

[0128] 3. Establish a dynamic knowledge calibration mechanism led by experts.

[0129] This embodiment aims to introduce an expert feedback-driven parameter calibration loop (EFPCL). This mechanism allows medical experts to directly modify and improve the concepts, causal relationships, and their strength in the evaluation model based on the latest clinical evidence or practical experience, ensuring that the model can keep pace with the times and achieve continuous and traceable evolution of knowledge.

[0130] 4. It enables multi-level and in-depth interpretation of risk assessment results.

[0131] This embodiment not only makes the model itself transparent, but also aims to provide a systematic interpretation mechanism. It quantifies the importance of each input indicator through structural perturbation counterfactual contribution analysis (SP-CCA), and clearly displays the complete logical chain of conclusions through decision path (i.e., maximum influence propagation path) tracing and concept activation visualization (DPT-CAV), providing comprehensive, in-depth, and reliable auxiliary information for clinical decision-making.

[0132] This embodiment addresses the needs of myocardial infarction risk assessment and interpretability by constructing a two-stage framework entirely driven by expert knowledge. This framework directly constructs model parameters through expert knowledge rather than data-driven training. The first stage (MPIR-KRA) utilizes Multimodal Physiological Indicator Refinement (MPIR) techniques, including adaptive baseline correction, multi-scale wavelet transform (ABC-MWT), and radiomics feature extraction (PRFE), to transform raw electrocardiogram (ECG), blood biomarkers, and imaging data into structured clinical indicators. These indicators are then input into a Dynamically Weighted Fuzzy Cognitive Map (DW-FCM) model for risk inference. The second phase (MIMA-KC) focuses on the output results of the DW-FCM model. It employs two complementary modules, Structural Perturbation Counterfactual Contribution Analysis (SP-CCA) and Decision Path Tracing Concept Activation Visualization (DPT-CAV), to quantify "which indicator is important" and "why it is so important." The two are then integrated to generate a comprehensive explanatory report through Semantic Guided Contribution Focusing (SGCF). Finally, an Expert Feedback-Driven Parameter Calibration Loop (EFPCL) is introduced to continuously optimize the DW-FCM structure and weights without data retraining, ensuring the accuracy and credibility of the system model in the evolution of medical knowledge.

[0133] The method in this embodiment specifically includes the following:

[0134] Phase 1: Multimodal physiological indicator refinement and knowledge-driven risk assessment (MPIR-KRA).

[0135] This phase constructs an end-to-end knowledge flow from raw signals to risk assessment. First, the Multimodal Physiological Indicator Extraction and Quantification (MPIR) module generates a unified indicator vector through steps such as adaptive baseline correction, multi-scale wavelet decomposition, clinical feature extraction, blood biomarker standardization, and radiomics extraction. Subsequently, the risk assessment engine based on Dynamically Weighted Fuzzy Cognitive Graph (DW-FCM) utilizes expert-predefined medical concepts and causal weights to iteratively activate and propagate between concept nodes, outputting myocardial infarction risk values, thus building the model skeleton and input / output anchors for subsequent interpretability analysis.

[0136] Step 1: Adaptive Harmonic Wavelet Packet Transform (AHWPT).

[0137] (1) Objective: Traditional multi-scale wavelet decomposition uses a fixed wavelet basis, which is difficult to optimally fit the complex and non-stationary pathological features in electrocardiogram (ECG) signals. This step uses an adaptive harmonic wavelet packet transform (AHWPT) to adaptively select the optimal decomposition basis (i.e., the optimal wavelet basis) that best represents the signal structure through a time-frequency entropy-guided mechanism, thereby capturing transient change information related to myocardial infarction more precisely.

[0138] (2) Formula and description:

[0139] Baseline correction: ,in, This represents the adaptive baseline correction operator.

[0140] Adaptive harmonic wavelet packet transform: First, the baseline-corrected ECG signal... Perform harmonic wavelet packet decomposition to obtain a complete wavelet packet decomposition tree. For any node in the decomposition tree... ( The number of decomposition layers, (This is the frequency band index of this layer), and its node coefficients are... .

[0141] Then, calculate the time-frequency entropy of each node. Time-frequency entropy is used to measure the uncertainty or complexity of a signal within a corresponding frequency band.

[0142] ;

[0143] in, , is the normalized energy probability distribution.

[0144] Finally, based on the principle of minimizing time-frequency entropy, pruning is performed on the wavelet packet decomposition tree to find a path that minimizes the sum of time-frequency entropy of all terminal nodes. This path constitutes the optimal wavelet basis.

[0145] .

[0146] in, Let represent a complete basis (i.e., a set of terminal nodes covering the entire signal frequency band) in a wavelet packet decomposition tree. The optimization process is performed on all possible bases. Find the basis in the set that minimizes the sum of the time-frequency entropy of the terminal nodes.

[0147] (3) Inputs / outputs and variables:

[0148] enter: .

[0149] Output: and optimal wavelet basis .

[0150] symbol: It is the raw ECG vector, representing the raw ECG signal vector collected at time t; It is the corrected vector.

