Construction method and system for realizing medical atlas through data attributes

Abstract

The invention relates to the technical field of medical data processing, and discloses a construction method and system for realizing a medical map by data attributes, and the method comprises the steps: carrying out the quality sublimation processing of medical medical record text data, obtaining target medical record data, carrying out the statement splitting of the target medical record data, and obtaining medical record split data; performing privacy purification processing on the medical record split data to obtain purified medical record data, calculating information interconnection strength between entity information, and analyzing an entity association relationship between medical record entities; evaluating the relationship confidence of the entity association relationship, performing relationship screening processing on the entity association relationship to obtain a credible entity relationship, and performing entity fusion on the medical record entities to obtain a medical record entity set; and setting a relationship gradient corresponding to the trusted entity relationship, and constructing a medical record medical map corresponding to the purified medical record data in combination with the relationship gradient, the medical record entity set and the entity information. The main purpose of the invention is to solve the problem of low construction accuracy of medical maps.

Description

Technical Field

[0001] The present invention relates to a method and system for constructing a medical atlas using data attributes, and belongs to the technical field of medical data processing. Background Art

[0002] Medical graphs play a vital role in the modern medical field. They can present complex medical knowledge in an intuitive and structured way, assisting medical personnel in diagnosis, treatment, and medical research. With the rapid growth of medical data and the continuous updating of medical knowledge, higher requirements are placed on the accuracy and efficiency of medical graph construction.

[0003] At present, the construction method of medical graphs is mainly based on traditional data processing technology, which classifies and annotates according to expert experience and preset rules, such as dividing diseases and symptoms into categories and annotating relationships, and then using algorithms to construct graphs from the processed data, with medical entities as nodes and relationships as edges. This method does not make full use of data attributes, which greatly affect the accuracy and reliability of medical graphs. However, traditional methods often simply integrate data without deeply exploring the value of data attributes to the construction of medical graphs. For example, for data from authoritative medical research institutions and ordinary medical institutions, there is no reasonable relationship evaluation based on their data attributes, which makes the constructed medical graph have deviations in reflecting the actual medical situation, thereby reducing the accuracy of the construction of the medical graph. Summary of the invention

[0004] The present invention provides a method and system for constructing a medical graph using data attributes, the main purpose of which is to solve the problem of low accuracy in constructing a medical graph.

[0005] To achieve the above-mentioned purpose, the present invention provides a method for constructing a medical atlas using data attributes, comprising: Acquire the medical record data to be processed, perform quality improvement processing on the medical record text data to obtain target medical record data, analyze the grammatical rules corresponding to the target medical record data, and perform sentence splitting on the target medical record data in combination with the grammatical rules to obtain medical record split data; Performing privacy purification processing on the medical record split data to obtain purified medical record data, analyzing the medical record data attributes corresponding to the purified medical record data, and extracting the medical record entities and their corresponding entity information in the purified medical record data, calculating the information interconnection strength between the entity information, and analyzing the entity association relationship between the medical record entities according to the information interconnection strength; Based on the medical record data attributes, evaluating the relationship confidence of the entity association relationship, based on the relationship credibility, performing relationship screening processing on the entity association relationship to obtain a credible entity relationship, and based on the credible entity relationship, performing entity fusion on the medical record entity to obtain a medical record entity set; Analyze the heterogeneous features corresponding to the medical record entity set, set the relationship gradient corresponding to the trusted entity relationship based on the heterogeneous features, and construct a medical record medical graph corresponding to the purified medical record data by combining the relationship gradient, the medical record entity set and the entity information.

[0006] Optionally, the performing quality improvement processing on the medical record text data to obtain target medical record data includes: Cleaning the medical record text data to obtain cleaned medical record data; Performing denoising processing on the cleaned medical record text data to obtain denoised medical record data; The denoised medical record text data is repaired to obtain target medical record data.

[0007] Optionally, the analyzing the grammatical rules corresponding to the target medical record data includes: Detecting the data identification symbol in the target medical record data, and analyzing the symbol efficiency corresponding to the data identification symbol; Identify grammatical linking words in the target medical record data, and analyze the grammatical logic corresponding to the grammatical linking words; The grammatical rules corresponding to the target medical record data are analyzed in combination with the grammatical logic and the symbol effectiveness.

[0008] Optionally, performing privacy purification on the medical record split data to obtain purified medical record data includes: Identifying medical record metadata in the medical record split data, and extracting key metadata in the medical record metadata; Performing character recognition on the key metadata to obtain metadata characters; Performing word segmentation processing on the metadata characters to obtain character word segmentations, and marking the privacy characters in the character word segmentations according to a preset privacy character list; Based on the privacy characters, the medical record split data is subjected to privacy purification processing to obtain purified medical record data.

[0009] Optionally, analyzing the medical record data attributes corresponding to the purified medical record data includes:

[0010] Extracting the medical record labels corresponding to the purified medical record data, and classifying the purified medical record data based on the medical record labels to obtain classified medical record data;

[0011] The medical record data features corresponding to the classified medical record data are extracted, and the Gini coefficient corresponding to the medical record data features is calculated by the following formula: ; Among them, A represents the Gini coefficient corresponding to the medical record data feature, a represents the feature sequence number of the medical record data feature, and r represents the feature number of the medical record data feature. represents the misjudgment probability corresponding to the ath feature in the medical record data features; Based on the Gini coefficient, feature selection processing is performed on the medical record data features to obtain key data features; Perform feature clustering processing on the key data features to obtain medical record data attributes corresponding to the purified medical record data.

