Surgical decision support method and system

Abstract

The invention discloses a surgical operation decision support method and system, and relates to the technical field of surgical operations, and the method comprises the steps: constructing a cloud database with a privacy protection mechanism, and uploading multi-source medical data to a cloud server after a data source end confirms that a privacy protection standard is satisfied; the encrypted clinical diagnosis and treatment information is uploaded to the cloud server through the data source end, a query matching request is sent to the cloud service platform, and the cloud service platform establishes an encryption and decryption engine and decrypts the clinical diagnosis and treatment information after the verification request passes; setting a weight factor of each feature for risk assessment based on the selected effective features, and constructing a risk assessment model; and outputting an evaluation result of the preoperative risk through model calculation. The method achieves the assessment and prediction of the preoperative risk of the data subject in the surgical operation, provides a scientific basis and an auxiliary decision for a doctor, achieves the comprehensive assessment, risk analysis and prediction of the data subject, and guarantees the privacy and safety of the data of the data subject.

Description

Technical Field

[0001] This invention relates to the field of surgical technology, and in particular to a surgical decision support method and system. Background Technology

[0002] Thoracic surgery is a medical field that specializes in the research and treatment of chest-related diseases, encompassing surgeries and treatments of organs such as the heart, lungs, and esophagus. As a common yet complex medical procedure, thoracic surgery presents significant risks and challenges. Preoperative risk assessment plays a crucial role in thoracic surgery. By systematically evaluating the patient's overall health, surgical suitability, and potential preoperative risks and complications, it provides essential information for physicians to develop the optimal treatment plan. Therefore, preoperative risk assessment is indispensable in clinical practice, its importance stemming not only from ensuring patient health and safety but also from improving the quality of medical care, optimizing the allocation of medical resources, and reducing medical costs.

[0003] Currently, surgical decision support technologies mainly include clinical assessment, physiological assessment, and surgical risk scoring systems. Clinical assessment is primarily based on the physician's clinical experience and expertise, combined with the patient's medical history, physical examination, and laboratory test results. Physiological assessment predicts surgical risk by evaluating the patient's physiological function, such as cardiopulmonary function and renal function. In addition, there are surgical risk scoring systems, such as the ASA classification and Lee classification, which assess and classify surgical risk by evaluating factors such as the patient's physical condition, type of surgery, and preoperative circumstances.

[0004] However, assessments based on clinical experience are easily influenced by subjective factors and lack objectivity and standardization; while physiological assessments can provide some physiological indicators, they often cannot fully reflect the patient's overall health status; and while other surgical risk scoring systems can provide quantitative risk assessments, they still have certain limitations, cannot make full use of massive amounts of clinical data, and lack personalization and precision.

[0005] In recent years, with the rise of cloud computing technology, it has played a significant role in risk assessment across various fields. It provides efficient computing and storage resources, supports large-scale data processing and analysis, and facilitates the effective integration and utilization of massive amounts of data. Furthermore, cloud computing offers secure and reliable data storage and transmission mechanisms, improving the efficiency of data exchange. However, directly applying cloud computing to preoperative risk assessment has some drawbacks, such as data privacy and security issues, which could lead to the risk of patient privacy and data breaches. In addition, the reliability and stability of the cloud computing platform are also crucial considerations. If the cloud service provider experiences a failure or service interruption, it may affect the conduct of the preoperative risk assessment and the reliability of the results. Therefore, when using cloud computing for preoperative risk assessment, it is essential to fully consider issues such as data privacy and security, and service reliability to ensure the safety and reliability of the assessment process. Summary of the Invention

[0006] Therefore, it is necessary to provide a surgical decision support method and system to address the aforementioned technical problems.

[0007] In a first aspect, the present invention provides a surgical decision support method, the method comprising:

[0008] S1. Acquire multi-source medical data. After the data source confirms that the data meets the privacy protection standards, upload the multi-source medical data to the cloud server and encrypt and store the corresponding attribute values ​​according to the decomposition results to build a cloud database with a privacy protection mechanism.

[0009] S2. Obtain clinical diagnosis and treatment information. Upload the encrypted clinical diagnosis and treatment information to the cloud server through the data source end. Send a query matching request to the cloud service platform. After the cloud service platform verifies the request, it establishes an encryption and decryption engine and decrypts the clinical diagnosis and treatment information. Then, through distributed computing nodes, it filters similar information data related to the clinical diagnosis and treatment information from the non-privacy attribute data of the cloud database.

[0010] S3. Extract effective features from similar information data and establish a risk assessment system. Based on the selected effective features, set the weight factors of each feature for risk assessment, construct a risk assessment model, and train and test the model using a training set and a test set composed of similar information data.

[0011] S4. Extract risk indicators that conform to the risk assessment system from clinical diagnosis and treatment information, use the risk indicators as input data for the risk assessment model, and calculate and output the preoperative risk assessment results through the model to provide scientific basis and auxiliary decision support.

[0012] Furthermore, after acquiring multi-source medical data and confirming that the data source meets privacy protection standards, the multi-source medical data is uploaded to the cloud server. The corresponding attribute values ​​are then encrypted and stored according to the decomposition results, constructing a cloud database with privacy protection mechanisms, including:

[0013] S11. The data source applies for an identity certificate from the certification center and uses the identity certificate to apply for database outsourcing services from the cloud service platform. After the service application is completed, the data source sets an attribute set according to the data types contained in the multi-source medical data, and then submits the attribute set and privacy constraints to the cloud service platform.

