Spinal arthritis medical information data processing system and method based on agent learning model

Abstract

The invention discloses a spinal arthritis medical information data processing system and method based on an agent learning model, and the system comprises a pre-inquiry agent module which is used for responding to pre-inquiry interaction data of a patient to generate an initial medical record; the medical data processing agent module is used for acquiring the initial medical record and the heterogeneous medical detection data of the patient, and extracting medical entity information data related to the spinal arthritis from the initial medical record and the heterogeneous medical detection data by utilizing a BERT-BiLSTM-CRF hybrid model; determining an entity information link formed by the medical entity information data by using an EMR-Graph entity relation graph model; the medical reasoning agent module is used for performing reasoning analysis on an entity information link by adopting an LSTM-TPP model to obtain an initial reasoning probability between a patient and the spondyloarthritis, and can fully consider a link relationship with medical entity information related to the spondyloarthritis in a disease probabilistic reasoning process; therefore, the accuracy of disease probabilistic reasoning is effectively improved.

Description

TECHNICAL FIELD

[0001] The present application generally relates to the technical field of medical artificial intelligence, in particular to a spondylarthritis medical information data processing system and method based on an agent learning model, and especially to a multi-agent collaborative data processing system and method for spondylarthritis diagnosis and treatment. BACKGROUND

[0002] Axial spondyloarthritis (axSpA) is a chronic inflammatory arthritis mainly manifested as axial spondyloarthritis.

[0003] With the development of medical platforms, related technologies have been developed to provide remote interactive medical platforms for axSpA patients. However, due to the data heterogeneity and unstructuredness of medical data, the related technologies usually only perform inference analysis on certain medical data, such as deep analysis of magnetic resonance imaging (MRI) or computed tomography (CT) of the sacroiliac joint or spinal joint of the patient to achieve more accurate probabilistic reasoning. However, although the probabilistic reasoning accuracy of single-type medical data is high, it lacks correlation analysis between multiple medical data, which is easy to cause misjudgment between similar diseases. SUMMARY

[0004] In view of the above-mentioned defects or deficiencies in the prior art, it is desirable to provide a spondylarthritis medical information data processing system and method based on an agent learning model, which can fully consider the link relationship between spondylarthritis-related medical entity information during disease probability reasoning, thereby effectively improving the accuracy of disease probability reasoning.

[0005] In a first aspect, an embodiment of the present application provides a spondylarthritis medical information data processing system based on an agent learning model, comprising:

[0006] A pre-diagnosis agent module is configured to generate an initial medical record in response to pre-diagnosis interaction data of a patient;

[0007] A medical data processing agent module is configured to obtain the initial medical record and heterogeneous medical detection data of the patient, extract spondylarthritis-related medical entity information data from the initial medical record and the heterogeneous medical detection data using a BERT-BiLSTM-CRF hybrid model, and determine an entity information link formed by the medical entity information data using an EMR-Graph entity relationship graph model;

[0008] The medical reasoning agent module is configured to perform inference analysis on the entity information link by using an LSTM-TPP model to obtain an initial reasoning probability between the patient and the spondyloarthritis.

[0009] In some embodiments, the medical data processing agent module is further configured to:

[0010] The medical entity information data related to spondyloarthritis extracted from the initial medical record and the heterogeneous medical detection data is respectively represented by a vector to obtain an original entity representation of each medical entity information data.

[0011] The original entity representation of each medical entity information data is projected into a corresponding relationship space based on a plurality of preset entity relationships.

[0012] The entity information link is generated based on the entity relationships between at least three medical entity information data.

[0013] In some embodiments, the medical reasoning agent module is further configured to:

[0014] Based on the entity information link formed by the medical entity information data, a symptom evolution prediction is performed on the patient to obtain a spondyloarthritis typical symptom evolution probability.

[0015] Based on the spondyloarthritis typical symptom evolution probability, an initial reasoning probability between the patient and the spondyloarthritis is determined.

[0016] In some embodiments, the spondyloarthritis typical symptoms include a plurality of, and the medical reasoning agent module is further configured to:

[0017] Based on the entity information link formed by the medical entity information data, an evolution prediction is performed on each typical symptom to obtain a symptom prediction probability corresponding to each typical symptom.

[0018] For each typical symptom, when the corresponding symptom prediction probability is greater than or equal to a preset threshold, the evolution probability of the typical symptom is determined as a first probability value.

[0019] Based on the sum of the evolution probabilities of a plurality of typical symptoms, an initial reasoning probability between the patient and the spondyloarthritis is determined.