[0012] Optionally, calculating the information interconnection strength between the entity information includes: Quantizing the entity information to obtain an information quantization value, and calculating an entity information mean corresponding to the entity information based on the information quantization value; Combining the information quantization value and the entity information mean, the information interconnection strength between the entity information is calculated by the following formula: ; Among them, B represents the information interconnection strength between entity information, Indicates the quantized value corresponding to the bth information in the information quantization value, represents the mean value of entity information, It represents the quantization value corresponding to the b+1th information in the information quantization value, and b and b+1 both represent the sequence number in the information quantization value.

[0013] Optionally, the evaluating the relationship confidence of the entity association relationship based on the medical record data attributes includes: Constructing an association judgment matrix between the attributes of the medical record data, calculating the distance values ​​between the attributes of the medical record data in the association judgment matrix, and obtaining a distance metric value; The attribute association frequencies corresponding to the medical record data attributes in the association judgment matrix are counted, and the attribute confidence between the medical record data attributes is calculated by combining the attribute association frequencies and the distance measurement value through the following formula: ; Among them, D represents the attribute confidence between the attributes of medical record data, represents the number of attributes of the medical record data attributes, e represents the attribute sequence number of the medical record data attributes, Represents the distance measurement value between the e-th attribute and the e+1-th attribute in the medical record data attributes. represents the attribute association frequency of the e-th attribute in the medical record data attributes, Represents the attribute association frequency of the e+1th attribute in the medical record data attributes; Based on the attribute confidence, the relationship confidence of the entity association relationship is evaluated.

[0014] Optionally, the entity fusion of the medical record entities based on the trusted entity relationship to obtain a medical record entity set includes: Based on the trusted entity relationship, the medical record entities are grouped to obtain grouped medical record entities; Performing vectorization processing on the grouped medical record entities to obtain medical record entity vectors; Performing dimensionality reduction processing on the medical record entity vector to obtain a dimensionality-reduced medical record entity vector; Performing vector fusion processing on the dimension-reduced medical record entity vector to obtain a fused medical record entity vector; The fused medical record entity vector is subjected to vector restoration processing to obtain a medical record entity set.

[0015] Optionally, the setting of the relationship gradient corresponding to the trusted entity relationship based on the heterogeneous feature includes: Calculating a feature gain value corresponding to the heterogeneous feature, and calculating a gain ratio corresponding to the heterogeneous feature according to the feature gain value; Calculating feature correlation coefficients between the heterogeneous features, and assigning feature weights corresponding to the heterogeneous features based on the feature correlation coefficients; Calculating the relationship priority corresponding to the trusted entity relationship by combining the feature weight and the gain ratio; Based on the relationship priority, a relationship gradient corresponding to the trusted entity relationship is set.

[0016] A system for constructing a medical atlas using data attributes, characterized in that the system comprises: A data splitting module is used to obtain the medical record data to be processed, perform quality sublimation processing on the medical record text data to obtain target medical record data, analyze the grammatical rules corresponding to the target medical record data, and perform sentence splitting on the target medical record data in combination with the grammatical rules to obtain medical record split data; An association relationship analysis module is used to perform privacy purification processing on the medical record split data to obtain purified medical record data, analyze the medical record data attributes corresponding to the purified medical record data, and extract the medical record entities and their corresponding entity information in the purified medical record data, calculate the information interconnection strength between the entity information, and analyze the entity association relationship between the medical record entities based on the information interconnection strength; An entity fusion module is used to evaluate the relationship confidence of the entity association relationship based on the medical record data attributes, perform relationship screening processing on the entity association relationship based on the relationship credibility to obtain a credible entity relationship, and perform entity fusion on the medical record entities based on the credible entity relationship to obtain a medical record entity set; A graph construction module is used to analyze the heterogeneous features corresponding to the medical record entity set, set the relationship gradient corresponding to the trusted entity relationship based on the heterogeneous features, and construct a medical record medical graph corresponding to the purified medical record data by combining the relationship gradient, the medical record entity set and the entity information.

[0017] Compared with the problems described in the background technology, the present invention performs quality sublimation processing on the medical record text data, which can improve the data accuracy of the medical record text data, reduce errors and ambiguous information, and make the results of subsequent analysis more accurate. The present invention performs privacy purification processing on the medical record split data, which can effectively ensure that the privacy information of the medical record split data is not leaked during the purification process, so that the purified medical record data can meet the privacy protection requirements while still providing valuable data for legal applications such as medical research and statistical analysis, ensuring the security and compliance of the entire medical data processing process. The present invention evaluates the entity association based on the medical record data attributes. The relationship confidence of the relationship can be used to understand the reliability of the entity association relationship according to the level of the relationship confidence, which lays an important foundation for the subsequent relationship screening and processing of the entity association relationship. The present invention sets the relationship gradient corresponding to the trusted entity relationship based on the heterogeneous characteristics, which can effectively distinguish the importance and closeness of different trusted entity relationships, thereby making the entity association in the medical map more hierarchical and accurate, and improving the representation ability of medical knowledge. The present invention combines the relationship gradient, the medical record entity set and the entity information to construct the medical record medical map corresponding to the purified medical record data, thereby obtaining an accurate visual knowledge map. Therefore, the data attributes proposed in the present invention realize the construction method and system of the medical map, thereby improving the accuracy of the construction of the medical map. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A schematic diagram of a process for constructing a medical graph using data attributes provided in one embodiment of the present invention; Figure 2 A functional module diagram of a system for constructing a medical atlas that implements the data attributes provided in one embodiment of the present invention.