[0151] Step 2: Extraction of clinical features from electrocardiogram (ECG).

[0152] (1) Objective: To obtain the optimal wavelet basis Clinically relevant physiological indicators (including ST segment deviation and T wave symmetry) are extracted and quantified to provide input for subsequent knowledge reasoning.

[0153] (2) Formula and description:

[0154] ST segment offset: ,in, Fixed time interval after J-level point Potential at the location, This indicates the time point from point J, pre-defined according to clinical standards, for measuring ST segment shift. This is the baseline potential.

[0155] T-wave symmetry: , in, , and These represent the time coordinates of the start, peak, and end points of the T-wave, respectively.

[0156] (3) Inputs / outputs and variables:

[0157] Input: The optimal wavelet basis generated in step 1 .

[0158] Output: Clinical feature vector This vector consists of one or more ECG clinical features ( and It is composed of splicing parts.

[0159] symbol: Indicates the ST segment offset; Indicates T-wave symmetry; J-level point time: The time coordinate of the turning point in the electrocardiogram where the QRS complex ends and the ST segment begins; Indicates the baseline potential. This indicates the number of sampling points in the original electrocardiogram signal vector.

[0160] Step 3: Standardization of blood biomarkers.

[0161] (1) Purpose: To eliminate dimensions and remove anomalies in order to improve the comparability of various markers.

[0162] (2) Formula and description:

[0163] ;

[0164] in, This is the raw data vector collected at the current moment, containing one or more blood biomarkers related to myocardial infarction assessment. To eliminate the influence of different biomarker dimensions and perform normalization, this embodiment uses the Z-score standardization method. The statistical parameters required by this method (mean) with standard deviation The result was obtained through statistical calculations on a dataset of similar blood biomarkers collected from a large number of historical cases. Finally, The function limits the standardized result to the range [-3,3] to handle potential extreme outliers.

[0165] (3) Inputs / outputs and variables:

[0166] enter: , This indicates the quantity of blood biomarkers collected. This represents the vector of raw blood biomarker data collected at time t.

[0167] Output: This represents the vector of raw blood biomarker data collected at time t.

[0168] symbol: This represents the statistical parameters of blood biomarkers calculated from a large amount of historical case data, used for Z-score standardization.

[0169] Step 4: Extraction of radiomics features from images.

[0170] (1) Objective: This step aims to extract quantified texture features from medical images. This is achieved by calculating the Gray-Level Co-occurrence Matrix (GLCM) and deriving multiple features based on this matrix. Elements in the GLCM Indicates the grayscale value in the image The pixel and grayscale values The probability that pixels co-occur under a specific spatial relationship.

[0171] (2) Formula and description:

[0172] A series of radiomics features can be extracted from GLCM, such as:

[0173] Contrast: Measures the degree of drastic local changes in an image, reflecting the sharpness of textures and the depth of grooves.

[0174] ;

[0175] Energy (Angular Second Moment, ASM): Measures the uniformity of image texture. The higher the energy value, the more regular and uniform the texture distribution.

[0176] ;

[0177] (3) Inputs / outputs and variables:

[0178] Input: Original medical images , This indicates the dimensions of the original medical image.

[0179] Output: Radiomics feature vector This vector is composed of one or more extracted texture features, for example... .

[0180] symbol: This indicates that the gray-level co-occurrence matrix consists of gray-level pairs. Defined element values ​​(probabilities); This represents a set (vector) containing all extracted image texture features.

[0181] Step 5: Adaptive feature fusion based on attention mechanism.

[0182] (1) Purpose: The indicative role of physiological indicators of different modalities in myocardial infarction risk is not constant, and their importance may vary due to individual differences. Simple feature splicing cannot reflect this dynamism. This step introduces an attention mechanism module to automatically learn the weights of features of each modality and realize adaptive weighted fusion of information.

[0183] (2) Formula and description:

[0184] The obtained electrocardiogram clinical features, blood biomarker features and imaging features Concatenate into a single feature vector .

[0185] Then, an attention score for each feature is calculated using a small neural network (such as a fully connected layer). :

[0186] ;

[0187] in, and Represents the learnable parameters. The function ensures that the sum of all scores is 1.

[0188] Final aggregated feature vector It is a weighted sum of the original eigenvectors:

[0189]

[0190] in, , and Attention scores are represented by ECG clinical features, blood biomarker features, and imaging features, respectively.

[0191] The parameters of this attention module can be obtained by pre-training on relevant datasets or optimized in the expert feedback loop.

[0192] (3) Inputs / outputs and variables:

[0193] enter: .

[0194] Output: Aggregated feature vector The comprehensive feature vector is formed by weighted summation of clinical features of electrocardiogram, blood biomarker features and imaging features.

[0195] Step 6: DW-FCM Risk Reasoning.

[0196] (1) Purpose: This step is the core of risk assessment, aiming to use a dynamically weighted fuzzy cognitive graph model (DW-FCM model) constructed from medical expert knowledge to comprehensively reason about the multimodal feature vectors aggregated in step 5. The DW-FCM model simulates the thought process of clinical experts when diagnosing myocardial infarction, transforming the input physiological indicators into the final quantitative risk value through the causal relationships and dynamic weights between conceptual nodes. Its purpose is to achieve a transparent, interpretable risk assessment process that does not require data training.

