Peritoneal dialysis question and answer method and system based on mapping knowledge domain

Abstract

The invention discloses a peritoneal dialysis question and answer method and system based on a knowledge graph. According to the method, the knowledge of multiple data sources is fused, so that the island type data is converted into the comprehensive knowledge graph, and the limitation of dependence on single or fragmented data is avoided. A systematic medical knowledge base is formed through a comprehensive knowledge graph covering disease mechanisms, operation specifications and individual cases. And secondly, through a deep reasoning engine constructed through the knowledge graph, manual intervention is not needed, responses including operation steps, risk prompts and medical advices can be automatically generated, and the response efficiency is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of biomedical engineering, and more particularly to a peritoneal dialysis question-and-answer method and system based on knowledge graphs. Background Technology

[0002] Peritoneal dialysis is an important renal replacement therapy for patients with chronic renal failure. This therapy allows patients to perform dialysis at home, improving their autonomy and quality of life while reducing the burden on hospitals. However, peritoneal dialysis involves many technical and management challenges, such as adjusting the frequency and duration of dialysis, equipment maintenance, and infection prevention. These challenges require patients to have certain professional knowledge and skills, and obtaining relevant information can be difficult.

[0003] Currently, several peritoneal dialysis devices on the market are equipped with corresponding software. This software features remote monitoring, automated operation, and doctor-patient communication capabilities, achieving significant results in improving patient management and treatment efficiency. However, existing technologies still have the following shortcomings:

[0004] Information fragmentation: While existing software can provide popular science articles, the content lacks comprehensiveness and systematicity. Patients need to obtain information from multiple sources, which not only increases learning costs and time consumption, but also makes it more difficult for patients to understand accurate peritoneal dialysis knowledge due to the possibility of contradictory information from different sources.

[0005] Limitations of passive response: Existing software mainly relies on patients to initiate inquiries or consultations, making it difficult for doctors to respond promptly. For example, when patients experience discomfort or operational abnormalities during peritoneal dialysis, they cannot receive timely guidance, leading to dialysis treatment failing to achieve the desired results.

[0006] Insufficient personalized support: Different patients have different conditions and needs, but most existing software only has dialysis recording functions and cannot provide customized suggestions and guidance based on the specific situation of the patient.

[0007] Poor user interaction experience: There are obvious deficiencies in user interface design and interaction experience, such as chaotic interface layout, complex operation process, and slow response speed.

[0008] Therefore, existing technologies still need improvement and development. Summary of the Invention

[0009] The technical problem to be solved by the present invention is to provide a peritoneal dialysis question-and-answer method and system based on knowledge graph, which addresses the above-mentioned deficiencies of the existing technology. The aim is to solve the problems that the existing peritoneal dialysis equipment software requires high levels of professional knowledge from users and relies on manual answers, resulting in untimely responses.

[0010] The technical solution adopted by this invention to solve the problem is as follows:

[0011] In a first aspect, embodiments of the present invention provide a peritoneal dialysis question-and-answer method based on a knowledge graph, the method comprising:

[0012] Medical data is collected from multiple data sources, including at least one of the following: medical literature, clinical guidelines, expert experience, and patient data; the medical data includes data related to peritoneal dialysis.

[0013] The medical-related data is subjected to entity extraction to obtain several tuple data sets; each tuple data set includes an entity and corresponding entity feature information; the entity feature information includes the relationship information between the entity and other entities, the attribute information of the entity, and / or event information related to the entity;

[0014] A knowledge graph is constructed based on the aforementioned medical-related data and the various tuple data.

[0015] A deep reasoning engine is built based on a knowledge graph, and the deep reasoning engine automatically responds to user interaction data; the user interaction data includes at least one of user questions, user health data, and user treatment data.

[0016] In one implementation, entity extraction is performed on the medical-related data, including:

[0017] Entity extraction is performed on the unstructured data in the aforementioned medical-related data.

[0018] In one implementation, a knowledge graph is constructed based on the medical-related data and each of the tuple data, including:

[0019] Based on the structured data in the aforementioned medical-related data, knowledge-integrated data is obtained;

[0020] Entity alignment and / or referential resolution are performed on each of the aforementioned tuple data to obtain knowledge fusion data;

[0021] Knowledge modeling is performed based on the knowledge integration data and the knowledge fusion data to obtain a knowledge graph; the knowledge modeling includes at least one of the following modeling processes: concept modeling, rule modeling, event modeling, and spatiotemporal modeling.

[0022] In one implementation, the concept modeling includes:

[0023] Based on the knowledge integration data and the knowledge fusion data, a concept extraction operation is performed to obtain several concepts related to peritoneal dialysis.

[0024] The concepts are classified, and the hierarchical relationships between the concepts are established based on the classification results to achieve concept modeling.

[0025] In one implementation, the deep inference engine automatically responds to user interaction data, including:

[0026] When the user interaction data is a user question, the knowledge graph is searched based on the user question to match several keywords;

[0027] Based on the keywords mentioned, the intent of the question can be analyzed;

[0028] For each keyword, the corresponding entity is matched in the knowledge graph, and the matched entities are filtered according to the question intent and entity feature information to obtain the target entity;

[0029] The search scope is expanded based on the relationship information of each target entity to obtain multiple target entities;

[0030] All the target entities and their respective entity feature information are integrated, and the answer to the user question is generated based on the integration result.

[0031] In one implementation, the deep inference engine automatically responds to user interaction data, including:

[0032] When the user interaction data is user health data, the deep inference engine makes personalized recommendations based on the user health data to obtain health guidance suggestions.