[0019] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0020] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0021] The embodiment of the present application provides a method for constructing a medical graph using data attributes. The execution subject of the method for constructing a medical graph using data attributes includes but is not limited to at least one of the electronic devices such as a server and a terminal that can be configured to execute the method provided by the embodiment of the present application. In other words, the method for constructing a medical graph using data attributes can be executed by software or hardware installed on a terminal device or a server device. The server includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0022] Embodiment 1: Reference Figure 1 FIG. 1 is a flow chart of a method for constructing a medical graph using data attributes according to an embodiment of the present invention. In this embodiment, the method for constructing a medical graph using data attributes includes: S1. Obtain the medical record data to be processed, perform quality improvement processing on the medical record text data to obtain target medical record data, analyze the grammatical rules corresponding to the target medical record data, and perform sentence splitting on the target medical record data in combination with the grammatical rules to obtain medical record split data.

[0023] The present invention performs quality improvement processing on the medical case record text data, which can improve the data accuracy of the medical case record text data, reduce errors and ambiguous information, and make the results of subsequent analysis more accurate. It should be explained that the medical case data refers to a collection of various types of information related to the patient case generated during the medical process, and the target medical record data is the high-quality data corresponding to the medical case record text data.

[0024] In detail, the quality improvement process of the medical record text data to obtain the target medical record data includes: Cleaning the medical record text data to obtain cleaned medical record data; Performing denoising processing on the cleaned medical record text data to obtain denoised medical record data; The denoised medical record text data is repaired to obtain target medical record data.

[0025] It should be explained that the cleaned medical record data is the medical record data obtained after removing redundant information in the medical record text data. The redundant information includes some administrative records that are not related to the diagnosis and treatment of the disease, repeated examination item records, etc., such as "the patient has taken medicine and eaten normally today", which are repeated every day and do not reflect changes in the disease. The denoised medical record data is the data obtained after removing noise interference in the cleaned medical record text data.

[0026] Furthermore, the cleaning process of the medical record text data can be achieved through a box plot method; the denoising process of the cleaned medical record text data can be achieved through a statistical method, such as the 3σ principle method. If the data approximately obeys a normal distribution, the outliers can be judged based on the mean and standard deviation. Data points that fall outside the mean ±3 times the standard deviation can be regarded as outliers. For example, in a patient's blood routine examination, the normal range of white blood cell count usually has a certain statistical distribution. If the white blood cell count of a patient exceeds this normal range too much (judged according to the 3σ principle), it can be marked as an outlier. For the identified outliers, the original data can be further checked for errors, and if they are erroneous data, they can be corrected. If the data is correct, it can be decided whether to retain it based on the specific situation; the repair processing of the denoised medical record text data can be achieved through the mean substitution method. When there are numerical missing values ​​or abnormal values ​​in the denoised medical record text data that need to be repaired, the mean substitution method is a simple and effective method. For a group of data with similar characteristics (such as a certain examination indicator of a group of patients with the same disease), calculate its mean, and then use the mean to replace the missing or abnormal values. For example, in a certain ward, the red blood cell count data of some patients are missing or obviously abnormal due to instrument failure. If the red blood cell count data of other patients are relatively stable and conform to the normal distribution, the mean of these valid data can be calculated, and the mean can be used to fill in the missing or replace the abnormal values.

[0027] By analyzing the grammatical rules corresponding to the target medical record data, the present invention can accurately understand the structure and logic of the target medical record data, which is helpful to efficiently split sentences. In combination with the grammatical rules, the target medical record data is split into sentences, which is convenient for the subsequent extraction and processing of medical record entities and their corresponding entity information in the purified medical record data. It should be explained that the grammatical rules are the organizational structure and internal logic corresponding to the target medical record data, and the medical record split data are the data obtained after the target medical record data is split according to the grammatical rules. Further, in combination with the grammatical rules, the target medical record data is split into sentences to obtain medical record split data. For example, different types of information such as the patient's basic information, disease symptoms, and test results are separated from the sentence respectively. This is like dividing a complex puzzle into several small pieces according to specific shape and color rules, which is convenient for subsequent separate processing and analysis of each type of information.

[0028] In detail, the grammatical rules corresponding to the target medical record data are analyzed, including: Detecting the data identification symbol in the target medical record data, and analyzing the symbol efficiency corresponding to the data identification symbol; Identify grammatical linking words in the target medical record data, and analyze the grammatical logic corresponding to the grammatical linking words; The grammatical rules corresponding to the target medical record data are analyzed in combination with the grammatical logic and the symbol effectiveness.

[0029] It should be explained that the data identification symbols are special symbol elements in the target medical record data, such as punctuation marks such as commas, colons, semicolons, quotation marks, brackets, and marking symbols for specific medical terms (such as special code symbols representing different examination items). The symbol effectiveness is the role and effect of the data identification symbols in information transmission and organization, such as commas are used to separate parallel information, colons are used to explain or introduce key content, quotation marks are used to emphasize special terms or medical advice, brackets are used to supplement additional information, etc.; the grammatical linking words are words or phrases in the target medical record data that are used to connect sentence components or different sentences and reflect logical relationships, such as "because", "so", "although", "but", "if", "when...", "according to...", "through...", etc.; the grammatical logic is the logical relationship at the semantic and syntactic levels corresponding to the grammatical linking words, such as causal relationships, conditional relationships, concession relationships, time sequence relationships, subordinate relationships, etc. These logical relationships help to construct the coherence and rationality of information in the medical record text, so that different medical information can be organized together according to certain rules and accurately convey the relationship between conditions, diagnoses, treatments, etc.

