A data query method for a medical experiment system

Abstract

The application relates to the technical field of data query, and particularly discloses a data query method for a medical experiment system, which comprises the following steps: step S1: establishing a medical experiment report library, binding structured stored experiment information for each piece of experiment data, including experiment object attributes, a detection index set, experiment condition parameters and experiment conclusion keywords; step S2: constructing a knowledge graph based on the experiment information, taking the experiment object attributes as root nodes, the detection index set as secondary nodes, the experiment condition parameters as tertiary nodes, and generating semantic relationship edges between nodes according to the experiment conclusion keywords; and step S3: analyzing an abstract query sentence input by a user, mapping the abstract query sentence to corresponding nodes and semantic edges of the knowledge graph, and returning matched experiment data through path traversal; the method solves the problems of semantic understanding difficulty and insufficient correlation analysis in traditional medical data query, and significantly improves the query efficiency and accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data query, in particular to a data query method for a medical experiment system. BACKGROUND

[0002] Traditional data storage and management mode seriously restricts the full mining of its value. Specifically, in the past, numerous medical experiment reports are only piled together in disorder, lacking necessary association and systematic integration means among each other. This makes the organization of data extremely chaotic, and when researchers try to extract a specific theme or explore the association between data, they have to face the tedious manual screening work of massive and chaotic data, which is not only time-consuming and laborious, but also extremely inefficient.

[0003] At the same time, in actual application scenarios, the query requirements put forward by users are often based on relatively abstract condition expressions, such as "experimental data related to a certain disease". However, due to the lack of intelligent analysis capability, the existing system is difficult to accurately understand such vague requirements, and it is also impossible to effectively locate and filter out the experimental data meeting the requirements from massive data quickly. SUMMARY

[0004] The purpose of the present application is to provide a data query method for a medical experiment system, which solves the following technical problems.

[0005] The purpose of the present application can be achieved by the following technical solutions:

[0006] A data query method for a medical experiment system, comprising the following steps:

[0007] Step S1: establishing a medical experiment report library, the medical experiment report library comprising all medical experiment reports; binding structured experimental information for each medical experiment report in the medical experiment report library, the experimental information comprising experimental object attributes, a detection index set, experimental condition parameters and experimental conclusion keywords;

[0008] Step S2: based on the experimental information of each medical experiment report, taking the experimental object attributes as the root node, the detection index set as the secondary node, the experimental condition parameters as the tertiary node, and generating semantic relationship edges between nodes based on the experimental conclusion keywords, to obtain a knowledge graph;

[0009] Step S3: obtaining an abstract query sentence input by a user, extracting an entity object, a target index and a relationship description word in the abstract query sentence; mapping the entity object to the root node of the knowledge graph, mapping the target index to the secondary node of the knowledge graph, and matching the relationship description word with the semantic relationship edges; traversing the paths in the knowledge graph that meet the constraints, and returning the end of the path to obtain the medical experiment report associated with the end.

[0010] As a further scheme of the present application: the structured stored experimental information specifically includes:

[0011] The experimental object attributes include several biological characteristics, the biological characteristics include species information, gender, age stage and health status indicators; the experimental condition parameters include temperature, humidity and light cycle; the experimental conclusion keywords include significant, trend difference, positive correlation and negative correlation.

[0012] As a further scheme of the present application: when the detection index set of any medical experiment report exists more than one detection index, the storage of the detection index set of the medical experiment report is dynamically grouped, specifically including:

[0013] A plurality of detection index groups are set, the detection index groups include a molecular detection group, an image feature group and a physiological parameter group; if the detection index in the detection index set of the medical experiment report is a molecular biology index, the molecular biology index is classified into the molecular detection group, the molecular biology index includes genes, proteins and metabolites; if the detection index in the detection index set of the medical experiment report is an imaging index, the imaging index is classified into the image feature group, the imaging index includes CT parameters and MRI texture parameters; if the detection index in the detection index set of the medical experiment report is a physiological signal index, the physiological signal index is classified into the physiological parameter group, the physiological signal index includes heart rate and blood pressure fluctuation value.

[0014] As a further scheme of the present application: the obtaining process of the experimental information includes:

[0015] The related text of the medical experiment to which the medical experiment report belongs is acquired, the related text includes the experimental report text, analysis summary and conclusion derivation of the medical experiment; the related text is split into a plurality of vocabulary units based on natural language processing technology, and each vocabulary unit is subjected to named entity recognition to obtain an entity recognition result, and the experimental information is extracted based on the entity recognition result.

