A disease diagnosis scheme generation method, electronic device, and medium

Abstract

This application provides a method, electronic device, and medium for generating a disease diagnosis plan. The method includes: acquiring symptoms input by a user; obtaining macro nodes with semantically identical to the symptoms, and macro links where the macro nodes with semantically identical to the symptoms are located, based on the macro subnet; determining, based on the meso subnet, a meso node matching the macro node at the starting position of the macro link, and a meso link where the matching meso node is located, wherein the matching meso node represents the pathology of the symptoms; determining, based on the micro subnet, a micro node matching the meso node at the starting position of the meso link, and a micro link where the matching micro node is located, wherein the matching micro node represents the molecular information of the pathology; and determining a diagnosis plan based on the micro link, the meso link, and the macro link.

Description

Technical Field

[0001] This invention relates to the field of knowledge graph technology, and more specifically, to a method for generating disease diagnosis schemes, an electronic device, and a medium. Background Technology

[0002] Current mainstream medical knowledge graphs mostly focus on constructing macro-level connections between "diseases, symptoms, and drugs." Their core limitation lies in only achieving a macro-level direct mapping between clinical phenotypes and treatment methods, lacking deep support for the logic of diagnosis and treatment. This leads to significant shortcomings in clinical application: First, such single macro-level connections fail to form a complete causal logic chain, and cannot answer core scientific questions in clinical diagnosis and treatment such as "the cause of symptoms," "the mechanism of disease occurrence," and "the principle of drug action," only constructing a superficial connection mapping. Second, diagnostic solutions generated based on macro-level connections are mostly universal solutions, which are difficult to adapt to the different needs of individual clinical diagnosis and treatment, and are disconnected from the progressive process of actual clinical diagnosis and treatment, failing to provide direct and accurate decision support for clinical practice. Summary of the Invention

[0003] In view of this, the purpose of this application is to provide a method for generating disease diagnosis schemes, which can improve the problem that the prior art can only construct superficial association mappings.

[0004] To achieve the above technical objectives, the technical solution adopted in this application is as follows:

[0005] In a first aspect, embodiments of this application provide a method for generating a disease diagnosis plan. The method is based on a preset knowledge graph, which includes a macro subnetwork, a meso subnetwork, and a micro subnetwork. The macro subnetwork includes several macro links, each macro link including several macro nodes connected sequentially in a first preset order. The meso subnetwork includes several meso links, each meso link including several meso nodes connected sequentially in a second preset order. The micro subnetwork includes several micro links, each micro link including several micro nodes connected sequentially in a third preset order. The method includes:

[0006] Obtain the symptoms input by the user;

[0007] Based on the macro subnet, obtain the macro nodes that have the same semantic meaning as the symptoms, and the macro links where the macro nodes with the same semantic meaning as the symptoms are located;

[0008] Based on the meso-level subnet, determine the meso-level node that matches the macro-level node at the starting position of the macro-level link, and the meso-level link where the matching meso-level node is located. The matching meso-level node represents the pathology of the symptom.

[0009] Based on the micro subnet, determine the micro node that matches the meso node at the starting position of the meso link, and the micro link where the matching micro node is located, wherein the matching micro node represents the molecular information of the pathology.

[0010] A diagnostic plan is determined based on the micro-link, meso-link, and macro-link in which the device is located.

[0011] Furthermore, the preset knowledge graph includes a first mapping table, which includes a number of pre-established correspondences between macro nodes and meso nodes and a first weight;

[0012] The step of determining, based on the meso-level subnet, the meso-level node matching the macro-level node at the starting position of the macro-level link, and the meso-level link where the matching meso-level node is located, includes:

[0013] Based on the first mapping table, determine the meso-level node corresponding to the macro-level node at the starting position;

[0014] When the number of corresponding meso nodes is greater than or equal to 2, and the difference between the largest first weight and the remaining first weights is less than or equal to a first preset difference, a first detection prompt is issued based on the corresponding meso node. After obtaining the first detection result corresponding to the first detection prompt, the matching meso node and the meso link where the matching meso node is located are determined.

[0015] Furthermore, after issuing the first detection prompt based on the corresponding meso-level node, the method further includes:

[0016] Upon receiving the first detection result, a pathology with the same semantic meaning as the first detection result is obtained;

[0017] Increase the first weight of the mesoscopic node corresponding to the pathology with the same semantics as the first detection result, so as to improve the reliability of the association between the macroscopic node and the corresponding mesoscopic node.

[0018] Furthermore, increasing the first weight of the mesoscopic nodes corresponding to the semantically identical pathologies in the first detection result includes:

[0019] Based on the number of historical first detection results, the time of the most recent historical first detection result, and the type of the first detection result;

[0020] Based on the preset model, determine the increased first weight;

[0021] The preset model is:

[0022]

[0023] This represents the first weight after the increase;

[0024] The contribution coefficient is obtained based on the type of the first detection result, within a pre-established correspondence between contribution levels and contribution coefficients.

[0025] This represents the sample size correction coefficient, calculated based on the number of historical first detection results, where the number of historical first detection results is related to... Positive correlation;

[0026] This represents the time decay coefficient, calculated based on the time of the most recent historical first detection result. It is negatively correlated with the time of the most recent historical first test result.

[0027] Furthermore, the method also includes:

[0028] Receive academic evidence input from users;

[0029] When the semantic representation of the academic evidence matches the macro node at the starting position of the macro link with the meso node, an increase in the corresponding first weight is determined based on the credibility score of the academic evidence. The credibility score is determined based on the type of the academic evidence and the publication time of the academic evidence. When the credibility score is lower than the score threshold, the increase is 0. When the credibility score is greater than or equal to the score threshold, the increase is positively correlated with the credibility score.

[0030] Furthermore, determining the diagnostic plan based on the micro-link, the meso-link, and the macro-link includes:

[0031] Based on the symptoms of the macro-level pathway in which it is located, determine the initial diagnosis and treatment path;

[0032] Based on the mesoscopic and macroscopic pathways in which the patient resides, the initial diagnosis and treatment path is revised.

[0033] The diagnostic scheme is obtained based on the revised initial treatment pathway.

[0034] Furthermore, the preset knowledge graph includes a second mapping table, which includes several pre-established correspondences between micro-nodes and meso-nodes and a second weight;

[0035] Based on the micro-subnet, determine the micro-node matching the meso-node at the starting position of the meso-link, and the micro-link where the matching micro-node is located, including:

[0036] Based on the second mapping table, determine the micro node corresponding to the meso node at the starting position;

[0037] When the number of corresponding micro-nodes is greater than or equal to 2, and the difference between the largest second weight and the remaining second weights is less than or equal to the second preset difference, a second detection prompt is issued based on the corresponding micro-node, so that the matching micro-node can be determined after obtaining the second detection result corresponding to the second detection prompt.