[0014] S12. The cloud service platform performs minimum encrypted attribute decomposition on the attribute set according to the privacy constraints and returns the decomposition results to the data source. The data source confirms whether the privacy protection standard is met. If yes, the multi-source medical data is uploaded to the cloud server. If not, the privacy constraints are reset and the data is resubmitted to the cloud service platform along with the attribute set.

[0015] S13. The cloud server encrypts the corresponding attribute values ​​in the multi-source medical data according to the decomposition results of the attribute set, stores the processed data in the cloud database, and retains the metadata.

[0016] S14. The data subject applies to the certification center for a personal identity certificate, and sends an access application to the cloud service platform with the certificate. After the cloud service platform verifies the application, the data subject can legally access the cloud database.

[0017] Furthermore, clinical diagnosis and treatment information is obtained by uploading encrypted clinical diagnosis and treatment information to a cloud server through the data source, and sending a query matching request to the cloud service platform. After the cloud service platform verifies the request, it establishes an encryption / decryption engine and decrypts the clinical diagnosis and treatment information. Then, through distributed computing nodes, it filters similar information data related to the clinical diagnosis and treatment information from the non-privacy attribute data of the cloud database, including:

[0018] S21. The data source end encrypts and uploads the clinical diagnosis and treatment information of the data subject to the cloud server. The clinical diagnosis and treatment information includes basic information, medical history information, examination results and surgical plan.

[0019] S22. The data source sends a query matching request to the cloud service platform, and after the request is verified, the cloud service platform grants the data source and the cloud server the query matching permission of the cloud database.

[0020] S23. The cloud server assigns data query and matching tasks to distributed computing nodes, and filters similar information data from non-privacy attribute data in the cloud database.

[0021] Furthermore, the data source sends a query matching request to the cloud service platform, and after the request is verified, the cloud service platform grants the data source and the cloud server query matching permissions for the cloud database, including:

[0022] S221. The data source sends an authentication request to the cloud server. After successful authentication, the cloud server sends a query matching request from the data source to the cloud service platform.

[0023] S222. The cloud service platform establishes an encryption and decryption engine between the cloud server and the cloud database. It uses the encryption and decryption engine to analyze the query matching request, extract the key of the clinical diagnosis and treatment information contained in the query matching request, and then detect the legality of the key. When the key legality detection is passed, the encrypted clinical diagnosis and treatment information is decrypted and stored in the database to be matched.

[0024] S223. The cloud service platform re-verifies the legality of the data source identity information contained in the query matching request. After the verification is passed, it grants query matching permissions to the data source and the cloud server, and opens up non-privacy attribute data in the cloud database for query matching.

[0025] Furthermore, the cloud server distributes data query and matching tasks to distributed computing nodes, filtering similar information data from non-privacy attribute data in the cloud database, including:

[0026] S231. The cloud server distributes data query matching tasks to available computing nodes through the task scheduling manager, and achieves parallel computing through several computing nodes.

[0027] S232. Based on the task requirements of the query matching task, the computing node reads a corresponding amount of non-privacy attribute data from the cloud database and obtains the decrypted clinical diagnosis and treatment data.

[0028] S233. Based on a predefined semantic framework, fill the corresponding positions within the semantic framework with clinical diagnosis and treatment data and non-privacy attribute data, and check for missing items.

[0029] S234. Based on the semantic framework after filling, calculate the similarity between the clinical diagnosis and treatment data and each non-privacy attribute data in turn.

[0030] S235. Pre-set a standard threshold for similarity and retain non-privacy attribute data that exceeds the standard threshold as similar information data related to clinical diagnosis and treatment data.

[0031] Furthermore, based on a predefined semantic framework, clinical diagnosis and treatment data and non-privacy attribute data are filled into the corresponding positions within the semantic framework, and missing items are checked, including:

[0032] S2331. Using the factors that need to be considered in the preoperative assessment of thoracic surgery as framework elements, a semantic framework is constructed by integrating multiple framework elements.

[0033] S2332. Fill the data contained in the clinical diagnosis and treatment data into the corresponding frame elements to obtain the clinical semantic frame. Then fill the data contained in the non-privacy attribute data into the corresponding frame elements in turn to obtain the matching semantic frame.

[0034] S2333. Check and count the number of frame elements that cannot be filled simultaneously between the clinical semantic frame and each matching semantic frame. If the number of frame elements that cannot be filled simultaneously is greater than or equal to a preset threshold, delete the non-privacy attribute data. If the number of frame elements that cannot be filled simultaneously is less than the preset threshold, retain the non-privacy attribute data.

[0035] Furthermore, based on the semantic framework after padding, the similarity between the clinical diagnosis and treatment data and each non-privacy attribute data is calculated sequentially, including:

[0036] S2341. According to the order of the frame elements within the semantic frame, compare the two word data located in the same frame element of the clinical semantic frame and the matching semantic frame in turn.

[0037] S2342. Set up synonym sets for two word data respectively, and represent the two synonym sets in the form of semantic trees. Record the height of the two semantic trees and the length of the edge tree connecting the nodes of the two word data. Calculate the similarity between the two word data using the word data similarity formula.

[0038] S2343. Assign weight values ​​to the similarity of each word data between the clinical semantic frame and the matching semantic frame, and use the weighted similarity formula to calculate the similarity between the clinical semantic frame and the matching semantic frame, which is used as the similarity between clinical diagnosis and treatment data and non-privacy attribute data.

[0039] Furthermore, the expression for the word data similarity formula is as follows:

[0040]

[0041] In the formula, S(x) i y i ) represents the similarity between the i-th word data x in the clinical semantic framework and the i-th word data y in the matching semantic framework; L represents the length of the edge tree connecting the two word data nodes; H represents the height of the semantic tree;

[0042] The formula for weighted similarity is:

[0043]

[0044] In the formula, Sim(AB) represents the similarity between clinical semantic frame A and matching semantic frame B; n represents the number of frame elements in the clinical semantic frame and the matching semantic frame; w i This represents the weight value of the i-th word data.