[0030] Furthermore, the detection of data identification symbols in the target medical record data can be achieved through regular expression matching in natural language processing technology. For example, regular expressions can be used to set matching patterns for common punctuation marks such as commas and colons, and these symbols can be quickly located in the text; the analysis of the symbol effectiveness corresponding to the data identification symbols can be achieved by constructing a symbol effectiveness analysis model. For example, a model is constructed to perform semantic analysis on the content separated by commas to determine whether it is a parallel relationship. Through statistics of a large number of medical record texts, the high frequency of use of colons in the context of eliciting diagnostic results is determined, thereby analyzing the effectiveness of its explanation; the recognition of grammatical connecting words in the target medical record data can be achieved through part-of-speech tagging tools in natural language processing. Part-of-speech tagging tools can mark the part of speech of words with a connecting function, and can recognize "because" and "so" " and other grammatical conjunctions. The analysis of the grammatical logic corresponding to the grammatical conjunctions can be achieved through dependency syntactic analysis. Dependency syntactic analysis can reveal the dependency relationship between different components in a sentence, so as to analyze the logical relationship constructed by the grammatical conjunctions. In combination with the grammatical logic and the symbol effectiveness, the grammatical rules corresponding to the target medical record data are analyzed. For example, the target medical record data is firstly subjected to grammatical conjunction recognition and grammatical logic analysis to determine the logical structure of sentences and passages, and then data identification symbol detection and symbol effectiveness analysis are performed to clarify the organization of information. Finally, the two are combined to sort out the grammatical rules in the use of punctuation marks, sentence construction and logical relationship expression of different parts such as symptom description, diagnosis and treatment. For example, in symptom description, parallel symptoms are often separated by commas and are often accompanied by adjectives indicating the degree of symptoms. In diagnostic sentences, grammatical conjunctions "according to... Diagnosed as..." are often used to reflect causal logic, and colons are used to introduce diagnostic results.

[0031] S2. Perform privacy purification on the medical record split data to obtain purified medical record data, analyze the medical record data attributes corresponding to the purified medical record data, and extract the medical record entities and their corresponding entity information in the purified medical record data, calculate the information interconnection strength between the entity information, and analyze the entity association relationship between the medical record entities based on the information interconnection strength.

[0032] The present invention can effectively ensure that the privacy information of the medical record split data is not leaked during the purification process by performing privacy purification processing on the medical record split data, so that the purified medical record data can meet the privacy protection requirements while still providing valuable data for legal applications such as medical research and statistical analysis, thereby ensuring the security and compliance of the entire medical data processing process. It should be explained that the purified medical record data is the data obtained after the privacy data in the medical record split data is removed.

[0033] Specifically, the privacy purification process is performed on the medical record split data to obtain the purified medical record data, including: Identifying medical record metadata in the medical record split data, and extracting key metadata in the medical record metadata; Performing character recognition on the key metadata to obtain metadata characters; Performing word segmentation processing on the metadata characters to obtain character word segmentations, and marking the privacy characters in the character word segmentations according to a preset privacy character list; Based on the privacy characters, the medical record split data is subjected to privacy purification processing to obtain purified medical record data.

[0034] It should be explained that the medical record metadata is a collection of information such as the source of medical records, collection time, collection equipment, patient category, etc. in the medical record split data; the key metadata is the core part of the medical record metadata used to determine the sensitivity, relevance and usage restrictions of medical record data; the metadata characters are the specific content of the key metadata presented in text form; the character segmentation is the individual words divided from the metadata characters according to language rules; the privacy character list is a pre-constructed list covering all vocabulary related to patient privacy information in the medical field; the privacy characters are specific words in the character segmentation that can be identified as being related to patient privacy protection.

[0035] Furthermore, the identification of medical record metadata in the medical record split data can be achieved through data mining technology; the extraction of key metadata in the medical record metadata can be achieved through a machine learning classifier; the character recognition of the key metadata can be achieved through optical character recognition (OCR) technology; the word segmentation processing of the metadata characters can be achieved through word segmentation tools in natural language processing (such as word segmenters based on dictionaries or statistical models); according to a preset list of privacy characters, the privacy characters in the character segmentation can be marked out through a string matching algorithm; based on the privacy characters, the medical record split data can be subjected to privacy purification processing by formulating targeted deletion rules to obtain purified medical record data.

[0036] The present invention can obtain content-related features of the purified medical record data by analyzing the medical record data attributes corresponding to the purified medical record data, so as to improve the accuracy of the subsequent evaluation of the relationship confidence of the entity association relationship, extract the medical record entities and their corresponding entity information in the purified medical record data, and obtain information descriptions of the entities and related entities in the data. It should be explained that the medical record data attributes are data features of the purified medical record data, such as patient basic information, medical information (diagnosis, treatment, examination results, etc.), and these content attributes that reflect medical conditions. The medical record entities are objects with independent meanings in the purified medical record data (such as patients, doctors, hospitals, diseases, etc.), and the entity information is used to describe the various attributes and related content of the medical record entities. Furthermore, the medical record entities in the purified medical record data and their corresponding entity information can be realized through named entity recognition models and information extraction models, such as Bi-LSTM-CRF (bidirectional long short-term memory network-conditional random field) models and convolutional neural network models based on deep learning.

[0037] In detail, the analysis of the medical record data attributes corresponding to the purified medical record data includes: Extracting the medical record labels corresponding to the purified medical record data, and classifying the purified medical record data based on the medical record labels to obtain classified medical record data; The medical record data features corresponding to the classified medical record data are extracted, and the Gini coefficient corresponding to the medical record data features is calculated by the following formula: ; Among them, A represents the Gini coefficient corresponding to the medical record data feature, a represents the feature sequence number of the medical record data feature, and r represents the feature number of the medical record data feature. represents the misjudgment probability corresponding to the ath feature in the medical record data features; Based on the Gini coefficient, feature selection processing is performed on the medical record data features to obtain key data features; Perform feature clustering processing on the key data features to obtain medical record data attributes corresponding to the purified medical record data.