[0197] (2) Formula and description:

[0198] DW-FCM is a directed graph consisting of a set of concept nodes. and the weighted edges connecting these nodes constitute.

[0199] Concept nodes (): Represents key concepts in the medical field, such as "ST segment abnormalities," "myocardial enzyme levels," "degree of coronary artery stenosis," and ultimately, the "risk of myocardial infarction." Each concept node... Each has an activation value. , indicating at time The activity or quantification level of this concept node typically ranges from [value range missing]. .

[0200] Weighted edges (Weights, Basic weights Representing concept nodes For concept nodes The strength and nature of the effect (positive promotion, negative inhibition, or no effect).

[0201] Context-adaptive weight modulation mechanism: Unlike traditional static FCM, the weights in this embodiment can be dynamically fine-tuned based on specific input indicators (such as patient age, gender, and past medical history, etc.), making it more personalized. However, its core reasoning process is still based on the following iterative calculation. Specifically,

[0202] To achieve personalized evaluation, this embodiment introduces a weight modulation mechanism into the standard DW-FCM inference. The fine-tuned weight matrix... Including fine-tuned weights, the system receives a context vector containing patient background information (such as age, gender, and past medical history). Basic weights Based on vector Fine-tuning is performed to obtain the adjusted weights:

[0203] ;

[0204] in, Representing concept nodes Pointing to concept nodes The base weights are predefined by experts and are the weights of the dynamically weighted fuzzy cognitive graph model itself. It is a modulation function (e.g., a small neural network) that modulates based on the context vector. Output a specific connection The modulation coefficients are used. This allows the strength of causal inference within the model to dynamically change for patients of different ages or genders, making it more consistent with clinical reality. Subsequent iterative calculations will use the modulated weights. .

[0205] (3) Risk reasoning process:

[0206] Risk reasoning is an iterative computational process that continuously updates the activation values ​​of each concept node to simulate the propagation of causal chains until the system model reaches a stable state. Specifically:

[0207] 1) Initialization: Initialize the aggregated feature vector generated in step 5. Each component of the aggregated feature vector is mapped to a corresponding input layer concept node in the DW-FCM graph. In other words, each component of the aggregated feature vector is treated as a node in the DW-FCM graph, and each component of the aggregated feature vector is used as the initial activation value. .

[0208] 2) Iterative calculation: in each iteration step Each concept node activation value Update according to the following formula:

[0209] ;

[0210] in, Representing concept nodes In the The new activation value after the next iteration This represents the total number of concept nodes. Indicates the first In this iteration, for concept nodes All other concept nodes that are affected activation value, Representing concept nodes Pointing to concept nodes The fine-tuned weights represent right The strength of causal influence Representing concept nodes The initial activation value (if it is an input node) ensures that external input information continues to play a role in each iteration. This represents a nonlinear transformation function (or activation function) used to compress the calculation results to a specific interval (e.g., ...). Common choices are the Sigmoid function or the tanh function, but other activation functions known to those skilled in the art can also be used; this embodiment does not impose specific limitations. For example, using the Sigmoid function:

[0211] ;

[0212] in, It is a parameter that controls the slope of the function, and its sensitivity can be adjusted.

[0213] 3) Termination condition: The iteration process continues until the change in the activation value of all nodes is less than a preset minimum threshold. Or reach the maximum number of iterations. At this point, the system is considered to have reached a stable state. The final "myocardial infarction risk" node is then obtained. The activation value is the final risk result of this assessment (i.e., the myocardial infarction risk assessment result). .

[0214] (4) Inputs / outputs and variables:

[0215] enter:

[0216] Aggregated feature vectors .

[0217] Predefined DW-FCM model, i.e., model This includes: a set of concept nodes. The fundamental weight matrix constituted by expert knowledge. .

[0218] Output:

[0219] Final myocardial infarction risk assessment value .

[0220] The activation vectors of all concept nodes when a stable state is reached. This vector will serve as input for the second phase of interpretability analysis.

[0221] symbol:

[0222] For concept nodes In the The activation value of the next iteration; Representing concept nodes To the concept node The basic weights. This represents an activation function (such as Sigmoid). This represents the final output myocardial infarction risk assessment value.

[0223] In summary, this embodiment has achieved the refinement and risk assessment process of multimodal data (such as electrocardiogram signals, blood biomarkers, and imaging data) in this stage.

[0224] Phase Two: Construction of Multi-Level Interpretability Mechanisms and Knowledge Calibration.

