Surgical Risk Prediction Method, Device, Equipment and Medium Based on Knowledge Graph

Abstract

The present invention discloses a surgical risk prediction method, device, equipment and medium based on a knowledge graph, including: extracting data to be processed from a plurality of data sources and performing data preprocessing on the data to be processed to obtain processed data; obtaining a plurality of triples and constructing a surgical risk knowledge graph; performing ontology alignment on the surgical risk knowledge graph based on a preset terminology system; performing risk assessment and update on multiple triples belonging to the same relationship to update the surgical risk knowledge graph; inputting target patient characteristics and a target surgical type into the surgical risk knowledge graph to obtain a target surgical risk. The present invention belongs to the field of surgical risk prediction. The present invention maps text, numerical values, and images into graph nodes uniformly to solve the problem of heterogeneous data integration.

Description

Technical Field

[0001] The present invention relates to the field of surgical risk prediction, and particularly to a surgical risk prediction method, device, equipment, and medium based on a knowledge graph. Background Art

[0002] A knowledge graph is a model for representing information and knowledge. The knowledge graph organizes and connects entities in the real world and the relationships between them in a structured manner. Specifically, a knowledge graph consists of a series of entities, attributes, and relationships. Each entity represents a specific concept, object, or event, and the relationships describe various connections between these entities.

[0003] Currently, the prediction of surgical risk usually uses neural networks. However, relying solely on neural networks for prediction lacks dynamic correlation analysis of elements such as patient individual characteristics, surgical types, and postoperative complications. Therefore, the present invention proposes a surgical risk prediction method based on a knowledge graph. Summary of the Invention

[0004] By providing a surgical risk prediction method, device, equipment, and medium based on a knowledge graph, the present invention solves the technical problem in the prior art of lacking dynamic correlation analysis of elements such as patient individual characteristics, surgical types, and postoperative complications, and achieves the technical effect of more comprehensively predicting the surgical risk of patients.

[0005] In a first aspect, the present invention provides a surgical risk prediction method based on a knowledge graph, the method comprising:

[0006] Extracting data to be processed from a plurality of data sources and performing data preprocessing on the data to be processed to obtain processed data;

[0007] Performing entity recognition and relationship extraction on the processed data to obtain a plurality of triples and constructing a surgical risk knowledge graph, the triples including patient characteristics, surgical types, and risk assessments;

[0008] Based on a preset terminology system, performing ontology alignment on the surgical risk knowledge graph;

[0009] According to the importance of the data sources, performing risk assessment update on multiple triples belonging to the same relationship to update the surgical risk knowledge graph;

[0010] Inputting target patient characteristics and a target surgical type into the surgical risk knowledge graph to obtain a target surgical risk.

[0011] Further, according to the importance of the data sources, performing risk assessment update on multiple triples belonging to the same relationship to update the surgical risk knowledge graph, including:

[0012] Classify each data source according to its importance level, where each level corresponds to a weight;

[0013] Trace the data source of the triple, and determine the first weight of the risk assessment in this triple according to the level of this data source;

[0014] Determine the second weight of the risk assessment in this triple according to the maximum update frequency in the data source and the update frequency of the data source of this triple;

[0015] Update the risk assessment of this triple according to the first weight, the second weight and the risk assessment in this triple.

[0016] Furthermore, update the risk assessment of this triple according to the first weight, the second weight and the risk assessment in this triple, including:

[0017]

[0018]

[0019] Among them, is the updated risk assessment of the th triple, is the first weight of the th triple, is the second weight of the th triple, is the risk assessment of the th triple before update, is the weight corresponding to the level of the th data source, is the weight corresponding to the maximum level of the data source.

[0020] Furthermore, extract the data to be processed from several data sources and perform data preprocessing on the data to be processed to obtain the processed data, including:

[0021] Segment the unstructured text in the data to be processed based on a preset medical dictionary;

[0022] Standardize the time series data in the data to be processed;

[0023] Perform privacy desensitization processing on the data to be processed based on differential privacy technology.