[0045] Furthermore, effective features are extracted from similar information data to establish a risk assessment system. Based on the selected effective features, weighting factors for each feature in risk assessment are set, a risk assessment model is constructed, and the model is trained and tested using a training set and a test set composed of similar information data, including:

[0046] S31. Based on correlation analysis, extract effective features from similar information data;

[0047] S32. Set risk indicators for thoracic surgery, establish a risk assessment system, and set weighting factors for each effective feature for risk assessment based on the selected effective features.

[0048] S33. Establish a machine learning model based on effective features, and train and test the model using a training set and a test set composed of similar information data to obtain a risk assessment model.

[0049] Secondly, the present invention also provides a surgical decision support system, the system comprising:

[0050] The data storage module is used to acquire multi-source medical data. After the data source confirms that the data meets the privacy protection standards, it uploads the multi-source medical data to the cloud server and encrypts and stores the corresponding attribute values ​​according to the decomposition results, thus building a cloud database with a privacy protection mechanism.

[0051] The data filtering module is used to obtain clinical diagnosis and treatment information. It uploads the encrypted clinical diagnosis and treatment information to the cloud server through the data source and sends a query matching request to the cloud service platform. After the cloud service platform verifies the request, it establishes an encryption and decryption engine and decrypts the clinical diagnosis and treatment information. It also filters similar information data related to the clinical diagnosis and treatment information from the non-privacy attribute data of the cloud database through distributed computing nodes.

[0052] The model building module is used to extract effective features from similar information data and establish a risk assessment system. Based on the selected effective features, the weight factors of each feature for risk assessment are set, a risk assessment model is built, and the model is trained and tested using a training set and a test set composed of similar information data.

[0053] The model evaluation module is used to extract risk indicators that conform to the risk assessment system from clinical diagnosis and treatment information. The risk indicators are used as input data for the risk assessment model. The model calculates and outputs the assessment results of preoperative surgical risks to provide scientific basis and auxiliary decision support.

[0054] The data storage module, data filtering module, model building module, and model evaluation module are sequentially connected.

[0055] The beneficial effects of this invention are as follows:

[0056] 1. By integrating massive amounts of multi-source medical data and storing it in a cloud database with privacy protection mechanisms, the security of medical data is effectively guaranteed. Coupled with a secure query matching strategy, similar information data related to the clinical diagnosis and treatment information of the data subject is quickly filtered from the cloud database, providing sufficient data support for subsequent risk assessment and achieving personalized data customization for specific diseases. Furthermore, by analyzing the effective characteristics of similar information data and using machine learning algorithms to build a risk assessment model, accurate assessment and prediction of the preoperative risks of thoracic surgery for the data subject are achieved, providing doctors with scientific basis and decision support. This enables a comprehensive assessment, risk analysis, and prediction of the data subject, providing crucial support for preoperative decision-making while ensuring the privacy and security of the data subject's data.

[0057] 2. This invention establishes a collaborative and mutually supportive regulatory framework among the data source, cloud service platform, certification center, and data subject to build a cloud computing platform for online medical services and data sharing. By leveraging the powerful data processing capabilities of cloud servers and their computing nodes, it maximizes the efficiency of data processing, model calculation, and risk assessment while ensuring data security, thereby achieving a comprehensive preoperative risk assessment for the data subject.

[0058] 3. This invention divides multi-source medical data into privacy-protected attribute data and non-privacy attribute data according to privacy constraints. While ensuring the data security of the data subject, it opens up relevant non-privacy attribute data in the medical field for data analysis and model building, effectively improving the processing accuracy and generalization ability of the risk assessment model. At the same time, it adopts a data query and matching method based on a semantic framework, which effectively improves the comprehensiveness and accuracy of data matching, retains the closest known medical data of the data subject to the greatest extent, and improves the accuracy of model evaluation. Attached Figure Description

[0059] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this invention, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention. In the drawings:

[0060] Figure 1 This is a flowchart of a surgical decision support method according to an embodiment of the present invention;

[0061] Figure 2 This is a system principle block diagram of a surgical decision support system according to an embodiment of the present invention.

[0062] The diagram shows the following modules: 1. Data storage module; 2. Data filtering module; 3. Model building module; 4. Model evaluation module. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0064] Please see Figure 1 This invention provides a surgical decision support method, the method comprising:

[0065] S1. Acquire multi-source medical data. After the data source confirms that the data meets the privacy protection standards, upload the multi-source medical data to the cloud server and encrypt and store the corresponding attribute values ​​according to the decomposition results to build a cloud database with a privacy protection mechanism.

[0066] In the description of this invention, acquiring multi-source medical data, uploading the multi-source medical data to a cloud server after the data source confirms that it meets privacy protection standards, and encrypting and storing the corresponding attribute values ​​according to the decomposition results to construct a cloud database with a privacy protection mechanism includes:

[0067] S11. The data source applies for an identity certificate from the certification authority and uses the identity certificate to apply for database outsourcing services from the cloud service platform. After the service application is approved, the data source sets an attribute set based on the data types contained in the multi-source medical data, and then submits the attribute set and privacy constraints to the cloud service platform.

[0068] Specifically, in the process of outsourcing the database to a cloud service platform, to ensure data security, privacy, and availability management, it is necessary to implement functions such as privacy constraint set generation, data content management, data query processing, and certificate and key management. Privacy constraint set generation generates privacy constraints based on privacy protection policies and the characteristics of data attributes. Data content management handles the initialization of the outsourced database and the submission of data and requests for update operations such as insertion, modification, and deletion. Data query processing handles database queries and data processing. Certificate and key management manages the data source's own identity certificates and related keys.