[0038] It should be explained that the medical record label is the mark corresponding to the classified medical record data for identifying different categories (such as disease type, treatment stage, etc.); the classified medical record data is a subset of the purified medical record data divided according to the medical record label (such as disease type, treatment method, etc.); the medical record data feature is the data element corresponding to the classified medical record data that can reflect the characteristics of the medical record (such as patient basic information, disease symptoms, examination results, etc.); the Gini coefficient is an indicator for measuring the degree of imbalance in the distribution of the medical record data features or the uncertainty of classification; the key data feature is the feature when the Gini coefficient corresponding to the medical record data feature is greater than the preset coefficient, which can be set to 0.8 or according to the actual business scenario; the probability of misjudgment corresponding to the feature.

[0039] Furthermore, the extraction of medical record labels corresponding to the purified medical record data can be achieved through a label extraction tool, which is compiled by a scripting language, such as JS scripting language; the classification processing of the purified medical record data can be achieved through a decision tree; the extraction of medical record data features corresponding to the classified medical record data can be achieved through feature engineering methods, such as principal component analysis; the feature clustering processing of the key data features can be achieved through a hierarchical clustering algorithm.

[0040] The present invention calculates the information interconnection strength between the entity information, and can use the information interconnection strength to understand the degree of association between the entity information, thereby laying an important basis for the subsequent analysis of the entity association relationship between the medical record entities. It should be explained that the information interconnection strength represents the degree of association between the entity information.

[0041] Specifically, the calculating the information interconnection strength between the entity information includes: Quantizing the entity information to obtain an information quantization value, and calculating an entity information mean corresponding to the entity information based on the information quantization value; Combining the information quantization value and the entity information mean, the information interconnection strength between the entity information is calculated by the following formula: ; Among them, B represents the information interconnection strength between entity information, Indicates the quantized value corresponding to the bth information in the information quantization value, represents the mean value of entity information, It represents the quantization value corresponding to the b+1th information in the information quantization value, and b and b+1 both represent the sequence number in the information quantization value.

[0042] It should be explained that the information quantization value is the information expression value corresponding to the entity information, and the entity information mean is the average quantization value corresponding to the entity information. Furthermore, the quantization processing of the entity information can be achieved through a coding algorithm, such as a binary coding algorithm; the calculation of the entity information mean corresponding to the entity information can be achieved through the arithmetic mean method.

[0043] The present invention can improve the accuracy of the entity association relationship analysis between the medical record entities based on the information interconnection strength. It should be explained that the entity association relationship is a description of the relationship between the medical record entities. If the information interconnection strength is large, it means that there is a strong association between the medical record entities. If the information interconnection strength is in the middle value, the medical record entities are moderately associated. When the information interconnection strength value is very small (close to 0), it means that the medical record entities are weakly associated or have no association.

[0044] S3. Based on the medical record data attributes, the relationship confidence of the entity association relationship is evaluated (the probability is calculated); based on the relationship credibility, the entity association relationship is subjected to relationship screening processing to obtain a credible entity relationship; based on the credible entity relationship, the medical record entities are subjected to entity fusion to obtain a medical record entity set.

[0045] The present invention evaluates the relationship confidence of the entity association relationship based on the medical record data attributes, and can understand the reliability of the entity association relationship according to the level of the relationship confidence, thereby laying an important foundation for the subsequent relationship screening processing of the entity association relationship. It should be explained that the relationship confidence represents the relationship reliability corresponding to the entity association relationship.

[0046] In detail, the evaluating the relationship confidence of the entity association relationship based on the medical record data attributes includes: Constructing an association judgment matrix between the attributes of the medical record data, calculating the distance values ​​between the attributes of the medical record data in the association judgment matrix, and obtaining a distance metric value; The attribute association frequencies corresponding to the medical record data attributes in the association judgment matrix are counted, and the attribute confidence between the medical record data attributes is calculated by combining the attribute association frequencies and the distance measurement value through the following formula: ; Among them, D represents the attribute confidence between the attributes of medical record data, represents the number of attributes of the medical record data attributes, e represents the attribute sequence number of the medical record data attributes, Represents the distance measurement value between the e-th attribute and the e+1-th attribute in the medical record data attributes. represents the attribute association frequency of the e-th attribute in the medical record data attributes, Represents the attribute association frequency of the e+1th attribute in the medical record data attributes; Based on the attribute confidence, the relationship confidence of the entity association relationship is evaluated.

[0047] It should be explained that the association judgment matrix is ​​a quantitative representation matrix of the association relationship between the medical record data attributes, the attribute association frequency is the occurrence frequency statistic corresponding to the medical record data attributes in the association judgment matrix, and the attribute confidence represents the reliability of the association relationship between the medical record data attributes.

[0048] Furthermore, the construction of the association judgment matrix between the medical record data attributes can be achieved through a matrix function, such as a zero matrix function; the calculation of the distance values ​​between the medical record data attributes in the association judgment matrix can be achieved through a distance measurement algorithm, such as a Manhattan distance algorithm; the attribute association frequency corresponding to the medical record data attributes in the association judgment matrix can be obtained by counting the number of simultaneous appearances; the relationship confidence can be obtained by summing the attribute confidences corresponding to the entity association relationship.

