A Method and System for Rescuing Aortic Dissection Based on Multi-Dimensional Data Fusion

Abstract

This invention relates to the field of medical information technology, specifically to a method and system for guiding the rescue of aortic dissection based on multi-dimensional data fusion. The method includes acquiring comprehensive medical data of the patient, including at least one of real-time vital signs data, imaging data, and medical history information; fusing and processing the comprehensive medical data, and constructing a patient profile and capturing the trend of disease changes based on the fused dataset to obtain a dynamic patient profile; constructing a rule base based on historical data, and matching the patient's profile with the rules in the rule base to obtain disease characteristics and risk indicators; establishing risk assessment standards based on disease characteristics and risk indicators, assessing the severity of the disease and predicting possible complications through the risk assessment standards, thereby providing personalized rescue suggestions and decision support for medical personnel.

Description

Technical Field

[0001] This invention relates to the field of medical information technology, specifically to a method and system for guiding the rescue of aortic dissection based on multi-dimensional data fusion. Background Technology

[0002] Aortic dissection is a critical cardiovascular disease characterized by rapid onset, rapid progression, high mortality, and the need for swift medical intervention. Its mortality rate is very high, approximately 40% at initial onset, and increases by 1% every hour. Furthermore, the misdiagnosis rate of aortic dissection is also very high; about one-third of patients ultimately diagnosed with aortic dissection are initially misdiagnosed. Patients suspected of having aortic dissection should undergo imaging examinations as early as possible; however, the imaging diagnosis of aortic dissection is complex and requires interpretation by experienced physicians, which increases the difficulty and time cost of diagnosis.

[0003] Currently, the main treatment guidelines for aortic dissection include drug therapy, surgical treatment, and interventional therapy. However, the selection of these treatment methods and the evaluation of their effectiveness all depend on accurate diagnosis and condition assessment. Therefore, improving the diagnostic accuracy and treatment efficiency of aortic dissection is crucial to reducing mortality and improving patient prognosis.

[0004] With the continuous advancement of medical technology, multi-dimensional data fusion technology is being applied more and more widely in the medical field. By integrating data from different medical imaging equipment, physiological monitoring equipment, and clinical information systems, a comprehensive view of patient information can be formed, providing doctors with more accurate and comprehensive diagnostic information. Summary of the Invention

[0005] To create a comprehensive view of patient information and provide doctors with more accurate and complete diagnostic information, this invention aims to provide a method and system for guiding the rescue of aortic dissection based on multi-dimensional data fusion. The specific technical solution adopted is as follows:

[0006] In a first aspect, this application discloses a method for guiding the rescue of aortic dissection based on multi-dimensional data fusion, the method comprising:

[0007] S1. Obtain the patient's comprehensive medical data, which includes at least one of real-time vital signs data, imaging data, and medical history information;

[0008] S2. The comprehensive medical data is fused and processed, and a disease profile is constructed and the trend of disease change is captured based on the fused dataset to obtain a dynamic disease profile of the patient.

[0009] S3. Construct a rule base based on historical data, and match the patient's condition profile with the rules in the rule base to obtain condition characteristics and risk indicators;

[0010] S4. Establish risk assessment criteria based on disease characteristics and risk indicators, assess the severity of the disease and predict possible complications through the risk assessment criteria, so as to provide medical staff with personalized rescue suggestions and decision support.

[0011] Furthermore, in step S2, the process of fusing the comprehensive medical data and constructing a patient's condition profile and capturing the trend of condition changes based on the fused dataset to obtain a dynamic patient condition profile includes:

[0012] S21. Using principal component analysis, the comprehensive medical data is fused using nonlinear dimensionality reduction techniques to obtain a fused dataset containing multi-dimensional information.

[0013] S22. Based on the fused dataset, construct a static portrait of the patient's condition using augmented reality and 3D modeling technology;

[0014] S23. Utilize time series analysis and machine learning algorithms to perform in-depth mining of the comprehensive dataset in order to capture the trend of changes in the patient's condition;

[0015] S24. Update the patient's three-dimensional virtual model and disease profile based on the disease change trend to achieve real-time updating and visualization of the patient's dynamic disease profile.

[0016] Furthermore, in step S21, the principal component analysis method is used to fuse the comprehensive medical data through nonlinear dimensionality reduction technology to obtain a comprehensive dataset containing multi-dimensional information, including:

[0017] S211. Preprocess the comprehensive medical data to obtain a preprocessed dataset;

[0018] S212. The preprocessed dataset is linearly reduced in dimensionality using principal component analysis to obtain a set of principal components.

[0019] S213. Based on feature importance assessment and data distribution characteristics, the principal component set is weighted and nonlinearly combined to obtain a weighted nonlinear principal component set.

[0020] S214. The weighted nonlinear principal component set is subjected to deep dimensionality reduction processing through nonlinear dimensionality reduction technology to capture the nonlinear relationships and complex structures in the data, and obtain a comprehensive dataset containing multi-dimensional information.

[0021] Furthermore, in step S212, the step of using principal component analysis to perform linear dimensionality reduction on the preprocessed dataset to obtain a set of principal components includes:

[0022] S2121. Calculate the covariance matrix between each feature in the preprocessed dataset;

[0023] S2122. Perform eigenvalue decomposition on the covariance matrix to obtain multiple eigenvalues ​​and corresponding eigenvectors;

[0024] S2123. Based on the magnitude of the eigenvalue, set the range of k values, and within the range of k values, select the k value that minimizes the reconstruction error as the optimal k value;

[0025] S2124. According to the optimal k value, select the eigenvectors corresponding to the k largest eigenvalues, and construct the principal component space based on these k eigenvectors;

[0026] S2125. Project the preprocessed dataset onto the principal component space to obtain the dimensionality-reduced principal component set.

[0027] Furthermore, in step S23, the process of using time series analysis and machine learning algorithms to deeply mine the comprehensive dataset to capture the trend of changes in the patient's condition includes:

[0028] S231. Based on the comprehensive dataset, determine the patient's key physiological indicators, and use time series analysis technology to process the key physiological indicators to obtain the corresponding dynamic trend graph.

[0029] S232. Based on the moving average and threshold settings, analyze the abnormal fluctuation points from the dynamic trend chart and mark the potential health risk areas.