[0016] As a further scheme of the present application: the node includes a root node, a secondary node and a tertiary node.

[0017] As a further scheme of the present application: the generation process of the semantic relationship edge includes:

[0018] A plurality of relationship rules are preset to generate a rule library, the relationship rules including object-index relationship, index-condition relationship and conclusion derivation relationship; the experimental conclusion keywords are substituted into the rule library, the relationship rules corresponding to the experimental conclusion keywords are matched to establish a relationship extraction model, and the semantic relationship of the experimental conclusion keywords is obtained through the relationship extraction model; the direction of the semantic relationship is defined, the strength of the semantic relationship is labeled, and the semantic relationship edges are created between nodes according to the direction and the strength.

[0019] As a further scheme of the present application, the process of mapping the target index to the secondary nodes of the knowledge graph comprises:

[0020] If the target index is a composite index, the target index is split into a plurality of single detection indexes, and each detection index is sequentially mapped to a corresponding secondary node.

[0021] As a further scheme of the present application, the process of determining the path satisfying the constraint comprises:

[0022] The confidence degree of each path is obtained, a confidence degree threshold is set, the paths whose confidence degrees exceed the confidence degree threshold are screened out, and are recorded as the paths satisfying the constraint.

[0023] The present application has the following beneficial effects:

[0024] The present application automatically converts the fuzzy query input by the user into a graph path search through structured experimental information binding and multi-level knowledge graph construction, solves the low accuracy problem caused by the dependence of the traditional technology on keyword matching, automatically classifies the cross-modal indexes such as molecular detection, image features and physiological parameters according to types, breaks the multi-source data barrier in traditional medical experiments, realizes the horizontal comparison of different experimental data through unit conversion and feature extraction, solves the problem of cross-research data incompatibility, converts the text conclusion into a quantitative relationship edge with a direction based on the rule library and the relationship extraction model, automatically expands the graph nodes and edges with new data, ensures the continuous evolution of the system without reconstruction, and labels a warning for the experimental data with contradictory conclusions to reduce the risk of user misjudgment. BRIEF DESCRIPTION OF DRAWINGS

[0025] The present application will be further described below with reference to the drawings.

[0026] Figure 1 is a flowchart of a data query method for a medical experiment system. DETAILED DESCRIPTION

[0027] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described, obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the protection scope of the present application.

[0028] Please refer to Figure 1 The present application is a data query method for a medical experiment system, comprising the following steps:

[0029] The medical experiment system in the present application has a report query function, deeply covers the whole life cycle management requirements of medical test report forms, and builds a multi-dimensional and refined query and processing system; on the query condition configuration, custom report time interval is supported, for example, accurate locking of time period, adaptation to laboratory review according to detection period, monthly statistics and other scenarios; at the same time, rich sample and personnel information retrieval dimensions are provided, input of "name or name initial letter" and "bar code" can quickly associate sample report forms, and realize sample whole process traceability; screening of "hospital", "department", "agent", "district manager" and "sales" can meet the multi-role business requirements of hospital docking and sales data analysis, and can open up the report form screening channel across organizations and processes; in addition, setting of "whether external reception", "report status (success)", "report form type (all report forms)" and other filtering conditions can clearly distinguish internal or external submission, reports in different completion states, help the laboratory to preferentially control abnormal state reports, and strengthen quality supervision;

[0030] The query result is presented in a structured table, and the core fields are covered comprehensively; the bar code and pathological number of the sample identification dimension are associated with the whole process information of the sample from collection to detection; the submission unit and primary or secondary agent of the business flow dimension clearly restore the report business path; the name, positive or negative, gender and age of the patient information dimension are subdivided into years, months and days, which are suitable for medical statistics and clinical analysis scenarios, support field sorting function, and can quickly sort out data rules, such as arranging different age group detection results distribution in ascending order of age; when there is no matching result, the explicit prompt of "no data" accurately feedbacks the query state, avoiding user misunderstanding;

[0031] At the result operation level, the basic functions are diverse; "preview" and "print" can directly view the electronic version or output the paper version of the report, meeting the clinical delivery and archiving needs; "merge" and "download" support batch processing of multiple reports, facilitating data backup and offline deep analysis; the precise export capability is flexible and practical, and "export results", "selected export" and "all export" can control the data output range as needed, such as exporting only positive sample reports for scientific research statistics, or connecting to third-party data platforms to adapt to multiple business scenarios. In addition, this function can also be coordinated with multiple tab pages such as "liver cancer detection result statistics" to realize data linkage of different business modules, build an integrated and intelligent report management ecosystem for the laboratory, and comprehensively improve the query efficiency and application value of medical experiment data.