[0020] In some embodiments, the medical reasoning agent module is further configured to:

[0021] When the sum of the evolution probabilities of the typical symptoms is greater than or equal to a reasoning threshold, the initial reasoning probability is determined as a third probability value, and when the sum of the evolution probabilities of the typical symptoms is less than the reasoning threshold, the initial reasoning probability is determined as a fourth probability value.

[0022] In some embodiments, the medical reasoning intelligent agent module is further configured to:

[0023] optimize the LSTM-TPP model in response to a doctor modifying the initial reasoning probability to obtain a target reasoning probability.

[0024] In some embodiments, the medical reasoning intelligent agent module is further configured to:

[0025] generate updated questionnaire information based on entity information links formed from the medical entity information data and the target reasoning probability;

[0026] The pre-consultation intelligent agent module is further configured to:

[0027] provide the updated questionnaire information to the patient and generate an initial medical record corresponding to the updated questionnaire information in response to pre-consultation interaction data of the patient.

[0028] In a second aspect, the embodiments of the present application provide a medical information data processing method for spondyloarthritis based on an intelligent agent learning model, comprising:

[0029] generate an initial medical record in response to pre-consultation interaction data of the patient;

[0030] obtain the initial medical record and heterogeneous medical detection data of the patient, and extract medical entity information data related to spondyloarthritis from the initial medical record and the heterogeneous medical detection data using a BERT-BiLSTM-CRF hybrid model;

[0031] determine entity information links formed from the medical entity information data using an EMR-Graph entity relationship graph model;

[0032] perform reasoning analysis on the entity information links using an LSTM-TPP model to obtain an initial reasoning probability between the patient and spondyloarthritis.

[0033] In some embodiments, the method further comprises:

[0034] optimize the LSTM-TPP model in response to a doctor modifying the initial reasoning probability to obtain a target reasoning probability.

[0035] In some embodiments, the method further comprises:

[0036] generate updated questionnaire information based on the initial reasoning probability; and

[0037] provide the updated questionnaire information to the patient and generate an initial medical record corresponding to the updated questionnaire information in response to pre-consultation interaction data of the patient.

[0038] In a third aspect, an electronic device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the method described in the embodiments of the present application when executing the program.

[0039] In a fourth aspect, a computer readable storage medium is provided, which stores a computer program, and the program is executed by a processor to implement the method described in the embodiments of the present application.

[0040] In a fifth aspect, a computer program product is provided, which includes a computer program, and the computer program is executed by a processor to implement the method described in the embodiments of the present application.

[0041] The spinal arthritis medical information data processing system and method based on the agent learning model provided in the embodiments of the present application can effectively break through the entity quantity limitation of information expression in the traditional medical graph by embedding the medical information data related to spinal arthritis into the EMR-Graph entity relationship graph model and generating entity information links based on the entity relationship between the medical entity information data, so that the entity information links for subsequent reasoning analysis can contain more medical entity information data and relationship data, improving the accuracy of subsequent reasoning analysis. At the same time, the entity data types and relationship types carried by the EMR-Graph entity relationship graph model are enriched, providing the possibility of applying the EMR-Graph entity relationship graph model to various medical analysis scenarios.

[0042] Additional aspects and advantages of the application will be set forth in part in the description which follows, and in part will become apparent to those skilled in the art upon examination of the following and / or can be learned by practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0043] Other features, objects, and advantages of the application will become more apparent from the following detailed description when read in conjunction with the accompanying drawings:

[0044] Figure 1 A flowchart of the medical information data processing method based on the agent learning model for spinal arthritis provided in an embodiment of the present application is shown;

[0045] Figure 2 A principle diagram of the medical information data processing method based on the agent learning model for spinal arthritis provided in an embodiment of the present application is shown;

[0046] Figure 3 A schematic diagram of a four-node entity information link provided in an embodiment of the present application is shown;

[0047] Figure 4 Fig. 1 shows a structural schematic diagram of a medical information data processing system for rheumatoid arthritis provided by an embodiment of the present application. DETAILED DESCRIPTION

[0048] The present application will be further described below in conjunction with the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the related application, and not to limit the application. In addition, it should be noted that only the parts related to the application are shown in the drawings for ease of description.