[0030] S233. Based on machine learning algorithms, abnormal fluctuation points and potential health risk areas are processed to predict the trend of changes in the patient's condition.

[0031] Furthermore, in step S232, the step of analyzing abnormal fluctuation points from the dynamic trend chart based on the moving average and threshold settings, and marking potential health risk areas, includes:

[0032] S2321. Smooth the dynamic trend chart based on the moving average line to obtain a basic trend chart that filters out random fluctuations.

[0033] S2322. For each data point in the basic trend chart, calculate the distance between the data point and the moving average line to determine the deviation between them;

[0034] S2323. Set a threshold based on historical data analysis, and when the deviation of a corresponding data point is determined to be greater than the threshold, the corresponding data point is regarded as an abnormal fluctuation point.

[0035] S2324. Based on the obtained abnormal fluctuation points, perform cluster analysis to identify and mark potential health risk areas.

[0036] Furthermore, in step S3, the step of constructing a rule base based on historical data and matching the patient's condition profile with the rules in the rule base to obtain condition characteristics and risk indicators includes:

[0037] S31. Use data mining algorithms to extract historical disease information from historical data;

[0038] S32. Based on data analysis techniques using statistical learning methods, generate rules to describe the relationship between historical disease information, disease characteristics, and risk indicators;

[0039] S33. Import the generated rules into the rule base to form the corresponding rule system;

[0040] S34. Extract real-time condition information from the patient's condition profile, and perform rule-based reasoning with the rules in the rule system to obtain target rules that conform to the current patient's condition.

[0041] S35. Analyze the target rule to obtain the disease characteristics and risk indicators that are consistent with the current patient's condition.

[0042] Furthermore, in step S4, the establishment of risk assessment criteria based on disease characteristics and risk indicators, and the assessment of disease severity and prediction of possible complications through these risk assessment criteria, includes:

[0043] S41. Based on the characteristics of the disease and risk indicators, risk factors are identified, and each risk factor is quantitatively scored using the risk matrix method. Based on the scoring results and expert experience, risk assessment standards are established. The risk assessment standards take into account the diversity of the disease, dynamic changes, and the potential for complications, and can automatically generate personalized risk assessment reports.

[0044] S42. Using the aforementioned risk assessment criteria, assess the severity of the condition based on real-time updated disease data, and predict possible complications.

[0045] Secondly, this application discloses an aortic dissection rescue guidance system based on multi-dimensional data fusion. The system includes a data acquisition module, a data fusion and processing module, a rule base construction and matching module, and a condition assessment and prediction module, wherein:

[0046] The data acquisition module is used to acquire the patient's comprehensive medical data, which includes at least one of real-time vital signs data, imaging data, and medical history information.

[0047] The data fusion and processing module is used to fuse and process the comprehensive medical data, and to construct a disease profile and capture the trend of disease changes based on the fused dataset, so as to obtain a dynamic disease profile of the patient.

[0048] The rule base construction and matching module is used to construct a rule base based on historical data and match the patient's condition profile with the rules in the rule base to obtain condition characteristics and risk indicators.

[0049] The disease assessment and prediction module is used to establish risk assessment standards based on disease characteristics and risk indicators. Through these risk assessment standards, the severity of the disease is assessed, and possible complications are predicted, so as to provide medical staff with personalized rescue suggestions and decision support.

[0050] Thirdly, this application discloses a computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, it implements the aforementioned aortic dissection rescue guidance method based on multi-dimensional data fusion.

[0051] The present invention has the following beneficial effects:

[0052] 1) By using integrated medical data after fusion processing to construct dynamic disease profiles, it is possible to capture the trend of disease changes in real time, thereby forming a comprehensive view of patient information, reflecting the real-time health status and potential risks of patients in a timely manner, and providing medical staff with more accurate and timely monitoring of patient status.

[0053] 2) By matching a rule base built on historical data with patient profiles, disease characteristics and risk indicators can be automatically identified, thereby establishing risk assessment standards. This allows for rapid and accurate assessment of disease severity and prediction of potential complications, guiding medical staff to develop more rational and effective treatment plans and improving the scientific rigor and efficiency of medical decision-making. Attached Figure Description

[0054] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1This is a flowchart of a method for guiding the rescue of aortic dissection based on multi-dimensional data fusion, provided in one embodiment of the present invention.

[0056] Figure 2 This is a system structure diagram of an aortic dissection rescue guidance system based on multi-dimensional data fusion, provided as an embodiment of the present invention. Detailed Implementation

[0057] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a method and system for guiding the rescue of aortic dissection based on multi-dimensional data fusion proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0059] The following description, in conjunction with the accompanying drawings, details the specific scheme of the aortic dissection rescue guidance method and system provided by the present invention.

[0060] Please see Figure 1 The diagram illustrates a flowchart of a method for guiding the rescue of aortic dissection based on multi-dimensional data fusion, according to an embodiment of the present invention. The method includes:

[0061] Step S1: Obtain the patient's comprehensive medical data, which includes at least one of real-time vital signs data, imaging data, and medical history information.

[0062] Step S2: The comprehensive medical data is fused and processed, and a disease profile is constructed and the trend of disease change is captured based on the fused dataset to obtain a dynamic disease profile of the patient.

[0063] Step S3: Construct a rule base based on historical data, and match the patient's condition profile with the rules in the rule base to obtain condition characteristics and risk indicators.

[0064] Step S4: Establish risk assessment criteria based on disease characteristics and risk indicators. Use the risk assessment criteria to assess the severity of the disease and predict possible complications, so as to provide medical staff with personalized rescue suggestions and decision support.

[0065] As described above, the aortic dissection rescue guidance method disclosed in this application, based on multi-dimensional data fusion, constructs a dynamic patient profile using integrated medical data after fusion processing. This allows for real-time capture of changes in the patient's condition, resulting in a comprehensive view of patient information. It promptly reflects the patient's real-time health status and potential risks, providing medical staff with more accurate and timely patient status monitoring. By matching a rule base built based on historical data with the patient's patient profile, it can automatically identify disease characteristics and risk indicators, thereby establishing risk assessment standards. This enables rapid and accurate assessment of the severity of the condition and prediction of potential complications, guiding medical staff to develop more rational and effective treatment plans and improving the scientific rigor and efficiency of medical decision-making.