[0069] S12. The cloud service platform performs minimum encrypted attribute decomposition on the attribute set according to the privacy constraints and returns the decomposition results to the data source. The data source confirms whether the privacy protection standard is met. If yes, the multi-source medical data is uploaded to the cloud server. If not, the privacy constraints are reset and the data is resubmitted to the cloud service platform along with the attribute set.

[0070] Specifically, the cloud service platform has trusted privacy data processing capabilities, and is responsible for functions such as identity authentication, access control and authorization, minimum encryption attribute decomposition, data content management, metadata management, key management and data query processing.

[0071] The system includes several functionalities: Identity Authentication, Access Control, and Authorization; Authentication, Access Control, and Permission Management; Least Encryption Attribute Decomposition; Data Content Management; Metadata Management; and Data Query Processing. The system comprises: Identity Authentication, Access Control, and Authorization functions to verify the identity of database administrators, data sources, and data subjects (patients or medical volunteers); Least Encryption Attribute Decomposition function to decompose the attribute set based on the attribute set submitted by the data source and privacy constraints using a least encryption attribute decomposition algorithm; Data Content Management to initialize, insert, modify, and delete data based on data and requests submitted by the data source; Metadata Management to store and manage metadata such as least encryption attribute decomposition data, encryption algorithms used for each attribute, and keys; Key Management to generate corresponding keys based on the encryption / decryption algorithm requirements of each attribute and update and destroy keys according to security policies; and Data Query Processing to respond to data query requests from tenants and users.

[0072] S13. The cloud server encrypts the corresponding attribute values ​​in the multi-source medical data according to the decomposition results of the attribute set, stores the processed data in the cloud database, and retains the metadata.

[0073] S14. The data subject applies to the certification center for a personal identity certificate, and sends an access application to the cloud service platform with the certificate. After the cloud service platform verifies the application, the data subject can legally access the cloud database.

[0074] Furthermore, in this invention, the cloud database divides multi-source medical data into privacy-protected attribute data and non-privacy attribute data according to attribute sets and privacy constraints. Based on these attribute sets and constraints, the medical data is categorized, with privacy-protected data encrypted and access permissions set to ensure that only authorized users can access and decrypt it. Non-privacy attribute data, on the other hand, can be directly stored in the database and queried by the data source or cloud server that has completed identity authentication and key verification.

[0075] S2. Obtain clinical diagnosis and treatment information. Upload the encrypted clinical diagnosis and treatment information to the cloud server through the data source. Send a query matching request to the cloud service platform. After the cloud service platform verifies the request, it establishes an encryption and decryption engine and decrypts the clinical diagnosis and treatment information. Then, through distributed computing nodes, it filters similar information data related to the clinical diagnosis and treatment information from the non-privacy attribute data of the cloud database.

[0076] In the description of this invention, clinical diagnosis and treatment information is obtained, encrypted clinical diagnosis and treatment information is uploaded to a cloud server through a data source, a query matching request is sent to a cloud service platform, and after the cloud service platform verifies the request, it establishes an encryption / decryption engine and decrypts the clinical diagnosis and treatment information. Then, through distributed computing nodes, it filters similar information data associated with the clinical diagnosis and treatment information from the non-privacy attribute data of the cloud database, including:

[0077] S21. The data source end encrypts and uploads the clinical diagnosis and treatment information of the data subject to the cloud server. The clinical diagnosis and treatment information includes basic information, medical history information, examination results and surgical plan.

[0078] S22. The data source sends a query matching request to the cloud service platform, and after the request is verified, the cloud service platform grants query matching permissions for the cloud database to the data source and the cloud server.

[0079] In the description of this invention, the data source sends a query matching request to the cloud service platform, and after the request is verified, the cloud service platform grants query matching permissions for the cloud database to the data source and the cloud server, including:

[0080] S221. The data source sends an authentication request to the cloud server. After successful authentication, the cloud server sends a query matching request from the data source to the cloud service platform.

[0081] S222. The cloud service platform establishes an encryption / decryption engine between the cloud server and the cloud database. The encryption / decryption engine is used to analyze query matching requests, extract the key containing clinical diagnosis and treatment information in the query matching requests, and then check the validity of the key. When the validity of the key passes the check, the encrypted clinical diagnosis and treatment information is decrypted and stored in the database to be matched.

[0082] Specifically, the encryption / decryption engine is a crucial component of a database encryption protection system. It refers to a software or hardware component used for encrypting and decrypting data. It sits between the cloud server's business application system and the cloud database server, responsible for encrypting and decrypting database information, and is transparent to application developers and operators. When medical data is sent from the data source to the cloud server, the encryption / decryption engine encrypts the data to ensure its security during transmission. Upon receiving the data on the cloud service platform, the encryption / decryption engine decrypts it for further processing and analysis. Encryption / decryption engines typically implement various encryption algorithms and key management mechanisms to provide efficient and secure data encryption and decryption services. In the processing of medical data, the encryption / decryption engine plays a vital role in ensuring the secure transmission and storage of the data subject's private information.

[0083] S223. The cloud service platform re-verifies the legality of the data source identity information contained in the query matching request. After the verification is passed, it grants query matching permissions to the data source and the cloud server, and opens up non-privacy attribute data in the cloud database for query matching.

[0084] S23. The cloud server assigns data query and matching tasks to distributed computing nodes, and filters similar information data from non-privacy attribute data in the cloud database.