[0038] Furthermore, after issuing the second detection prompt based on the corresponding micro-node, the method further includes:

[0039] Upon receiving the second detection result, molecular information with the same semantics as the second detection result is obtained;

[0040] Increase the second weight of the micro-nodes corresponding to the semantically identical molecular information in the second detection result to improve the correlation between the meso-nodes and the corresponding micro-nodes.

[0041] Secondly, embodiments of this application also provide an electronic device, which includes a processor and a memory coupled to each other. The memory stores a computer program, and when the computer program is executed by the processor, the electronic device performs the above-described method.

[0042] Thirdly, embodiments of this application also provide a computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, which, when run on a computer, causes the computer to perform the above-described method.

[0043] The invention employing the above technical solution has the following advantages:

[0044] In the technical solution provided in this application, after obtaining the symptoms input by the user, the macro subnetwork matches semantically consistent macro nodes and corresponding macro links to clarify the initial direction of clinical investigation. Subsequently, based on the meso subnetwork, the macro symptoms are anchored to the specific organ pathological state by associating the starting nodes of the macro links with the meso subnetwork corresponding to the meso nodes and meso links. Then, the micro subnetwork matches the starting nodes of the meso links with the micro nodes and micro links corresponding to the micro links to further trace the root molecular mechanism of pathological changes and clarify the molecular abnormalities that cause the pathology. Finally, the complete causal logic of the micro, meso, and macro three-layer links is integrated, and combined with the clinical guideline nodes and treatment path nodes in the macro subnetwork, a diagnostic plan that combines scientific causal support and clinical operability is generated. This solution, by building and matching multi-scale knowledge graph association links of macro, meso, and micro levels, makes up for the shortcomings of traditional medical knowledge graphs that only focus on macro clinical element associations and do not cover pathological and molecular level associations. Attached Figure Description

[0045] This application can be further illustrated by the non-limiting embodiments given in the accompanying drawings. It should be understood that the following drawings only illustrate some embodiments of this application and should not be considered as limiting the scope. For those skilled in the art, other related drawings can be obtained from these drawings without any inventive effort.

[0046] Figure 1 A flowchart illustrating the method for generating a disease diagnosis scheme provided in this application embodiment.

[0047] Figure 2 This is a sub-flowchart of S150 provided in an embodiment of this application. Detailed Implementation

[0048] The present application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that similar or identical parts are referred to by the same reference numerals in the drawings or description. Implementations not shown or described in the drawings are forms known to those skilled in the art. In the description of this application, terms such as "first" and "second" are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0049] This application provides an electronic device that may include a processing module and a storage module. The storage module stores a computer program, which, when executed by the processing module, enables the electronic device to perform the corresponding steps in the disease diagnosis plan generation method described below. The electronic device may be a PC, tablet computer, etc.

[0050] Please refer to Figure 1 This application also provides a method for generating a disease diagnosis plan, which can be executed by the aforementioned electronic device. The method for generating a disease diagnosis plan may include the following steps:

[0051] S110, obtain the symptoms input by the user;

[0052] S120, Based on the macro subnet, obtain the macro nodes with the same symptom semantics as the macro nodes with the same symptom semantics as the macro links where the macro nodes with the same symptom semantics are located;

[0053] S130, based on the meso-level subnet, determine the meso-level node that matches the macro-level node at the starting position of the macro-level link, and the meso-level link where the matching meso-level node is located, wherein the matching meso-level node represents the pathology of the symptom;

[0054] S140, based on the micro subnet, determine the micro node that matches the meso node at the starting position of the meso link, and the micro link where the matching micro node is located, wherein the matching micro node represents the molecular information of the pathology.

[0055] S150, determine a diagnostic scheme based on the micro-link, the meso-link, and the macro-link in which the device is located.

[0056] The method proposed in the above implementation is based on a pre-built knowledge graph, which includes macroscopic, mesoscopic, and microscopic sub-networks and a structured medical knowledge system with built-in node mapping relationships and weights. The pre-built knowledge graph is a pre-constructed "medical knowledge network" that covers all dimensions of association from molecular to clinical levels.

[0057] In this embodiment, the construction of the pre-defined knowledge graph is based on a three-tiered collaborative architecture from macro to meso to micro levels. It relies on multi-source authoritative medical data and a logical system for the diagnosis and treatment of hematological diseases. The core data sources cover clinical diagnosis and treatment data (electronic medical records from hematology departments, guidelines for the diagnosis and treatment of acute promyelocytic leukemia (APL), treatment guidelines for hematological diseases, and APL clinical pathway databases from tertiary hospitals), pathological medical data (hematological pathology textbooks, hematopoietic histology research literature, clinical bone marrow pathology testing databases, and WHO classification standards for hematological diseases), and molecular medical data (TCGA hematologic oncology subsets, hematologic oncology pathway databases, APL molecular biology experimental research results, and O...). The database includes the ncoKB Hematologic Oncology Driver Gene Database, authoritative academic resources (APL-related papers in top journals such as *Blood* and *Leukemia*, meta-analyses, evidence-based medicine databases (such as the Cochrane Library Hematologic Oncology Database)), and medical expert consensus (multidisciplinary expert consultation opinions from hematology, clinical pathology, and molecular pathology departments, and minutes of meetings on the formulation of APL diagnosis and treatment industry standards). The specific construction process is as follows: First, the core components and association rules of each subnetwork are clarified. The macro-subnetwork revolves around the clinical phenotypes and treatment processes of APL, selecting symptoms (fever with gingival bleeding, skin ecchymosis, fatigue) and diagnoses (blood...) based on data from hematology electronic medical records and APL treatment guidelines. The macro-level network, encompassing macro-level nodes such as hematologic system abnormalities (unexplained), acute promyelocytic leukemia (APL), guideline entries, and treatment pathways, constructs a macro-level link according to a first-preset sequence: "symptoms to preliminary diagnosis to confirmed diagnosis to clinical guidelines to treatment pathways." For example, it might be structured as "fever with gingival bleeding and skin ecchymosis to hematologic system abnormalities (unexplained), to APL, to the 'Chinese Guidelines for the Diagnosis and Treatment of Acute Promyelocytic Leukemia (2024 Edition)', to the retinoic acid combined with arsenic targeted therapy pathway." The meso-level sub-network focuses on hematopoietic tissue and systemic pathological changes, selecting meso-level nodes characterizing the core pathological state of APL (such as abnormal proliferation of bone marrow promyelocytic cells, coagulation dysfunction, and other factors) based on data from hematologic pathology textbooks and clinical bone marrow pathology databases. The mesoscopic pathway is constructed according to the second preset order of "pathological progression," such as "abnormal proliferation of promyelocytes in the bone marrow to abnormal release of cytoplasmic granules from promyelocytes to coagulation dysfunction." The microscopic subnetwork focuses on the molecular pathogenesis of APL, and selects microscopic nodes with characteristic molecular abnormalities (characterizing molecular information, such as PML to RARA fusion gene, PI3K / AKT pathway activation, and MAPK pathway abnormalities) based on data such as the TCGA hematologic malignancy subset and APL molecular biology literature. The microscopic pathway is constructed according to the third preset order of "molecular action," such as "PML to RARA fusion gene to PI3K / AKT pathway activation to impaired promyelocyte differentiation."Based on this, a cross-subnet mapping table is established: the first mapping table pre-determines the correspondence between macro-level nodes and meso-level nodes related to APL and the initial first weight (e.g., "fever accompanied by gingival bleeding, skin ecchymosis to abnormal proliferation of promyelocytes in bone marrow" initial weight 0.68), and the second mapping table pre-determines the correspondence between meso-level nodes and micro-level nodes and the initial second weight (e.g., "abnormal proliferation of promyelocytes in bone marrow to PML to RARA fusion gene" initial weight 0.92). The initial weights are determined by comprehensively considering the statistical results of APL clinical big data, authoritative academic research conclusions, and multidisciplinary expert consensus. At the same time, semantic matching rules based on hematologic malignancy literature and APL clinical practice summary (e.g., the symptom-to-pathological semantic matching logic of the BERT fine-tuning model), a weight update auxiliary coefficient system (e.g., the pre-determined correspondence of contribution coefficient α, sample size correction coefficient β, and time decay coefficient γ, adapted to the characteristics of APL molecular detection and clinical data), and academic evidence grading standards (focusing on the research type level of hematologic malignancy) are embedded, ultimately forming a structured knowledge system that combines APL clinical logic, hematopoietic pathology science, and molecular mechanism correlation, and supports dynamic optimization.