[0066] In one embodiment, step S2, which involves fusing the comprehensive medical data and constructing a patient's condition profile and capturing trends in condition changes based on the fused dataset to obtain a dynamic patient condition profile, includes:

[0067] Step S21: Using principal component analysis, the comprehensive medical data is fused using nonlinear dimensionality reduction techniques to obtain a fused dataset containing multi-dimensional information.

[0068] Specifically, this application preprocesses the comprehensive medical dataset, including data cleaning, missing value handling, and outlier detection, to ensure data accuracy. Then, principal component analysis (PCA) is used to perform linear dimensionality reduction on the preprocessed dataset to extract the principal components, resulting in a principal component set. Next, to better reflect the nonlinear relationships between key information and features in the data, this application also incorporates data distribution characteristics and feature importance assessment to weight and nonlinearly combine the obtained principal component set. Finally, to capture the nonlinear relationships and complex structures in the data, nonlinear dimensionality reduction techniques are introduced. By mapping the weighted nonlinearly combined dataset to a lower-dimensional space, key information and nonlinear relationships are preserved, resulting in a comprehensive dataset containing multi-dimensional information.

[0069] Step S22: Based on the fused dataset, construct a static portrait of the patient's condition using augmented reality and 3D modeling techniques.

[0070] Specifically, this application extracts key information related to the patient's condition from the fused dataset, such as the location, size, shape, surrounding tissue condition, and possible pathological type of the lesion. Then, using 3D modeling software, a 3D model of the lesion and its surrounding tissue is constructed based on the extracted key information. Finally, the constructed 3D model is imported into an augmented reality application. Through AR technology, the 3D model is fused with the patient's actual imaging data (such as CT scans and MRI images). This fusion can be presented in real-time from the doctor's perspective, allowing the doctor to intuitively see a static portrait of the patient's condition.

[0071] Step S23: Use time series analysis and machine learning algorithms to perform in-depth mining on the comprehensive dataset in order to capture the trend of changes in the patient's condition.

[0072] Specifically, this application first identifies key physiological indicators of patients from a comprehensive dataset. These indicators reflect the patients' health status and changes in their condition. Then, time series analysis is used to process these key physiological indicators, revealing their patterns and trends over time by calculating moving averages and plotting dynamic trend graphs. Next, based on moving averages and threshold settings, abnormal fluctuation points are analyzed from the dynamic trend graphs. These abnormal fluctuation points may indicate significant changes in the patient's health status or potential risks. Simultaneously, this application incorporates cluster analysis to mark potential health risk areas, which may correspond to specific disease states or health problems. Finally, machine learning algorithms are used to process the abnormal fluctuation points and potential health risk areas. Through model training, feature extraction, and pattern recognition, the system can learn the patterns of changes in the patient's condition and predict future trends in their condition.

[0073] Step S24: Update the patient's three-dimensional virtual model and disease profile based on the disease change trend to achieve real-time updating and visualization of the patient's dynamic disease profile.

[0074] Specifically, this application adjusts relevant parameters in the 3D virtual model based on the trend of disease progression, such as the size, shape, location, and functional status of organs, to ensure that the model accurately reflects the patient's latest health status. It also identifies key changes in the patient's health status based on the trend, including shifts in the disease or newly emerging pathological features. Finally, the disease profile is updated synchronously based on these key changes, reflecting real-time changes in the patient's health status and providing accurate data support for subsequent calculations of disease characteristics and risk indicators.

[0075] In the above embodiments, by deeply mining the comprehensive dataset through time series analysis and machine learning algorithms, the changing trends of the patient's condition can be captured. Based on this trend, the patient's 3D virtual model and condition profile can be updated in real time. This real-time updating and visualization helps to accurately reflect the patient's latest health status and provides accurate data support for subsequent calculations of condition characteristics and risk indicators.

[0076] In one embodiment, step S21 involves using principal component analysis to fuse the comprehensive medical data using nonlinear dimensionality reduction techniques to obtain a comprehensive dataset containing multi-dimensional information, including:

[0077] Step S211: Preprocess the comprehensive medical data to obtain a preprocessed dataset.

[0078] Specifically, this application will perform data cleaning, data normalization, and feature selection processing on the comprehensive medical data. Data cleaning includes removing duplicate data, handling missing values ​​(e.g., filling missing values ​​with the mean or median), and deleting invalid data (e.g., data that does not conform to medical logic). Data normalization includes scaling numerical data to the range of 0-1 to eliminate differences between different units; text-based medical history information undergoes text preprocessing and vectorization to normalize the information; and image-based imaging data is normalized using pixel value normalization (for grayscale images), channel normalization (for color images), and normalization for specific physical quantity ranges (for medical images such as CT and MRI). Feature selection includes selecting feature variables that are significantly correlated with the analysis target from the original feature set based on correlation analysis (e.g., Pearson correlation coefficient).

[0079] Step S212: Perform linear dimensionality reduction on the preprocessed dataset using principal component analysis to obtain a set of principal components.

[0080] Specifically, this application calculates the covariance matrix among the features in the preprocessed dataset to understand the correlation between features. Then, eigenvalue decomposition is performed on the covariance matrix to obtain multiple eigenvalues ​​and corresponding eigenvectors. The eigenvalues ​​represent the importance of each principal component, and the eigenvectors indicate the orientation of each principal component in the original feature space. Next, based on the magnitude of the eigenvalues, the eigenvectors corresponding to the k largest eigenvalues ​​are selected, and these k eigenvectors constitute the principal component space. Finally, the original dataset is projected onto this principal component space to obtain the dimensionality-reduced set of principal components. This process can be achieved by calculating the dot product between each sample (i.e., observation) in the original dataset and each principal component eigenvector. Ultimately, a k-dimensional dataset is obtained.

[0081] In one embodiment, k is not a fixed value determined according to preset rules or experience. This application considers the inherent structural characteristics of the data and the needs of the analysis purpose, and determines the value of k through dynamic evaluation and optimization strategies. Specifically, after obtaining eigenvalues ​​and eigenvectors through eigenvalue decomposition, this application first sets a corresponding range of k values ​​based on the magnitude of the eigenvalues. Then, within this range of k values, a data reconstruction error analysis method is used to evaluate the data reconstruction error under different k values, quantifying the degree of information loss during dimensionality reduction.