[0225] This phase, building upon the DW-FCM risk assessment engine constructed in the first phase, focuses on providing comprehensive, actionable, and interpretable risk assessment reports, and establishing a non-data-driven system maintenance mechanism. The core is an interpretability mechanism that integrates feature contribution and decision path analysis. This mechanism is achieved through two complementary analytical techniques: first, Structural Perturbation Counterfactual Contribution Analysis (SP-CCA) quantifies the impact of each extracted input physiological indicator (i.e., the various features extracted in the first phase), precisely identifying which indicator is more important; second, Decision Path Tracing and Concept Activation Visualization (DPT-CAV) deeply analyzes the internal reasoning process of the constructed DW-FCM model, clearly revealing the medical logic chain of important causes. The analytical results from these two dimensions are intelligently integrated through an innovative Semantic Guided Contribution Focusing (SGCF) mechanism to generate a comprehensive explanatory report (i.e., a myocardial infarction risk assessment report) that connects quantitative attribution and symbolic logic. Finally, to ensure the long-term effectiveness and accuracy of the DW-FCM model, this embodiment introduces an Expert Feedback-Based Parameter Calibration Loop (EFPCL). This allows medical experts to directly and evidence-based manually adjust the weights of weighted edges and concept nodes in the DW-FCM model based on clinical feedback, thereby achieving adaptive optimization of the system during the evolution of medical knowledge. This process is completely independent of data retraining. Specifically:

[0226] 1. An interpretability mechanism that integrates feature contribution and decision path analysis.

[0227] 1.1 Counterfactual contribution analysis based on structural perturbation (SP-CCA).

[0228] (1) Objective: To obtain the final risk assessment value in the first stage (MPIR-KRA). Subsequently, the primary interpretative requirement is to quantify the contribution of each raw input physiological indicator to the outcome. This step systematically assesses the marginal effect of each indicator by introducing counterfactual thinking, thereby identifying the risk drivers of key indicators.

[0229] (2) Formula and description:

[0230] The core of this embodiment lies in comparing the risk assessment results under the original input with the risk assessment results after removing or neutralizing the influence of a specific indicator. This is achieved using aggregated feature vectors. The final risk value obtained is ,in, Represents aggregated feature vectors Include The first physiological indicator. To calculate the first... Physiological indicators Contribution Construct a counterfactual input vector , of which Each indicator is compared to its clinical baseline value (or neutral value, such as the median of the normal range). Replacement is performed while other indicators remain unchanged. The specific calculation formula is as follows:

[0231] ;

[0232] Subsequently, the counterfactual input vector The data is re-entered into the DW-FCM risk assessment engine for a complete iterative calculation until it stabilizes, yielding a new counterfactual risk value. . No. Contribution of each physiological indicator Defined as the difference between the original risk and the counterfactual risk:

[0233] ;

[0234] The output value of the formula It intuitively quantifies physiological indicators. The presence of this indicator has a net impact on the final risk value. A positive value indicates that the indicator increases the risk, while a negative value indicates that it decreases the risk. The absolute value represents the strength of the physiological indicator's influence.

[0235] 3. Inputs / outputs and variables:

[0236] enter:

[0237] The aggregated original feature vector .

[0238] The predefined DW-FCM model includes a set of concept nodes C and a basic weight matrix W.

[0239] Final risk assessment value The final activation value of a concept node in a stable state. .

[0240] Output:

[0241] An M-dimensional contribution vector Each element corresponds to the contribution of an input physiological indicator.

[0242] 1.2 Decision Path Tracing and Concept Activation Visualization (DPT-CAV).

[0243] (1) Purpose: SP-CCA reveals "which indicator is important," but fails to elucidate the internal logical pathway by which it affects risk assessment. This step aims to address this issue from the steady state of the DW-FCM model. Starting with a novel nonlinear influence propagation model, this study traces back and identifies the key causal chain from high-contribution input nodes to the final risk output node, thereby answering the deep mechanism of "why it matters".

[0244] (2) Formula and description:

[0245] This embodiment seeks the maximum influence propagation path on the DW-FCM graph structure. Instead of using simple linear multiplication to calculate the influence flow, it defines a nonlinear influence propagation function. This function better simulates complex, nonlinear causal interactions in medicine, such as threshold and saturation effects. (From concept nodes...) (That is, it can be regarded as the source node) passed to the concept node The influence flow (which can be viewed as the target node) can be quantified as follows:

[0246] ;

[0247] in, It is the source node The final activation value, It is a concept node For concept nodes The intensity of the influence, i.e. and The connection weights between them (i.e., the fine-tuned weights). This represents the sensitivity coefficient of a conceptual node, which is preset by experts based on the physiological characteristics of the node (for example, even a small change in certain myocardial enzyme markers can trigger a drastic response). The value is relatively high, and it is used to adjust the contribution of the activation value to the flow. The introduction of the hyperbolic tangent function causes the influence flow to grow non-linearly and tend to saturate when the activation value is very high, which is consistent with the physiological limit.

[0248] A complete decision-making path It is a sequence of concept nodes ,in It is usually a high-contribution input concept node (that is, it can be regarded as a sub-source node). It is the final risk concept node (i.e., it can be viewed as a sub-objective node), the decision path. Include A concept node, Indicates the first Concept Nodes , Indicates the first Concept Nodes Total influence of the path Defined as a weighted summation of the influence flows of each segment along the path, for example, by adding them together:

[0249] ;

[0250] in, express Path context decay factor for each concept node This factor makes in the causal chain The farther away a concept node is, the less its contribution to the total impact will be, which is more in line with the loss characteristics of information transmission in medical logic.