[0024] Furthermore, perform ontology alignment on the surgical risk knowledge graph based on a preset term system, including:

[0025] Based on the SNOMED CT-based terminology system, perform ontology alignment on the vocabulary in the surgical risk knowledge graph and eliminate the vocabulary with concept alignment;

[0026] For the vocabulary without concept alignment, calculate the edit distance with the vocabulary in the SNOMED CT-based terminology system based on the fuzzy matching algorithm, and perform ontology alignment on the pair of vocabulary with the shortest distance;

[0027] After the surgical risk knowledge graph completes ontology alignment, determine whether the ontology alignment between each vocabulary is reasonable based on the attribute relationships in the SNOMED CT-based terminology system;

[0028] If it is unreasonable, send verification information to the client, and the verification information includes the vocabulary groups with unreasonable ontology alignment.

[0029] Further, input the target patient characteristics and the target surgical type into the surgical risk knowledge graph to obtain the target surgical risk, including:

[0030] Input the target patient characteristics and the target surgical type into the surgical risk knowledge graph to obtain several surgical risk results;

[0031] According to the several surgical risk results, construct a risk range and use it as the target surgical risk.

[0032] Further, according to the several surgical risk results, construct a risk range and use the risk range as the target surgical risk, including:

[0033] Use the minimum value of the several surgical risk results as the minimum value of the risk range;

[0034] Use the maximum value of the several surgical risk results as the maximum value of the risk range.

[0035] In a second aspect, the present invention provides a surgical risk prediction device based on a knowledge graph, and the device includes:

[0036] An extraction module for extracting data to be processed from several data sources and performing data preprocessing on the data to be processed to obtain processed data;

[0037] A graph module for performing entity recognition and relationship extraction on the processed data to obtain several triples and constructing a surgical risk knowledge graph, where the triples include patient characteristics, surgical types, and risk assessments;

[0038] An alignment module for performing ontology alignment on the surgical risk knowledge graph based on a preset terminology system;

[0039] An evaluation and update module, configured to perform risk assessment and update on multiple triples belonging to the same relationship according to the importance degree of data sources, so as to update the surgical risk knowledge graph;

[0040] A risk module, configured to input target patient characteristics and a target surgical type into the surgical risk knowledge graph to obtain a target surgical risk.

[0041] In a third aspect, the present invention provides an electronic device, including:

[0042] A processor;

[0043] A memory for storing executable instructions of the processor;

[0044] Wherein, the processor is configured to execute to implement the surgical risk prediction method based on a knowledge graph provided in the first aspect.

[0045] In a fourth aspect, the present invention provides a non-transitory computer-readable storage medium, when instructions in the storage medium are executed by a processor of an electronic device, enabling the electronic device to execute and implement the surgical risk prediction method based on a knowledge graph provided in the first aspect.

[0046] One or more technical solutions provided in the present invention have at least the following technical effects or advantages:

[0047] The present invention maps text (medical records), numerical values (vital signs), and images (surgical videos) into graph nodes uniformly, solving the problem of heterogeneous data integration. Based on a streaming computing framework, a closed loop of "data input - graph update - model optimization" is realized; a "risk conduction weight" algorithm is proposed to quantify the cascading effect between risk factors.

[0048] The present invention can distinguish the authority and reliability of different sources by considering the importance degree of data sources, assigns high weights to more credible data sources, thereby improving the accuracy of overall evaluation; secondly, by considering the update frequency of data sources, it ensures that risk assessment reflects the latest medical discoveries and clinical practices, enhances the information freshness of the knowledge graph, and also improves its dynamic adaptation ability, enabling it to quickly respond to newly emerging data and changes; in addition, combining these two for risk assessment and update helps to optimize resource allocation, giving priority to processing high-quality and up-to-date data and avoiding the influence of outdated information.

[0049] The present invention constructs a risk range through the prediction results of multiple triples, making the expectations of relevant personnel for surgical risks more accurate and reducing the doctor-patient risk. Description of the Drawings

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for description in the embodiments. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0051] Figure 1 It is a schematic flowchart of the surgical risk prediction method based on a knowledge graph provided by the present invention;

[0052] Figure 2 It is a schematic structural diagram of the surgical risk prediction device based on a knowledge graph provided by the present invention. Detailed implementation manners

[0053] By providing a surgical risk prediction method based on a knowledge graph in the embodiments of the present invention, the technical problem in the prior art of lacking dynamic correlation analysis of elements such as patient individual characteristics, surgical types, and postoperative complications is solved.