[0033] When the user interaction data is user treatment data, the deep reasoning engine performs personalized recommendations based on the user treatment data to obtain treatment plan suggestions.

[0034] In one implementation, the system interface corresponding to the deep inference engine supports multimodal interaction methods; the multimodal interaction methods include at least one of text and voice interaction methods.

[0035] Secondly, embodiments of the present invention also provide a peritoneal dialysis question-and-answer system based on a knowledge graph, the system comprising:

[0036] The data collection module is used to collect medical-related data from multiple data sources, including at least one of the following: medical literature, clinical guidelines, expert experience, and patient data; the medical-related data includes data related to peritoneal dialysis.

[0037] The data processing module is used to perform entity extraction operations on the medical-related data to obtain several tuple data sets; each tuple data set includes an entity and corresponding entity feature information; the entity feature information includes the relationship information between the entity and other entities, the attribute information of the entity, and / or event information related to the entity;

[0038] The knowledge graph construction module is used to construct a knowledge graph based on the medical-related data and each of the tuple data.

[0039] The deep reasoning module is used to build a deep reasoning engine based on the knowledge graph and to automatically respond to user interaction data through the deep reasoning engine; the user interaction data includes at least one of user questions, user health data, and user treatment data.

[0040] Thirdly, embodiments of the present invention also provide a terminal, the terminal including a memory and one or more processors; the memory stores one or more programs; the programs include instructions for executing the knowledge graph-based peritoneal dialysis question-and-answer method as described above; the processor is used to execute the programs.

[0041] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a plurality of instructions adapted to be loaded and executed by a processor to implement the steps of the knowledge graph-based peritoneal dialysis question-and-answer method as described above.

[0042] The beneficial effects of this invention are as follows: By integrating knowledge from multiple data sources, the embodiments of this invention transform isolated data into a comprehensive knowledge graph, avoiding the limitations of relying on single or fragmented data. A systematic medical knowledge base is formed through a comprehensive knowledge graph covering disease mechanisms, operational procedures, and individual cases. Furthermore, the deep reasoning engine built upon the knowledge graph can automatically generate responses containing operational steps, risk warnings, and medical advice without human intervention, significantly improving response efficiency. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is a flowchart illustrating the peritoneal dialysis question-and-answer method based on knowledge graphs provided in this embodiment of the invention.

[0045] Figure 2This is a schematic diagram of the process for constructing a knowledge graph provided in an embodiment of the present invention.

[0046] Figure 3 This is a schematic diagram of the internal modules of the peritoneal dialysis question-and-answer system based on knowledge graphs provided in an embodiment of the present invention.

[0047] Figure 4 This is a schematic diagram of the terminal provided in the embodiment of the present invention. Detailed Implementation

[0048] This invention discloses a peritoneal dialysis question-and-answer method and system based on knowledge graphs. To make the objectives, technical solutions, and effects of this invention clearer and more explicit, the invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention.

[0049] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” “the,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we say an element is “connected” or “coupled” to another element, it can be directly connected or coupled to the other element, or there may be intermediate elements. Furthermore, “connected” or “coupled” as used herein can include wireless connections or wireless coupling. The term “and / or” as used herein includes all or any units and all combinations of one or more associated listed items.

[0050] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0051] To address the aforementioned deficiencies in existing technologies, this invention provides a knowledge graph-based question-and-answer method for peritoneal dialysis. The method includes: collecting medical-related data from multiple data sources; these multiple data sources include at least one of medical literature, clinical guidelines, expert experience, and patient data; the medical-related data includes data related to peritoneal dialysis; performing entity extraction on the medical-related data to obtain several tuples of data; each tuple of data includes an entity and corresponding entity feature information; the entity feature information includes relationship information between the entity and other entities, attribute information of the entity, and / or event information related to the entity; constructing a knowledge graph based on the medical-related data and each tuple of data; establishing a deep reasoning engine based on the knowledge graph, and automatically responding to user interaction data through the deep reasoning engine; the user interaction data includes at least one of user questions, user health data, and user treatment data. This invention, by integrating knowledge from multiple data sources, transforms isolated data into a comprehensive knowledge graph, avoiding the limitations of relying on single or fragmented data. A systematic medical knowledge base is formed through a comprehensive knowledge graph covering disease mechanisms, operational procedures, and individual cases. Secondly, the deep reasoning engine built through knowledge graphs can automatically generate responses containing operation steps, risk warnings, and medical advice without human intervention, significantly improving response efficiency.

[0052] like Figure 1 As shown, the method specifically includes the following steps:

[0053] Step S100: Collect medical-related data through multiple data sources; the multiple data sources include at least one of the following: medical literature, clinical guidelines, expert experience, and patient data; the medical-related data includes data related to peritoneal dialysis.

[0054] Specifically, in order to construct a comprehensive knowledge graph covering the entire peritoneal dialysis process, this embodiment requires the collection and integration of relevant data from multiple data sources, including but not limited to: medical literature, clinical guidelines, expert experience, and patient data. This ensures that the constructed knowledge graph not only includes basic medical knowledge but also specific operational steps, common problems, and solutions.

[0055] For example, data can be collected from multiple sources, including domestic and international medical literature databases, professional websites, medical guidelines, and clinical research data, to obtain knowledge on various aspects of peritoneal dialysis, such as basic concepts, indications, operating procedures, and complication management.

[0056] Regarding data source selection: it can integrate multiple public databases such as specific official websites and PubMed, as well as data from specific peritoneal dialysis societies, specific nephrology societies, and other professional societies and associations to ensure the comprehensiveness and timeliness of the knowledge graph information.