[0251] This step employs a method starting from the risk output node. The initial greedy best-first search algorithm is used in reverse tracing. At each step, the algorithm selects the flow with the greatest influence. Precursor concept node This process continues from the previous concept node in the path, tracing back to a specific input concept node. This method can identify one or more decision paths that contribute most to the final risk activation value. If only one path from the source node to the target node needs to be found, there will be a decision path that contributes the most to the final risk activation value. If multiple paths from the source node to the target node need to be found, there will be multiple decision paths that contribute the most to the final risk activation value.

[0252] (3) Inputs / outputs and variables:

[0253] enter:

[0254] Fine-tuned weight matrix and concept node sensitivity coefficient .

[0255] The final activation value of a concept node in a stable state .

[0256] High-contribution input indicators and their corresponding concept nodes identified by SP-CCA.

[0257] Output:

[0258] One or more of the most influential propagation paths The set. Each path consists of an ordered list of medical concept nodes and their corresponding influence flow values.

[0259] 1.3 Semantic-guided contribution focus (SGCF) mechanism.

[0260] (1) Motivation: The separate contribution values ​​(from SP-CCA) and logical pathways (from DPT-CAV) are still fragmented information for clinicians. This step aims to deeply integrate the quantitative attribution results of SP-CCA with the qualitative logical pathways of DPT-CAV, and transform the analysis results into a structured, semantically coherent, and clinically readable comprehensive explanatory report through a standardized report generation process.

[0261] (2) Formula and description:

[0262] The report generation process of this mechanism begins with the acquisition of the contribution vector. Process by its absolute value Sort in descending order and based on a preset threshold. Screening out the set of key physiological indicators The process then sequentially builds the report's three core modules: First, it generates the "Quantitative Attribution Summary" module, which is designed for... Each key physiological indicator clearly lists its name and origin from the input vector. The specific values ​​and contribution scores calculated by SP-CCA Following this, to provide in-depth logical analysis, the system created the "core logic path" module, which... Each key physiological indicator in the decision path set Search and match the total influence The largest path (i.e., a key physiological indicator corresponds to a path with the greatest impact) is presented in the form of a textual causal chain, clearly indicating the names of each intermediate concept node in the path and its final activation value. To further provide a global perspective, the system continues to generate a "Visualized Decision Graph" module. This module renders the complete DW-FCM diagram and utilizes visual encoding (such as mapping activation values ​​by concept node size or color intensity). The risk propagation process is visually illustrated using path highlighting (highlighting identified core decision paths with special colors and line thicknesses). Finally, the three modules—quantitative attribution summary, core logical path, and visualized decision map—are integrated into a single final risk assessment value. Output a comprehensive explanatory report with a standardized title.

[0263] (3) Inputs / outputs and variables:

[0264] enter:

[0265] Contribution vector .

[0266] Set of paths with the greatest impact .

[0267] Original input vector The basic weight matrix of DW-FCM and the final activation value of the concept node in a stable state. .

[0268] Output:

[0269] A standardized, comprehensive, and interpretable report that includes quantitative attribution, textual logical paths, and visual graphs.

[0270] 2. Expert knowledge calibration loop based on Bayesian fusion.

[0271] (1) Motivation: Any model based on prior knowledge must be able to evolve with the development of medical knowledge. In order to make the knowledge update process more rigorous and mathematically grounded, this embodiment proposes an expert knowledge calibration mechanism based on Bayesian fusion.

[0272] (2) Formula and description:

[0273] Within this framework, each basic weight in the dynamically weighted fuzzy cognitive graph model It is no longer a fixed value, but is regarded as a random variable that follows a specific prior probability distribution (such as a Gaussian distribution): ,in, This represents the mean of the prior distribution of the current weights. This represents the variance of the prior distribution of the current weights.

[0274] When experts provide feedback (e.g., "I think indicator A should have a stronger impact on indicator B"), this feedback is formalized as a likelihood function. ,in, Feedback evidence representing experts.

[0275] According to Bayes' theorem, prior knowledge and new evidence can be combined to obtain the posterior probability distribution of the weights:

[0276] ;

[0277] in, This represents the posterior probability distribution of the weights. This indicates a direct proportion. This represents the prior probability distribution of the weights.

[0278] Updated weight distribution mean This will be used as a new weight value in the dynamically weighted fuzzy cognitive graph model. The advantages of this approach are:

[0279] 1) Quantifying uncertainty: The dynamic weighted fuzzy cognitive graph model has an uncertainty measure for its own knowledge.

[0280] 2) Incremental update: Through Bayesian inference, expert feedback is regarded as new evidence and combined with the model's existing prior knowledge, so that the model weights are adjusted incrementally and quantifiable in the form of posterior probabilities, rather than being updated abruptly.

[0281] 3) Traceability: Each update has a clear record of the Bayesian inference process.