[0058] Based on this, the pre-defined knowledge graph includes a macro subnetwork, a meso subnetwork, and a micro subnetwork. The macro subnetwork includes several macro links, each containing several macro nodes connected sequentially in a first pre-defined order; the meso subnetwork includes several meso links, each containing several meso nodes connected sequentially in a second pre-defined order; and the micro subnetwork includes several micro links, each containing several micro nodes connected sequentially in a third pre-defined order.

[0059] The steps involved in generating a disease diagnosis protocol will be explained in detail below:

[0060] In S110, symptoms are generally described by the patient or determined by the doctor based on the patient's physical condition and professional judgment. For example, if the patient says that he or she is weak and has bleeding gums, and the doctor determines after questioning and physical examination that the patient also has persistent fever and skin ecchymosis, then fever, bleeding gums, and skin ecchymosis are considered symptoms. The user inputs the symptoms into the electronic device, and the electronic device generates a diagnostic plan based on a preset knowledge graph.

[0061] In S120, a macro node with semantic consistency with the input symptom is found through a semantic matching algorithm of the macro subnet (such as the BERT model), and the macro link to which the node belongs. For example, if the macro node is "fever with gingival bleeding and skin ecchymosis", the macro link it belongs to is from fever with gingival bleeding and skin ecchymosis to blood system abnormalities to acute promyelocytic leukemia to clinical guideline recommended treatment plan.

[0062] In S130 and S140, the first mapping table and the second mapping table are applied respectively. The first mapping table is a structured data table, which is the core association carrier connecting the macro subnetwork and the meso subnetwork in the pre-set knowledge graph. It clearly records the correspondence between "macro nodes (such as symptoms, preliminary diagnosis)" and "meso nodes (such as pathological state)". At the same time, each pair of correspondences is assigned a first weight, which is used to characterize the credibility of the association between macro nodes and meso nodes.

[0063] The first mapping table contains three key fields: macro node ID / name, meso node ID / name, and initial first weight. It addresses the correlation between "macro clinical manifestations (such as fever with gingival bleeding and skin ecchymosis) and which organ / tissue level pathological changes (such as abnormal bone marrow hematopoietic function and coagulation dysfunction)" and provides a basis for subsequent "symptom-to-pathology" link matching. At the same time, the weights distinguish the priority of different pathologies' contribution to symptoms.

[0064] The core of building the first mapping table is "based on clinical diagnosis and treatment logic, integrating multi-source authoritative data, and quantifying the credibility of the association between macro symptoms and pathology". The specific steps are as follows:

[0065] Step 1: Define the scope of related objects (node ​​screening). First, extract the core macro nodes that need to be associated with mesoscopic pathology from the macro subnet: mainly common clinical symptoms (such as fever with gingival bleeding, skin ecchymosis, fatigue) and preliminary diagnoses (such as blood system abnormalities to be investigated, bleeding tendency to be investigated). The screening criteria are "clinical manifestations that occur frequently in electronic medical records and require pathological investigation" (data source: electronic medical record database of tertiary hospitals, clinical diagnosis and treatment guidelines).

[0066] Mesoscopic nodes related to the macroscopic nodes mentioned above are extracted from the mesoscopic subnetwork: that is, the pathological state at the organ / tissue level that the symptom / preliminary diagnosis may correspond to (such as fever with gingival bleeding and skin ecchymosis corresponding to "abnormal proliferation of promyelocytes in the bone marrow" and "coagulation dysfunction caused by coagulation factor deficiency"). The screening criteria are pathology textbooks, WHO classification of hematological diseases, and clinical hematological pathology test databases.

[0067] Step 2: Collect multi-source data to support the correspondence.

[0068] Core data sources: ① Clinical big data statistics (such as bone marrow aspiration and coagulation function test results of patients with symptoms of "fever accompanied by gingival bleeding and skin ecchymosis" in the hematology department of a tertiary hospital in the past 5 years, statistically analyzing the proportion of different hematologic pathologies, and focusing on screening APL-related pathologies); ② Authoritative academic evidence (such as meta-analysis and RCT studies on "symptoms to pathological mechanisms of acute promyelocytic leukemia" in the journals "Chinese Journal of Hematology" and "Blood", and the pathological association discussion in the APL diagnosis and treatment guidelines); ③ Medical expert consensus (consultation of experts from hematology and clinical pathology departments to determine a clear causal relationship between "fever accompanied by bleeding symptoms and abnormal proliferation of bone marrow promyelocytic cells and coagulation dysfunction", excluding interference from non-hematologic pathologies).

[0069] Data processing: NLP tools (such as BERT fine-tuning model) were used to extract the "symptom-to-pathology" association statements from hematology literature and electronic medical records of APL patients. Associations without causal relationship (such as "fever with bleeding and diabetes") were removed, and core association pairs (such as "fever with gingival bleeding, skin ecchymosis to abnormal proliferation of promyelocytes in bone marrow" and "fever with bleeding to coagulation dysfunction") were retained.

[0070] Step 3: Determine the initial first weight

[0071] The core basis for weighting is: ① Clinical incidence (e.g., 68% of patients with fever accompanied by gingival bleeding and skin ecchymosis have bone marrow aspiration pathology indicating "abnormal proliferation of promyelocytes in the bone marrow", then the initial weight of this association is set at 0.68; 55% of such patients have "coagulation dysfunction", so the initial weight is set at 0.55); ② Academic evidence level (associations supported by APLRCT studies in top journals such as *Blood* and *Leukemia*, with an additional weight of 0.05 to 0.1; associations supported by observational studies in the *Chinese Journal of Hematology*, with an additional weight of 0.02 to 0.03); ③ Expert consensus (symptom-to-pathology associations recognized by ≥80% of hematology and pathology experts, such as "fever accompanied by bleeding and abnormal proliferation of promyelocytes in the bone marrow", with an additional weight of 0.03).