[0057] Regarding data collection methods, various approaches can be employed, such as manual collection, automated web crawling, machine learning, and expert methods, to ensure the comprehensiveness and timeliness of the data.

[0058] For data preprocessing: it is necessary to clean, label, and standardize the collected medical data to ensure data quality.

[0059] For data quality assessment, the focus should be on evaluating the completeness, consistency, and accuracy of the data.

[0060] For data storage: a combination of local file storage and database storage can be used to ensure secure and efficient data access.

[0061] Step S200: Perform entity extraction on the medical-related data to obtain several tuple data sets; each tuple data set includes an entity and corresponding entity feature information; the entity feature information includes the relationship information between the entity and other entities, the attribute information of the entity, and / or event information related to the entity.

[0062] Specifically, to enrich the content of the knowledge graph, this embodiment uses natural language processing technology to perform entity extraction on the collected medical-related data, extracting key entities (such as disease names, drug names, etc.), relationship information (such as etiology-effect, drug-mechanism of action), attribute information (such as dosage, frequency), and / or event information (operation time or abnormal events), laying the foundation for subsequent knowledge fusion. Among these, relationship information, attribute information, and event information serve as entity feature information corresponding to the entities.

[0063] Relational information is used to reflect the logical connections between entities, such as hierarchical relationships, operational relationships, and causal relationships. Extracted relational information allows scattered entities to form knowledge networks. Constructing knowledge graphs using relational information enables systems to respond quickly through the relational chains of entities.

[0064] Attribute information is used to reflect the inherent characteristics or state of an entity, adding details to the entity and supporting accurate reasoning in subsequent systems.

[0065] Event information is used to reflect the specific events in which an entity participates and their context. By combining static knowledge with real-world application scenarios, it can support subsequent system reasoning based on events.

[0066] In practical applications, the extracted <entity-relationship-attribute> or <entity-relationship-attribute-event> is a tuple of data. A large amount of tuple data can be obtained through medical-related data.

[0067] For example, entity types include: diseases, symptoms, drugs, side effects, procedures, equipment, and laboratory tests. The following main relation types are defined during entity extraction:

[0068] Diseases and symptoms: such as chronic kidney disease and abdominal pain;

[0069] Disease-drug: such as acute kidney injury and icodextrin;

[0070] Drug side effects: such as icodextrin and peritonitis;

[0071] Operating procedures - Equipment: such as dialysis fluid and tubing;

[0072] Laboratory tests - diseases: such as serum creatinine and chronic kidney disease;

[0073] Procedures - Symptoms: such as peritoneal dialysis and abdominal pain.

[0074] For relation extraction: Extract the relationships between entities, such as "peritoneal dialysis requires the use of dialysis fluid" and "catheter insertion location".

[0075] For attribute extraction: obtain the attribute information of the entity, such as "type of dialysis fluid" and "type of catheter".

[0076] In one implementation, entity extraction is performed on the medical-related data, including:

[0077] Entity extraction is performed on the unstructured data in the aforementioned medical-related data.

[0078] Specifically, medical data can be categorized into structured and unstructured data. Structured data consists of data with a fixed format that can be directly stored in tables / databases, such as test results, medication dosages, and time / frequency data in dialysis records. Unstructured data, on the other hand, lacks a fixed format and is primarily free text. Computers struggle to process unstructured data directly and require entity extraction to transform it into machine-understandable structured data, thereby systematically integrating fragmented clinical knowledge.

[0079] For example, structured data can include dialysis records in electronic medical records (such as ultrafiltration volume and number of dialysis sessions) and laboratory test results (such as serum potassium and creatinine levels). Unstructured data can include semi-structured data and / or completely unstructured data, such as: handwritten medical records from doctors (such as "The patient had poor drainage during dialysis today, considering catheter displacement") and patient self-reports (such as "I worry about infection when the peritoneal dialysis fluid is cloudy").

[0080] Step S300: Construct a knowledge graph based on the medical-related data and each of the tuple data.

[0081] Specifically, a knowledge graph is a structured form of knowledge storage that organizes and represents data in the form of a graph. Nodes represent entities, and edges represent relationships between entities. Knowledge graphs can effectively integrate and manage large-scale heterogeneous data, supporting complex query and reasoning tasks. In the healthcare field, knowledge graphs can be used for disease diagnosis, personalized treatment recommendations, and drug interaction analysis, greatly improving the quality and efficiency of chronic disease management. This embodiment is specifically designed for the peritoneal dialysis field, constructing a highly specialized and segmented knowledge graph that enables the system to more deeply understand and process complex information and specific needs related to peritoneal dialysis.

[0082] In one implementation, a knowledge graph is constructed based on the medical-related data and each of the tuple data, including:

[0083] Based on the structured data in the aforementioned medical-related data, knowledge-integrated data is obtained;

[0084] Entity alignment and / or referential resolution are performed on each of the aforementioned tuple data to obtain knowledge fusion data;

[0085] Knowledge modeling is performed based on the knowledge integration data and the knowledge fusion data to obtain a knowledge graph; the knowledge modeling includes at least one of the following modeling processes: concept modeling, rule modeling, event modeling, and spatiotemporal modeling.

[0086] Specifically, such as Figure 2 As shown, data integration, knowledge fusion, and knowledge modeling are key steps in knowledge graph construction. The goal is to transform scattered structured medical data into machine-understandable graph-structured knowledge, and to solve the problems of heterogeneity, redundancy, and ambiguity of multi-source structured data, thereby forming a unified, accurate, and interconnected knowledge system.