[0082] It should be noted that the application of data reconstruction error analysis further quantifies the degree of information loss during dimensionality reduction. For each k value, this application projects the original dataset into a principal component space consisting of k principal components to obtain the dimensionality-reduced dataset. Then, by calculating the reconstruction error between the original dataset and the dimensionality-reduced dataset (e.g., using mean squared error as a quantification of the reconstruction error), the degree of information loss during dimensionality reduction is further determined. By comparing the reconstruction errors under different k values, the k value that minimizes the reconstruction error is taken as the optimal k value.

[0083] Step S213: Based on feature importance assessment and data distribution characteristics, the principal component set is weighted and nonlinearly combined to obtain a weighted nonlinear principal component set.

[0084] Specifically, this application determines the weight of each principal component in the nonlinear combination based on the magnitude of its corresponding eigenvalue using a proportional allocation method. See the formula for details: Among them, w i t represents the weight of the i-th principal component. i Let represent the eigenvalue of the i-th principal component, and n represent the total number of principal components in the principal component set.

[0085] Furthermore, based on the data distribution characteristics and considering the potential nonlinear relationships and trends between the data, this application introduces logarithmic functions, periodic kernel functions, sigmoid functions, and polynomial functions to perform nonlinear transformations on the principal components, wherein:

[0086] For data with an exponential growth trend (such as the changing trend of certain physiological indicators under specific disease states), this application will apply a logarithmic function to perform nonlinear transformation processing.

[0087] Logarithmic functions can effectively compress numbers with exponential growth trends. This application will apply logarithmic functions to perform nonlinear transformation processing on them.

[0088] For data with periodic variations (such as the diurnal rhythm of certain vital signs), this application will use a periodic kernel function for nonlinear transformation processing to capture the periodic characteristics of the data.

[0089] For data that needs to be normalized to a specific range (such as quantitative indicators of imaging data), this application will apply the sigmoid function for nonlinear transformation to map the data to the desired output range;

[0090] For data exhibiting polynomial relationships (such as the relationship between certain physiological indicators and disease severity), this application will apply polynomial functions for nonlinear transformation to capture the complex nonlinear relationships between the data.

[0091] Furthermore, the principal components obtained after nonlinear transformation (i.e., nonlinear principal components) are weighted and combined according to their corresponding weights to obtain a weighted nonlinear principal component set. Each element in this set is the result of the original data undergoing nonlinear transformation and weighting. See the formula for details: Among them, z i Let f(*) represent the i-th principal component, and f(*) represent the nonlinear mapping function.

[0092] Step S214: Perform deep dimensionality reduction on the weighted nonlinear principal component set using nonlinear dimensionality reduction technology to capture the nonlinear relationships and complex structures in the data, and obtain a comprehensive dataset containing multi-dimensional information.

[0093] Specifically, this application uses kernel principal component analysis (KPCA) to perform deep dimensionality reduction on the weighted nonlinear principal component set to capture the nonlinear relationships and complex structures in the data, resulting in a comprehensive dataset containing multidimensional information. The application of KPCA maps the original data to a high-dimensional feature space, where previously linearly inseparable data becomes linearly separable. Then, linear principal component analysis is performed in this space, calculating the covariance matrix of the data and finding its largest eigenvector to capture the nonlinear characteristics of the data. These nonlinear principal components reflect the inherent structure of the data in the low-dimensional space, thus achieving dimensionality reduction.

[0094] In the above embodiments, on the one hand, based on feature importance assessment and data distribution characteristics, the principal component set is weighted and nonlinearly combined to highlight important features and suppress unimportant features, thereby extracting more critical information. On the other hand, nonlinear dimensionality reduction techniques are used to perform deep dimensionality reduction on the weighted nonlinear principal component set, enabling the capture of nonlinear relationships and complex structures in the data, and extracting more refined and in-depth information.

[0095] In one embodiment, step S212, which involves using principal component analysis to perform linear dimensionality reduction on the preprocessed dataset to obtain a set of principal components, includes:

[0096] Step S2121: Calculate the covariance matrix between each feature in the preprocessed dataset.

[0097] Step S2122: Perform eigenvalue decomposition on the covariance matrix to obtain multiple eigenvalues ​​and corresponding eigenvectors.

[0098] Step S2123: Based on the magnitude of the feature value, set the range of k values, and within the range of k values, select the k value that minimizes the reconstruction error as the optimal k value.

[0099] Step S2124: According to the optimal k value, select the eigenvectors corresponding to the k largest eigenvalues, and construct the principal component space based on these k eigenvectors.

[0100] Step S2125: Project the preprocessed dataset onto the principal component space to obtain the dimensionality-reduced principal component set.

[0101] Based on steps S2121 to S2125, it should be noted that this application employs principal component analysis, a classic dimensionality reduction technique. By calculating the covariance matrix among the features in the preprocessed dataset, the linear correlation between features is captured, and the principal components in the data are identified through eigenvalue decomposition of the covariance matrix. These principal components are represented by eigenvectors, which constitute a new, low-dimensional feature space. In selecting the optimal k value, this application employs a data reconstruction error analysis method. This method not only helps determine how many principal components to retain to minimize information loss but also ensures that the dimensionality-reduced dataset retains as much of the original data's structure and information as possible. By comparing the reconstruction errors under different k values, a balance point can be found, achieving effective dimensionality reduction while ensuring data quality.

[0102] In the above embodiments, by setting the range of k values ​​and calculating the reconstruction error, the k value that minimizes the reconstruction error can be automatically selected as the optimal k value. This avoids the subjectivity and uncertainty caused by manually selecting the k value, improves the accuracy and objectivity of parameter selection, and ensures the accuracy of data dimensionality reduction.

[0103] In one embodiment, step S23, which involves using time series analysis and machine learning algorithms to perform deep mining on the comprehensive dataset to capture the trend of changes in the patient's condition, includes:

[0104] Step S231: Based on the comprehensive dataset, determine the patient's key physiological indicators, and use time series analysis technology to process the key physiological indicators to obtain the corresponding dynamic trend graph.

[0105] Specifically, this application first selects key physiological indicators directly related to the patient's health status from a comprehensive dataset, such as heart rate, blood pressure, blood glucose, and blood oxygen saturation. These indicators reflect the patient's physiological state and its changing trends. Next, time series analysis technology is used to perform time series analysis on the selected key physiological indicators to capture the patterns and trends of their changes over time, thereby generating corresponding dynamic trend charts. These dynamic trend charts visually demonstrate the fluctuations in the patient's physiological indicators, providing a foundation for subsequent detection of abnormal fluctuation points and prediction of disease progression trends.