[0072] Weight calibration: The weights are adjusted through expert review (e.g., if the clinical incidence of a certain association is 0.63, but there is strong support from the latest RCT study in Blood, the weight is set to 0.7 after expert calibration), and finally ensure that the weight range is between 0.1 and 0.95 (below 0.1 is a weak association, such as "fever with bleeding and bone marrow megakaryopenia", which is not included in the mapping table).

[0073] Step 4: Structured Storage and Rule Embedding

[0074] The filtered "macro-node to meso-node to initial first weight" is organized into a structured data table (supporting subsequent dynamic updates of weights based on clinical test results and academic evidence of APL patients). At the same time, weight update triggering rules are embedded (such as "when 8 new APL patients are tested and verified to verify the association between 'fever with bleeding and abnormal proliferation of promyelocytes in the bone marrow', the weight update is triggered"). The data is stored in a graph database of the knowledge graph (such as NebulaGraph), and associated with the IDs of the corresponding macro-nodes (such as "fever with gingival bleeding and skin ecchymosis") and meso-nodes (such as "abnormal proliferation of promyelocytes in the bone marrow") to ensure the accuracy of the link association.

[0075] The second mapping table is the core linking carrier connecting the meso-level subnetwork and the micro-level subnetwork. It is also a structured data table that records the correspondence between "meso-level nodes (APL-related pathological states)" and "micro-level nodes (APL characteristic molecular information)" and assigns an initial second weight to each pair of correspondences (characterizing the credibility of the association between pathological and molecular abnormalities).

[0076] The second mapping table mainly includes three key fields: meso-level node ID / name, micro-level node ID / name, and initial second weight. It is designed to solve the problem of "which molecular-level abnormalities (such as fusion genes and pathway activation) cause pathological changes at the organ / tissue level (such as abnormal proliferation of promyelocytes in the bone marrow)". It provides a basis for tracing the link from "pathology to molecular root" in APL and is a core prerequisite for target matching in targeted therapy (such as retinoic acid combined with arsenic).

[0077] The core of building the second mapping table is "based on the molecular pathological mechanism of APL, integrating molecular medical data of hematological malignancies, and quantifying the credibility of the association between pathology and molecular abnormalities". The specific steps are as follows:

[0078] Step 1: Define the scope of related objects (node ​​filtering)

[0079] The core mesoscopic nodes related to the molecular mechanisms to be identified were extracted from the mesoscopic subnetwork. These mainly consist of key clinical pathological states of APL (such as abnormal proliferation of promyelocytes in the bone marrow, coagulation dysfunction, impaired differentiation of promyelocytes, and abnormal release of cytoplasmic granules from promyelocytes). The screening criteria were "pathological findings related to APL disease progression, targeted therapy targets, and prognosis" (data source: Clinical Hematologic Oncology Pathology Database, "Chinese Guidelines for the Diagnosis and Treatment of Acute Promyeloid Leukemia (2024 Edition)"). Microscopic nodes related to the above mesoscopic nodes were extracted from the microscopic subnetwork. These are the characteristic molecular abnormalities of APL corresponding to the pathology (such as fusion genes, signaling pathway activation, and abnormal molecular expression). The screening criteria were genomic databases (such as the TCGA hematologic oncology subset, HGMD hematologic disease-related entries), APL molecular biology experimental research results, and the OncoKB hematologic oncology driver gene database.

[0080] Step 2: Collect multi-source data to support the correspondence.

[0081] Core data sources: ① Molecular pathology databases (such as the fusion gene types and detection frequencies of "abnormal proliferation of myeloid promyelocytes" samples in the TCGA hematologic malignancy subset); ② Molecular biological experimental evidence (such as cell line experiments and animal model experiments showing that "PML to RARA fusion gene drives abnormal proliferation and differentiation arrest of myeloid promyelocytes"); ③ Precision medicine guidelines (such as the "correspondence between APL pathological features and characteristic fusion genes" in the NCCN hematologic malignancy guidelines and the "Chinese APL Diagnosis and Treatment Guidelines (2024 Edition)"); ④ Molecular pathology expert consensus (experts from hematology, molecular pathology, and hematopoietic stem cell transplantation departments determine "which molecular abnormalities have a direct driving relationship with APL pathology"). Data processing: Bioinformatics tools (such as DAVID, String, Cytoscape) were used to analyze the regulatory relationship from APL pathology to molecules, removing indirect associations (such as "abnormal expression of a certain molecule has no direct driving effect on APL pathology"), and retaining core driving association pairs (such as "abnormal proliferation of bone marrow promyelocytes to PML to RARA fusion gene" and "abnormal release of cytoplasmic granules from promyelocytes to activation of the PI3K / AKT pathway").

[0082] Step 3: Determine the initial second weights

[0083] The core basis for weighting is: ① Molecular driver incidence (e.g., if 98% of samples with "abnormal proliferation of promyelocytes in the bone marrow" detect the PML to RARA fusion gene, the initial weight is set to 0.98; if 85% of samples with "abnormal release of cytoplasmic granules from promyelocytes" detect PI3K / AKT pathway activation, the initial weight is set to 0.85); ② Strength of experimental evidence (for APL molecular-pathological associations verified by both in vivo and in vitro experiments, an additional weight of 0.05 to 0.1 is added; for those supported only by in vitro experiments, an additional weight of 0.02 to 0.03 is added); ③ Guideline recommendation level (if APL diagnosis and treatment guidelines explicitly list this molecular abnormality as a core pathological driver and a target for targeted therapy, an additional weight of 0.05 is added). Weight calibration: The weights are adjusted through expert review of hematologic malignancy molecular pathology (e.g., if the incidence of a certain associated molecular factor is 0.95 and it is verified by multiple top-tier experiments published in *Blood* and *Leukemia*, the weight is set to 0.98 after expert calibration), ensuring that the weight range is between 0.1 and 0.95.

[0084] Step 4: Structured Storage and Rule Embedding

[0085] The data from "APL-related meso-level nodes to micro-level nodes to the initial second weight" is organized into a structured data table, and weight update trigger rules are embedded (such as "adding 5 new APL molecular detections to verify the association, or adding 1 APL molecular mechanism study in top journals such as Blood or Leukemia to trigger weight update"). The data is stored in a graph database and associated with the IDs of the corresponding meso- and micro-level nodes to achieve precise link association.