[0087] Regarding data integration: Structured data is data organized according to a preset format. The purpose of data integration is to solve problems such as data heterogeneity and eliminate data redundancy. Structured data from different data sources may have inconsistent field naming and formatting issues. Specifically, the data integration process can be achieved by establishing a unified data schema, mapping fields from different data sources to unified standard fields. For example, the unit can be unified as: times / day. Secondly, information about the same entity may be stored repeatedly in multiple data sources, requiring deduplication and retention of the latest or most accurate version.

[0088] For knowledge fusion, the main methods are entity alignment and reference resolution. The former addresses the issue of different representations of the same entity, such as different data sources using different names to refer to the same entity. The latter resolves ambiguous references of the same entity, such as the same name referring to different entities, or the same entity having different references in different contexts. This process eliminates entity ambiguity, merges duplicate entities, and unifies entity representations in the dataset, ensuring that each node in the knowledge graph is unique and accurate.

[0089] Conceptual modeling involves constructing a conceptual model for the peritoneal dialysis domain, defining various concepts and their relationships. The role of the conceptual model is to define the core concepts and their interrelationships within the domain during knowledge graph construction. In the peritoneal dialysis domain, these concepts include disease names, treatment plans, drug names, etc., and the relationships between them may be causal (e.g., a symptom is caused by a specific disease), therapeutic (e.g., a drug is used to treat a specific disease), etc. The data used in conceptual modeling comes from multiple data sources, and the entities, attributes, and relationships are obtained after data processing. The result of conceptual modeling is a structured knowledge system, i.e., a knowledge graph, which clearly defines each concept and its connections. The conceptual model provides the framework and rules for the knowledge graph, guiding the knowledge graph construction process. Conceptual modeling is the foundation of the knowledge graph, determining which entities and attributes are included in the knowledge graph and how their relationships are organized.

[0090] For rule-based modeling: Incorporate the knowledge of medical experts to establish clinical decision-making rules.

[0091] For spatiotemporal modeling: The system records and analyzes temporal and spatial information during peritoneal dialysis. Temporal information may include treatment frequency and catheter change times, while spatial information may include catheter insertion location. This temporal and spatial information helps the system perform spatiotemporal modeling, which is crucial for providing personalized medical advice. For example, based on a patient's specific circumstances (such as treatment history), the system can recommend the most suitable treatment plan or remind the patient of the appropriate catheter change time.

[0092] The above modeling process completes the knowledge modeling and outputs the final knowledge graph. The graph structure of the knowledge graph includes:

[0093] Node: An instance of an entity category defined in conceptual modeling. For example, peritoneal dialysis is an instance of the treatment method category;

[0094] Edges: Relationships between entities defined by rule modeling, event associations defined by event modeling (e.g., peritoneal dialysis causing peritonitis), and time / space attributes defined by spatiotemporal modeling (e.g., patient a's dialysis start time is XXXX year-XX month-XX day).

[0095] The knowledge graph provides the foundation for subsequent deep reasoning engines. For example, a patient might input, "My stomach hurts after peritoneal dialysis yesterday, what should I do?". The knowledge graph can determine this through entity alignment: peritoneal dialysis = peritoneal dialysis, and stomach pain = abdominal pain; through rule modeling: triggering the relationship between abdominal pain, possible complications, and peritonitis; through event modeling: querying the patient's historical dialysis records (whether there is a history of infection); and through spatiotemporal modeling: combining the current time (e.g., if the pain occurs at night, is emergency treatment required). Ultimately, it generates an answer corresponding to the information input by the patient.

[0096] In one implementation, entity alignment is performed on each of the said tuple data, including:

[0097] For any two sets of data, the similarity between entities in each set of data is calculated based on the entity feature information of the two sets of data.

[0098] If the similarity is greater than a preset threshold, the entities of the two tuple data sets are merged to achieve entity alignment.

[0099] Specifically, the goal of entity alignment is to determine whether entities in different tuples of data point to the same real object, and to merge duplicate entities to eliminate data redundancy and ambiguity. Entity feature information includes the corresponding relationship information, attribute information, and event information. This embodiment quantifies the similarity between two entities from multiple dimensions based on their entity feature information. A similarity threshold is set; if the similarity scores of two entities exceed the threshold, they are considered to point to the same object. The two entities pointing to the same object are merged, i.e., their entity identifiers are unified to avoid duplicate nodes in the knowledge graph, and their feature information is integrated, deduplicating and retaining more complete entity feature information. The merged entity has more comprehensive feature information, making the nodes in the knowledge graph more complete and the relationships more accurate, providing a reliable foundation for subsequent deep reasoning.

[0100] For example, in practical applications, the entity alignment process mainly includes the following steps:

[0101] Data Collection and Labeling: Peritoneal dialysis-related data were collected from multiple sources. The same entity may be represented in different ways in these data. For example, "peritoneal dialysis fluid" may be represented as "dialysis fluid" in some data and "peritoneal dialysis fluid" in others.

[0102] Feature extraction and similarity calculation: Extracting the attributes, relationships, and other features of entities, and calculating the similarity between entities from different data sources. Taking the entity related to "dialysis fluid" as an example, it compares its attributes such as composition, function, and applicable diseases, as well as its relationship features with other entities such as "catheter" and "dialysis operation steps," and calculates the similarity using algorithms such as cosine similarity.

[0103] Entity merging and unification: Entities with similarity exceeding a set threshold are merged and represented in a unified manner. After confirming that "dialysis fluid" and "peritoneal dialysis fluid" both refer to "peritoneal dialysis fluid," the term "peritoneal dialysis fluid" is used uniformly in the knowledge graph to integrate related attributes and relationships, avoiding data redundancy and conflicts.