[0106] Step S232: Based on the moving average and threshold settings, analyze the abnormal fluctuation points from the dynamic trend chart and mark the potential health risk areas.

[0107] Specifically, this application first smooths the dynamic trend chart based on a moving average to reduce random fluctuations. The moving average is generated as a smooth curve based on the average value of data points within a window. This curve reflects the overall trend of the data sequence while filtering out short-term random fluctuations. Next, for each data point in the trend chart, a deviation index is obtained by calculating the distance between the data point and the moving average. This deviation index reflects the degree of deviation of the data point from the moving average. Then, a threshold is set based on historical data analysis. When the deviation index of a data point exceeds the set threshold, that data point is identified as an abnormal fluctuation point and recorded. Finally, based on the recorded abnormal fluctuation points, this application further analyzes their distribution, using cluster analysis to group similar abnormal fluctuation points together to identify potential health risk areas.

[0108] Step S233: Process abnormal fluctuation points and potential health risk areas based on machine learning algorithms to predict the trend of changes in the patient's condition.

[0109] Specifically, this application collects data related to abnormal fluctuation points and potential health risk areas, including the values ​​of key physiological indicators, timestamps, and abnormality types, and divides the data into training and validation sets. Considering the temporal and complex nature of the data, this application selects a machine learning model based on a Long Short-Term Memory (LSTM) network and inputs the training set into the model for training. During training, the model will predict the patient's disease progression trend based on the characteristics of abnormal fluctuation points and potential health risk areas in historical data, as well as their correlation with the disease progression trend, and output the confidence interval of the prediction result. The confidence interval reflects the uncertainty of the prediction result and provides a reference for the reliability of the prediction result.

[0110] In the above embodiments, abnormal fluctuation points and potential health risk areas are processed based on machine learning algorithms to predict the trend of changes in the patient's condition. This intelligent prediction method can provide more scientific decision support for subsequent updates to the dynamic condition profile of the patient.

[0111] In one embodiment, step S232, which involves analyzing abnormal fluctuation points from the dynamic trend chart based on moving averages and threshold settings, and marking potential health risk areas, includes:

[0112] Step S2321: Smooth the dynamic trend chart based on the moving average to obtain a basic trend chart that filters out random fluctuations.

[0113] Specifically, this application first determines a suitable moving average window size based on the characteristics of the data and the purpose of the analysis, such as a period of 5 days, 10 days, or longer. This window size determines the number of data points involved in the average calculation. Then, the window is slid along the time axis, and the average value of the data points within each window is calculated. The smooth curve formed by connecting these average values ​​is the moving average line. Next, the original data points in the dynamic trend chart are combined with the moving average line using linear interpolation (e.g., using the linear interpolation formula y = y0 + α(y1 - y0) to calculate the interpolation point values, where y0 and y1 are the values ​​of the known data points, and α is an interpolation coefficient between 0 and 1, determined based on the positional ratio between the original data points and the corresponding points on the moving average line). This results in a smoother base trend chart that filters out most random fluctuations.

[0114] Step S2322: For each data point in the base trend chart, calculate the distance between the data point and the moving average line to determine the deviation between them.

[0115] Specifically, this application uses the vertical distance method to calculate the distance between each data point and the moving average line. Specifically, for each data point in the trend chart, this application finds its corresponding point on the moving average line and calculates the vertical distance between the two points; the result is the deviation between them. It should be noted that this deviation reflects the degree to which the data point deviates from the overall trend.

[0116] Step S2323: Set a threshold based on historical data analysis, and when the deviation of a corresponding data point is determined to be greater than the threshold, the corresponding data point is regarded as an abnormal fluctuation point.

[0117] Step S2324: Perform cluster analysis based on the obtained abnormal fluctuation points to identify and mark potential health risk areas.

[0118] Specifically, this application will perform cluster analysis on the obtained abnormal fluctuation points to obtain multiple cluster groups. Each cluster group represents a potential health risk area, and this area contains similar abnormal fluctuation points.

[0119] Subsequently, to facilitate subsequent machine learning processing, this application will perform vector labeling on potential health risk areas based on the location, size, and density of data points within clusters, including at least one statistical feature among mean, variance, maximum, and minimum. Wherein:

[0120] The mean reflects the average level of abnormal fluctuation points within a cluster group, which helps to understand the overall fluctuation trend of the region.

[0121] Variance measures the dispersion of abnormal fluctuation points within a cluster group. The larger the variance, the more dispersed the fluctuation points in that region, which may indicate a higher degree of instability in the condition.

[0122] The maximum value records the highest value of the abnormal fluctuation points within the cluster group, and the minimum value records the lowest value of the abnormal fluctuation points within the cluster group. Together with the maximum value, they can form a comprehensive description of the range of fluctuations in the condition.

[0123] This processing provides rich input features for subsequent machine learning algorithms. These features can comprehensively describe the characteristics of potential health risk areas. Machine learning algorithms can use these features to train predictive models, thereby achieving accurate prediction of the patient's condition trend.

[0124] In one embodiment, step S3, which involves constructing a rule base based on historical data and matching the patient's condition profile with the rules in the rule base to obtain condition characteristics and risk indicators, includes:

[0125] Step S31: Use data mining algorithms to extract historical disease information from historical data.

[0126] Specifically, this application uses a data mining algorithm based on association rules to mine frequent itemsets and calculate confidence scores in historical data. The support score of frequent itemsets measures their prevalence in the dataset, while the confidence score is used to assess the strength and reliability of the association between itemsets (e.g., calculating conditional probability to quantify the association strength from one itemset (e.g., a specific combination of symptoms) to another itemset (e.g., a potential disease). This allows the identification of associations between key information such as medical history, symptoms, and examination results, thereby extracting valuable historical medical information from historical data.

[0127] Step S32: Based on data analysis techniques using statistical learning methods, generate rules to describe the correlation between historical disease information, disease characteristics, and risk indicators.