[0032] Step S1: establishing a medical experiment report library, the medical experiment report library including all medical experiment reports; binding structured stored experiment information for each medical experiment report in the medical experiment report library, the experiment information including experiment object attributes, detection index set, experiment condition parameters and experiment conclusion keywords;

[0033] Specifically, a medical experiment report library is established, which includes all medical experiment reports generated in all scenarios from basic medical research, clinical verification to multi-center cooperation, covering different levels of data such as cell experiments, animal model experiments and human clinical trials; in the medical experiment report library, structured stored experiment information is bound for each medical experiment report to build a standardized data correlation system;

[0034] Specifically, the structured storage process includes:

[0035] Experiment object attributes: stored in the format {biological category: enumeration value, disease code: string, sample size: integer};

[0036] Detection index set: stored in the format {index name: string, measurement unit: string, numerical range: float array};

[0037] The experiment condition parameters are stored as a key-value pair dictionary, with the key name being the standardized label of the intervention type, such as {dose: 10 mg / kg, temperature: 37℃};

[0038] As a preferred embodiment of the present application, the structured stored experiment information specifically includes:

[0039] The experiment object attributes include several biological characteristics, including species information, gender, age stage and health status indicators; the experiment condition parameters include temperature, humidity and light cycle; the experiment conclusion keywords include significant, trend difference, positive correlation and negative correlation;

[0040] As a preferred embodiment of the present application, when the detection index set of any medical experiment report currently exists more than one detection index, the storage of the detection index set of the medical experiment report is dynamically grouped, specifically including:

[0041] A plurality of detection index groups are set, including a molecular detection group, an image feature group and a physiological parameter group; if the detection index in the detection index set of the medical experiment report is a molecular biology index, the molecular biology index is classified into the molecular detection group, and the molecular biology index includes genes, proteins and metabolites; if the detection index in the detection index set of the medical experiment report is an imaging index, the imaging index is classified into the image feature group, and the imaging index includes CT parameters and MRI texture parameters; if the detection index in the detection index set of the medical experiment report is a physiological signal index, the physiological signal index is classified into the physiological parameter group, and the physiological signal index includes heart rate and blood pressure fluctuation value;

[0042] Normalization processing is performed on the grouped detection indexes, specifically including:

[0043] The same type of indexes with different units are converted into standard units, the gray scale histogram statistical value of the image feature group index is extracted as a comparable parameter, and the time sequence variance of the physiological parameter group index is calculated as a dynamic characteristic value;

[0044] As a preferred embodiment of the present application, the obtaining process of the experiment information includes:

[0045] The related text of the medical experiment to which the medical experiment report belongs is obtained, the related text including the experiment report text, analysis summary and conclusion derivation of the medical experiment; the related text is split into a plurality of lexical units based on natural language processing technology, and each lexical unit is subjected to named entity recognition to obtain an entity recognition result, and the experiment information is extracted based on the entity recognition result;

[0046] Step S2: based on the experiment information of each medical experiment report, taking the experiment object attribute as the root node, taking the detection index set as the secondary node, taking the experiment condition parameter as the tertiary node, and generating the semantic relationship edge between nodes based on the experiment conclusion keyword, a knowledge graph is obtained;

[0047] Specifically, based on the experiment information of each medical experiment report, a multi-level knowledge graph architecture is constructed, taking the experiment object attribute as the root node, refining the root node branches according to the medical classification system, covering the disease subtypes of human subjects, the modeling characteristics of animal models, and the gene editing background of cell samples, forming a precise classification tree of experiment objects;

[0048] The detection index set is a secondary node, and based on the medical index standardization coding, the multi-modal indexes such as molecular detection, image detection and pathological detection are associated and indexed across experimental methods, and the horizontal comparison of indexes of different experimental data is supported.

[0049] The experimental condition parameters are the third nodes, the experimental variable dimensions are disassembled, the environmental conditions are refined into the real-time temperature and humidity curves of the constant temperature and humidity box, the gas concentration fluctuation of the CO2 incubator, the reagent condition is associated with the reagent production batch traceability, the pH value and the osmotic pressure parameters in the preparation process, and the instrument condition is bound with the calibration log of the equipment, the voltage and the acquisition rate setting during detection, so that the whole chain of the experimental conditions is traced back.