[0282] The update process of the dynamically weighted fuzzy cognitive graph model can be represented as:

[0283] ;

[0284] in, It is the current dynamic weighted fuzzy cognitive graph model (defined by weight distribution). It is the experts Evidence provided at all times.

[0285] Compared with the prior art, the technical solution of this embodiment has the following advantages:

[0286] 1. By utilizing various modal data processing techniques such as adaptive baseline correction (ABC), adaptive harmonic wavelet packet transform, clinical feature extraction, blood biomarker standardization, and radiomics feature extraction, a unified quantitative representation of raw electrocardiogram signals, blood biomarkers, and medical images is achieved, constructing a comprehensive and structured clinical indicator input, providing a reliable foundation for subsequent risk assessment.

[0287] 2. An innovative fuzzy cognitive graph with dynamic weight adjustment mechanism is designed. Through concept nodes and causal weights predefined by medical experts, it enables iterative reasoning of complex nonlinear causal relationships among multimodal indicators, outputting accurate and interpretable myocardial infarction risk assessment results, and possessing knowledge-driven characteristics that do not depend on training data.

[0288] 3. A multi-level interpretability mechanism combining Structural Perturbation Counterfactual Contribution Analysis (SP-CCA) and Decision Path Source Tracing Activation Visualization (DPT-CAV). By counterfactually perturbing the input features of the risk assessment model, the marginal contribution of each indicator is quantified. Combined with graph-based optimal path backward search and node activation visualization, it reveals "which indicator is important" and "why it is important," achieving a deep and transparent interpretation of the risk assessment results.

[0289] 4. By integrating contribution analysis with path activation information, and weighting and ranking key indicators based on medical semantics, a comprehensive explanatory report that is clear, logically consistent with clinical practice is generated, thereby enhancing the system model's ability to support medical staff in making decisions.

[0290] 5. An innovative expert knowledge calibration loop mechanism based on Bayesian fusion is introduced, which eliminates the need to retrain the model and ensures that the system model can continuously optimize risk assessment performance and interpretation accuracy as medical knowledge is updated, thereby improving the long-term adaptability and reliability of the system model.

[0291] Reference Figure 2 This application also provides a myocardial infarction risk assessment system based on multimodal interpretability. The system includes a data extraction unit 201, a feature splicing unit 202, a feature aggregation unit 203, a model building unit 204, a risk assessment unit 205, a contribution calculation unit 206, a path finding unit 207, and an assessment report construction unit 208, wherein:

[0292] Data extraction unit 201 is used to extract multimodal physiological indicators including electrocardiogram clinical features, blood biomarker features and radiomics features;

[0293] The feature splicing unit 202 is used to splice the features in the multimodal physiological index features to obtain the spliced ​​feature vector;

[0294] The feature aggregation unit 203 is used to calculate the attention score of each feature in the multimodal physiological index features based on the spliced ​​feature vector, and to perform a weighted summation of each feature in the multimodal physiological index features based on the attention score to obtain the aggregated feature vector.

[0295] Model building unit 204 is used to build a dynamic weighted fuzzy cognitive graph model for myocardial infarction risk assessment. The dynamic weighted fuzzy cognitive graph model includes a weight adjustment mechanism, multiple concept nodes, weighted edges and activation functions. The weight adjustment mechanism is used to fine-tune the basic weights based on the context vector to obtain the fine-tuned weights. The weighted edges are the edges connecting two concept nodes and are used to characterize the influence strength of one concept node on another concept node.

[0296] The risk assessment unit 205 is used to input the aggregated feature vector into the dynamic weighted fuzzy cognitive graph model, update the activation value of each concept node with the fine-tuned weights and activation function, and obtain the final activation value of each concept node after the preset conditions are met, and use the final activation value of the myocardial infarction risk concept node as the myocardial infarction risk assessment result.

[0297] The contribution calculation unit 206 is used to calculate the contribution of each physiological indicator in the aggregated feature vector based on the myocardial infarction risk assessment results.

[0298] The path finding unit 207 is used to find the maximum influence propagation path between each source node and the target node based on the preset concept node sensitivity coefficient, path context decay factor, multiple concept nodes and weighted edges, so as to obtain at least one maximum influence propagation path.

[0299] Assessment report construction unit 208 is used to construct an interpretable myocardial infarction risk assessment report based on the contribution of each physiological indicator and at least one maximum impact propagation path.

[0300] It should be noted that since the myocardial infarction risk assessment system based on multimodal interpretability in this embodiment is based on the same inventive concept as the myocardial infarction risk assessment method based on multimodal interpretability described above, the corresponding content in the method embodiment is also applicable to this system embodiment, and will not be described in detail here.

[0301] Reference Figure 3 This application also provides an electronic device, which includes:

[0302] At least one memory;

[0303] At least one processor;

[0304] At least one program;

[0305] The program is stored in memory, and the processor executes at least one program to implement the multimodal interpretability-based myocardial infarction risk assessment method described above in this disclosure.

[0306] This electronic device can be any smart terminal, including mobile phones, tablets, personal digital assistants (PDAs), and in-vehicle computers.

[0307] The electronic devices according to embodiments of this application will now be described in detail.