[0049] The present invention can obtain a trusted entity relationship by performing relationship screening processing on the entity association relationship based on the relationship credibility, and can enhance the data breadth of the purified medical record data by performing entity fusion on the medical record entities based on the trusted entity relationship, thereby laying a foundation for subsequently improving the accuracy of medical record medical map construction. It should be explained that the trusted entity relationship is the relationship with the highest relationship confidence among the entity association relationships, and the medical record entity set is the entity obtained after the medical record entities are fused according to the trusted entity relationship. Furthermore, the relationship screening processing of the entity association relationship can be implemented by a screening function, and the screening function is compiled by a programming language.

[0050] Specifically, based on the trusted entity relationship, the medical record entities are entity fused to obtain a medical record entity set, including: Based on the trusted entity relationship, the medical record entities are grouped to obtain grouped medical record entities; Performing vectorization processing on the grouped medical record entities to obtain medical record entity vectors; Performing dimensionality reduction processing on the medical record entity vector to obtain a dimensionality-reduced medical record entity vector; Performing vector fusion processing on the dimension-reduced medical record entity vector to obtain a fused medical record entity vector; The fused medical record entity vector is subjected to vector restoration processing to obtain a medical record entity set.

[0051] It should be explained that the grouped medical record entities are medical record entities divided into different groups according to the trusted entity relationship, the medical record entity vector is the expression vector corresponding to the grouped medical record entity, the reduced dimensionality medical record entity vector is the feature vector obtained after removing the redundant vectors in the medical record entity vector, and the fused medical record entity vector is the new vector obtained by merging the reduced dimensionality medical record entity vectors together.

[0052] Furthermore, the grouping processing of the medical record entities can be achieved through manual operation; the vectorization processing of the grouped medical record entities can be achieved through a one-hot encoding algorithm; the dimensionality reduction processing of the medical record entity vectors can be achieved through a linear dimensionality reduction algorithm; the vector fusion processing of the reduced dimensionality medical record entity vectors can be achieved through vector splicing; the row vector restoration processing of the fused medical record entity vectors can be achieved through reverse encoding rules. If one-hot encoding or other encoding methods are used in the vectorization process, then reverse operations can be performed during restoration based on the encoding mapping relationship.

[0053] S4. Analyze the heterogeneous features corresponding to the medical record entity set, set the relationship gradient corresponding to the trusted entity relationship based on the heterogeneous features, and construct a medical record medical graph corresponding to the purified medical record data by combining the relationship gradient, the medical record entity set and the entity information.

[0054] The present invention sets the relationship gradient corresponding to the trusted entity relationship based on the heterogeneous features, which can effectively distinguish the importance and closeness of different trusted entity relationships, thereby making the entity associations in the medical atlas more hierarchical and precise, and improving the ability to represent medical knowledge. It should be explained that the heterogeneous features are the multi-source attributes of the medical record entity set, and the relationship gradient represents the priority level of the trusted entity relationship. Furthermore, the analysis of the heterogeneous features corresponding to the medical record entity set can be achieved through a multi-layer perceptron.

[0055] In detail, the setting of the relationship gradient corresponding to the trusted entity relationship based on the heterogeneous features includes: Calculating a feature gain value corresponding to the heterogeneous feature, and calculating a gain ratio corresponding to the heterogeneous feature according to the feature gain value; Calculating feature correlation coefficients between the heterogeneous features, and assigning feature weights corresponding to the heterogeneous features based on the feature correlation coefficients; Calculating the relationship priority corresponding to the trusted entity relationship by combining the feature weight and the gain ratio; Based on the relationship priority, a relationship gradient corresponding to the trusted entity relationship is set.

[0056] It should be explained that the feature gain value is the degree to which the uncertainty of the classification result is reduced by the heterogeneous features; the gain ratio is a measure of the correction of the information gain corresponding to the heterogeneous features to balance the influence of the feature values; the feature correlation coefficient represents the degree of linear correlation between the heterogeneous features; the feature weight represents the quantification of the influence of the heterogeneous features on the target variable; the relationship priority represents the importance and processing order of the trusted entity relationship.

[0057] Furthermore, the calculation of the feature gain value corresponding to the heterogeneous feature can be implemented by xx; the ratio of each gain value of the feature gain value to the total gain value is calculated to obtain the gain ratio corresponding to the heterogeneous feature; the calculation of the feature correlation coefficient between the heterogeneous features can be implemented by xx; the feature correlation coefficient is used as the feature weight corresponding to the heterogeneous feature; the feature weight and the corresponding gain ratio are multiplied and summed to obtain the relationship priority corresponding to the trusted entity relationship; based on the relationship priority, the relationship gradient corresponding to the trusted entity relationship is set. For example, in a medical diagnosis system, for the "disease-typical symptoms" relationship, such as "pneumonia-cough, fever, sputum", since these typical symptoms are extremely critical to the diagnosis of pneumonia, their relationship priority can be set to high, and the corresponding relationship gradient value is set to 0.8, while for a relatively minor relationship such as "pneumonia-occasional fatigue", the priority is lower and the relationship gradient value is set to 0.3.

[0058] The present invention constructs a medical record graph corresponding to the purified medical record data by combining the relationship gradient, the medical record entity set and the entity information, thereby obtaining an accurate visualized knowledge graph. It should be explained that the medical record graph is a visualized graph corresponding to the purified medical record data. Furthermore, according to the structural features corresponding to the medical record entity set, the graph structure is determined, and the medical record entity set and the entity information are filled in the graph structure to obtain a filled graph. According to the relationship gradient, the structure in the filled graph is adjusted. The higher the relationship gradient, the larger the font of the corresponding graph content and the color can be adjusted to a bright color, thereby obtaining a medical record graph corresponding to the purified medical record data.