[0086] The specific technical solution of S130 based on the first mapping table is as follows: Based on the first mapping table, S130 can be: determining the meso-level node corresponding to the macro-level node at the starting position according to the first mapping table; when the number of corresponding meso-level nodes is greater than or equal to 2, and the difference between the largest first weight and the remaining first weights is less than or equal to a first preset difference, then issuing a first detection prompt based on the corresponding meso-level node, so as to determine the matching meso-level node and the meso-level link where the matching meso-level node is located after obtaining the first detection result corresponding to the first detection prompt. When the user inputs symptoms (such as "fever with gingival bleeding and skin ecchymosis", which is the macro-level node at the starting position), the electronic device will directly extract all the meso-level nodes bound to this symptom from the first mapping table, and quickly provide all the hematopoietic tissue and systemic pathological directions that may lead to this symptom through preset medical knowledge, replacing the process of doctors listing and investigating directions based on experience. For example, from the first mapping table, the two core meso-level pathologies corresponding to "fever with gingival bleeding and skin ecchymosis" can be directly obtained: abnormal proliferation of promyelocytes in the bone marrow and coagulation dysfunction. When there are two or more such meso-level nodes, it indicates that the symptom indeed has multiple possible pathologies (for example, fever accompanied by bleeding could be due to bone marrow hematopoietic abnormalities or a simple coagulation system problem caused by a lack of coagulation factors), creating a basis for ambiguity. The first preset difference is a pre-set "probability gap threshold" (e.g., preset to 0.05) used to determine "whether the probabilities of these pathologies are similar." For example, if the first weight for abnormal proliferation of promyelocytes in the bone marrow is 0.68, and the first weight for coagulation dysfunction is 0.65, a difference of 0.03 ≤ 0.05 means that, based on preset medical knowledge, the probabilities of these two pathologies are almost identical, and it is impossible to determine which is the true pathology based on weight alone. In such scenarios, one cannot blindly choose, otherwise the subsequent micro-level pathways and diagnostic plans will be incorrect. At this point, a first detection prompt needs to be issued, indicating the need for clinical testing. Then, combined with the first test result (clinical test result) corresponding to the first detection prompt, the matching meso-level node is determined. For example, if the candidate meso-level nodes are "abnormal proliferation of promyelocytes in the bone marrow" and "coagulation dysfunction", the system will issue prompts such as "bone marrow aspiration + bone marrow morphology analysis (to verify the proliferation status of promyelocytes in the bone marrow)" and "four coagulation function tests (to verify whether coagulation function is abnormal)". These tests can directly verify the authenticity of each candidate pathology and are tests targeting the pathological state at the meso-level (abnormalities in hematopoietic tissue function and coagulation system function), which perfectly match the definition of meso-level nodes.Because clinical test results are objective "gold standard evidence," they are closer to the patient's actual situation than the preset weights in the first mapping table. For example, if the results of bone marrow aspiration and bone marrow morphology analysis show that the proportion of promyelocytes in the bone marrow is ≥30%, supporting the pathology of "abnormal proliferation of promyelocytes in the bone marrow," and the abnormalities in the four coagulation function tests are secondary to the bone marrow pathology, then the meso-level node of "abnormal proliferation of promyelocytes in the bone marrow" will be identified as the matching node. At the same time, its corresponding meso-level link (such as "abnormal proliferation of promyelocytes in the bone marrow leading to abnormal release of promyelocyte cytoplasmic granules leading to coagulation dysfunction") will be found, preparing for subsequent investigation of the molecular-level etiology.

[0087] The technical solution of this embodiment utilizes the efficiency of a pre-defined knowledge graph while avoiding the limitations of pre-defined knowledge, ensuring the accuracy of mesoscopic pathological nodes. Mesoscopic nodes are the core bridge connecting macroscopic symptoms and microscopic molecular causes. Only when this node is selected correctly can subsequent precise diagnosis and treatment be based.

[0088] Simultaneously, upon receiving the first detection result, the first weight should be updated accordingly. This involves increasing the first weight between meso-level nodes and macro-level nodes that share the same semantics as the first detection result, and decreasing the first weight between meso-level nodes and macro-level nodes that contradict the semantics of the first detection result. Specifically:

[0089] Upon receiving the first detection result, a pathology with semantically identical meaning to the first detection result is obtained; the first weight of the mesoscopic node corresponding to the pathology with semantically identical meaning to the first detection result is increased to improve the reliability of the association between the macroscopic node and the corresponding mesoscopic node. The specific method for increasing the corresponding first weight is as follows:

[0090] Based on the number of historical first detection results, the time of the most recent historical first detection result, and the type of the first detection result;

[0091] Based on the preset model, determine the increased first weight;

[0092] The preset model is:

[0093]

[0094] This represents the first weight after the increase;

[0095] The contribution coefficient is obtained based on the type of the first detection result, within a pre-established correspondence between contribution levels and contribution coefficients. This represents the sample size correction coefficient, calculated based on the number of historical first detection results, where the number of historical first detection results is related to... Positive correlation; N represents the number of historical first test results.

[0096] This represents the time decay coefficient, calculated based on the time of the most recent historical first detection result. It is negatively correlated with the time of the most recent historical first test result.

[0097] e is the natural constant, k is the attenuation coefficient of 0.2, and t represents the time of the most recent historical first detection result in years.

[0098] In this embodiment, This is used to control the upper limit of the first weight, preventing the first weight from exceeding 1, because the first weight is "association confidence", and its maximum value is 1.

[0099] The credibility of medical evidence directly determines the magnitude of weight updates. We first pre-define the contribution coefficients corresponding to different test results (evidence types). The higher the coefficient, the greater the impact on the weight. Therefore, we can pre-establish the relationship between the type and value of the first test result, and then determine the value based on the first test result. For example:

[0100] Bone marrow pathological sections and PML to RARA fusion gene detection were performed, with a value of 0.9.

[0101] Bone marrow aspiration + bone marrow morphology analysis, coagulation function tests (four items), with a value of 0.6;

[0102] The blood routine test and hematology symptom score scale were set at 0.3; the case report, expert consensus, single case record, and expert experience judgment were set at 0.1.

[0103] The choice of indicates that the smaller the sample size, the more the impact of new detection results on the weight is "suppressed"; the larger the sample size, the stronger the persuasiveness of the new detection, and the more significant the impact, even preventing a few individual cases from causing drastic fluctuations in the weight.

[0104] The negative correlation with "the time of the most recent test" means that the credibility of older test results depreciates over time, allowing the weights to reflect the latest clinical evidence first, thus avoiding the problem of "outdated evidence continuing to affect the weights".

[0105] and The three parameters are determined by multiplication, and they correspond to different reliability conditions for the new test results. Only when the test quality is high ( Large, sufficient sample size Large), results are new enough ( The persuasive power of new evidence is strong only when all four dimensions (high, medium, and high) are simultaneously satisfied; if any one dimension is weak (e.g., low detection quality, small sample size, outdated results), it will directly reduce the overall impact, and there will be no problem of "one strong dimension overshadowing other weak dimensions". For example, even if the detection is the gold standard (high, medium, and high), =0.9), but the sample size was only 1 case ( =0.5), multiplication reduces the combined effect of the two to 0.45, which intuitively shows that "insufficient sample size weakens the persuasiveness of the gold standard test".