[0104] In one implementation, the referencing of each of the tuple data is resolved, including:

[0105] For any of the tuple data, if the text corresponding to the tuple data contains ambiguous information whose referent is unclear, then the explicit referent of the ambiguous information can be inferred based on the context information.

[0106] The ambiguous information is replaced with the explicit referent to achieve referential resolution.

[0107] Specifically, ambiguous information refers to content in text that cannot be directly identified due to unclear expression, such as pronoun references, omissions, and abbreviations. The purpose of referential resolution is to determine the unique, explicit referent (such as an entity, event, or attribute) of the ambiguous information by analyzing the context, and to replace the ambiguous expression in the original text, thus making the semantics of the tuple data more precise. First, the ambiguous information is identified. Then, the contextual information corresponding to the ambiguous information is extracted, including local and / or global context. Next, the explicit referent is inferred from the contextual information. For example, several candidate referents are inferred from the contextual information, and then the semantic relevance, distance, and co-occurrence frequency between each candidate referent and the ambiguous information are calculated. Based on the calculation results, one candidate referent is selected as the explicit referent. The ambiguous expression is replaced with a specific entity name, ensuring the semantic uniqueness of the tuple data.

[0108] For example, in practical applications, the process of resolution mainly includes the following steps:

[0109] Text parsing and referential recognition: The system parses the input text and identifies the referential relationships within it. For example, in the sentence "If the peritoneal fluid becomes cloudy during peritoneal dialysis, it should be treated promptly," the system identifies that "it" refers to "dialysis fluid."

[0110] Contextual Analysis and Reasoning: Combining textual context information with knowledge from the knowledge graph for reasoning. Based on the preceding description of peritoneal dialysis procedures and the close relationship between "dialysis fluid" and "peritoneal dialysis" in the knowledge graph, determine the specific meaning of "it".

[0111] Identify and replace the referent: After identifying the referent, replace the referential word with a specific entity to make the meaning of the text clearer and facilitate subsequent processing and application of the knowledge graph. For example, replacing "it" in the above sentence with "dialysis fluid" enhances the accuracy and comprehensibility of the information.

[0112] In one implementation, the conceptual modeling includes:

[0113] Based on the knowledge integration data and the knowledge fusion data, a concept extraction operation is performed to obtain several concepts related to peritoneal dialysis.

[0114] The concepts are classified, and the hierarchical relationships between the concepts are established based on the classification results to achieve concept modeling.

[0115] Specifically, the knowledge integration data and knowledge fusion data are processed data, both derived from medical-related data collected from multiple data sources. Concepts related to peritoneal dialysis are extracted from this processed data and categorized. The inherent logical relationships between concepts within the same category are analyzed based on the classification results, and the existence of hierarchical relationships is determined based on these relationships. Through the classification results and the determination of hierarchical relationships, hierarchical relationships between concepts are established.

[0116] For example, in practical applications, the process of generating conceptual context relationships mainly includes the following steps:

[0117] Concept Extraction and Classification: Peritoneal dialysis-related concepts, such as "dialysis complications," "peritonitis," and "catheter infection," were extracted from the collected data and preliminarily classified. "Peritonitis" and "catheter infection" were categorized under the "dialysis complications" category.

[0118] Relationship Judgment and Hierarchical Construction: Analyze the inherent logical relationships between concepts to determine whether a hierarchical relationship exists. Based on medical knowledge, "peritonitis" is a type of "dialysis complication," establishing a hierarchical relationship where "dialysis complication" is the superordinate concept and "peritonitis" is the subordinate concept.

[0119] Relationship Improvement and Knowledge Graph Update: The generated hierarchical relationships between concepts are added to the knowledge graph to further improve its structure and semantic expression. Directed edges are established in the knowledge graph from "dialysis complications" to "peritoneitis" to identify hierarchical relationships, enabling the knowledge graph to more accurately reflect the knowledge system in the field of peritoneal dialysis.

[0120] In one implementation, the method further includes: periodically obtaining the latest information from designated medical resources, and automatically updating the knowledge graph based on the latest information.

[0121] Specifically, this embodiment also establishes an automatic update mechanism to periodically obtain the latest information from designated medical resources (such as designated authoritative medical resources) and automatically update the knowledge graph to ensure that the information provided by the system is always the latest and most accurate.

[0122] Step S400: Build a deep reasoning engine based on the knowledge graph, and automatically respond to user interaction data through the deep reasoning engine; the user interaction data includes at least one of user questions, user health data, and user treatment data.

[0123] Specifically, in this embodiment, the user of the deep reasoning engine can be the patient themselves or other people related to the patient, such as family members. This embodiment utilizes structured data and relationships in a knowledge graph to develop a deep reasoning engine (or intelligent question-answering system), capable of logical reasoning and multi-step operational guidance for complex questions, effectively answering patients' questions, providing scientific guidance and suggestions, thereby improving patient treatment adherence and reducing the incidence of complications. Furthermore, the deep reasoning engine can not only respond to user questions but also automatically respond to the user's health data and / or treatment data, achieving a diversified data interaction process.

[0124] For example, when a patient asks about adjusting dialysis frequency, the system can provide detailed treatment plan adjustments, steps, and precautions. The deep reasoning engine not only helps patients resolve problems that arise during peritoneal dialysis, but also reduces the workload of medical staff.