[0054] The technical solution of the present invention for solving the above technical problem has the following general idea:

[0055] A surgical risk prediction method based on a knowledge graph, the method includes: extracting data to be processed from a plurality of data sources and performing data preprocessing on the data to be processed to obtain processed data; performing entity recognition and relationship extraction on the processed data to obtain a plurality of triples and constructing a surgical risk knowledge graph, the triples include patient characteristics, surgical types, and risk assessments; performing ontology alignment on the surgical risk knowledge graph based on a preset terminology system; updating risk assessments for multiple triples belonging to the same relationship according to the importance of the data sources to update the surgical risk knowledge graph; inputting target patient characteristics and target surgical types into the surgical risk knowledge graph to obtain a target surgical risk.

[0056] To better understand the above technical solution, the following will describe the above technical solution in detail in combination with the accompanying drawings of the specification and specific implementation manners.

[0057] First, it should be noted that the term "and / or" appearing in this article is merely a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this article generally represents an "or" relationship between the preceding and following associated objects.

[0058] The present invention provides a surgical risk prediction method based on a knowledge graph as Figure 1 shown, the method includes:

[0059] S11. Extract the data to be processed from several data sources and perform data preprocessing on the data to be processed to obtain the processed data.

[0060] Specifically, it includes: segmenting the unstructured text in the data to be processed based on a preset medical dictionary; standardizing the time series data in the data to be processed; and performing privacy desensitization processing on the data to be processed based on differential privacy technology.

[0061] The data sources can be electronic medical records (EMR), PubMed literature, surgical video reports, medical device logs, and patient physiological signals.

[0062] In the medical field, data (such as medical records, doctor diagnoses, patient descriptions, etc.) can exist in the form of unstructured text and usually needs to be processed before further analysis or modeling.

[0063] The preset medical dictionary can be a medical-specific dictionary enhancement, which includes professional terms, disease names, drug names, symptom descriptions, and other vocabulary in the medical field.

[0064] Segmenting the unstructured text means splitting the continuous text into individual words or phrases. When segmenting, the preset medical dictionary will be referred to to ensure that medical terms can be correctly identified and retained. Converting the unstructured text into a structured vocabulary set facilitates subsequent natural language processing (NLP) tasks, such as information extraction, classification, clustering, etc.

[0065] Time series data (such as a patient's body temperature changes, heart rate monitoring data, medication time records, etc.) is an important part of medical data. However, different devices, systems, or institutions may record this data in different formats.

[0066] The timestamp can be converted to a unified format (such as the ISO 8601 standard); the numerical values of the time series data can be standardized (such as Z-score standardization or Min-Max normalization) to make them have the same dimension or distribution range; if the sampling frequencies of time series data from different sources are inconsistent (such as once per second or once per minute), they can be aligned through interpolation or resampling.

[0067] Standardization eliminates the differences in data format and scale, enabling time series data to be analyzed and modeled within a unified framework.

[0068] Medical data usually contains sensitive information (such as patient names, ID numbers, medical histories, etc.), and direct use may lead to privacy leakage.

[0069] Differential privacy is a mathematically rigorous privacy protection method that conceals the true information of an individual by adding random noise to the original data while ensuring the accuracy of the overall statistical properties.

[0070] In the data preprocessing stage, differential privacy algorithms can be applied to sensitive fields (such as age, gender, diagnosis results, etc.) to add appropriate noise. Differential privacy technology preserves the statistical characteristics of the data as much as possible while protecting personal privacy, thereby supporting subsequent analysis and machine learning tasks.

[0071] S12, perform entity recognition and relationship extraction on the processed data to obtain several triplets and construct a surgical risk knowledge graph. The triplets include patient characteristics, surgical type, and risk assessment.