[0106] It is understandable that when there is only one matching meso node, the unique meso node is selected. When the difference between the largest first weight and the remaining first weights is greater than the first preset difference, the meso node with the largest first weight is selected as the matching meso node.

[0107] In this embodiment, authoritative academic research results are transformed into credibility support for the "macro-to-meta-level connections" in the knowledge graph by having users upload relevant academic evidence to their electronic devices. Specifically:

[0108] The system receives academic evidence input by the user. When the semantic representation of the academic evidence matches the macro node at the starting position of the macro link, the system determines an increase in the corresponding first weight based on the credibility score of the academic evidence. The credibility score is determined based on the type of the academic evidence and the publication time of the academic evidence. When the credibility score is lower than the score threshold, the increase is 0. When the credibility score is greater than or equal to the score threshold, the increase is positively correlated with the credibility score.

[0109] First, the system receives academic evidence (such as journal articles and research findings) provided by users. Then, it uses semantic matching tools to verify whether the core conclusions of this evidence correspond to a macro-level node (such as symptom) and a meso-level node (such as pathology) in the knowledge graph. Next, based on the authority level and timeliness of the evidence type (such as RCT studies and case reports), a credibility score is calculated. The system then determines whether the score reaches a preset threshold. If the score is insufficient, it indicates weak evidence quality or timeliness, and the first weight of the corresponding association is not adjusted. If the score meets the threshold, the weight is increased based on the score (the higher the score, the larger the increase). Finally, the first weight of the macro-to-meso-level node association is updated, allowing the credibility of the knowledge graph's associations to be combined with both clinical practice and authoritative academic research.

[0110] The credibility score calculation formula in this embodiment is as follows:

[0111]

[0112] The contribution coefficient, representing the type of academic evidence, is used to determine the contribution factor. This represents the time decay factor calculated based on the publication date. ).

[0113] This is obtained from the pre-established table below.

[0114] Academic Evidence Types and Levels Research Type Description Corresponding δ value Level I (Top-Tier Evidence) RCT studies and meta-analyses (sample size ≥ 1000) published in top journals (such as Nature and Lancet). 0.8 to 1.0 Level II (High-Quality Evidence) RCT studies and cohort studies (sample size ≥ 500) published in core journals (such as the Chinese Medical Journal) 0.6 to 0.7 Level III (General Evidence) Case-control studies and cross-sectional studies (sample size ≥ 100) published in general core journals. 0.4 to 0.5 Level IV (Limited Evidence) Observational studies and single case reports in regular journals 0.2 to 0.3 Level V (Weak Evidence) Overview and expert consensus (no original data) 0.05 to 0.15

[0115] In S140, the specific technical solution may include the following: based on the micro subnet, determining the micro node matching the meso node at the starting position of the meso link, and the micro link where the matching micro node is located, including: determining the micro node corresponding to the meso node at the starting position according to the second mapping table; when the number of the corresponding micro nodes is greater than or equal to 2, and the difference between the largest second weight and the remaining second weights is less than or equal to a second preset difference, then issuing a second detection prompt based on the corresponding micro node, so as to determine the matching micro node after obtaining the second detection result corresponding to the second detection prompt.

[0116] After receiving the second detection result, obtain molecular information that is semantically identical to the second detection result; increase the second weight of the micro-nodes corresponding to the molecular information that is semantically identical to the second detection result, so as to improve the correlation between the meso-nodes and the corresponding micro-nodes.

[0117] First, based on the pre-built second mapping table (which records the association between meso-level nodes and micro-level nodes and their second weights), all candidate micro-level nodes corresponding to the current meso-level node (such as "abnormal proliferation of promyelocytes in the bone marrow") are located. For example, the micro-level node corresponding to this pathology may contain "PML to RARA fusion gene" and "FLT3 to ITD mutation", with second weights of 0.98 and 0.95, respectively.

[0118] When the number of candidate micro-nodes is ≥2, and the difference between the maximum weight and the other weights is ≤ the second preset difference (e.g., the preset difference is 0.05, and the difference between 0.98 and 0.95 is 0.03, which meets the condition), it indicates that these molecular abnormalities are similar in probability. The preset weights alone cannot determine the true molecular target. Therefore, a targeted second detection prompt will be issued (e.g., "It is recommended to supplement bone marrow chromosome karyotype analysis, fluorescence in situ hybridization (FISH) to detect PML to RARA fusion genes, and next-generation sequencing to detect FLT3 to ITD mutations"). This type of detection is a molecular-level examination of blood tumors specifically for candidate micro-nodes, ensuring that the results can directly verify the candidate molecular abnormalities.

[0119] After obtaining the second test result, the corresponding molecular information is extracted from the result (e.g., the test shows "PML to RARA fusion gene positive"). This allows us to determine that the micro-node matching the meso-level node is the "PML to RARA fusion gene". At the same time, we can associate it with the corresponding micro-link (e.g., "PML to RARA fusion gene to PI3K / AKT pathway activation to promyelocytic cell differentiation inhibition"). This step is the core basis for subsequent precision targeted therapy for acute promyelocytic leukemia and can directly correspond to suitable treatment regimens such as retinoic acid and arsenic preparations.

[0120] Finally, based on the results of this test, the second weight corresponding to this micro-node is increased (for example, the original weight of 0.98 is increased to 0.99) to strengthen the correlation between "abnormal proliferation of promyelocytes in bone marrow to PML to RARA fusion gene". This will allow the knowledge graph to prioritize matching micro-nodes that have been validated by clinical molecular testing when it encounters the same meso-pathology in the future, thereby achieving dynamic optimization of the credibility of the meso-to-micro correlation.

[0121] Understandably, after obtaining the second detection result, the corresponding second weights can be updated according to the aforementioned preset model. Furthermore, based on academic evidence, the corresponding second weights are updated, specifically:

[0122] Receive academic evidence input from users;

[0123] When the semantic representation of the academic evidence matches the meso-level node at the starting position of the meso-level link with the micro-level node, an increase value for the corresponding second weight is determined based on the credibility score of the academic evidence. The credibility score is determined based on the type of the academic evidence and the publication time of the academic evidence. When the credibility score is lower than the score threshold, the increase value is 0. When the credibility score is greater than or equal to the score threshold, the increase value is positively correlated with the credibility score.

[0124] In this embodiment, in S150, as follows: Figure 2 As shown, the specific steps may include the following:

[0125] S151: Determine the initial treatment path based on the symptoms of the macro-link mentioned above;

[0126] S152: Based on the mesoscopic link and the macroscopic link, revise the initial diagnosis and treatment path;

[0127] S153: Based on the revised initial treatment pathway, the diagnostic scheme is obtained.