[0085] In the description of this invention, the cloud server assigns data query and matching tasks to distributed computing nodes, and filters similar information data from non-privacy attribute data in the cloud database, including:

[0086] S231. The cloud server distributes data query matching tasks to available computing nodes through the task scheduling manager, and achieves parallel computing through several computing nodes.

[0087] S232. Based on the task requirements of the query matching task, the computing node reads a corresponding amount of non-privacy attribute data from the cloud database and obtains the decrypted clinical diagnosis and treatment data.

[0088] S233. Based on a predefined semantic framework, clinical diagnosis and treatment data and non-privacy attribute data are filled into the corresponding positions within the semantic framework, and missing items are checked.

[0089] In the description of this invention, based on a predefined semantic framework, clinical diagnosis and treatment data and non-privacy attribute data are respectively filled into the corresponding positions within the semantic framework, and missing items are checked, including:

[0090] S2331. Using the factors to be considered in the preoperative assessment of thoracic surgery as framework elements, a semantic framework is constructed by integrating multiple framework elements.

[0091] S2332. Fill the data contained in the clinical diagnosis and treatment data into the corresponding frame elements to obtain the clinical semantic frame. Then, fill the data contained in the non-privacy attribute data into the corresponding frame elements in turn to obtain the matching semantic frame.

[0092] S2333. Check and count the number of frame elements that cannot be filled simultaneously between the clinical semantic frame and each matching semantic frame. If the number of frame elements that cannot be filled simultaneously is greater than or equal to a preset threshold, delete the non-privacy attribute data. If the number of frame elements that cannot be filled simultaneously is less than the preset threshold, retain the non-privacy attribute data.

[0093] S234. Based on the semantic framework after filling, calculate the similarity between the clinical diagnosis and treatment data and each non-privacy attribute data in turn.

[0094] In the description of this invention, based on the semantic framework after padding, the similarity between clinical diagnosis and treatment data and each non-privacy attribute data is calculated sequentially, including:

[0095] S2341. According to the order of the frame elements within the semantic frame, compare the two word data located in the same frame element of the clinical semantic frame and the matching semantic frame in turn.

[0096] S2342. Set up synonym sets for two word data respectively, and represent the two synonym sets in the form of semantic trees. Record the height of the two semantic trees and the length of the edge tree connecting the nodes of the two word data. Calculate the similarity between the two word data using the word data similarity formula.

[0097] Specifically, let the clinical semantic frame be A = (x1, x2, x3, ..., x...). i Matching semantic frame B = (y1, y2, y3, ..., y i Since each word has at least one polysemous word, it belongs to the multiple synonym set method. This invention uses the following steps to calculate the similarity between two words:

[0098] (1) Use a join query in the Word Net dictionary database to find the IDs of all possible synonym sets of the two words to be compared.

[0099] (2) If word X has m meanings, that is, it belongs to m sets of synonyms, and word Y has n meanings, then the sets of synonyms to which X and Y belong have m×n pairs of combinations.

[0100] (3) Represent the two sets of synonyms in the form of a semantic tree.

[0101] (4) Find the intersection of the two semantic trees. If the synonyms of X and Y do not intersect, continue to search and expand the synonyms to generate an intersection.

[0102] In the description of this invention, the expression for the word data similarity formula is:

[0103]

[0104] In the formula, S(x) i y i Let represent the similarity between the i-th word data x in the clinical semantic framework and the i-th word data y in the matching semantic framework, where L represents the length of the edge tree connecting the two word data nodes, and H represents the height of the semantic tree.

[0105] S2343. Assign weight values ​​to the similarity of each word data between the clinical semantic frame and the matching semantic frame, and use the weighted similarity formula to calculate the similarity between the clinical semantic frame and the matching semantic frame, which is used as the similarity between clinical diagnosis and treatment data and non-privacy attribute data.

[0106] Specifically, the formula for calculating the weight w can be adjusted according to the number of synonym combinations num, such that w∈[0.5, 0.9]. When num is smaller, the weight w of the first pair of synonyms will be higher. However, the value of w decreases as num increases. When num is less than or equal to 3, w is at its maximum, with a maximum value of 0.9. When num is greater than 23, w is at its minimum, with a minimum value of 0.5.

[0107] When num is between 3 and 23, the formula for calculating w is 0.9 - 0.02 × (num - 3).

[0108] The expression for the weighted similarity formula is as follows:

[0109]

[0110] In the formula, Sim(AB) represents the similarity between clinical semantic frame A and matching semantic frame B, n represents the number of frame elements in the clinical semantic frame and the matching semantic frame, and w i This represents the weight value of the i-th word data.

[0111] S235. Pre-set a standard threshold for similarity and retain non-privacy attribute data that exceeds the standard threshold as similar information data related to clinical diagnosis and treatment data.

[0112] S3. Extract effective features from similar information data and establish a risk assessment system. Based on the selected effective features, set the weight factors of each feature for risk assessment, construct a risk assessment model, and train and test the model using a training set and a test set composed of similar information data.

[0113] In the description of this invention, the process of extracting effective features from similar information data and establishing a risk assessment system, setting weight factors for each feature based on the selected effective features, constructing a risk assessment model, and training and testing the model using a training set and a test set composed of similar information data includes:

[0114] S31. Based on correlation analysis, extract effective features from similar information data.

[0115] Specifically, correlation analysis is used to screen features with a high correlation to the risk of thoracic surgery by calculating the correlation coefficient or other correlation indicators between features. Generally, the larger the absolute value of the correlation coefficient, the stronger the correlation between the two features. Based on the results of the correlation analysis, the features are ranked, and those with a high correlation to the risk of thoracic surgery are selected as valid features. This selection and ranking requires the use of sorting or threshold setting methods.