[0050] Based on the node construction, the semantic relationship edges between nodes are generated based on the experimental conclusion keywords: the cause-effect logic, the association relationship and the condition constraint in the experimental conclusion are analyzed through a medical natural language processing model; meanwhile, a medical domain ontology library is introduced to verify the rationality of the semantic relationship, and standardized terminology mapping is supplemented, so that a medical experiment knowledge graph with clear node hierarchy and accurate semantic relationship is finally formed, which provides a structured knowledge network support for subsequent data query, and potential rules across experiments can be quickly mined based on the node association of the knowledge graph.

[0051] As a preferred embodiment of the present application, the nodes include root nodes, secondary nodes and tertiary nodes.

[0052] As a preferred embodiment of the present application, the knowledge graph is a multi-level knowledge graph, which specifically includes:

[0053] The root nodes are layered according to the biological category labels and disease classification codes in the experimental object attributes, the biological category labels include humans, animals and cell lines, the secondary nodes are divided into sub-trees according to the measurement types of the detection index set, the measurement types include molecules, images and physiology, and the tertiary nodes are generated into branches according to the intervention types of the experimental condition parameters, the experimental condition parameters include drugs, physics and genetic operations.

[0054] As a preferred embodiment of the present application, the generation process of the semantic relationship edges includes:

[0055] A plurality of relationship rules are preset to generate a rule library, the rule library supports extension and can be dynamically supplemented in combination with domain expert knowledge, the relationship rules include object-index relationship, index-condition relationship and conclusion deduction relationship, the experimental conclusion keywords are substituted into the rule library, the relationship rules corresponding to the experimental conclusion keywords are matched to establish a relationship extraction model, the semantic relationship of the experimental conclusion keywords is obtained through the relationship extraction model, the direction of the semantic relationship is defined, the strength of the semantic relationship is labeled, and the semantic relationship edges are created between nodes according to the direction and the strength.

[0056] The object-index relationship, such as: "show" (lung cancer patients show abnormal IL-6 levels), "associated with" (a gene locus is associated with a tumor marker);

[0057] The index-condition relationship, such as: "affected by" (IL-6 test results are affected by sample storage temperature), "based on detection" (IL-6 levels are detected by ELISA experiments);

[0058] The conclusion derivation relationship, such as: "verified" (experimental data verify that IL-6 is related to lung cancer staging), "supported" (animal experiment results support IL-6 as a lung cancer marker);

[0059] The process of generating the semantic relationship edge also includes:

[0060] When the experimental conclusion keyword contains up-regulation, increase type vocabulary, generate a positive influence edge containing an upward arrow symbol; when the experimental conclusion keyword contains down-regulation, inhibition type vocabulary, generate a negative influence edge containing a downward arrow symbol; when the experimental conclusion contains a correlation coefficient, generate a statistical association edge with a numerical label;

[0061] The process of defining the direction of the semantic relationship and labeling the strength of the semantic relationship includes:

[0062] Define the direction of the relationship edge: distinguish between one-way or two-way relationships, and identify them with directed edges or undirected edges in the knowledge graph; such as "experimental conditions → influence → detection indicators" is one-way; "gene A ← interacts → gene B" is two-way;

[0063] Label the relationship strength: combine experimental evidence to assign weights to relationship edges to assist in relevance sorting during subsequent queries;

[0064] Specifically, preset medical field semantic rules, such as experimental object attributes and detection indicator sets, the common relationship between the two is: "show" (such as lung cancer patients showing CT image features) "carry" (such as lung cancer patients carrying EGFR gene mutations);

[0065] Train the model with medical text corpus, let the model learn the implicit semantics between keywords, supplement the relationships not covered by the rules; such as identifying the association between lung cancer patient survival period and EGFR gene mutation, improving the accuracy of relationship edges; according to the identified semantic relationship, establish edges between the corresponding nodes in the knowledge graph, label the relationship type, and connect discrete nodes with semantic associations to form a knowledge network;

[0066] It is worth noting that when generating semantic relationship edges between nodes based on experimental conclusions, a medical domain ontology library is introduced to assist semantic alignment, and pre-defined concept association rules in the ontology library are used to correct the rationality of the semantic relationship edges. After generating the semantic relationship edges, the relationship edges are verified and optimized through manual annotation of a sample set, the confidence of the relationship edges is calculated, and the incorrect relationship edges with a confidence lower than a preset threshold are filtered out.

