Model training method, and hospital medical information processing method and device
By training semantic processing models and fully connected neural network models, the system recommends examination items and medical information, solving the problem of long medical treatment processes and improving medical efficiency and experience.
Patent Information
- Application Number
- CN202610084545.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-21
- Publication Date
- 2026-05-15
AI Technical Summary
The current medical treatment process is lengthy, requires multiple visits and has unpredictable timing, resulting in low efficiency and a poor experience.
By training a target semantic processing model and a fully connected neural network model, the system extracts the user's fused disease features, recommends multiple examination items, and recommends medical information to the user based on a medical knowledge base.
It shortens the time required for medical treatment, improves the efficiency and experience of medical treatment, and reduces the waiting time for registration and examination.
Smart Images
Figure CN122050776A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of smart healthcare technology, and in particular to a model training method, a hospital medical information processing method, and an apparatus. Background Technology
[0002] Currently, during the medical treatment process, the doctor first orders multiple tests based on the user's medical condition. After the user has received the results of all the tests, they go to the hospital again for a diagnosis.
[0003] Existing methods suffer from low efficiency, low level of intelligence, and poor patient experience due to lengthy procedures, multiple visits, and unpredictable timing. Summary of the Invention
[0004] In view of this, embodiments of the present invention provide a model training method, a hospital medical information processing method and apparatus, capable of receiving user information and medical condition information; extracting fused disease features of the user from the user information and medical condition information using a target semantic processing model; outputting multiple recommended examination items for the user using a target fully connected neural network model and the fused disease features; sending each recommended examination item to the user; and recommending medical information (medical department information, doctor information, etc.) to the user based on the recommended examination items and a medical knowledge base. Embodiments of the present invention, through an artificial intelligence model, recommend examination items and medical information to users, shortening the user's medical process time, improving the user's medical efficiency, and enhancing the user's medical experience.
[0005] To achieve the above objectives, according to one aspect of the present invention, a model training method is provided, comprising: acquiring user information, historical medical condition information, and historical medical visit information for training; training a semantic processing model to be trained based on the user information and the historical medical condition information, so as to extract fused disease features of the user using the semantic processing model to be trained; training a fully connected neural network model to be trained using the fused disease features, so as to output multiple recommended examination items matching the user using the fully connected neural network model to be trained; in each iteration of training, acquiring the user's actual examination items from the historical medical visit information, calculating the similarity between the recommended examination item information and the actual examination items, and obtaining a target semantic processing model and a target fully connected neural network model when the error of the similarity is determined to be lower than a set error threshold.
[0006] Optionally, training the semantic processing model to be trained based on the user information and the historical medical condition information includes: using the user information and the historical medical condition information to train the semantic processing model to be trained to obtain a first semantic processing model; and / or, obtaining domain training data from a medical domain knowledge base, and using the domain training data to train the first semantic processing model to obtain a second semantic processing model; obtaining the target semantic processing model includes: fusing the first semantic processing model and / or the second semantic processing model to obtain the target semantic processing model.
[0007] Optionally, the step of extracting the user's fused disease features using the semantic processing model to be trained includes: extracting user features corresponding to the user information and extracting disease features corresponding to the historical medical condition information using the semantic processing model to be trained; concatenating the user features and the disease features, or fusing the user features and the disease features using a cross-network; and using the concatenated features or the fused features as the fused disease features.
[0008] Optionally, the step of extracting user features corresponding to the user information and extracting disease features corresponding to the historical medical condition information using the semantic processing model to be trained further includes: performing a weight allocation operation on the disease features based on the user features through an attention mechanism.
[0009] Optionally, if it is determined that the disease feature belongs to a preset disease category, the disease feature is semantically expanded to generate expanded training data; the user information, the historical medical information, and the expanded training data are combined to train the semantic processing model to be trained.
[0010] Optionally, if the error in determining the similarity exceeds the set error threshold, an incremental training step is triggered for the semantic processing model to be trained and the fully connected neural network model to be trained.
[0011] To achieve the above objectives, according to a second aspect of the present invention, a hospital medical information processing method based on the model training method of the first aspect is provided, comprising: receiving user information and medical condition information of a user; extracting fused disease features of the user from the user information and the medical condition information using a target semantic processing model; outputting multiple recommended examination items information of the user using a target fully connected neural network model and the fused disease features; sending each of the recommended examination items information to the user; and recommending medical information to the user based on the recommended examination items and a medical knowledge base.