[0116] In surgical decision support, effective features include the data subject's age, gender, body mass index (BMI), past medical history, type of surgery, surgical site, and preoperative examination results. These features are crucial for assessing surgical risk and can provide key information to help physicians determine the data subject's surgical risk level.

[0117] In addition, the selected effective features can be verified by using professional medical knowledge and statistical analysis to verify their correlation and reliability with the risk of thoracic surgery.

[0118] S32. Set risk indicators for thoracic surgery, establish a risk assessment system, and set weighting factors for each effective feature based on the selected effective features.

[0119] Specifically, the first step is to identify key risk indicators for assessing the risks of thoracic surgery, including the incidence of surgical complications, postoperative recovery time, and postoperative hospital stay, to comprehensively reflect the relationship between the data subject's preoperative condition and surgical risk. Based on these identified risk indicators, a risk assessment system for thoracic surgery is then established, encompassing various factors influencing surgical risk, such as the data subject's basic information, medical history, type of surgery, and preoperative examination results.

[0120] Based on the established risk assessment system, weights are assigned to selected effective features. Data-driven methods, specifically feature importance analysis in machine learning algorithms, can be used to automatically determine feature weights. The established risk assessment system is then validated and adjusted. By comparing the model with actual data, its accuracy and stability are evaluated. Furthermore, the feature weights are adjusted and optimized based on actual conditions to improve the model's predictive performance.

[0121] S33. Establish a machine learning model based on effective features, and train and test the model using a training set and a test set composed of similar information data to obtain a risk assessment model.

[0122] Specifically, the selected effective features and corresponding thoracic surgery risk indicators are compiled into a dataset, which is then divided into training and testing sets using a common split ratio, such as 70% for training and 30% for testing. The selected machine learning model is trained using the training set. During training, the model learns patterns and regularities in the data to establish relationships between features and risk indicators. The training process includes parameter optimization and model fitting.

[0123] The trained model is then evaluated using a test set. Performance metrics such as accuracy, precision, recall, and F1 score are calculated based on the predictions made on the test set to assess the model's predictive and generalization abilities. Based on the evaluation results, the model is fine-tuned and improved by experimenting with different parameter combinations, feature selection methods, or algorithms to enhance its performance and stability. Finally, the best-performing model is selected and applied to practical surgical decision support. By inputting clinical diagnosis and treatment information and characteristics of the data subjects, the model can output corresponding risk assessment results, providing a scientific basis for doctors' clinical decisions.

[0124] S4. Extract risk indicators that conform to the risk assessment system from clinical diagnosis and treatment information, use the risk indicators as input data for the risk assessment model, and calculate and output the preoperative risk assessment results through the model to provide scientific basis and auxiliary decision support.

[0125] By using the established risk assessment model and inputting the clinical diagnosis and treatment information of the data subject, the preoperative risk assessment results of the data subject's thoracic surgery are obtained. These assessment results help doctors to conduct a scientific and objective assessment of the data subject's surgical risk, guiding clinical decision-making and surgical management.

[0126] The evaluation results include the following aspects:

[0127] Risk Level: The assessment results categorize data subjects into different risk levels, such as low risk, medium risk, and high risk, to indicate the overall risk level of the data subjects.

[0128] Risk probability: The assessment results provide an estimate of the probability of the data subject experiencing specific complications or adverse events before surgery, such as postoperative infection or bleeding. This can help doctors more accurately assess the surgical risk of the data subject.

[0129] Risk factors: The assessment results may provide information about specific risk factors present in the data subject, such as hypertension, diabetes, and lung disease, helping doctors to have a more comprehensive understanding of the data subject's preoperative condition.

[0130] Recommendations and warnings: The assessment results provide recommendations or warnings based on the risk level of the data subject, such as recommendations to take additional examination measures, adjust surgical plans, or warnings about surgical risks.

[0131] Please see Figure 2 It also provides a surgical decision support system, which includes:

[0132] Data storage module 1 is used to acquire multi-source medical data. After the data source confirms that the data meets the privacy protection standards, it uploads the multi-source medical data to the cloud server and encrypts and stores the corresponding attribute values ​​according to the decomposition results, thus building a cloud database with a privacy protection mechanism.

[0133] Data filtering module 2 is used to obtain clinical diagnosis and treatment information. It uploads the encrypted clinical diagnosis and treatment information to the cloud server through the data source and sends a query matching request to the cloud service platform. After the cloud service platform verifies the request, it establishes an encryption and decryption engine and decrypts the clinical diagnosis and treatment information. Then, it filters similar information data related to the clinical diagnosis and treatment information from the non-privacy attribute data of the cloud database through distributed computing nodes.

[0134] Model building module 3 is used to extract effective features from similar information data and establish a risk assessment system. Based on the selected effective features, the weight factors of each feature for risk assessment are set, a risk assessment model is constructed, and the model is trained and tested using a training set and a test set composed of similar information data.

[0135] Model evaluation module 4 is used to extract risk indicators that conform to the risk assessment system from clinical diagnosis and treatment information. The risk indicators are used as input data for the risk assessment model. The model calculates and outputs the assessment results of preoperative surgical risks to provide scientific basis and auxiliary decision support.

[0136] The data storage module, data filtering module, model building module, and model evaluation module are sequentially connected.