[0072] Pre-trained medical language models (such as BioBERT+CRF) can be used to extract entities, including: patient characteristics (age, underlying diseases, allergy history); surgical elements (operative procedures, incision types, anesthesia methods); risk factors (bleeding volume, infection indicators, complications).

[0073] After entity recognition, a "patient-surgery-risk" triplet can be constructed based on a graph neural network (GNN), for example: (patient-orthopedic surgery-venous thrombosis risk), where the risk of venous thrombosis is expressed as a percentage. Patient refers to patient data, which includes the patient's physical signs, such as body temperature, blood pressure, emotions, etc. Orthopedic surgery refers to orthopedic surgery data, which includes data such as the steps, drugs, and duration required for the surgery.

[0074] S13, based on the preset terminology system, ontology alignment of surgical risk knowledge graph is performed.

[0075] Specifically, it includes: based on the terminology system of SNOMED CT, the vocabulary in the surgical risk knowledge graph is ontologically aligned, and the vocabulary with concept alignment is eliminated; for the vocabulary that is not conceptually aligned, the edit distance is calculated with the vocabulary in the terminology system of SNOMEDCT based on the fuzzy matching algorithm, and the pair of vocabulary with the shortest distance is ontologically aligned; after the surgical risk knowledge graph completes the ontology alignment, based on the attribute relationship in the terminology system of SNOMED CT, it is determined whether the ontology alignment between the vocabulary is reasonable; if it is unreasonable, verification information is sent to the client, and the verification information includes the vocabulary group with unreasonable ontology alignment.

[0076] By comparing the vocabulary in the knowledge graph with the terminology system of SNOMED CT, semantically consistent concepts are identified; once a match is found, the vocabulary is considered to have completed ontology alignment.

[0077] For terms without concept alignment, calculate the edit distance (or Euclidean distance) between them and the terms in the SNOMED CT terminology system based on the fuzzy matching algorithm, and perform ontology alignment on the pair of terms with the shortest distance.

[0078] Specifically, the edit distance (such as the Levenshtein distance) can be used as a measurement criterion to calculate the similarity between the unaligned terms and all potentially relevant terms in the SNOMED CT terminology system. Then, select the pair of terms with the highest similarity (i.e., the shortest edit distance) for pairing and attempt to complete the ontology alignment.

[0079] After the ontology alignment of the surgical risk knowledge graph is completed, based on the attribute relationships in the SNOMED CT terminology system, determine whether the ontology alignment between each pair of terms is reasonable to verify whether the relationships between the terms with completed ontology alignment conform to the logic and structure defined within SNOMED CT.

[0080] According to the attributes and relationships defined in SNOMED CT (such as parent-child relationships, synonyms, etc.), check whether the relationships between each pair of terms with ontology alignment are reasonable.

[0081] If it is unreasonable, send a verification message to the client. The verification message includes the group of terms with unreasonable ontology alignment, and by transmitting the message to the client, relevant personnel can make judgments based on experience.

[0082] S14. According to the importance of the data source, perform risk assessment updates on multiple triples belonging to the same relationship to update the surgical risk knowledge graph.

[0083] Specifically, it includes: classifying each data source according to its importance, where each level corresponds to a weight; tracing the data source of the triple, and determining the first weight of the risk assessment in this triple according to the level of this data source; determining the second weight of the risk assessment in this triple according to the maximum update frequency in the data source and the update frequency of the data source of this triple; updating the risk assessment of this triple according to the first weight, the second weight, and the risk assessment in this triple.

[0084] It also includes:

[0085]

[0086]

[0087] Among them, is the updated risk assessment for the th triple, is the first weight of the th triple, is the second weight of the th triple, is the risk assessment before updating the th triple, is the weight corresponding to the level of the th data source, is the weight corresponding to the maximum level of the data source.

[0088] First, according to the importance of the data sources, each data source is classified into levels. Each level corresponds to a weight, and the weight setting for each level can be determined according to historical experience values, such as 0.9, 0.8, 0.6, 0.5, etc. from high to low.

[0089] For each triple, trace the data source it comes from, and determine the first weight of the triple according to the level of the data source. The first weight is directly equal to the weight of the level of the data source.

[0090] In addition, weighted fusion can also be performed according to the importance of the data sources.