[0012] Optionally, the medical information includes the department and / or doctor information; recommending medical information to the user based on the recommended examination items and the medical knowledge base includes: obtaining the mapping relationship between the recommended examination items and the medical information from the medical knowledge base, determining the department and / or doctor information corresponding to the recommended examination items according to the mapping relationship; and / or, inputting the recommended examination items into a target medical network model, and outputting the medical information using the target medical network model; wherein, the target medical network model is obtained by training a fully connected network based on the set of examination items and the medical knowledge base; and recommending the medical information to the user.
[0013] To achieve the above objectives, according to a third aspect of the present invention, a model training apparatus is provided, comprising: a training data acquisition module, a semantic model training module, a network model training module, and an evaluation module; wherein, The training data acquisition module is used to acquire user information, historical medical condition information, and historical medical visit information for training purposes. The training semantic model module is used to train a semantic processing model to be trained based on the user information and the historical medical information, so as to extract the user's fused disease features using the semantic processing model to be trained. The training network model module is used to train a fully connected neural network model to be trained using the fused disease features, so as to output multiple recommended examination items information matching the user using the fully connected neural network model to be trained; The evaluation module is used to obtain the user's actual examination items from the historical medical information during iterative training, calculate the similarity between the recommended examination item information and the actual examination items, and obtain the target semantic processing model and the target fully connected neural network model when the error of the similarity is determined to be lower than a set error threshold.
[0014] Optionally, the model training device is used to train a semantic processing model to be trained based on the user information and the historical medical condition information, including: using the user information and the historical medical condition information to train the semantic processing model to be trained to obtain a first semantic processing model; and / or, obtaining domain training data from a medical domain knowledge base, and using the domain training data to train the first semantic processing model to obtain a second semantic processing model; obtaining the target semantic processing model includes: fusing the first semantic processing model and / or the second semantic processing model to obtain the target semantic processing model.
[0015] Optionally, the model training device is used to extract fused disease features of the user using the semantic processing model to be trained, including: extracting user features corresponding to the user information and extracting disease features corresponding to the historical medical condition information using the semantic processing model to be trained; concatenating the user features and the disease features, or fusing the user features and the disease features using a cross-network; and using the concatenated features or the fused features as the fused disease features.
[0016] Optionally, the model training device, used to extract user features corresponding to the user information and extract disease features corresponding to the historical medical condition information using the semantic processing model to be trained, further includes: performing a weight allocation operation on the disease features based on the user features through an attention mechanism.
[0017] Optionally, the model training device is used to semantically expand the disease features when it is determined that the disease features belong to a preset disease category, generate expanded training data, and train the semantic processing model to be trained by combining the user information, the historical disease information, and the expanded training data.
[0018] Optionally, the model training apparatus is configured to trigger an incremental training step for the semantic processing model to be trained and the fully connected neural network model to be trained if the error in determining the similarity exceeds the set error threshold.
[0019] To achieve the above objectives, according to a fourth aspect of the present invention, a hospital medical information processing apparatus is provided, characterized in that it includes: an information receiving module, an item determination module, and a recommendation module; wherein, The information receiving module is used to receive user information and medical condition information; The project determination module is used to extract the user's fused disease features from the user information and the disease information using a target semantic processing model; and to output multiple recommended examination items for the user using a target fully connected neural network model and the fused disease features. The recommendation module is used to send information on each of the recommended examination items to the user; and to recommend medical information to the user based on the recommended examination items and the medical knowledge base.
[0020] Optionally, a hospital medical information processing device is used to process medical information including the department and / or doctor information; the hospital medical information processing device is used to recommend medical information to the user based on the recommended examination items and the medical knowledge base, including: obtaining the mapping relationship between the recommended examination items and the medical information from the medical knowledge base, determining the department and / or doctor information corresponding to the recommended examination items according to the mapping relationship; and / or, inputting the recommended examination items into a trained medical network model, and outputting the medical information using the medical network model; wherein, the medical network model is obtained by training a fully connected network based on the set of examination items and the medical knowledge base; and recommending the medical information to the user.