[0137] In summary, by leveraging the technical solutions described above, this invention integrates massive amounts of multi-source medical data and stores it in a cloud database with privacy protection mechanisms, effectively ensuring the security of medical data. Coupled with a secure query and matching strategy, similar information data related to the clinical diagnosis and treatment information of the data subject is quickly filtered from the cloud database, providing sufficient data support for subsequent risk assessment and achieving personalized data customization for specific diseases. Furthermore, by analyzing the effective characteristics of similar information data and utilizing machine learning algorithms to construct a risk assessment model, accurate assessment and prediction of the preoperative risk of thoracic surgery for the data subject can be achieved, providing doctors with scientific basis and decision support. This enables comprehensive assessment, risk analysis, and prediction of the data subject, providing crucial support for preoperative decision-making while ensuring the privacy and security of the data subject's data. This invention constructs a cloud computing platform for online medical services and data sharing by establishing a collaborative and mutually supportive regulatory mechanism among the data source, cloud service platform, authentication center, and data subject. Utilizing the powerful data processing capabilities of cloud servers and their computing nodes, it maximizes the efficiency of data processing, model calculation, and risk assessment while ensuring data security, achieving a comprehensive assessment of the preoperative risk of the data subject. This invention categorizes multi-source medical data into privacy-protected attribute data and non-privacy attribute data according to privacy constraints. While ensuring the data security of the data subject, it opens up relevant non-privacy attribute data in the medical field for data analysis and model building, effectively improving the processing accuracy and generalization ability of the risk assessment model. At the same time, it adopts a data query and matching method based on a semantic framework, which effectively improves the comprehensiveness and accuracy of data matching, retains the closest known medical data of the data subject to the greatest extent, and improves the accuracy of model evaluation.

[0138] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

Claims

1. A surgical decision support method, characterized in that, The method includes: S1. Acquire multi-source medical data. After the data source confirms that the data meets the privacy protection standards, upload the multi-source medical data to the cloud server and encrypt and store the corresponding attribute values ​​according to the decomposition results to build a cloud database with a privacy protection mechanism. S2. Obtain clinical diagnosis and treatment information. Upload the encrypted clinical diagnosis and treatment information to the cloud server through the data source end. Send a query matching request to the cloud service platform. After the cloud service platform verifies the request, it establishes an encryption and decryption engine and decrypts the clinical diagnosis and treatment information. Then, through distributed computing nodes, it filters similar information data related to the clinical diagnosis and treatment information from the non-privacy attribute data of the cloud database. S3. Extract effective features from similar information data and establish a risk assessment system. Based on the selected effective features, set the weight factors of each feature for risk assessment, construct a risk assessment model, and train and test the model using a training set and a test set composed of similar information data. S4. Extract risk indicators that conform to the risk assessment system from clinical diagnosis and treatment information, use the risk indicators as input data for the risk assessment model, and calculate and output the preoperative risk assessment results through the model to provide scientific basis and auxiliary decision support.

2. The surgical decision support method according to claim 1, characterized in that, The process of acquiring multi-source medical data, whereby the data source confirms that the data meets privacy protection standards, uploads the multi-source medical data to the cloud server, and encrypts and stores the corresponding attribute values ​​according to the decomposition results to construct a cloud database with privacy protection mechanisms, includes: S11. The data source applies for an identity certificate from the certification center and applies for database outsourcing services from the cloud service platform based on the identity certificate. After the service application is completed, the data source sets an attribute set according to the data types contained in the multi-source medical data, and then submits the attribute set and privacy constraints to the cloud service platform. S12. The cloud service platform performs minimum encrypted attribute decomposition on the attribute set according to the privacy constraints and returns the decomposition result to the data source. The data source confirms whether the privacy protection standard is met. If yes, the multi-source medical data is uploaded to the cloud server. If no, the privacy constraints are reset and the data is resubmitted to the cloud service platform along with the attribute set. S13. The cloud server encrypts the corresponding attribute values ​​in the multi-source medical data according to the decomposition results of the attribute set, stores the processed data in the cloud database, and retains the metadata. S14. The data subject applies to the certification center for a personal identity certificate, and sends an access application to the cloud service platform with the certificate. After the cloud service platform verifies the application, the data subject can legally access the cloud database.

3. The surgical decision support method according to claim 2, characterized in that, The process of acquiring clinical diagnosis and treatment information involves uploading encrypted clinical diagnosis and treatment information to a cloud server via a data source, sending a query matching request to a cloud service platform, and, after verifying the request, the cloud service platform establishing an encryption / decryption engine to decrypt the clinical diagnosis and treatment information. It then uses distributed computing nodes to filter similar information data related to the clinical diagnosis and treatment information from the non-privacy attribute data of the cloud database, including: S21. The data source end encrypts and uploads the clinical diagnosis and treatment information of the data subject to the cloud server, wherein the clinical diagnosis and treatment information includes basic information, medical history information, examination results and surgical plan; S22. The data source sends a query matching request to the cloud service platform, and after the request is verified, the cloud service platform grants the data source and the cloud server the query matching permission of the cloud database. S23. The cloud server assigns the data query and matching task to the distributed computing nodes and filters similar information data from the non-privacy attribute data in the cloud database.

4. The surgical decision support method according to claim 3, characterized in that, The data source sends a query matching request to the cloud service platform, and after the request is verified, the cloud service platform grants the data source and the cloud server query matching permissions for the cloud database, including: S221. The data source sends an authentication request to the cloud server. After successful authentication, the cloud server sends a query matching request from the data source to the cloud service platform. S222. The cloud service platform establishes an encryption / decryption engine between the cloud server and the cloud database, uses the encryption / decryption engine to analyze the query matching request, extracts the key of the clinical diagnosis and treatment information contained in the query matching request, and then checks the validity of the key. When the validity of the key passes the check, the encrypted clinical diagnosis and treatment information is decrypted and stored in the database to be matched. S223. The cloud service platform verifies the legality of the data source identity information contained in the query matching request again. After the verification is successful, it grants query matching permissions to the data source and the cloud server, and opens up the non-privacy attribute data in the cloud database for query matching.