[0049] It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0050] In order to further illustrate the technical solutions provided by the embodiments of the present application, the following will be described in detail in conjunction with the accompanying drawings and specific embodiments. Although the present application provides the method operation instruction steps as shown in the following embodiments or drawings, more or less operation instruction steps can be included in the method based on conventional or non-creative labor. The execution order of the steps is not limited to the execution order provided by the embodiments of the present application in the logical sense. The method can be executed in sequence or in parallel when the method is executed in actual processing process or by the device, as shown in the embodiments or drawings.

[0051] It should be noted that the data obtained or used by the embodiments of the present application needs to be agreed by the user, and the relevant data can be obtained after the user's authorization permission, and the data obtained or used complies with the relevant legal regulations.

[0052] Please refer to Figure 1 , Figure 1 Fig. 1 shows a structural schematic diagram of a medical information data processing system for rheumatoid arthritis provided by an embodiment of the present application. As shown in Figure 1 the method comprises:

[0053] Step 101, generating an initial medical record in response to pre-consultation interaction data of a patient.

[0054] It should be noted that the pre-consultation interaction data is the question data and reply data provided by the patient based on the preset questionnaire information. Optionally, the preset questionnaire information can be questionnaire information of fixed questions, or questionnaire information generated according to the target reasoning probability corrected by the doctor afterwards, which is not limited by the present application.

[0055] Optionally, the pre-consultation interaction data includes, but is not limited to, patient data, symptom data, and medication data. The patient data can include basic information of the patient, medical history, etc., the symptom data can be the user's feeling data, such as pain degree, pain time, morning stiffness time, etc., and the medication data can be the medication data determined by the doctor based on the previous diagnosis of the user or the medication data of other diseases, etc., which is not limited in the present application.

[0056] In a feasible embodiment, as shown in Figure 2 After obtaining the pre-consultation interaction data submitted by the patient for the questionnaire information, the pre-consultation intelligent agent module generates an initial medical record according to the pre-consultation interaction data through a preset medical record generation rule.

[0057] Optionally, the initial medical record can be generated according to the mapping relationship between the question data and the reply data in the pre-consultation interaction data. For example, the question data in the pre-consultation interaction data includes question terms such as patient name, gender, age, medical history, etc., the reply data in the pre-consultation interaction data includes reply information of the corresponding question terms, including name, gender, age, no medical history, etc., and the initial medical record includes the question data and the corresponding reply data in the pre-consultation interaction data, and is presented according to the preset position relationship.

[0058] In step 102, the initial medical record and the heterogeneous medical detection data of the patient are obtained, and the BERT-BiLSTM-CRF hybrid model is used to extract medical entity information data related to spondyloarthritis from the initial medical record and the heterogeneous medical detection data.

[0059] It should be noted that the heterogeneous medical detection data is different types of medical detection data, including but not limited to blood detection data, image detection data, etc. For example, as shown in Figure 2 The heterogeneous medical detection data can be the heterogeneous medical detection data obtained by the patient after the pre-consultation interaction. The medical detection items performed by the patient can be fixed items, such as blood detection and image detection, and can also include related detection items recommended by the pre-consultation intelligent agent module according to the initial medical record.

[0060] In a feasible embodiment, before the BERT-BiLSTM-CRF hybrid model is used to extract medical entity information data related to spondyloarthritis from the initial medical record and the heterogeneous medical detection data, it further includes: constructing an initial BERT-BiLSTM-CRF hybrid model, training the initial BERT-BiLSTM-CRF hybrid model using label samples related to spondyloarthritis, and obtaining a BERT-BiLSTM-CRF hybrid model for extracting medical entity information data related to spondyloarthritis from the initial medical record and the heterogeneous medical detection data.

[0061] The BERT-BiLSTM-CRF hybrid model includes an embedding layer BERT (Bidirectional Encoder Representation from Transformers), a bidirectional long short-term memory network BiLSTM (Bi-directional Long Short-Term Memory), and a conditional random field CRF (Conditional Random Field). The embedding layer BERT is used to convert text data into word embedding vectors, the bidirectional long short-term memory network BiLSTM is used to represent the hidden state of the relationship before and after the word, and the conditional random field CRF is used to predict the medical entity information data related to spondyloarthritis in the input data.

[0062] In a feasible embodiment, the CRF decoding layer adopts the following prediction function:

[0063]

[0064] where y = (y1,..., y L ) is the predicted label sequence, X is the input data sequence, such as the initial medical record and text type medical detection data, W crf is the label weight matrix, h t is the word before and after relationship hidden state obtained based on the BiLSTM context encoding layer, b crf is the label bias vector, Trans(y t-1 , y t ) is the label transition score.