[0128] Specifically, this application employs correlation analysis techniques to calculate the association between historical medical information (such as medical history, symptoms, and examination results) and disease characteristics (such as disease type) and risk indicators (such as the risk of complications). By calculating correlation coefficients (the calculation of correlation coefficients is prior art, and this application does not limit the specific calculation formula), the strength of the association between these variables is quantified, and rules describing the association between historical medical information, disease characteristics, and risk indicators are generated based on the magnitude of the correlation coefficients. The general logic for generating these rules includes:

[0129] If the absolute value of the correlation coefficient is greater than the preset correlation threshold (e.g., 0.7), then it is considered that there is a significant correlation between historical disease information and corresponding disease characteristics and risk indicators.

[0130] If the absolute value of the correlation coefficient is close to 0, it is considered that there is no significant correlation between historical medical information and corresponding medical characteristics and risk indicators.

[0131] Step S33: Import the generated rules into the rule base to form the corresponding rule system.

[0132] Specifically, this application will categorize, organize, and store the generated rules to form a structured rule base. Each rule will be assigned a unique identifier and description to facilitate subsequent indexing and matching.

[0133] Step S34: Extract real-time condition information from the patient's condition profile, and perform rule-based reasoning with the various rules in the rule system to obtain target rules that conform to the current patient condition.

[0134] Step S35: Analyze the target rule to obtain the disease characteristics and risk indicators that match the current patient's condition.

[0135] Based on steps S34 and S35, it should be noted that this application will compare and reason with each rule in the rule system, based on the extracted real-time patient information. This reasoning process is based on rule matching and logical judgment, aiming to find the rule that best matches or is most relevant to the current patient's condition. During the reasoning process, the similarity or matching degree between the real-time patient information and the historical patient information described in the rule will be considered. Once a target rule that matches the current patient's condition is found, the patient's condition characteristics and risk indicators will be fed back, and the patient's current risk status will be assessed based on the condition characteristics and risk indicators mentioned in the target rule.

[0136] In the above embodiments, on the one hand, rule generation based on statistical learning methods can quantify the correlation strength between historical medical information and disease characteristics and risk indicators, thereby improving the accuracy of rule matching. On the other hand, reasoning between real-time medical information and rules in the rule system can find the most suitable target rule for each patient's specific situation, thereby achieving personalized risk assessment and early warning, and helping doctors to more accurately judge the patient's health status and potential risks.

[0137] In one embodiment, step S4, which involves establishing risk assessment criteria based on disease characteristics and risk indicators, and using these criteria to assess disease severity and predict potential complications, includes:

[0138] Step S41: Based on the characteristics of the disease and risk indicators, risk factors are identified, and each risk factor is quantitatively scored using the risk matrix method. Based on the scoring results and expert experience, risk assessment standards are established. These risk assessment standards integrate the diversity of the disease, dynamic changes, and the potential for complications, and can automatically generate personalized risk assessment reports.

[0139] Specifically, this application first identifies and determines risk factors relevant to the assessed subject based on disease characteristics and risk indicators. Then, using a risk matrix method, each risk factor is quantitatively scored based on its probability and severity, assigning it a specific quantitative value. Finally, based on the quantitative scoring results and expert experience, and comprehensively considering the diversity of the disease, its dynamic characteristics, and the potential threat of complications, corresponding risk assessment standards are established.

[0140] Step S42: Using the risk assessment criteria, assess the severity of the condition based on real-time updated disease data, and predict possible complications.

[0141] Specifically, this application will input real-time updated patient data into the risk assessment model, and calculate the patient's risk score and risk level according to the risk assessment criteria. Then, by combining the risk score and risk level and comparing them with preset thresholds, the severity of the condition will be assessed, and possible complications will be predicted based on the correlation analysis of risk factors and the statistical patterns of historical data.

[0142] Please refer to Figure 2 This application discloses a multi-dimensional data fusion-based emergency guidance system for aortic dissection. The system includes a data acquisition module, a data fusion and processing module, a rule base construction and matching module, and a condition assessment and prediction module, wherein:

[0143] The data acquisition module is used to acquire the patient's comprehensive medical data, which includes at least one of real-time vital signs data, imaging data, and medical history information.

[0144] The data fusion and processing module is used to fuse and process the comprehensive medical data, and to construct a patient's condition profile and capture the trend of condition changes based on the fused dataset, thereby obtaining a dynamic condition profile of the patient.

[0145] The rule base construction and matching module is used to construct a rule base based on historical data and match the patient's condition profile with the rules in the rule base to obtain condition characteristics and risk indicators.

[0146] The disease assessment and prediction module is used to establish risk assessment standards based on disease characteristics and risk indicators. Through these risk assessment standards, the severity of the disease is assessed, and possible complications are predicted, so as to provide medical staff with personalized rescue suggestions and decision support.

[0147] As described above, the aortic dissection rescue guidance system disclosed in this application, based on multi-dimensional data fusion, constructs a dynamic patient profile using integrated medical data after fusion processing. This allows for real-time capture of changes in the patient's condition, resulting in a comprehensive view of patient information. It promptly reflects the patient's real-time health status and potential risks, providing medical staff with more accurate and timely patient status monitoring. By matching a rule base built based on historical data with the patient's patient profile, it can automatically identify disease characteristics and risk indicators, thereby establishing risk assessment standards. This enables rapid and accurate assessment of the severity of the condition and prediction of potential complications, guiding medical staff to develop more rational and effective treatment plans and improving the scientific rigor and efficiency of medical decision-making.

[0148] In one embodiment, the data fusion and processing module is further configured to use principal component analysis to perform fusion processing on the comprehensive medical data through nonlinear dimensionality reduction technology to obtain a fused dataset containing multi-dimensional information; based on the fused dataset, to construct a static patient condition profile using augmented reality and 3D modeling technology; to perform in-depth mining on the comprehensive dataset using time series analysis and machine learning algorithms to capture the patient's condition change trend; and to update the patient's 3D virtual model and condition profile based on the condition change trend to achieve real-time updating and visualization of the patient's dynamic condition profile.

[0149] In one embodiment, the data fusion and processing module is further configured to preprocess the comprehensive medical data to obtain a preprocessed dataset; perform linear dimensionality reduction on the preprocessed dataset using principal component analysis to obtain a set of principal components; based on feature importance assessment and data distribution characteristics, weight and nonlinear combination of the set of principal components are performed to obtain a weighted nonlinear principal component set; and perform deep dimensionality reduction on the weighted nonlinear principal component set using nonlinear dimensionality reduction techniques to capture nonlinear relationships and complex structures in the data, thereby obtaining a comprehensive dataset containing multidimensional information.