[0308] The processor 1600 can be implemented using a general-purpose central processing unit (CPU), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this disclosure.

[0309] The memory 1700 can be implemented as a read-only memory (ROM), static storage device, dynamic storage device, or random access memory (RAM). The memory 1700 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1700 and is called and executed by the processor 1600 to execute the multimodal interpretability-based myocardial infarction risk assessment method of this disclosure.

[0310] The input / output interface 1800 is used to implement information input and output.

[0311] The communication interface 1900 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0312] Bus 2000 transmits information between various components of the device (e.g., processor 1600, memory 1700, input / output interface 1800, and communication interface 1900);

[0313] The processor 1600, memory 1700, input / output interface 1800 and communication interface 1900 are connected to each other within the device via bus 2000.

[0314] This disclosure also provides a storage medium, which is a computer-readable storage medium storing computer-executable instructions for causing a computer to execute the above-described myocardial infarction risk assessment method based on multimodal interpretability.

[0315] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0316] The embodiments described in this disclosure are for the purpose of more clearly illustrating the technical solutions of this disclosure and do not constitute a limitation on the technical solutions provided by this disclosure. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by this disclosure are also applicable to similar technical problems.

[0317] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this disclosure, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0318] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0319] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0320] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0321] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0322] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0323] The units described as separate components may or may not be physically separate. The 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 the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0324] Furthermore, the functional units in the various embodiments of this 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 integrated unit can be implemented in hardware or as a software functional unit.

[0325] If the integrated unit is implemented as 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 this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause an electronic device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks. The embodiments of this application have been described in detail above with reference to the accompanying drawings, but this application is not limited to the above embodiments. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of this application.

[0326] The embodiments of this application have been described in detail above with reference to the accompanying drawings. However, this application is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of this application.

Claims

1. A method for assessing myocardial infarction risk based on multimodal interpretability, characterized in that, The method includes: Extract multimodal physiological indicators that include electrocardiogram clinical features, blood biomarker features, and radiomics features; The features in the multimodal physiological index features are concatenated to obtain a concatenated feature vector; Based on the spliced ​​feature vector, the attention score of each feature in the multimodal physiological indicator features is calculated, and based on the attention score, the weighted sum of each feature in the multimodal physiological indicator features is performed to obtain the aggregated feature vector. A dynamic weighted fuzzy cognitive graph model for myocardial infarction risk assessment is constructed. The dynamic weighted fuzzy cognitive graph model includes a weight adjustment mechanism, multiple concept nodes, weighted edges, and an activation function. The weight adjustment mechanism is used to fine-tune the basic weights based on the context vector to obtain the fine-tuned weights. The weighted edges are edges connecting two concept nodes and are used to characterize the influence strength of one concept node on another concept node. The aggregated feature vector is input into the dynamic weighted fuzzy cognitive graph model, and the activation value of each concept node is updated using the fine-tuned weights and activation function until the preset conditions are met to obtain the final activation value of each concept node. The final activation value of the myocardial infarction risk concept node is used as the myocardial infarction risk assessment result. Based on the myocardial infarction risk assessment results, the contribution of each physiological indicator in the aggregated feature vector is calculated; Based on the preset concept node sensitivity coefficient, path context decay factor, the multiple concept nodes and the weighted edges, find the maximum influence propagation path between each source node and the target node, and obtain at least one maximum influence propagation path; Based on the contribution of each physiological indicator and the at least one maximum impact propagation path, construct an interpretable myocardial infarction risk assessment report.

2. The myocardial infarction risk assessment method based on multimodal interpretability according to claim 1, characterized in that, The extraction of multimodal physiological indicators, including electrocardiogram clinical features, blood biomarker features, and radiomics features, includes: Baseline correction is performed on the electrocardiogram (ECG) signals obtained from the acquired ECG to obtain the corrected ECG signals. The corrected electrocardiogram signal is subjected to harmonic wavelet packet decomposition to obtain a wavelet packet decomposition tree; Based on the node coefficients in the wavelet packet decomposition tree, the time-frequency entropy of each node in the wavelet packet decomposition tree is calculated, and the optimal wavelet basis is obtained by minimizing the time-frequency entropy. Extract ECG clinical features including ST segment shift and T wave symmetry from the optimal wavelet basis; The extracted blood biomarkers were standardized to obtain their characteristics; Radiomic features containing texture features are extracted from medical images to obtain image radiomic features; The electrocardiogram clinical features, the blood biomarker features, and the radiomics features are constructed into a multimodal physiological indicator feature.

3. The myocardial infarction risk assessment method based on multimodal interpretability according to claim 1, characterized in that, The step of updating the activation value of each concept node using fine-tuned weights and activation functions includes: ； in, Representing concept nodes In the The new activation value after the next iteration Indicates the first In this iteration, for concept nodes All other concept nodes that are affected activation value, Representing concept nodes Pointing to concept nodes The fine-tuned weights, Representing concept nodes The initial activation value, This represents the activation function. This represents the total number of concept nodes.