[0125] In one implementation, the deep inference engine automatically responds to user interaction data, including:

[0126] When the user interaction data is a user question, the knowledge graph is searched based on the user question to match several keywords;

[0127] Based on the keywords mentioned, the intent of the question can be analyzed;

[0128] For each keyword, the corresponding entity is matched in the knowledge graph, and the matched entities are filtered according to the question intent and entity feature information to obtain the target entity;

[0129] The search scope is expanded based on the relationship information of each target entity to obtain multiple target entities;

[0130] All the target entities and their respective entity feature information are integrated, and the answer to the user question is generated based on the integration result.

[0131] Specifically, the process begins by searching the knowledge graph based on the user's question to identify relevant keywords. These keywords are then analyzed in depth to accurately pinpoint the question's intent. Next, matching entities are searched within the knowledge graph based on these keywords. The attributes of these matched entities are then meticulously filtered to obtain relevant target entities. The search scope is further expanded based on the relationships between entities to yield more relevant information. Finally, all relevant information from the filtered entities, relationships, and attributes is systematically integrated to form a complete answer.

[0132] For example, the process of generating answers based on user questions mainly includes the following steps:

[0133] Knowledge Retrieval: The system searches for information related to the user's question within the knowledge graph, matches keywords, and finds the most suitable answer. Keywords are not necessarily identical to the question's intent, but they are important clues for understanding that intent. By analyzing keywords in the question and combining them with natural language processing technology, the system can more accurately capture the user's actual needs and intent.

[0134] Keyword analysis: Utilizing natural language processing technology, the system deeply analyzes the keywords entered by the user to accurately identify the intent behind the question. For example, if a patient enters "questions related to dialysis fluid change frequency," the system will extract key information from the keywords to determine that the question revolves around the frequency of dialysis fluid changes.

[0135] Entity matching: Based on the parsed keywords, the system actively searches for matching entities within the knowledge graph. For example, for "dialysis fluid replacement frequency," the system will search for the entity "dialysis fluid." The "dialysis fluid" entity in the knowledge graph contains various attributes, such as "type," "dosage," and "replacement frequency." During the search, the system will focus on attribute values ​​related to "replacement frequency."

[0136] Relationship Expansion: After finding relevant entities, the search scope can be further expanded based on the preset relationships between entities. For example, the entity "dialysis fluid" and the entity "operation steps" have an "operation steps-equipment" relationship. After identifying the "dialysis fluid" entity, this relationship can be used to find the associated "operation steps" entity, thereby obtaining more relevant information, such as the specific operation procedures for changing dialysis fluid.

[0137] Attribute filtering: Based on the question intent, the system performs detailed filtering of the matched entity attributes. For the question "dialysis fluid change frequency," the system filters out the specific values ​​of the key attribute "change frequency." If the knowledge graph records the dialysis fluid change frequencies of patients with different conditions and stages, the system will accurately filter out the most suitable change frequency information based on the specific patient information determined in the subsequent personalized recommendation process.

[0138] Information integration: This involves systematically integrating relevant information obtained from entities, relationships, and attributes, organizing it according to a certain logic, and generating detailed and accurate answers. For example, information such as the frequency of dialysis fluid changes, the procedures for changing the fluid, and precautions can be integrated to form a complete answer and provided to the user.

[0139] The process of generating answers based on search results mainly includes the following steps:

[0140] Information filtering: From the vast amount of information retrieved from the knowledge graph, the system filters out the most relevant parts based on the core and focus of the question. For example, if a patient asks "how to prevent peritonitis caused by peritoneal dialysis," the system will filter out information directly related to peritonitis prevention, such as the correct cleaning methods for dialysis equipment and aseptic requirements during dialysis operations, while excluding irrelevant information such as dialysis fluid formulation.

[0141] Content Organization: Organize the selected information logically. For example, regarding peritonitis prevention in peritoneal dialysis, first introduce the importance and specific methods of maintaining a clean dialysis environment, then explain how to strictly adhere to aseptic principles during dialysis procedures, and finally mention precautions for patients' personal hygiene habits, forming a clear and well-structured content.

[0142] Personalized approach: The prepared content is tailored to the individual patient's information, such as age, severity of illness, and dialysis history. For elderly patients, the language is made more accessible and easier to understand, taking into account their comprehension and physical condition. Emphasis is placed on preventative measures that are particularly important for elderly patients, such as the need to keep warm due to lower immunity.

[0143] Answer Verification: The system verifies the accuracy and completeness of the generated answers. It re-checks relevant information in the knowledge graph to ensure the accuracy of operational steps, medical knowledge, etc., in the answer, avoiding incorrect guidance. Simultaneously, it checks whether the answer covers the key aspects of the question, ensuring completeness.

[0144] Answer output: Present the validated answer to the user in an appropriate format. If the system supports multimodal interaction, it can display the answer in text form on the interface or output it via voice broadcast, making it convenient for patients to obtain information.

[0145] In one implementation, the deep inference engine automatically responds to user interaction data, including:

[0146] When the user interaction data is user health data, the deep inference engine makes personalized recommendations based on the user health data to obtain health guidance suggestions.

[0147] When the user interaction data is user treatment data, the deep reasoning engine performs personalized recommendations based on the user treatment data to obtain treatment plan suggestions.

[0148] Specifically, this embodiment also provides a personalized recommendation algorithm: combining the user's personal information (such as age, gender, medical condition, etc.) and historical records, a personalized recommendation algorithm is developed to provide tailored suggestions and guidance for each patient. In practical application scenarios, after obtaining the user's personal information, the system first performs data cleaning and standardization, and then combines the user's historical consultation records and the content of the current question to generate suggestions using the personalized recommendation algorithm. These suggestions may include, but are not limited to, the most suitable dialysis time and frequency for the patient, dietary and exercise recommendations, etc., aiming to improve patient treatment adherence and quality of life. For example, based on the patient's dialysis history and lifestyle habits, the system can recommend the most suitable dialysis time and frequency.