[0091] Specifically: The weights corresponding to the levels of each data source can be obtained, and multiple triples belonging to the same relationship can be weighted and fused according to the corresponding weights. It can be understood that even under the same relationship, the risks may be different.

[0092] Taking the above as an example, patient - orthopedic surgery - venous thromboembolism risk, where the information in "patient" and "orthopedic surgery" is the same, but the "venous thromboembolism risk" may be different (due to the different credibility of the data sources, and the credibility is positively correlated with the importance). The results are weighted and fused, and the average is taken to obtain the fused venous thromboembolism risk.

[0093] Of course, if the weighted fusion method is not adopted, the above method provided by the present invention can also be used to update the risk assessment of the triple.

[0094] The present invention considers the importance of the data sources to distinguish the authority and reliability of different sources, assigns high weights to more credible data sources, thereby improving the accuracy of the overall assessment; secondly, by considering the update frequency of the data sources, it ensures that the risk assessment reflects the latest medical discoveries and clinical practices. This method not only enhances the information freshness of the knowledge graph, but also improves its dynamic adaptation ability, enabling it to quickly respond to newly emerging data and changes. In addition, combining these two for risk assessment update helps to optimize resource allocation, prioritize the processing of high-quality and up-to-date data, and avoid the influence of outdated information.

[0095] S15, input the target patient characteristics and the target surgical type into the surgical risk knowledge graph to obtain the target surgical risk.

[0096] Specifically, it includes: inputting the target patient characteristics and the target surgical type into the surgical risk knowledge graph to obtain several surgical risk results; constructing a risk range based on the several surgical risk results and using it as the target surgical risk.

[0097] It includes: constructing a risk range based on the several surgical risk results and using the risk range as the target surgical risk, including: taking the minimum value of the several surgical risk results as the minimum value of the risk range; taking the maximum value of the several surgical risk results as the maximum value of the risk range.

[0098] It can be understood that after inputting the target patient characteristics and the target surgical type, there may be the same information in multiple triples at the same time, and each triple corresponds to a surgical risk result. Since the surgical risk results have all been updated, it can be considered that the execution confidence of each triple corresponding to a surgical risk result is high.

[0099] The present invention constructs a risk range through the prediction results of multiple triples, making the expectations of relevant personnel for surgical risks more accurate and reducing the doctor-patient risk.

[0100] In addition, the present invention adopts an attribute graph model (Neo4j), which supports complex relationship queries; can access postoperative follow-up data in real time to trigger graph updates; and automatically eliminates obsolete knowledge based on the node importance evaluation of reinforcement learning.

[0101] It is also possible to generate a risk causation chain through graph path backtracking (for example: diabetes → delayed wound healing → infection risk). Through the risk causation chain, the cause of risk formation can be further explained, improving the trust between patients and medical staff and reducing the doctor-patient risk.

[0102] Taking laparoscopic cholecystectomy as an example:

[0103] The system extracts the patient's "obesity, hypertension" entities from the medical record;

[0104] Associating the risk rules in the literature: obesity + pneumoperitoneum pressure > 15 mmHg → the risk of deep vein thrombosis increases by 40%;

[0105] It is recommended to use an intermittent pneumatic device during the operation.

[0106] Effect:

[0107] Risk assessment accuracy: The AUC reaches 0.92 on the test set (compared with 0.78 of the traditional logistic regression model);

[0108] Support real-time early warning: When the intraoperative monitoring data is abnormal, automatically push the risks associated with the graph (such as a sudden increase in heart rate → indicating the possibility of air embolism).