[0067] In addition, the knowledge graph also includes a real-time updating mechanism:

[0068] When a new medical experiment report is added, the structured experiment information thereof is automatically extracted, and the graph is updated according to the following rules:

[0069] 1. If the root node does not exist, a new disease classification branch is created and associated data is mounted; 2. If the detection index is a new type, a sub-tree is expanded under the existing secondary node; 3. If the semantic relationship edges conflict, the high-frequency conclusion edges are retained and the suspicious edges are marked;

[0070] Step S3: obtaining an abstract query statement input by a user, extracting an entity object, a target index and a relationship description word in the abstract query statement, mapping the entity object to a root node of the knowledge graph, mapping the target index to a secondary node of the knowledge graph, matching the relationship description word with a semantic relationship edge, traversing a path in the knowledge graph that meets a constraint, and returning an end of the path to obtain a medical experiment report associated with the end;

[0071] Specifically, the extraction of the entity object, the target index and the relationship description word in the abstract query statement is based on natural language, professional search formula and other forms of input, and the entity object, the target index and the relationship description word are extracted by using a pre-trained medical dialogue model to analyze the semantics of the statement through a medical natural language understanding method;

[0072] The matched graph path is displayed in the form of a visual subgraph, wherein nodes and edges corresponding to the entity object, the target index and the relationship description word in the user query are highlighted;

[0073] As a preferred embodiment of the present application, the process of mapping the entity object to the root node of the knowledge graph includes:

[0074] If there is a synonymous expression of the entity object, the entity object is normalized and mapped through a pre-constructed medical entity synonym library, and the medical entity synonym library covers the corresponding relationship between the general name, the abbreviation, the alias and the professional abbreviation of the experimental object attribute;

[0075] For the entity object with a fuzzy expression, the corresponding attribute interval of the sub-node in the root node of the knowledge graph is matched in combination with the age stratification standard in the medical field, and accurate mapping is realized;

[0076] As a preferred embodiment of the present application, the process of mapping the target index to the secondary nodes of the knowledge graph comprises:

[0077] If the target index is a composite index, the target index is split into several single detection indexes, and each detection index is sequentially mapped to the corresponding secondary node;

[0078] It should be noted that fuzzy matching is performed when mapping entity objects, specifically:

[0079] If the user inputs Alzheimer's disease, it is automatically expanded to map to all root nodes containing Alzheimer's disease and cognitive impairment in the knowledge graph; if the user inputs anticancer drugs, it is automatically associated with the tertiary nodes of the intervention type of chemotherapy drugs and targeted inhibitors;

[0080] As a preferred embodiment of the present application, the process of matching the relationship descriptor with the semantic relationship edge comprises:

[0081] A multi-level relationship matching priority is constructed, and the precise medical semantic relationship is matched first, and then the general association relationship is matched, the precise medical semantic relationship includes direct regulation and significant correlation, and the general association relationship includes correlation and influence; when there are multiple semantic relationship edges matching, the high confidence path is preferentially traversed by combining the confidence of the relationship edge;

[0082] As a preferred embodiment of the present application, when traversing the knowledge graph path, the invalid circular path is filtered by the path depth limit and the circular detection algorithm, and the query efficiency is improved;

[0083] As a preferred embodiment of the present application, the process of determining the path satisfying the constraint comprises:

[0084] The confidence of each path is obtained, a confidence threshold is set, and the path whose confidence exceeds the confidence threshold is selected as the path satisfying the constraint;

[0085] The process of traversing the path further comprises: if the amount of associated data at the end of the path exceeds a threshold N, the first K paths are returned in descending order of the amount of experimental samples; and before returning the query result, a conclusion verification is performed to check whether the experimental conclusion keywords of the data at the end of the matched path contain negative words, if there are negative words, a conclusion conflict warning is marked in the result and the amount of contradictory data is displayed; for the experimental data returned at multiple ends, the data is filtered and sorted according to the data quality dimension, and high-quality experimental data is preferentially returned to assist the user in quickly locating effective information;

[0086] As a preferred embodiment of the present application, the abstract query performs the following operations:

[0087] When the query sentence contains the relationship description word "correlation", only the path with the quantified relationship edge is matched, for example, the edge attribute contains the correlation coefficient value; when the query sentence contains the mechanism class vocabulary, the path with the causal semantic edge is preferentially matched, for example, the edge label is activation, inhibition.