5. The surgical decision support method according to claim 3, characterized in that, The cloud server assigns data query and matching tasks to distributed computing nodes, and filters similar information data from the non-privacy attribute data in the cloud database, including: S231. The cloud server distributes data query matching tasks to available computing nodes through a task scheduling manager, and achieves parallel computing through several computing nodes. S232. Based on the task requirements of the query matching task, the computing node reads a corresponding amount of non-privacy attribute data from the cloud database and obtains the decrypted clinical diagnosis and treatment data. S233. Based on a predefined semantic framework, the clinical diagnosis and treatment data and non-privacy attribute data are filled into the corresponding positions within the semantic framework, and missing items are checked. S234. Based on the semantic framework after filling, calculate the similarity between the clinical diagnosis and treatment data and each non-privacy attribute data in sequence. S235. A standard threshold for similarity is preset, and non-privacy attribute data that is greater than the standard threshold is retained as similar information data related to the clinical diagnosis and treatment data.

6. The surgical decision support method according to claim 5, characterized in that, Based on a predefined semantic framework, the clinical diagnosis and treatment data and non-privacy attribute data are respectively filled into the corresponding positions within the semantic framework, and missing items are checked, including: S2331. Using the factors that need to be considered in the preoperative assessment of thoracic surgery as framework elements, a semantic framework is constructed by integrating multiple framework elements. S2332. Fill the data contained in the clinical diagnosis and treatment data into the corresponding frame elements to obtain the clinical semantic frame, and then fill the data contained in the non-privacy attribute data into the corresponding frame elements in turn to obtain the matching semantic frame. S2333. Check and count the number of frame elements that cannot be filled simultaneously between the clinical semantic frame and each of the matching semantic frames. If the number of frame elements that cannot be filled simultaneously is greater than or equal to a preset number threshold, delete the non-privacy attribute data. If the number of frame elements that cannot be filled simultaneously is less than the preset number threshold, retain the non-privacy attribute data.

7. The surgical decision support method according to claim 6, characterized in that, The similarity calculation between the clinical diagnosis and treatment data and each non-privacy attribute data is performed sequentially based on the semantic framework after filling in the information, including: S2341. According to the order of the frame elements within the semantic frame, compare the two word data located in the same frame element of the clinical semantic frame and the matching semantic frame in turn. S2342. Set up synonym sets for two word data respectively, and represent the two synonym sets in the form of semantic trees. Record the height of the two semantic trees and the length of the edge tree connecting the nodes of the two word data. Calculate the similarity between the two word data using the word data similarity formula. S2343. Assign weight values ​​to the similarity of each word data between the clinical semantic framework and the matching semantic framework, and calculate the similarity between the clinical semantic framework and the matching semantic framework using a weighted similarity formula, which is used as the similarity between the clinical diagnosis and treatment data and the non-privacy attribute data.

8. The surgical decision support method according to claim 7, characterized in that, The expression for the word data similarity formula is: In the formula, S(x) i y i () represents the similarity between the i-th word data x in the clinical semantic frame and the i-th word data y in the matching semantic frame; L represents the length of the edge tree connecting two word data nodes; H represents the height of the semantic tree; The expression for the weighted similarity formula is: In the formula, Sim(AB) represents the similarity between clinical semantic frame A and matching semantic frame B; n represents the number of frame elements within the clinical semantic frame and the matching semantic frame; w i This represents the weight value of the i-th word data.

9. A surgical decision support method according to claim 2, characterized in that, The process of extracting effective features from similar information data and establishing a risk assessment system, setting weight factors for each feature based on the selected effective features, constructing a risk assessment model, and training and testing the model using a training set and a test set composed of similar information data includes: S31. Based on correlation analysis, extract effective features from the similar information data; S32. Set risk indicators for thoracic surgery, establish a risk assessment system, and set weighting factors for each of the selected effective features for risk assessment based on the selected effective features. S33. Based on the effective features, establish a machine learning model, and train and test the model using the training set and test set composed of similar information data to obtain a risk assessment model.

10. A surgical decision support system for implementing the surgical decision support method according to any one of claims 1-9, characterized in that, The system includes: The data storage module is used to acquire multi-source medical data. After the data source confirms that the data meets the privacy protection standards, it uploads the multi-source medical data to the cloud server and encrypts and stores the corresponding attribute values ​​according to the decomposition results, thus building a cloud database with a privacy protection mechanism. The data filtering module is used to obtain clinical diagnosis and treatment information. It uploads the encrypted clinical diagnosis and treatment information to the cloud server through the data source and sends a query matching request to the cloud service platform. After the cloud service platform verifies the request, it establishes an encryption and decryption engine and decrypts the clinical diagnosis and treatment information. It also filters similar information data related to the clinical diagnosis and treatment information from the non-privacy attribute data of the cloud database through distributed computing nodes. The model building module is used to extract effective features from similar information data and establish a risk assessment system. Based on the selected effective features, the weight factors of each feature for risk assessment are set, a risk assessment model is built, and the model is trained and tested using a training set and a test set composed of similar information data. The model evaluation module is used to extract risk indicators that conform to the risk assessment system from clinical diagnosis and treatment information. The risk indicators are used as input data for the risk assessment model. The model calculates and outputs the assessment results of preoperative surgical risks to provide scientific basis and auxiliary decision support. The data storage module, the data filtering module, the model building module, and the model evaluation module are sequentially connected.

Images