[0128] In this embodiment, the mechanism of determining the initial treatment path in S151 is as follows: First, based on the symptoms of the macro-link, the general path in the clinical guidelines / treatment specifications for hematological diseases is anchored. The macro-link records the standardized clinical logic of "symptoms to diagnosis to guideline recommendation". For example, in the macro-link corresponding to "fever with gingival bleeding and skin ecchymosis", the initial treatment path is the general path for suspected hematological abnormalities in the "Diagnostic and Efficacy Standards for Hematological Diseases" and the "Chinese Guidelines for the Diagnosis and Treatment of Acute Promyelocytic Leukemia (2024 Edition)" (such as hemostasis, antipyretics and symptomatic treatment + basic blood tests + empirical coagulation support). The core of this step is to ensure the compliance of the plan and avoid blind treatment that deviates from the clinical consensus of hematology.

[0129] The mechanism of S152 in revising the initial treatment pathway is as follows: using the established mesopathological pathway (the patient's true pathological root cause) + the patient's individual clinical characteristics in the macropathological pathway, the general pathway is adjusted to be specific to the mesopathological pathway, which is the "pathological essence" of the symptoms (e.g., it has been determined that fever with bleeding originates from "abnormal proliferation of promyelocytes in the bone marrow leading to abnormal release of cytoplasmic granules from promyelocytes leading to coagulation dysfunction," rather than simply idiopathic coagulation factor deficiency). The macropathological pathway includes the patient's specific clinical state (e.g., the initial state of "hematological abnormality to be investigated," without extramedullary infiltration). Therefore, the mismatched parts in the general pathway will be corrected: for example, the step of empirically supplementing coagulation factors will be removed and replaced with preparation for targeted therapy against abnormal proliferation of promyelocytes in the bone marrow. At the same time, the follow-up frequency will be adjusted (from the general weekly blood routine test to a blood routine test + coagulation function test every 3 days to match the dynamic monitoring needs of coagulation dysfunction).

[0130] The mechanism by which S153 generates the final diagnostic plan is to transform the modified pathway into a feasible clinical implementation plan, systematize the modified treatment, examination, follow-up, and risk control, and at the same time associate the molecular targets of the micro-link (for example, the PML to RARA fusion gene of the micro-node will be combined later to supplement the specific usage requirements of retinoic acid + arsenic). The final plan not only meets the norms of hematology clinical guidelines, but also fully adapts to the pathological state and clinical characteristics of APL patients, avoiding the problem of "treating the symptoms but not the cause" of general plans.

[0131] For example: Macroscopic link: fever accompanied by gingival bleeding and skin ecchymosis to hematological abnormalities to be investigated to acute promyelocytic leukemia (confirmed diagnosis); Mesoscopic link: abnormal proliferation of promyelocytic cells in the bone marrow to abnormal release of cytoplasmic granules from promyelocytic cells to coagulation dysfunction; Microscopic link: PML to RARA fusion gene to PI3K / AKT pathway activation to impaired promyelocytic differentiation.

[0132] S151: The initial treatment pathway is determined based on the core symptoms of the macro-link: fever accompanied by gingival bleeding and skin ecchymosis (symptoms related to acute and critical illness in hematology, with bleeding tendency). The initial treatment pathway (general type, without considering individual pathology) is obtained by matching the general guidelines of the "Chinese Guidelines for the Diagnosis and Treatment of Acute Promyelocytic Leukemia (2024 Edition)" and the "Standards for Diagnosis and Efficacy of Hematological Diseases" from the macro-subnetwork.

[0133] Symptomatic treatment: tranexamic acid for hemostasis, ibuprofen for fever reduction, and intravenous infusion of fresh frozen plasma for empirical coagulation support;

[0134] Screening and examination: Complete blood routine, coagulation function tests (four items), bone marrow aspiration and bone marrow morphology analysis within 3 days;

[0135] Empirical treatment: If coagulation function indicates factor deficiency, supplement coagulation factors VIII and IX;

[0136] Follow-up: Weekly blood routine and coagulation function tests (four items) are required.

[0137] S152: The initial treatment pathway was revised by combining the established meso-level pathway (the root cause of fever with bleeding is secondary coagulation dysfunction caused by abnormal proliferation of promyelocytes in the bone marrow, rather than a simple idiopathic coagulation factor deficiency) and macro-level pathway (the patient's condition points to an unexamined hematologic abnormality, with no initial extramedullary infiltration), to make targeted modifications to the initial pathway:

[0138] The steps of empirically supplementing coagulation factors VIII and IX have been removed (the meso-level pathway clearly shows that coagulation disorders are a pathological secondary result, and simply supplementing factors cannot solve the root problem and may easily aggravate coagulation disorders).

[0139] Adjustments to symptomatic treatment and risk control: Tranexamic acid for hemostasis and ibuprofen for fever reduction are retained, and low molecular weight heparin for anticoagulation is added (to address the high risk of DIC caused by promyelocytic granule release in the meso-scale pathway and prevent thrombosis). Coagulation support is optimized to "dynamic infusion of fresh frozen plasma + platelets based on coagulation function" (to match the clinical characteristics of coagulation disorders secondary to APL).

[0140] New pre-targeted therapy tests: Bone marrow chromosome karyotype analysis and fluorescence in situ hybridization (FISH) detection of the PML-RARA fusion gene have been added to the screening program (corresponding to abnormal proliferation of promyelocytic cells in the meso-level bone marrow, identifying characteristic molecular targets of APL in advance); Follow-up plan adjusted: The usual weekly blood routine and coagulation function tests have been changed to blood routine + four coagulation function tests every 3 days (adapting to the dynamic changes in coagulation dysfunction in the meso-level pathway and adjusting the coagulation support plan in a timely manner); Prioritization of core tests strengthened: Bone marrow aspiration and bone marrow morphology analysis have been listed as urgent tests to be completed within 24 hours (the macro-level pathway highly points to hematologic malignancies, requiring rapid determination of bone marrow pathology).

[0141] S153: Obtain the final diagnostic solution by organizing the revised path into a directly executable personalized diagnostic solution, while supplementing details with molecular target logic of the microscopic links:

[0142] Symptomatic treatment and risk control: Intravenous infusion of tranexamic acid 0.5g / time, twice daily to stop bleeding; ibuprofen 0.3g / time, taken orally to reduce fever when body temperature ≥38.5℃;

[0143] Subcutaneous injection of low molecular weight heparin 4000 IU / day for anticoagulation and prevention of DIC;

[0144] Based on complete blood count and coagulation function results, administer fresh frozen plasma (200ml each time) and apheresis platelets (platelet count <20×10⁻⁶) via dynamic transfusion. 9 / L infusion);

[0145] Precise and urgent examination: Bone marrow aspiration and bone marrow morphology analysis (to determine the promyelocyte proliferation status in the bone marrow), complete blood count and peripheral blood smear (to observe the proportion of promyelocytes) within 24 hours.