[0088] The above has carried out the detailed description to one embodiment of the application, but the content described is only the preferred embodiment of the application, and cannot be considered as being used to limit the implementation range of the application. Any equivalent change and improvement made according to the application scope of the application should still belong to the application coverage range of the application.

Claims

1. A data query method for a medical laboratory system, characterized by, The method comprises the following steps: Step S1: establishing a medical experiment report library comprising all medical experiment reports; binding structured stored experiment information for each medical experiment report in the medical experiment report library, the experiment information comprising experiment object attributes, a detection index set, experiment condition parameters and experiment conclusion keywords; Step S2: based on the experiment information of each medical experiment report, taking the experiment object attributes as root nodes, the detection index set as secondary nodes, the experiment condition parameters as tertiary nodes, and generating semantic relationship edges between nodes based on the experiment conclusion keywords to obtain a knowledge graph; Step S3: obtaining a user inputted abstract query statement, extracting entity objects, target indexes and relationship description words in the abstract query statement; mapping the entity objects to root nodes of the knowledge graph, mapping the target indexes to secondary nodes of the knowledge graph, and matching the relationship description words with semantic relationship edges; traversing paths in the knowledge graph that satisfy constraints, and returning an end of the path to obtain a medical experiment report associated with the end; In step S2, the generation process of the semantic relationship edges comprises: Predefining a plurality of relationship rules to generate a rule library, the relationship rules comprising object-index relationships, index-condition relationships and conclusion derivation relationships; Substituting the experiment conclusion keywords into the rule library to match the relationship rules corresponding to the experiment conclusion keywords, establishing a relationship extraction model, and obtaining semantic relationships of the experiment conclusion keywords through the relationship extraction model; defining directions of the semantic relationships, and labeling strengths of the semantic relationships, and creating semantic relationship edges between nodes according to the directions and strengths. In step S1, the structured stored experiment information specifically comprises:

2. The data query method for a medical experiment system according to claim 1, wherein, The experiment object attributes comprise a plurality of biological characteristics, the biological characteristics comprising species information, gender, age stage and health status indicators; the experiment condition parameters comprise temperature, humidity and light cycle; and the experiment conclusion keywords comprise significance, trend difference, positive correlation and negative correlation. In step S1, when any detection index set of a medical experiment report has more than one detection index, dynamic grouping is performed on the storage of the detection index set of the medical experiment report, specifically comprising:

3. The data query method for a medical experiment system according to claim 1, wherein, A plurality of detection index groups are set, the detection index groups comprising a molecular detection group, an image feature group and a physiological parameter group; if a detection index in the detection index set of the medical experiment report is a molecular biology index, the molecular biology index is classified into the molecular detection group, the molecular biology index comprising genes, proteins and metabolites; if the detection index in the detection index set of the medical experiment report is an imaging index, the imaging index is classified into the image feature group, the imaging index comprising CT parameters and MRI texture parameters; and if the detection index in the detection index set of the medical experiment report is a physiological signal index, the physiological signal index is classified into the physiological parameter group, the physiological signal index comprising heart rate and blood pressure fluctuation values. In step S1, the experiment information obtaining process comprises:

4. The data query method for a medical laboratory system of claim 1, wherein, ​ The related text of the medical experiment to which the medical experiment report belongs is acquired, the related text including a medical experiment report main text, analysis and summary, and conclusion deduction; the related text is split into a plurality of vocabulary units based on a natural language processing technology, and each vocabulary unit is subjected to named entity recognition to obtain an entity recognition result, and experiment information is extracted based on the entity recognition result.

5. The data query method for a medical laboratory system of claim 1, wherein, In step S2, the nodes include root nodes, secondary nodes, and tertiary nodes.

6. The data query method for a medical laboratory system of claim 1, wherein, In step S3, the process of mapping the target index to the secondary nodes of the knowledge graph includes: If the target index is a composite index, the target index is split into a plurality of single detection indexes, and each detection index is sequentially mapped to a corresponding secondary node.

7. The data query method for a medical laboratory system of claim 1, wherein, In step S3, the process of determining the path satisfying the constraint includes: The confidence of each path is acquired, a confidence threshold is set, and paths whose confidence exceeds the confidence threshold are screened out and recorded as paths satisfying the constraint.

Images