[0065] Therefore, the trained BERT-BiLSTM-CRF hybrid model can effectively extract the medical entity information data of the patient in the initial medical record and the heterogeneous medical detection data, such as the medical entity information data (morning stiffness, pain, etc.) describing the patient's symptoms in the initial medical record, the numerical information of blood detection in the heterogeneous medical detection data, or the detection description provided by the doctor in the image detection, which is not limited by the present application.

[0066] In a preferred embodiment, the present application also uses other existing analysis models to extract medical entity information data from heterogeneous medical detection data, such as using an image analysis model to extract information data such as soft tissue changes and bone changes in patient images, so as to use multi-channel and multi-type medical entity information data for subsequent entity relationship graph establishment, and improve the comprehensiveness and reliability of medical entity information data in the entity relationship graph related to the patient.

[0067] Step 103, determining the entity information link formed by the medical entity information data by using the EMR-Graph entity relationship graph model.

[0068] It should be noted that the EMR-Graph entity relationship graph model is a medical knowledge graph of spondyloarthritis constructed based on spondyloarthritis medical information data, and is obtained by hybrid training of an electronic medical record system EMR (Electronic Medical Record) and a knowledge graph. The EMR-Graph entity relationship graph model is composed of graph nodes formed by spondyloarthritis medical information data entities, including but not limited to direct medical entity information data such as patients, symptoms, examinations, behaviors, medications, or attribute information, and indirect medical entity information data, and edges between nodes formed by relationships between spondyloarthritis medical information data entities, such as performance relationships between patients and symptoms, use relationships between patients and medications, etc.

[0069] In a feasible embodiment, the medical entity information data related to spondyloarthritis extracted from the initial medical record and the heterogeneous medical detection data are respectively represented by vectors to obtain original entity representations of the medical entity information data, the original entity representations of the medical entity information data are projected into corresponding relationship spaces based on a plurality of preset entity relationships, and at least three medical entity information data are generated into entity information links according to the entity relationships between the medical entity information data.

[0070] As shown in Figure 3 , the EMR-Graph entity relationship graph model gives a four-node entity information link related to a certain patient. The four nodes include a certain patient, A drug, morning stiffness, and night pain. The relationship between the certain patient and the A drug is a use relationship, the relationship between the certain patient and the morning stiffness is a performance relationship, the relationship between the A drug and the morning stiffness is a drug efficacy relationship (relief), and the relationship between the morning stiffness and the night pain is an association relationship.

[0071] Therefore, the embodiments of the present application embed the spondyloarthritis-related medical information data into the EMR-Graph entity relationship graph model by constructing the EMR-Graph entity relationship graph model, and generate entity information links based on the entity relationships between the medical entity information data. This effectively breaks through the entity quantity limitation of traditional medical graph information expression using triplets or quadruplets, so that the entity information links for subsequent reasoning analysis can contain more medical entity information data and relationship data, improving the accuracy of subsequent reasoning analysis. At the same time, the types of entity data and relationship types carried by the EMR-Graph entity relationship graph model are enriched, providing the possibility for the EMR-Graph entity relationship graph model to be applied to various medical analysis scenarios.

[0072] In a preferred embodiment, as Figure 2The medical data processing intelligent agent module shown can also correspondingly extract the time nodes of the medical entity information data related to spondylarthritis, such as the time node at which the patient has morning stiffness, or the interaction time at which the patient describes morning stiffness in the pre-consultation interaction data, or the medication time of the patient, and the corresponding symptom relief time, and the like. When the medical entity information data includes time information, the EMR-Graph entity relationship graph model can further form an entity information link according to the time relationship of the medical entity information data, that is, the entity information link can further represent the time relationship between the medical entity information data of the patient, so as to effectively deduce the evolution of the symptoms of the patient when the entity information link is analyzed by the LSTM-TPP model, and provide more dimensional data information for subsequent symptom evolution prediction and reasoning analysis.

[0073] In step 104, the entity information link is analyzed by the LSTM-TPP model to obtain an initial reasoning probability between the patient and spondylarthritis.

[0074] The LSTM-TPP model is a hybrid model of a long short-term memory network LSTM (Long Short-Term Memory) and a temporal point process TPP (Temporal Point Processes), which is used to predict future symptom occurrence events by the LSTM model based on the medical entity information data and the time relationship in the entity information link, and analyze discrete symptom occurrence events and frequencies by the TPP model, so as to determine the evolution probability of the typical symptoms of spondylarthritis.