[0146] Within 72 hours, complete four coagulation function tests, bone marrow chromosome karyotype analysis, and FISH detection of the PML to RARA fusion gene (targeting APL characteristic molecular targets).

[0147] Targeted therapy plan: If the PML to RARA fusion gene test is positive, immediately initiate standard APL induction therapy, with oral all-trans retinoic acid 25 mg / (m²). 2 (d) Administered in two divided doses, combined with intravenous infusion of arsenic trioxide 10 mg / day, continuing until bone marrow remission;

[0148] Follow-up and monitoring plan: Complete blood count and four coagulation function tests will be repeated every 3 days until coagulation function returns to normal;

[0149] Bone marrow aspiration and bone marrow morphology analysis were performed on the 14th and 28th days of treatment to assess bone marrow remission.

[0150] Liver and kidney function and electrolytes are checked weekly, and adverse reactions to targeted therapy drugs are monitored.

[0151] In this embodiment, the processing module can be an integrated circuit chip with signal processing capabilities. The processing module can be a general-purpose processor. For example, the processor can be a Central Processing Unit (CPU), a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application.

[0152] The storage module can be, but is not limited to, random access memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, etc.

[0153] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the electronic device described above can be referred to the corresponding steps in the aforementioned method, and will not be elaborated further here.

[0154] This application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program that, when run on a computer, causes the computer to perform the disease diagnosis scheme generation method as described in the above embodiments.

[0155] Based on the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by hardware or by using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application can be embodied in the form of a software product. This software product can be stored in a non-volatile storage medium (such as CD to ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, electronic device, or network device, etc.) to execute the methods described in the various implementation scenarios of this application.

[0156] In the embodiments provided in this application, it should be understood that the disclosed methods can also be implemented in other ways. The method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, program segment, or part of code, which includes one or more executable instructions for implementing a specified logical function. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions. Furthermore, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0157] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A disease diagnosis protocol generation method characterized by, The method is based on a preset knowledge graph, which includes macro subnetworks, meso subnetworks, and micro subnetworks. The macro subnetwork includes several macro links, each containing several macro nodes connected in a first preset order. The meso subnetwork includes several meso links, each containing several meso nodes connected in a second preset order. The micro subnetwork includes several micro links, each containing several micro nodes connected in a third preset order. The method includes: Obtain the symptoms input by the user; Based on the macro subnet, obtain the macro nodes that have the same semantic meaning as the symptoms, and the macro links where the macro nodes with the same semantic meaning as the symptoms are located; Based on the meso-level subnet, determine the meso-level node that matches the macro-level node at the starting position of the macro-level link, and the meso-level link where the matching meso-level node is located. The matching meso-level node represents the pathology of the symptom. Based on the micro subnet, determine the micro node that matches the meso node at the starting position of the meso link, and the micro link where the matching micro node is located, wherein the matching micro node represents the molecular information of the pathology. A diagnostic plan is determined based on the micro-link, meso-link, and macro-link in which the device is located.

2. The method of claim 1, wherein, The preset knowledge graph includes a first mapping table, which includes a number of pre-established correspondences between macro nodes and meso nodes and a first weight. The step of determining, based on the meso-level subnet, the meso-level node matching the macro-level node at the starting position of the macro-level link, and the meso-level link where the matching meso-level node is located, includes: Based on the first mapping table, determine the meso-level node corresponding to the macro-level node at the starting position; When the number of corresponding meso nodes is greater than or equal to 2, and the difference between the largest first weight and the remaining first weights is less than or equal to a first preset difference, a first detection prompt is issued based on the corresponding meso node. After obtaining the first detection result corresponding to the first detection prompt, the matching meso node and the meso link where the matching meso node is located are determined.

3. The method of claim 2, wherein, After issuing the first detection prompt based on the corresponding meso-level node, the method further includes: Upon receiving the first detection result, a pathology with the same semantics as the first detection result is obtained; Increase the first weight of the mesoscopic node corresponding to the pathology with the same semantics as the first detection result, so as to improve the reliability of the association between the macroscopic node and the corresponding mesoscopic node.

4. The method of claim 3, wherein, The step of increasing the first weight of the mesoscopic nodes corresponding to the semantically identical pathology of the first detection result includes: Based on the number of historical first detection results, the time of the most recent historical first detection result, and the type of the first detection result; Based on the preset model, determine the increased first weight; The preset model is: ； denotes the increased first weight; contribution coefficient, based on the type of the first detection result, in a correspondence relationship between contribution levels and contribution coefficients that is set in advance; This represents the sample size correction coefficient, calculated based on the number of historical first detection results, where the number of historical first detection results is related to... Positive correlation; This represents the time decay coefficient, calculated based on the time of the most recent historical first detection result. It is negatively correlated with the time of the most recent historical first test result.

5. The method according to claim 2, characterized in that, The method further includes: Receive academic evidence input from users; When the semantic representation of the academic evidence matches the macro node at the starting position of the macro link with the meso node, an increase in the corresponding first weight is determined based on the credibility score of the academic evidence. The credibility score is determined based on the type of the academic evidence and the publication time of the academic evidence. When the credibility score is lower than the score threshold, the increase is 0. When the credibility score is greater than or equal to the score threshold, the increase is positively correlated with the credibility score.

6. The method according to claim 1, characterized in that, The step of determining a diagnostic plan based on the micro-link, the meso-link, and the macro-link in which the diagnosis is performed includes: Based on the symptoms of the macro-level pathway in which it is located, determine the initial diagnosis and treatment path; Based on the mesoscopic and macroscopic pathways in which the patient resides, the initial diagnosis and treatment path is revised. The diagnostic scheme is obtained based on the revised initial treatment pathway.

7. The method according to claim 1, characterized in that, The preset knowledge graph includes a second mapping table, which includes a number of pre-established correspondences between micro-nodes and meso-nodes and a second weight. Based on the micro-subnet, determine the micro-node matching the meso-node at the starting position of the meso-link, and the micro-link where the matching micro-node is located, including: Based on the second mapping table, determine the micro node corresponding to the meso node at the starting position; When the number of corresponding micro-nodes is greater than or equal to 2, and the difference between the largest second weight and the remaining second weights is less than or equal to the second preset difference, a second detection prompt is issued based on the corresponding micro-node, so that the matching micro-node can be determined after obtaining the second detection result corresponding to the second detection prompt.

8. The method according to claim 7, characterized in that, After issuing the second detection prompt based on the corresponding micro-node, the method further includes: Upon receiving the second detection result, molecular information with the same semantics as the second detection result is obtained; Increase the second weight of the micro-nodes corresponding to the semantically identical molecular information in the second detection result to improve the correlation between the meso-nodes and the corresponding micro-nodes.

9. An electronic device, characterized in that, The electronic device includes a processor and a memory coupled together, the memory storing a computer program that, when executed by the processor, causes the electronic device to perform the method as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when run on a computer, causes the computer to perform the method as described in any one of claims 1-8.

Images