[0109] In summary, the present invention provides a surgical risk prediction method based on a knowledge graph. The method includes: extracting data to be processed from several data sources and performing data preprocessing on the data to be processed to obtain processed data; performing entity recognition and relationship extraction on the processed data to obtain a number of triples and constructing a surgical risk knowledge graph, where the triples include patient characteristics, surgical types, and risk assessments; performing ontology alignment on the surgical risk knowledge graph based on a preset terminology system; updating the risk assessment of multiple triples belonging to the same relationship according to the importance of the data sources to update the surgical risk knowledge graph; inputting target patient characteristics and a target surgical type into the surgical risk knowledge graph to obtain a target surgical risk. The present invention maps text (medical records), numerical values (vital signs), and images (surgical videos) into graph nodes uniformly, solving the problem of heterogeneous data integration. Based on a streaming computing framework, a closed loop of "data input - graph update - model optimization" is realized; a "risk conduction weight" algorithm is proposed to quantify the cascading effect between risk factors. The present invention can distinguish the authority and reliability of different sources by considering the importance of the data sources, assigning high weights to more credible data sources, thereby improving the accuracy of the overall assessment; secondly, by considering the update frequency of the data sources, it is ensured that the risk assessment reflects the latest medical discoveries and clinical practices, not only enhancing the information freshness of the knowledge graph, but also improving its dynamic adaptation ability, enabling it to quickly respond to newly emerging data and changes; combining the two for risk assessment update helps optimize resource allocation, giving priority to processing high-quality and up-to-date data and avoiding the influence of outdated information. The present invention constructs a risk range through the prediction results of multiple triples, making the expectations of relevant personnel for surgical risks more accurate and reducing the risks of doctors and patients.

[0110] Based on the same inventive concept, the present invention provides a surgical risk prediction device as shown in Figure 2 which includes:

[0111] An extraction module 21, configured to extract data to be processed from several data sources and perform data preprocessing on the data to be processed to obtain processed data;

[0112] A graph module 22, configured to perform entity recognition and relationship extraction on the processed data to obtain a number of triples and construct a surgical risk knowledge graph, where the triples include patient characteristics, surgical types, and risk assessments;

[0113] An alignment module 23, configured to perform ontology alignment on the surgical risk knowledge graph based on a preset terminology system;

[0114] An evaluation and update module 24, configured to update the risk assessment of multiple triples belonging to the same relationship according to the importance of the data sources to update the surgical risk knowledge graph;

[0115] A risk module 25, configured to input target patient characteristics and a target surgical type into a surgical risk knowledge graph to obtain a target surgical risk.

[0116] Based on the same inventive concept, the present invention further provides an electronic device as shown, including:

[0117] A processor;

[0118] A memory for storing instructions executable by the processor;

[0119] Wherein, the processor is configured to execute to implement the surgical risk prediction method based on a knowledge graph provided as described above.

[0120] Based on the same inventive concept, the present invention further provides a non-transitory computer-readable storage medium. When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute and implement the surgical risk prediction method based on a knowledge graph provided as described above.

[0121] Since the electronic device introduced in this embodiment is the electronic device adopted for implementing the information processing method in the embodiments of the present invention, based on the information processing method introduced in the embodiments of the present invention, those skilled in the art can understand the specific implementation manners and various variations of the electronic device in this embodiment. Therefore, the specific implementation of how this electronic device implements the method in the embodiments of the present invention will not be described in detail here. As long as the electronic device adopted by those skilled in the art to implement the information processing method in the embodiments of the present invention falls within the scope of protection of the present invention.

[0122] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program code.

[0123] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, and the combination of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate for implementing in the process Figure 1 a process or multiple processes and / or blocksFigure 1 means for the functions specified in one or more blocks.

[0124] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction means that implements the functions specified in one Figure 1 process or processes and / or one Figure 1 block or blocks.

[0125] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus, such that a series of operational steps are performed on the computer or other programmable apparatus to produce a computer-implemented process, so that the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one Figure 1 process or processes and / or one Figure 1 block or blocks.

[0126] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made by those skilled in the art once they learn of the basic inventive concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications that fall within the scope of the present invention.