[0075] In one possible embodiment, the patient can be predicted for symptom evolution based on the information link formed by the medical entity information data, to obtain the evolution probability of the typical symptoms of spondylarthritis, and based on the evolution probability of spondylarthritis, the initial reasoning probability between the patient and spondylarthritis is determined.

[0076] That is, in the embodiments of the present application, the LSTM-TPP model is used to predict the future symptom evolution of the patient based on the information link formed by the historical medical entity information data of the patient, and obtain the evolution probability of the typical symptoms of spondylarthritis, that is, the probability prediction of the natural development of the information link formed by the current medical entity information data of the patient to appear the typical symptoms of spondylarthritis, and then the initial reasoning probability between the patient and spondylarthritis is determined according to the probability of the patient appearing the typical symptoms of spondylarthritis.

[0077] Optionally, the initial probability between the patient and the spondyloarthritis is a binary probability value, for example, when the initial reasoning probability between the patient and the spondyloarthritis probability is 1, it means that the patient has a high probability of suffering from spondyloarthritis in the future, and when the initial reasoning probability between the patient and the spondyloarthritis probability is 0, it means that the patient has similar symptoms but is more likely not to suffer from spondyloarthritis.

[0078] In a feasible embodiment, the information link formed based on the medical entity information data is used to predict the symptom evolution of the patient, and the typical symptom evolution probability of spondyloarthritis is obtained, including: forming an entity information link based on the medical entity information data, respectively predicting the evolution of each typical symptom to obtain the symptom prediction probability corresponding to each typical symptom, for each typical symptom, when the corresponding symptom prediction probability is greater than a preset threshold, determining the evolution probability of the typical symptom as a first probability value, and determining the initial reasoning probability between the patient and the spondyloarthritis based on the sum of the evolution probabilities of the multiple typical symptoms.

[0079] For example, the following formula can be used to predict the evolution of each typical symptom, which is specifically represented as:

[0080]

[0081] wherein λ s (t|H0) is the evolution prediction value of the symptom s in the symptom prediction time window t according to the entity information link formed based on the medical entity information data, s is a typical symptom such as morning stiffness and sacroiliac joint pain, K is the number of clinical stages of symptom development (which can be determined according to the typical course of spondyloarthritis, for example, k=1 represents bone marrow edema, k=2 represents cartilage destruction, and k=3 represents bony ankylosis), w s,k is the progression tendency of symptom s at stage k, μ s,k is the typical progression time of symptom s at stage k, σ s,k is the individual variability of symptom progression time (such as genetic heterogeneity, treatment response difference or comorbidity influence), t is the prediction time span from the current to the occurrence of the symptom, for example, 90 days, and H0 is the entity information link formed based on the medical entity information data.

[0082] It should be noted that w s,k is determined according to the prediction weight h0 obtained by the symptom s exclusive weight matrix W s and the LSTM model, and can be specifically represented as:

[0083] w s,k =W s ·h0+b s .

[0084] wherein b sis the weight bias term of symptom s.

[0085] Furthermore, the symptom prediction probability corresponding to each typical symptom can be calculated using the following formula:

[0086]

[0087] Among them, P s is the symptom prediction probability corresponding to symptom s, λ s (t|H0) is the predicted evolution value of the entity information link formed by symptom s based on the medical entity information data in the symptom prediction time window t.

[0088] Specifically, based on the entity information link formed by the medical entity information data, the evolution of each typical symptom can be predicted separately to obtain the symptom prediction probability corresponding to each typical symptom, such as the symptom prediction probability corresponding to morning stiffness, the symptom prediction probability corresponding to sacroiliac joint pain, the symptom prediction probability corresponding to enthesitis, etc., and then determine whether the symptom prediction probability corresponding to each typical symptom is greater than a preset threshold. If it is greater than or equal to, it means that the patient will have the symptom in the future. If it is less than, it means that the patient will not have the symptom in the future. The evolution probability corresponding to the typical symptom with a prediction probability greater than or equal to the preset threshold is set to a first probability value, such as 1. It can be understood that other prediction probabilities less than the preset threshold are set. The evolution probability corresponding to the typical symptoms of a preset threshold is set to a second probability value, such as 0, and then the sum of the evolution probabilities of multiple typical symptoms is calculated, and the initial inference probability between the patient and spinal arthritis is determined based on the sum of the evolution probabilities. For example, when the sum of the evolution probabilities is greater than or equal to the inference threshold, it is determined that the patient may have multiple typical symptoms at the same time. At this time, the initial inference probability between the patient and spinal arthritis is determined to be a third probability value, such as 1, that is, the patient is more likely to suffer from spinal arthritis. When the sum of the evolution probabilities is less than the inference threshold, it is determined that the patient may have typical symptoms but are only similar to the symptoms of spinal arthritis. At this time, the initial inference probability between the patient and spinal arthritis is determined to be a fourth probability value, such as 0, that is, the patient is more likely to not suffer from spinal arthritis.