[0150] In one embodiment, the data fusion and processing module is further configured to calculate the covariance matrix among the features in the preprocessed dataset; perform eigenvalue decomposition on the covariance matrix to obtain multiple eigenvalues ​​and corresponding eigenvectors; set a range of k values ​​according to the magnitude of the eigenvalues, and select the k value that minimizes the reconstruction error within the range of k values ​​as the optimal k value; select the eigenvectors corresponding to the k largest eigenvalues ​​according to the optimal k value, and construct a principal component space based on these k eigenvectors; project the preprocessed dataset onto the principal component space to obtain a dimensionality-reduced principal component set.

[0151] In one embodiment, the data fusion and processing module is further configured to determine the patient's key physiological indicators based on the comprehensive dataset, and process the key physiological indicators using time series analysis technology to obtain a corresponding dynamic trend map; based on the moving average and threshold setting, analyze abnormal fluctuation points from the dynamic trend map and mark potential health risk areas; and process the abnormal fluctuation points and potential health risk areas based on machine learning algorithms to predict the patient's condition change trend.

[0152] In one embodiment, the data fusion and processing module is further configured to smooth the dynamic trend chart based on a moving average to obtain a basic trend chart that filters out random fluctuations; for each data point in the basic trend chart, calculate the distance between the data point and the moving average to determine the deviation between the two; set a threshold based on historical data analysis, and when the deviation of the corresponding data point is determined to be greater than the threshold, the corresponding data point is identified as an abnormal fluctuation point; and perform cluster analysis based on the obtained abnormal fluctuation points to identify and mark potential health risk areas.

[0153] In one embodiment, the rule base construction and matching module is further configured to: extract historical medical information from historical data using data mining algorithms; generate rules describing the correlation between historical medical information, medical characteristics, and risk indicators based on statistical learning-based data analysis techniques; import the generated rules into the rule base to form a corresponding rule system; extract real-time medical information from the patient's medical profile and perform rule-based reasoning with the rules in the rule system to obtain target rules that match the current patient's condition; and parse the target rules to obtain medical characteristics and risk indicators that match the current patient's condition.

[0154] In one embodiment, the disease assessment and prediction module is further used to determine risk factors based on disease characteristics and risk indicators, quantify and score each risk factor using a risk matrix method, and establish risk assessment standards based on the scoring results and expert experience. The risk assessment standards integrate the diversity, dynamic changes, and potential complications of the disease, and can automatically generate personalized risk assessment reports. Through the risk assessment standards, the severity of the disease is assessed based on real-time updated disease data, and possible complications are predicted.

[0155] Furthermore, this application also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned method for guiding the rescue of aortic dissection based on multi-dimensional data fusion.

[0156] As can be seen from the above, the computer-readable storage medium disclosed in this application utilizes integrated medical data after fusion processing to construct a dynamic patient profile. This profile can capture real-time trends in patient condition changes, enabling the formation of a comprehensive view of patient information. It promptly reflects the patient's real-time health status and potential risks, providing medical staff with more accurate and timely patient status monitoring. By matching a rule base built based on historical data with the patient's patient profile, it can automatically identify disease characteristics and risk indicators, thereby establishing risk assessment standards. In this way, the severity of the condition can be quickly and accurately assessed, and the risk of potential complications can be predicted, thus guiding medical staff to develop more rational and effective treatment plans and improving the scientific nature and efficiency of medical decision-making.

[0157] Furthermore, this application also discloses an aortic dissection rescue device based on multi-dimensional data fusion, including a communication interface, a memory, a communication bus, and a processor, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0158] The memory is used to store computer programs;

[0159] When the processor executes the program stored in the memory, it implements the steps of the aortic dissection rescue guidance method based on multi-dimensional data fusion.

[0160] As can be seen from the above, the computer-readable storage medium disclosed in this application utilizes integrated medical data after fusion processing to construct a dynamic patient profile. This profile can capture real-time trends in patient condition changes, enabling the formation of a comprehensive view of patient information. It promptly reflects the patient's real-time health status and potential risks, providing medical staff with more accurate and timely patient status monitoring. By matching a rule base built based on historical data with the patient's patient profile, it can automatically identify disease characteristics and risk indicators, thereby establishing risk assessment standards. In this way, the severity of the condition can be quickly and accurately assessed, and the risk of potential complications can be predicted, thus guiding medical staff to develop more rational and effective treatment plans and improving the scientific nature and efficiency of medical decision-making.

[0161] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0162] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A method for guiding the rescue of aortic dissection based on multi-dimensional data fusion, characterized in that, The method includes: S1. Obtain the patient's comprehensive medical data, which includes at least one of real-time vital signs data, imaging data, and medical history information; S2. The comprehensive medical data is fused and processed, and a disease profile is constructed and the trend of disease change is captured based on the fused dataset to obtain a dynamic disease profile of the patient. S3. Construct a rule base based on historical data, and match the patient's condition profile with the rules in the rule base to obtain condition characteristics and risk indicators; S4. Establish risk assessment criteria based on disease characteristics and risk indicators, assess the severity of the disease and predict possible complications through the risk assessment criteria, so as to provide medical staff with personalized rescue suggestions and decision support; In step S2, the process of fusing the comprehensive medical data and constructing a patient's condition profile and capturing the trend of condition changes based on the fused dataset to obtain a dynamic patient condition profile includes: S21. Using principal component analysis, the comprehensive medical data is fused using nonlinear dimensionality reduction techniques to obtain a fused dataset containing multi-dimensional information. S22. Based on the fused dataset, construct a static portrait of the patient's condition using augmented reality and 3D modeling technology; S23. Utilize time series analysis and machine learning algorithms to perform in-depth mining of the comprehensive dataset in order to capture the trend of changes in the patient's condition; S24. Update the patient's three-dimensional virtual model and disease profile based on the disease change trend to achieve real-time updating and visualization of the patient's dynamic disease profile; In step S21, the comprehensive medical data is fused using principal component analysis and nonlinear dimensionality reduction techniques to obtain a comprehensive dataset containing multi-dimensional information, including: S211. Preprocess the comprehensive medical data to obtain a preprocessed dataset; S212. The preprocessed dataset is linearly reduced in dimensionality using principal component analysis to obtain a set of principal components. S213. Based on feature importance assessment and data distribution characteristics, the principal component set is weighted and nonlinearly combined to obtain a weighted nonlinear principal component set. S214. The weighted nonlinear principal component set is subjected to deep dimensionality reduction processing through nonlinear dimensionality reduction technology to capture the nonlinear relationships and complex structures in the data, and obtain a comprehensive dataset containing multi-dimensional information.