[0149] In one implementation, the system interface corresponding to the deep inference engine supports multimodal interaction methods; the multimodal interaction methods include at least one of text and voice interaction methods.

[0150] Specifically, this embodiment also provides a user-friendly interface that supports natural language processing and speech recognition, enabling patients to easily ask questions via text or voice and receive intuitive and easy-to-understand answers. The user-friendly interface and multimodal interaction support multiple interaction methods such as text and voice, optimizing the cumbersome text input process and allowing direct interaction with the system through voice Q&A, helping middle-aged and elderly patients more easily obtain the information and support they need, thus improving the user experience.

[0151] In other implementations, multiple existing general-purpose large language model question-answering systems can be integrated to implement a question-answering model for the peritoneal dialysis field. For example, multiple models such as ChatGPT can be combined to implement a question-answering model for a specific medical field.

[0152] In other implementations, different knowledge retrieval algorithms can be used, such as graph-based search algorithms, vector-based similarity search, etc.

[0153] In other implementations, different answer generation strategies can be designed, such as template-based methods and generative model-based methods.

[0154] In other implementation methods, different interactive interfaces can be designed, such as hardware voice applications, web applications, etc.

[0155] Application scenarios of this invention include, but are not limited to:

[0156] Patient education: Provide patients with basic knowledge and operating guidelines for peritoneal dialysis.

[0157] Complication management: Help patients identify and manage common complications, such as peritonitis and catheter infection, and remind patients to seek medical attention promptly when necessary.

[0158] Daily management: Providing advice on diet, exercise, etc., to help patients better manage their daily lives.

[0159] Emergency Response: In emergency situations, provide immediate guidance to help patients take the correct response measures.

[0160] Specifically, in practical applications, the method of this invention can be deployed on a cloud service platform: the entire system is deployed in the cloud, utilizing cloud computing resources to provide highly available and high-performance services, allowing patients to access the system on mobile devices. It can also be deployed on a local server: for institutions requiring high confidentiality, such as hospitals, deploying the system on a local server ensures data security and privacy protection. Furthermore, it supports multimodal input: in addition to traditional text input, the system also supports voice input and voice broadcast functions, improving the user experience for middle-aged and elderly patients and enhancing their treatment adherence. It also supports multilingual interaction: by expanding the natural language processing module, the system can support input and output in multiple languages, meeting the needs of users in different regions and countries. Finally, it can provide personalized recommendations: combining the user's personal health records and historical consultation records, the system can provide personalized medical services and dietary suggestions.

[0161] The advantages of this invention include, but are not limited to:

[0162] Improving the accuracy of medical consultations: By constructing a detailed knowledge graph of peritoneal dialysis, the system can provide more precise medical consultations and patient education services, ensuring that patients follow correct operating procedures during dialysis and avoid potential risks. This not only helps improve treatment outcomes but also enhances patients' self-management abilities and confidence.

[0163] Enhancing User Experience: The natural language processing module can understand and parse users' natural language questions, enabling users to interact with the system in a more natural way and improving the user experience.

[0164] Enhancing the system's scalability and flexibility: The knowledge graph's construction and updating mechanism enables the system to continuously absorb new medical knowledge, maintaining its advanced nature and practicality.

[0165] Reduce healthcare costs: By providing efficient online consultation and education services, reduce unnecessary hospital visits for patients and lower healthcare costs.

[0166] Based on the above embodiments, the present invention also provides a peritoneal dialysis question-and-answer system based on a knowledge graph, such as... Figure 3 As shown, the system includes:

[0167] The data collection module is used to collect medical-related data from multiple data sources, including at least one of the following: medical literature, clinical guidelines, expert experience, and patient data; the medical-related data includes data related to peritoneal dialysis.

[0168] The data processing module is used to perform entity extraction operations on the medical-related data to obtain several tuple data sets; each tuple data set includes an entity and corresponding entity feature information; the entity feature information includes the relationship information between the entity and other entities, the attribute information of the entity, and / or event information related to the entity;

[0169] The knowledge graph construction module is used to construct a knowledge graph based on the medical-related data and each of the tuple data.

[0170] The deep reasoning module is used to build a deep reasoning engine based on the knowledge graph and to automatically respond to user interaction data through the deep reasoning engine; the user interaction data includes at least one of user questions, user health data, and user treatment data.

[0171] Based on the above embodiments, the present invention also provides a terminal, the principle block diagram of which can be as follows: Figure 4As shown, the terminal includes a processor, memory, network interface, and display screen connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a knowledge graph-based peritoneal dialysis question-and-answer method. The display screen can be an LCD screen or an e-ink screen.

[0172] Those skilled in the art will understand that Figure 4 The schematic diagram shown is merely a partial structural diagram related to the present invention and does not constitute a limitation on the terminal to which the present invention is applied. A specific terminal may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0173] In one implementation, the terminal's memory stores one or more programs, and these programs are configured to be executed by one or more processors, and the programs contain instructions for performing a knowledge graph-based peritoneal dialysis question-and-answer method.