[0089] In some embodiments, feedback optimization is further performed in response to the doctor's correction of the initial probability to obtain the target inference probability.

[0090] That is, in the embodiment of the present application, the initial reasoning probability obtained by the medical reasoning intelligent agent module is predicted, which needs to be corrected by the doctor to obtain the final target reasoning probability, in other words, the initial reasoning probability is only used for prediction assistance for doctor diagnosis, and finally the doctor needs to make a final diagnosis according to the initial reasoning probability. Then, after obtaining the target reasoning probability corrected by the doctor from the initial reasoning probability, the LSTM-TPP model is further optimized for feedback to improve the reliability of the prediction of the evolution of the typical symptoms, thereby improving the reliability of the initial reasoning probability.

[0091] In a feasible embodiment, the updated questionnaire information can also be generated based on the entity information link formed by the medical entity information data and the target reasoning probability, and the updated questionnaire information is provided to the patient, and the initial medical record corresponding to the updated questionnaire information is generated in response to the pre-consultation interaction data of the patient.

[0092] That is, after the target reasoning probability is obtained by the doctor for the final diagnosis each time, the medical reasoning intelligent agent module can predict the evolution of the subsequent symptoms by using the LSTM-TPP model according to the disease progression stage and the entity information link formed by the medical entity information data that has been generated, and generate questionnaire information related to the subsequent symptoms, that is, generate updated questionnaire information, and provide the updated questionnaire information to the patient, so as to generate an initial medical record corresponding to the updated questionnaire information in response to the pre-consultation interaction data of the patient, thereby realizing an iterative process of new round of medical record generation and initial reasoning according to new pre-consultation interaction data of the patient.

[0093] Therefore, the medical information data processing method for spondyloarthritis provided in the embodiment of the present application based on the intelligent agent learning model extracts the medical entity information data related to spondyloarthritis from the initial medical record and the heterogeneous medical detection data by using the BERT-BiLSTM-CRF hybrid model, improves the accuracy of obtaining the medical entity information data, determines the entity information link formed by the medical entity information data by using the EMR-Graph entity relationship graph model, effectively enhances the association relationship between the entity information, provides longer time span and more types of entity information data for reasoning analysis of the entity information link by using the LSTM-TPP model, and improves the reliability of the prediction result of the LSTM-TPP model. At the same time, the LSTM-TPP model fully considers the irregular time interval characteristics and multi-peak characteristics of the disease symptoms, and adopts the improved log Gaussian mixture model to fit the complex time distribution to realize the reliability of the evolution prediction of the typical symptoms.

[0094] It should be noted that although the operations of the present method are described in a particular order in the drawings, this does not require or imply that the operations must be performed in this particular order, or that all illustrated operations must be performed to achieve desirable results.

[0095] Figure 3 FIG1 shows a schematic diagram of the structure of a spinal arthritis medical information data processing system based on an agent learning model provided by an embodiment of the present application. Figure 3 As shown, the spinal arthritis medical information data processing system 10 based on the intelligent learning model includes:

[0096] a pre-diagnosis agent module 11 for generating an initial medical record in response to the patient's pre-diagnosis interaction data;

[0097] a medical data processing agent module 12, configured to obtain the initial medical records and the patient's heterogeneous medical test data, extract medical entity information data related to spondyloarthritis from the initial medical records and the heterogeneous medical test data using a BERT-BiLSTM-CRF hybrid model, and determine entity information links formed by the medical entity information data using an EMR-Graph entity relationship graph model;

[0098] The medical reasoning agent module 13 is used to use the LSTM-TPP model to perform reasoning analysis on the entity information link to obtain the initial reasoning probability between the patient and spinal arthritis.

[0099] In some embodiments, the medical data processing agent module 12 is further configured to:

[0100] Performing vector representation on the medical entity information data related to spinal arthritis extracted from the initial medical records and the heterogeneous medical test data to obtain original entity representations of each medical entity information data;

[0101] Projecting the original entity representation of each medical entity information data into a corresponding relationship space based on a plurality of preset entity relationships;

[0102] The entity information link is generated based on the entity relationship between at least three medical entity information data.