4. The myocardial infarction risk assessment method based on multimodal interpretability according to claim 1, characterized in that, The step of calculating the contribution of each physiological indicator in the aggregated feature vector based on the myocardial infarction risk assessment results includes: Each physiological indicator in the aggregated feature vector is replaced with a clinical baseline value to construct a counterfactual input vector corresponding to each physiological indicator; The counterfactual input vector is input into the dynamic weighted fuzzy cognitive graph model to obtain the counterfactual risk value corresponding to each physiological indicator; The difference between the counterfactual risk value and the myocardial infarction risk assessment result is calculated to obtain the contribution of each physiological indicator in the aggregated feature vector.

5. The myocardial infarction risk assessment method based on multimodal interpretability according to claim 1, characterized in that, The step of finding the maximum influence propagation path between each source node and the target node based on the preset concept node sensitivity coefficient, path context decay factor, the multiple concept nodes, and the weighted edges, to obtain at least one maximum influence propagation path, includes: Construct at least one path between each source node and the target node; If each path contains multiple concept nodes and weighted edges, the influence flow between two adjacent concept nodes is calculated based on the preset concept node sensitivity coefficient, the final activation value of the sub-source node in two adjacent concept nodes, and the fine-tuned weight of the weighted edge between two adjacent concept nodes. Based on the path context decay factor, the influence flow between all two adjacent concept nodes in each path is weighted and summed to obtain the total influence of each path; Based on the total influence, find the maximum influence propagation path between each source node and the target node, and obtain at least one maximum influence propagation path.

6. The myocardial infarction risk assessment method based on multimodal interpretability according to claim 1, characterized in that, The method for constructing an interpretable myocardial infarction risk assessment report based on the contribution of each physiological indicator and the at least one maximum impact propagation path includes: By sorting the contribution of each physiological indicator and filtering out multiple key physiological indicators, a quantitative attribution summary containing the indicator name corresponding to each key physiological indicator, the specific value derived from the aggregated feature vector, and the contribution is obtained. Match the maximum impact propagation path corresponding to each of the key physiological indicators, and obtain the core logic path containing the name and final activation value of each concept node in the maximum impact propagation path; The final activation value of each concept node and the maximum impact propagation path corresponding to each key physiological indicator are visually encoded to obtain a visualized decision map. By integrating the quantitative attribution summary, the core logical path, and the visualized decision graph, an interpretable myocardial infarction risk assessment report is constructed.

7. The myocardial infarction risk assessment method based on multimodal interpretability according to claim 1, characterized in that, After constructing an interpretable myocardial infarction risk assessment report, the method further includes: The basic weights are constructed as a prior probability distribution of the weights; Based on the prior probability distribution of the weights, the posterior probability distribution of the weights is obtained using Bayes' theorem. The mean of the weight distribution obtained from the posterior probability distribution of the weights is used as the new weights in the dynamically weighted fuzzy cognitive graph model to update the model.

8. A myocardial infarction risk assessment system based on multimodal interpretability, characterized in that, The system includes: The data extraction unit is used to extract multimodal physiological indicators that include electrocardiogram clinical features, blood biomarker features, and radiomics features. The feature splicing unit is used to splice the features in the multimodal physiological index features to obtain a spliced ​​feature vector; The feature aggregation unit is used to calculate the attention score of each feature in the multimodal physiological index features according to the spliced ​​feature vector, and to perform a weighted summation of each feature in the multimodal physiological index features based on the attention score to obtain the aggregated feature vector. The model building unit is used to build a dynamic weighted fuzzy cognitive graph model for myocardial infarction risk assessment. The dynamic weighted fuzzy cognitive graph model includes a weight adjustment mechanism, multiple concept nodes, weighted edges, and an activation function. The weight adjustment mechanism is used to fine-tune the basic weights based on the context vector to obtain the fine-tuned weights. The weighted edges are edges connecting two concept nodes and are used to characterize the influence strength of one concept node on another concept node. The risk assessment unit is used to input the aggregated feature vector into the dynamic weighted fuzzy cognitive graph model, update the activation value of each concept node with fine-tuned weights and activation functions, and obtain the final activation value of each concept node after reaching the preset conditions, and use the final activation value of the myocardial infarction risk concept node as the myocardial infarction risk assessment result. The contribution calculation unit is used to calculate the contribution of each physiological indicator in the aggregated feature vector based on the myocardial infarction risk assessment results. The path finding unit is used to find the maximum influence propagation path between each source node and the target node based on the preset concept node sensitivity coefficient, path context decay factor, the multiple concept nodes and the weighted edge, so as to obtain at least one maximum influence propagation path. The assessment report construction unit is used to construct an interpretable myocardial infarction risk assessment report based on the contribution of each physiological indicator and the at least one maximum impact propagation path.

9. An electronic device, characterized in that, It includes at least one control processor and a memory for communicatively connecting to the at least one control processor; the memory stores instructions executable by the at least one control processor to enable the at least one control processor to perform the myocardial infarction risk assessment method based on multimodal interpretability as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions for causing a computer to perform the myocardial infarction risk assessment method based on multimodal interpretability as described in any one of claims 1 to 7.

Images