[0059] Compared with the problems described in the background technology, the present invention performs quality sublimation processing on the medical record text data, which can improve the data accuracy of the medical record text data, reduce errors and ambiguous information, and make the results of subsequent analysis more accurate. The present invention performs privacy purification processing on the medical record split data, which can effectively ensure that the privacy information of the medical record split data is not leaked during the purification process, so that the purified medical record data can meet the privacy protection requirements while still providing valuable data for legal applications such as medical research and statistical analysis, ensuring the security and compliance of the entire medical data processing process. The present invention evaluates the entity association based on the medical record data attributes. The relationship confidence of the relationship can be used to understand the reliability of the entity association relationship according to the level of the relationship confidence, which lays an important foundation for the subsequent relationship screening and processing of the entity association relationship. The present invention sets the relationship gradient corresponding to the trusted entity relationship based on the heterogeneous characteristics, which can effectively distinguish the importance and closeness of different trusted entity relationships, thereby making the entity association in the medical map more hierarchical and accurate, and improving the representation ability of medical knowledge. The present invention combines the relationship gradient, the medical record entity set and the entity information to construct the medical record medical map corresponding to the purified medical record data, thereby obtaining an accurate visual knowledge map. Therefore, the data attribute proposed in the present invention realizes the method for constructing a medical map, thereby improving the accuracy of medical map construction.

[0060] Embodiment 2: like Figure 2 As shown, it is a functional module diagram of a system for constructing a medical graph by implementing data attributes provided by one embodiment of the present invention.

[0061] The system 100 for constructing a medical graph using data attributes of the present invention can be installed in an electronic device. According to the functions implemented, the system 100 for constructing a medical graph using data attributes can include a data splitting module 101, an association relationship analysis module 102, an entity fusion module 103, and a graph construction module 104. The module of the present invention can also be referred to as a unit, which refers to a series of computer program segments that can be executed by a processor of an electronic device and can complete fixed functions, and is stored in the memory of the electronic device.

[0062] In this embodiment, the functions of each module / unit are as follows: The data splitting module 101 is used to obtain the medical record data to be processed, perform quality sublimation processing on the medical record text data to obtain target medical record data, analyze the grammatical rules corresponding to the target medical record data, and perform sentence splitting on the target medical record data in combination with the grammatical rules to obtain medical record split data; The association relationship analysis module 102 is used to perform privacy purification processing on the medical record split data to obtain purified medical record data, analyze the medical record data attributes corresponding to the purified medical record data, and extract the medical record entities and their corresponding entity information in the purified medical record data, calculate the information interconnection strength between the entity information, and analyze the entity association relationship between the medical record entities according to the information interconnection strength; The entity fusion module 103 is used to evaluate the relationship confidence of the entity association relationship based on the medical record data attributes, perform relationship screening processing on the entity association relationship based on the relationship credibility to obtain a credible entity relationship, and perform entity fusion on the medical record entities based on the credible entity relationship to obtain a medical record entity set; The graph construction module 104 is used to analyze the heterogeneous features corresponding to the medical record entity set, set the relationship gradient corresponding to the trusted entity relationship based on the heterogeneous features, and construct a medical record medical graph corresponding to the purified medical record data by combining the relationship gradient, the medical record entity set and the entity information.

[0063] In detail, each module described in the system 100 for constructing a medical atlas using data attributes described in the embodiment of the present application is used in the same manner as described above. Figure 1 The data attribute described in the text realizes the same technical means as the method for constructing a medical graph and can produce the same technical effects, so I will not go into details here.

[0064] It is obvious to those skilled in the art that the present invention is not limited to the details of the above exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for constructing a medical graph using data attributes, characterized in that: The method comprises: Acquire the medical record data to be processed, perform quality improvement processing on the medical record text data to obtain target medical record data, analyze the grammatical rules corresponding to the target medical record data, and perform sentence splitting on the target medical record data in combination with the grammatical rules to obtain medical record split data; Performing privacy purification processing on the medical record split data to obtain purified medical record data, analyzing the medical record data attributes corresponding to the purified medical record data, and extracting the medical record entities and their corresponding entity information in the purified medical record data, calculating the information interconnection strength between the entity information, and analyzing the entity association relationship between the medical record entities according to the information interconnection strength; Based on the medical record data attributes, evaluating the relationship confidence of the entity association relationship, based on the relationship credibility, performing relationship screening processing on the entity association relationship to obtain a credible entity relationship, and based on the credible entity relationship, performing entity fusion on the medical record entity to obtain a medical record entity set; Analyze the heterogeneous features corresponding to the medical record entity set, set the relationship gradient corresponding to the trusted entity relationship based on the heterogeneous features, and construct a medical record medical graph corresponding to the purified medical record data by combining the relationship gradient, the medical record entity set and the entity information.

2. The method for constructing a medical atlas using data attributes as claimed in claim 1, characterized in that: The step of performing quality improvement processing on the medical record text data to obtain target medical record data includes: Cleaning the medical record text data to obtain cleaned medical record data; Performing denoising processing on the cleaned medical record text data to obtain denoised medical record data; The denoised medical record text data is repaired to obtain target medical record data.

3. The method for constructing a medical atlas using data attributes as claimed in claim 1, characterized in that: The analyzing the grammatical rules corresponding to the target medical record data includes: Detecting the data identification symbol in the target medical record data, and analyzing the symbol efficiency corresponding to the data identification symbol; Identify grammatical linking words in the target medical record data, and analyze the grammatical logic corresponding to the grammatical linking words; The grammatical rules corresponding to the target medical record data are analyzed in combination with the grammatical logic and the symbol effectiveness.