[0103] In some embodiments, the medical reasoning agent module 13 is further configured to:

[0104] Based on the entity information link formed by the medical entity information data, the patient's symptom evolution is predicted to obtain the probability of typical symptom evolution of spondyloarthritis;

[0105] Based on the evolution probability of the typical symptoms of the spondyloarthritis, an initial reasoning probability between the patient and the spondyloarthritis is determined.

[0106] In some embodiments, the typical symptoms of the spondyloarthritis include a plurality, and the medical reasoning intelligent agent module 13 is further configured to:

[0107] Based on the entity information link formed by the medical entity information data, an evolution prediction is performed on each typical symptom respectively, to obtain a symptom prediction probability corresponding to each typical symptom;

[0108] For each typical symptom, when the corresponding symptom prediction probability is greater than or equal to a preset threshold, the evolution probability of the typical symptom is determined as a first probability value;

[0109] Based on the sum of the evolution probabilities of the plurality of typical symptoms, the initial reasoning probability between the patient and the spondyloarthritis is determined.

[0110] In some embodiments, the medical reasoning intelligent agent module 13 is further configured to:

[0111] When the sum of the evolution probabilities of the typical symptoms is greater than or equal to a reasoning threshold, the initial reasoning probability is determined as a third probability value, and when the sum of the evolution probabilities of the typical symptoms is less than the reasoning threshold, the initial reasoning probability is determined as a fourth probability value.

[0112] In some embodiments, the medical reasoning intelligent agent module 13 is further configured to:

[0113] In response to a target reasoning probability obtained by the doctor correcting the initial reasoning probability, the LSTM-TPP model is fed back and optimized.

[0114] In some embodiments, the medical reasoning intelligent agent module 13 is further configured to:

[0115] Based on the entity information link formed by the medical entity information data and the target reasoning probability, updated questionnaire information is generated;

[0116] The pre-diagnosis intelligent agent module 11 is further configured to:

[0117] The updated questionnaire information is provided to the patient, and an initial medical record corresponding to the updated questionnaire information is generated in response to pre-diagnosis interaction data of the patient.

[0118] It should be understood that the modules or modules described in the spondyloarthritis medical information data processing system 10 based on the intelligent agent learning model are based on the intelligent agent learning model Figure 1The various steps in the described methods correspond. Thus, the operations and features described above for the method also apply to the agent-based learning model-based medical information data processing system for spinal arthritis and the modules contained therein, which will not be described here. The agent-based learning model-based medical information data processing system for spinal arthritis can be pre- implemented in the browser or other secure application of the electronic device, or can be loaded into the browser or secure application thereof of the electronic device by downloading or the like. The corresponding modules in the agent-based learning model-based medical information data processing system for spinal arthritis can cooperate with the modules in the electronic device to realize the schemes of the embodiments of the present application.

[0119] In the foregoing detailed description, several modules or units mentioned are not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, a module or unit described above can be further divided into a plurality of modules or units to perform the features and functions thereof.

[0120] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architectures, functions and operation instructions of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowcharts or block diagrams can represent a module, a program segment or a part of code, which contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions noted in the blocks can also occur in different order from that noted in the accompanying drawings. For example, two blocks connected by a connecting line can actually be executed substantially in parallel, and sometimes they can be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and the combination of blocks in the block diagrams and / or flowcharts, can be implemented by a dedicated hardware-based system that performs the specified functions or operation instructions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0121] As another aspect, the present application also provides a computer readable storage medium, which can be contained in the electronic device described in the above embodiments, or can exist separately without being assembled into the electronic device. The above computer readable storage medium stores one or more programs, and the above programs are used by one or more processors to execute the agent-based learning model-based medical information data processing method for spinal arthritis of the present application.

[0122] The above description is merely exemplary of the application and of the application of the principles thereof and the application is not limited to the disclosed technical features merely in the specific combination described. Rather, the scope of the disclosure is both the technical solutions and the disclosed technical features in any possible combination, as well as any other technical solutions which, without departing from the disclosed concept, can be attained by the technical features disclosed herein or their equivalent features. For example, technical solutions formed by replacing the disclosed technical features with technical features having similar functions (not limited to the disclosed technical features) disclosed in the application.