2. The method according to claim 1, characterized in that, In step S212, the step of using principal component analysis to perform linear dimensionality reduction on the preprocessed dataset to obtain a set of principal components includes: S2121. Calculate the covariance matrix between each feature in the preprocessed dataset; S2122. Perform eigenvalue decomposition on the covariance matrix to obtain multiple eigenvalues ​​and corresponding eigenvectors; S2123. Based on the magnitude of the eigenvalue, set the range of k values, and within the range of k values, select the k value that minimizes the reconstruction error as the optimal k value; S2124. According to the optimal k value, select the eigenvectors corresponding to the k largest eigenvalues, and construct the principal component space based on these k eigenvectors; S2125. Project the preprocessed dataset onto the principal component space to obtain the dimensionality-reduced principal component set.

3. The method according to claim 1, characterized in that, In step S23, the step of using time series analysis and machine learning algorithms to deeply mine the comprehensive dataset in order to capture the trend of changes in the patient's condition includes: S231. Based on the comprehensive dataset, determine the patient's key physiological indicators, and use time series analysis technology to process the key physiological indicators to obtain the corresponding dynamic trend graph. S232. Based on the moving average and threshold settings, analyze the abnormal fluctuation points from the dynamic trend chart and mark the potential health risk areas. S233. Based on machine learning algorithms, abnormal fluctuation points and potential health risk areas are processed to predict the trend of changes in the patient's condition.

4. The method according to claim 3, characterized in that, In step S232, the step of analyzing abnormal fluctuation points from the dynamic trend chart based on moving averages and threshold settings, and marking potential health risk areas, includes: S2321. Smooth the dynamic trend chart based on the moving average line to obtain a basic trend chart that filters out random fluctuations. S2322. For each data point in the basic trend chart, calculate the distance between the data point and the moving average line to determine the deviation between them; S2323. Set a threshold based on historical data analysis, and when the deviation of a corresponding data point is determined to be greater than the threshold, the corresponding data point is regarded as an abnormal fluctuation point. S2324. Based on the obtained abnormal fluctuation points, perform cluster analysis to identify and mark potential health risk areas.

5. The method according to claim 1, characterized in that, In step S3, the step of constructing a rule base based on historical data and matching the patient's condition profile with the rules in the rule base to obtain condition characteristics and risk indicators includes: S31. Use data mining algorithms to extract historical disease information from historical data; S32. Based on data analysis techniques using statistical learning methods, generate rules to describe the relationship between historical disease information, disease characteristics, and risk indicators; S33. Import the generated rules into the rule base to form the corresponding rule system; S34. Extract real-time condition information from the patient's condition profile, and perform rule-based reasoning with the rules in the rule system to obtain target rules that conform to the current patient's condition. S35. Analyze the target rule to obtain the disease characteristics and risk indicators that are consistent with the current patient's condition.

6. The method according to claim 1, characterized in that, In step S4, the establishment of risk assessment criteria based on disease characteristics and risk indicators, and the assessment of disease severity and prediction of possible complications through these risk assessment criteria, includes: S41. Based on the characteristics of the disease and risk indicators, risk factors are identified, and each risk factor is quantitatively scored using the risk matrix method. Based on the scoring results and expert experience, risk assessment standards are established. The risk assessment standards take into account the diversity of the disease, dynamic changes, and the potential for complications, and can automatically generate personalized risk assessment reports. S42. Using the aforementioned risk assessment criteria, assess the severity of the condition based on real-time updated disease data, and predict possible complications.

7. A rescue guidance system for aortic dissection based on multi-dimensional data fusion, characterized in that, The system includes a data acquisition module, a data fusion and processing module, a rule base construction and matching module, and a disease assessment and prediction module, wherein: The data acquisition module is used to acquire the patient's comprehensive medical data, which includes at least one of real-time vital signs data, imaging data, and medical history information. The data fusion and processing module is used to fuse and process the comprehensive medical data, and to construct a disease profile and capture the trend of disease changes based on the fused dataset, so as to obtain a dynamic disease profile of the patient. The rule base construction and matching module is used to construct a rule base based on historical data and match the patient's condition profile with the rules in the rule base to obtain condition characteristics and risk indicators. The disease assessment and prediction module is used to establish risk assessment standards based on disease characteristics and risk indicators, assess the severity of the disease through the risk assessment standards, and predict possible complications, so as to provide medical staff with personalized rescue suggestions and decision support. The data fusion and processing module fuses and processes the comprehensive medical data, and constructs a patient profile and captures the trend of disease changes based on the fused dataset, resulting in a dynamic patient profile, including: Principal component analysis was used to fuse the comprehensive medical data using nonlinear dimensionality reduction techniques, resulting in a fused dataset containing multi-dimensional information. Based on the fused dataset, a static profile of the patient's condition is constructed using augmented reality and 3D modeling techniques. The comprehensive dataset was analyzed in depth using time series analysis and machine learning algorithms to capture the trend of changes in patients' conditions. The patient's three-dimensional virtual model and disease profile are updated based on the aforementioned disease change trend to achieve real-time updating and visualization of the patient's dynamic disease profile; The data fusion and processing module utilizes principal component analysis and nonlinear dimensionality reduction techniques to fuse the comprehensive medical data, resulting in a comprehensive dataset containing multi-dimensional information, including: The comprehensive medical data is preprocessed to obtain a preprocessed dataset; The preprocessed dataset is linearly dimensionality-reduced using principal component analysis to obtain a set of principal components. Based on feature importance assessment and data distribution characteristics, the principal component set is weighted and nonlinearly combined to obtain a weighted nonlinear principal component set. The weighted nonlinear principal component set is subjected to deep dimensionality reduction using nonlinear dimensionality reduction techniques to capture the nonlinear relationships and complex structures in the data, thereby obtaining a comprehensive dataset containing multi-dimensional information.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the aortic dissection rescue guidance method based on multi-dimensional data fusion as described in any one of claims 1-6.

Images