4. The method for constructing a medical atlas using data attributes as claimed in claim 1, characterized in that: The privacy purification process is performed on the medical record split data to obtain purified medical record data, including: Identifying medical record metadata in the medical record split data, and extracting key metadata in the medical record metadata; Performing character recognition on the key metadata to obtain metadata characters; Performing word segmentation processing on the metadata characters to obtain character word segmentations, and marking the privacy characters in the character word segmentations according to a preset privacy character list; Based on the privacy characters, the medical record split data is subjected to privacy purification processing to obtain purified medical record data.

5. The method for constructing a medical atlas using data attributes as claimed in claim 1, characterized in that: The analyzing the medical record data attributes corresponding to the purified medical record data includes: Extracting the medical record labels corresponding to the purified medical record data, and classifying the purified medical record data based on the medical record labels to obtain classified medical record data; The medical record data features corresponding to the classified medical record data are extracted, and the Gini coefficient corresponding to the medical record data features is calculated by the following formula: ； Among them, A represents the Gini coefficient corresponding to the medical record data feature, a represents the feature sequence number of the medical record data feature, and r represents the feature number of the medical record data feature. represents the misjudgment probability corresponding to the ath feature in the medical record data features; Based on the Gini coefficient, feature selection processing is performed on the medical record data features to obtain key data features; Perform feature clustering processing on the key data features to obtain medical record data attributes corresponding to the purified medical record data.

6. The method for constructing a medical atlas using data attributes as claimed in claim 1, characterized in that: The calculating the information interconnection strength between the entity information includes: Quantizing the entity information to obtain an information quantization value, and calculating an entity information mean corresponding to the entity information based on the information quantization value; Combining the information quantization value and the entity information mean, the information interconnection strength between the entity information is calculated by the following formula: ； Among them, B represents the information interconnection strength between entity information, Indicates the quantized value corresponding to the bth information in the information quantization value, represents the mean value of entity information, It represents the quantization value corresponding to the b+1th information in the information quantization value, and b and b+1 both represent the sequence number in the information quantization value.

7. The method for constructing a medical atlas using data attributes as claimed in claim 1, characterized in that: The step of evaluating the relationship confidence of the entity association relationship based on the medical record data attributes includes: Constructing an association judgment matrix between the attributes of the medical record data, calculating the distance values ​​between the attributes of the medical record data in the association judgment matrix, and obtaining a distance metric value; The attribute association frequencies corresponding to the medical record data attributes in the association judgment matrix are counted, and the attribute confidence between the medical record data attributes is calculated by combining the attribute association frequencies and the distance measurement value through the following formula: ； Among them, D represents the attribute confidence between the attributes of medical record data, represents the number of attributes of the medical record data attributes, e represents the attribute sequence number of the medical record data attributes, Represents the distance measurement value between the e-th attribute and the e+1-th attribute in the medical record data attributes. represents the attribute association frequency of the e-th attribute in the medical record data attributes, Represents the attribute association frequency of the e+1th attribute in the medical record data attributes; Based on the attribute confidence, the relationship confidence of the entity association relationship is evaluated.

8. The method for constructing a medical atlas using data attributes as claimed in claim 1, characterized in that: The step of performing entity fusion on the medical record entities based on the trusted entity relationship to obtain a medical record entity set includes: Based on the trusted entity relationship, the medical record entities are grouped to obtain grouped medical record entities; Performing vectorization processing on the grouped medical record entities to obtain medical record entity vectors; Performing dimensionality reduction processing on the medical record entity vector to obtain a dimensionality-reduced medical record entity vector; Performing vector fusion processing on the dimension-reduced medical record entity vector to obtain a fused medical record entity vector; The fused medical record entity vector is subjected to vector restoration processing to obtain a medical record entity set.

9. The method for constructing a medical atlas using data attributes as claimed in claim 1, characterized in that: The setting of the relationship gradient corresponding to the trusted entity relationship based on the heterogeneous features includes: Calculating a feature gain value corresponding to the heterogeneous feature, and calculating a gain ratio corresponding to the heterogeneous feature according to the feature gain value; Calculating feature correlation coefficients between the heterogeneous features, and assigning feature weights corresponding to the heterogeneous features based on the feature correlation coefficients; Calculating the relationship priority corresponding to the trusted entity relationship by combining the feature weight and the gain ratio; Based on the relationship priority, a relationship gradient corresponding to the trusted entity relationship is set.

10. The data attribute realizes the construction system of medical atlas, characterized by: The system comprises: A data splitting module is used to obtain the medical record data to be processed, perform quality sublimation processing on the medical record text data to obtain target medical record data, analyze the grammatical rules corresponding to the target medical record data, and perform sentence splitting on the target medical record data in combination with the grammatical rules to obtain medical record split data; An association relationship analysis module is used to perform privacy purification processing on the medical record split data to obtain purified medical record data, analyze the medical record data attributes corresponding to the purified medical record data, and extract the medical record entities and their corresponding entity information in the purified medical record data, calculate the information interconnection strength between the entity information, and analyze the entity association relationship between the medical record entities based on the information interconnection strength; An entity fusion module is used to evaluate the relationship confidence of the entity association relationship based on the medical record data attributes, perform relationship screening processing on the entity association relationship based on the relationship credibility to obtain a credible entity relationship, and perform entity fusion on the medical record entities based on the credible entity relationship to obtain a medical record entity set; A graph construction module is used to analyze the heterogeneous features corresponding to the medical record entity set, set the relationship gradient corresponding to the trusted entity relationship based on the heterogeneous features, and construct a medical record medical graph corresponding to the purified medical record data by combining the relationship gradient, the medical record entity set and the entity information.

Images