Claims

1. A spinal arthritis medical information data processing system based on an agent learning model, characterized in that the system include: a pre-consultation agent module for generating an initial medical record in response to the patient's pre-consultation interaction data; a medical data processing agent module, configured to obtain the initial medical records and the patient's heterogeneous medical test data, extract medical entity information data related to spondyloarthritis from the initial medical records and the heterogeneous medical test data using a BERT-BiLSTM-CRF hybrid model, and determine entity information links formed by the medical entity information data using an EMR-Graph entity relationship graph model; The medical reasoning agent module is used to use the LSTM-TPP model to perform reasoning analysis on the entity information link to obtain the initial reasoning probability between the patient and spinal arthritis.

2. The spinal arthritis medical information data processing system based on the agent learning model according to claim 1 is characterized in that: The medical data processing intelligent agent module is further used to: Performing vector representation on the medical entity information data related to spinal arthritis extracted from the initial medical records and the heterogeneous medical test data to obtain original entity representations of each medical entity information data; Projecting the original entity representation of each medical entity information data into a corresponding relationship space based on a plurality of preset entity relationships; The entity information link is generated based on the entity relationship between at least three medical entity information data.

3. The spinal arthritis medical information data processing system based on the agent learning model according to claim 1 is characterized in that: The medical reasoning agent module is further used to: Based on the entity information link formed by the medical entity information data, the patient's symptom evolution is predicted to obtain the probability of typical symptom evolution of spondyloarthritis; Based on the probability of the typical symptoms of spondyloarthritis evolving, an initial inference probability between the patient and spondyloarthritis is determined.

4. The spinal arthritis medical information data processing system based on the agent learning model according to claim 3 is characterized in that: The typical symptoms of spinal arthritis include multiple symptoms, and the medical reasoning agent module is also used to: Based on the entity information link formed by the medical entity information data, each typical symptom is respectively predicted to evolve, and the symptom prediction probability corresponding to each typical symptom is obtained; For each of the typical symptoms, when the corresponding symptom prediction probability is greater than or equal to a preset threshold, determining the evolution probability of the typical symptom as a first probability value; Based on the sum of the evolution probabilities of the multiple typical symptoms, an initial inference probability between the patient and spinal arthritis is determined.

5. The spinal arthritis medical information data processing system based on the agent learning model according to claim 4 is characterized in that: The medical reasoning agent module is further used to: When the sum of the evolution probabilities of the typical symptoms is greater than or equal to the inference threshold, the initial inference probability is determined to be the third probability value; when the sum of the evolution probabilities of the typical symptoms is less than the inference threshold, the initial inference probability is determined to be the fourth probability value.

6. The spinal arthritis medical information data processing system based on the agent learning model according to claim 1 is characterized in that: The medical reasoning agent module is further used to: Feedback optimization is performed on the LSTM-TPP model in response to the target inference probability obtained by the doctor's correction of the initial inference probability.

7. The spinal arthritis medical information data processing system based on the agent learning model according to claim 6 is characterized in that: The medical reasoning agent module is further used to: generating updated questionnaire information based on the entity information link formed by the medical entity information data and the target inference probability; The pre-diagnosis agent module is further used to: The updated questionnaire information is provided to the patient, and an initial medical record corresponding to the updated questionnaire information is generated in response to the patient's pre-consultation interaction data.

8. A method for processing medical information data related to spinal arthritis based on an agent-based learning model, characterized in that: include: generating an initial medical record in response to the patient's pre-consultation interaction data; Obtaining the initial medical records and the patient's heterogeneous medical test data, and extracting medical entity information data related to spondyloarthritis from the initial medical records and the heterogeneous medical test data using a BER T-BiLSTM-CRF hybrid model; Determining the entity information link formed by the medical entity information data using the EMR-Graph entity relationship graph model; The LSTM-TPP model was used to perform reasoning analysis on the entity information link to obtain the initial reasoning probability between the patient and spondyloarthritis.

9. The method for processing medical information data of spinal arthritis based on an agent learning model according to claim 8, characterized in that: Also includes: The LSTM-TPP model is feedback optimized in response to the target inference probability obtained by the doctor's correction of the initial inference probability.

10. The method for processing medical information data of spinal arthritis based on an agent learning model according to claim 9, characterized in that: Also includes: generating updated questionnaire information based on the initial inference probability; as well as The updated questionnaire information is provided to the patient, and an initial medical record corresponding to the updated questionnaire information is generated in response to the patient's pre-consultation interaction data.

Images