[0174] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0175] In summary, this invention discloses a peritoneal dialysis question-answering method and system based on a knowledge graph. The method includes: collecting medical-related data from multiple data sources; these multiple data sources include at least one of medical literature, clinical guidelines, expert experience, and patient data; the medical-related data includes data related to peritoneal dialysis; performing entity extraction on the medical-related data to obtain several tuples of data; each tuple includes an entity and corresponding entity feature information; the entity feature information includes relationship information between the entity and other entities, attribute information of the entity, and / or event information related to the entity; constructing a knowledge graph based on the medical-related data and each tuple; establishing a deep reasoning engine based on the knowledge graph, and automatically responding to user interaction data through the deep reasoning engine; the user interaction data includes at least one of user questions, user health data, and user treatment data. This invention, by integrating knowledge from multiple data sources, transforms isolated data into a comprehensive knowledge graph, avoiding the limitations of relying on single or fragmented data. A systematic medical knowledge base is formed through a comprehensive knowledge graph covering disease mechanisms, operational procedures, and individual cases. Secondly, the deep reasoning engine built through knowledge graphs can automatically generate responses containing operation steps, risk warnings, and medical advice without human intervention, significantly improving response efficiency.

[0176] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A knowledge graph-based question-and-answer method for peritoneal dialysis, characterized in that, The method includes: Medical data is collected from multiple data sources, including at least one of the following: medical literature, clinical guidelines, expert experience, and patient data; the medical data includes data related to peritoneal dialysis. The medical-related data is subjected to entity extraction to obtain several tuple data sets; each tuple data set includes an entity and corresponding entity feature information; the entity feature information includes the relationship information between the entity and other entities, the attribute information of the entity, and / or event information related to the entity; A knowledge graph is constructed based on the aforementioned medical-related data and the various tuple data. A deep reasoning engine is built based on a knowledge graph, and the deep reasoning engine automatically responds to user interaction data; the user interaction data includes at least one of user questions, user health data, and user treatment data.

2. The peritoneal dialysis question-and-answer method based on knowledge graphs according to claim 1, characterized in that, The entity extraction operation for the aforementioned medical-related data includes: Entity extraction is performed on the unstructured data in the aforementioned medical-related data.

3. The peritoneal dialysis question-and-answer method based on knowledge graphs according to claim 1, characterized in that, Based on the aforementioned medical-related data and the various tuple data, a knowledge graph is constructed, including: Based on the structured data in the aforementioned medical-related data, knowledge-integrated data is obtained; Entity alignment and / or referential resolution are performed on each of the aforementioned tuple data to obtain knowledge fusion data; Knowledge modeling is performed based on the knowledge integration data and the knowledge fusion data to obtain a knowledge graph; the knowledge modeling includes at least one of the following modeling processes: concept modeling, rule modeling, event modeling, and spatiotemporal modeling.

4. The peritoneal dialysis question-and-answer method based on knowledge graphs according to claim 3, characterized in that, The conceptual modeling includes: Based on the knowledge integration data and the knowledge fusion data, a concept extraction operation is performed to obtain several concepts related to peritoneal dialysis. The concepts are classified, and the hierarchical relationships between the concepts are established based on the classification results to achieve concept modeling.

5. The peritoneal dialysis question-and-answer method based on knowledge graphs according to claim 1, characterized in that, The deep inference engine automatically responds to user interaction data, including: When the user interaction data is a user question, the knowledge graph is searched based on the user question to match several keywords; Based on the keywords mentioned, the intent of the question can be analyzed; For each keyword, the corresponding entity is matched in the knowledge graph, and the matched entities are filtered according to the question intent and entity feature information to obtain the target entity; The search scope is expanded based on the relationship information of each target entity to obtain multiple target entities; All the target entities and their respective entity feature information are integrated, and the answer to the user question is generated based on the integration result.

6. The peritoneal dialysis question-and-answer method based on knowledge graphs according to claim 1, characterized in that, The deep inference engine automatically responds to user interaction data, including: When the user interaction data is user health data, the deep inference engine makes personalized recommendations based on the user health data to obtain health guidance suggestions. When the user interaction data is user treatment data, the deep reasoning engine performs personalized recommendations based on the user treatment data to obtain treatment plan suggestions.

7. The peritoneal dialysis question-and-answer method based on knowledge graphs according to claim 1, characterized in that, The system interface corresponding to the deep inference engine supports multimodal interaction methods; the multimodal interaction methods include at least one of text and voice interaction methods.

8. A peritoneal dialysis question-and-answer system based on knowledge graphs, characterized in that, The system includes: The data collection module is used to collect medical-related data from multiple data sources, including at least one of the following: medical literature, clinical guidelines, expert experience, and patient data; the medical-related data includes data related to peritoneal dialysis. The data processing module is used to perform entity extraction operations on the medical-related data to obtain several tuple data sets; each tuple data set includes an entity and corresponding entity feature information; the entity feature information includes the relationship information between the entity and other entities, the attribute information of the entity, and / or event information related to the entity; The knowledge graph construction module is used to construct a knowledge graph based on the medical-related data and each of the tuple data. The deep reasoning module is used to build a deep reasoning engine based on the knowledge graph and to automatically respond to user interaction data through the deep reasoning engine; the user interaction data includes at least one of user questions, user health data, and user treatment data.

9. A terminal, characterized in that, The terminal includes a memory and one or more processors; the memory stores one or more programs; the programs contain instructions for executing the knowledge graph-based peritoneal dialysis question-and-answer method as described in any one of claims 1-7; the processor is used to execute the programs.

10. A computer-readable storage medium storing a plurality of instructions thereon, characterized in that, The instructions are applicable to be loaded and executed by a processor to implement the steps of the knowledge graph-based peritoneal dialysis question-and-answer method as described in any one of claims 1-7.

Images