[0127] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A surgical risk prediction method based on knowledge graph, characterized in that: The method comprises: Extracting data to be processed from a number of data sources and performing data preprocessing on the data to be processed to obtain processed data; Performing entity recognition and relationship extraction on the processed data to obtain a number of triples and construct a surgical risk knowledge graph, wherein the triples include patient characteristics, surgical type, and risk assessment; Based on a preset terminology system, ontology alignment is performed on the surgical risk knowledge graph; According to the importance of the data source, risk assessment is updated for multiple triples belonging to the same relationship to update the surgical risk knowledge graph; Input the target patient characteristics and the target surgery type into the surgical risk knowledge graph to obtain the target surgical risk; wherein, According to the importance of the data source, multiple triples belonging to the same relationship are updated with risk assessment to update the surgical risk knowledge graph, including: According to the importance of data sources, each data source is graded, where each grade corresponds to a weight; Tracing back the data source of the triplet, and determining the first weight of the risk assessment in the triplet according to the level of the data source; Determining a second weight of risk assessment in the triplet according to a maximum value of update frequencies in the data source and an update frequency of the data source of the triplet; The risk assessment of the triplet is updated according to the first weight and the second weight of the risk assessment in the triplet and the risk assessment.

2. The method for predicting surgical risks based on knowledge graph according to claim 1, characterized in that: The risk assessment of the triplet is updated according to the first weight and the second weight of the risk assessment in the triplet and the risk assessment, including: in, For the The updated risk assessment of the triples is For the The first weight of the triples, For the The second weight of the triplet, For the Risk assessment before triple update, For the The update frequency of each data source, The maximum update frequency of the data source.

3. The method for predicting surgical risks based on knowledge graph according to claim 1, characterized in that: Extract the data to be processed from several data sources and perform data preprocessing on the data to obtain processed data, including: Segment the unstructured text in the data to be processed based on the preset medical dictionary; Standardize the time series data in the data to be processed; Based on differential privacy technology, the data to be processed is desensitized.

4. The method for predicting surgical risks based on knowledge graph according to claim 1, characterized in that: Based on the preset terminology system, ontology alignment is performed on the surgical risk knowledge graph, including: Based on the terminology system of SNOMED CT, the vocabulary in the surgical risk knowledge graph is aligned ontology, and the vocabulary with concept alignment is removed; For the words that are not conceptually aligned, the edit distance is calculated with the words in the terminology system of SNOMED CT based on the fuzzy matching algorithm, and the pair of words with the shortest distance is aligned with the ontology; After the surgical risk knowledge graph completes the ontology alignment, based on the attribute relationship in the terminology system of the SNOMED CT, determine whether the ontology alignment between the terms is reasonable; If it is unreasonable, a verification message is sent to the client, which includes the vocabulary group with unreasonable ontology alignment.

5. The method for predicting surgical risks based on knowledge graph according to claim 1, characterized in that: Input the target patient characteristics and the target surgery type into the surgical risk knowledge graph to obtain the target surgical risk, including: Inputting target patient characteristics and target surgery type into the surgical risk knowledge graph to obtain several surgical risk results; Based on the results of several surgical risks, a risk range is constructed and used as the target surgical risk.

6. The method for predicting surgical risks based on knowledge graph according to claim 5, characterized in that: Based on the results of several surgical risks, a risk range is constructed and used as the target surgical risk, including: The minimum value of several surgical risk results is taken as the minimum value of the risk range; The maximum value of several surgical risk results is taken as the maximum value of the risk range.

7. A surgical risk prediction device based on knowledge graph, characterized in that: The method for predicting surgical risks based on a knowledge graph applied to any one of claims 1 to 6, wherein the device comprises: An extraction module is used to extract the data to be processed from several data sources and perform data preprocessing on the data to be processed to obtain processed data; A graph module, used for performing entity recognition and relationship extraction on the processed data, obtaining a number of triples and constructing a surgical risk knowledge graph, wherein the triples include patient characteristics, surgical type, and risk assessment; An alignment module, used for ontology alignment of the surgical risk knowledge graph based on a preset terminology system; An evaluation and updating module, used to perform risk evaluation and update on multiple triples belonging to the same relationship according to the importance of the data source, so as to update the surgical risk knowledge graph; The risk module is used to input target patient characteristics and target surgery type into the surgery risk knowledge graph to obtain target surgery risks.

8. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; Wherein, the processor is configured to execute to implement the surgical risk prediction method based on knowledge graph as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium, characterized in that: When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to implement the surgical risk prediction method based on a knowledge graph as described in any one of claims 1 to 6.

Images