[0021] To achieve the above objectives, according to a fifth aspect of the present invention, an electronic device for model training or information processing is provided, characterized in that it includes: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement any of the methods described above, such as the model training method or the hospital medical information processing method.
[0022] To achieve the above objectives, according to a sixth aspect of the present invention, a computer-readable medium is provided having a computer program stored thereon, characterized in that, when the program is executed by a processor, it implements any of the methods described above, such as the model training method or the hospital medical information processing method.
[0023] To achieve the above objectives, according to a seventh aspect of the present invention, a computer program product is provided, comprising a computer program, characterized in that, when executed by a processor, the computer program implements any one of the above-described model training method or hospital medical information processing method.
[0024] One embodiment of the above invention has the following advantages or beneficial effects: it can receive user information and medical condition information; extract the user's fused disease features from the user information and medical condition information using a target semantic processing model; output multiple recommended examination items for the user using a target fully connected neural network model and the fused disease features; send each recommended examination item to the user; and recommend medical information to the user based on the recommended examination items and a medical knowledge base. This embodiment of the invention recommends examination items and medical information to users through an artificial intelligence model, shortening the user's medical process time (eliminating the need for registration, obtaining examination orders, and re-registering after examination), improving the user's medical efficiency, and enhancing the user's medical experience.
[0025] The further effects of the aforementioned unconventional alternative methods will be explained below in conjunction with specific implementation methods. Attached Figure Description
[0026] The accompanying drawings are provided to better understand the invention and are not intended to unduly limit the scope of the invention. Wherein: Figure 1 This is a schematic flowchart of a model training method provided in an embodiment of the present invention; Figure 2 is a schematic diagram of a model training process provided by an embodiment of the present invention; Figure 3A This is a flowchart illustrating a hospital medical information processing method according to an embodiment of the present invention; Figure 3B This is a schematic diagram of the processing flow of an appointment inspection system provided in one embodiment of the present invention; Figure 4A This is a schematic diagram of the structure of a model training device provided in an embodiment of the present invention; Figure 4B This is a schematic diagram of the structure of a hospital medical information processing device provided in an embodiment of the present invention; Figure 5 is an exemplary system architecture diagram in which embodiments of the present invention can be applied; Figure 6 This is a schematic diagram of the structure of a computer system suitable for implementing terminal devices or servers of the present invention. Detailed Implementation
[0027] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These details should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.
[0028] It should be noted that the collection, use, storage, sharing and transfer of user personal information involved in the technical solution of the present invention all comply with the provisions of relevant laws and regulations, and require notification to users and obtaining their consent or authorization. When applicable, user personal information is subjected to de-identification and / or anonymization and / or encryption technical processing.
[0029] like Figure 1 As shown, this embodiment of the invention provides a model training method, which may include the following steps: Step S101: Obtain user information, historical medical condition information, and historical medical visit information for training.
[0030] Step S102: Train a semantic processing model based on the user information and the historical medical condition information, so as to extract the user's fused disease features using the semantic processing model.
[0031] Step S103: Train the fully connected neural network model to be trained using the fused disease features, and then use the fully connected neural network model to be trained to output multiple recommended examination items information matching the user.
[0032] Step S104: In each iteration of training, the user's actual examination items are obtained from the historical medical information, the similarity between the recommended examination item information and the actual examination items is calculated, and if the error of the similarity is determined to be lower than the set error threshold, the target semantic processing model and the target fully connected neural network model are obtained.
[0033] Specifically, in the current hospital setting, patients typically need to first "register" to see a doctor. The doctor then prescribes relevant examinations (such as blood tests and ultrasound scans) based on the patient's condition. The patient undergoes these examinations step by step, and the appointment and examination time varies depending on the type of examination. Therefore, there is a significant gap between the time it takes for the results to be available and the time it takes for the first visit. After receiving all the examination results, the patient needs to "register" again, which also usually takes some time. As a result, the entire process is time-consuming and may potentially worsen the condition, significantly impacting the patient's overall medical experience.
[0034] In view of this, embodiments of the present invention provide a method for recommending examination items based on the patient's information using an artificial intelligence model, which significantly shortens the time required for medical treatment and improves the patient's medical experience.
[0035] In this embodiment of the invention, the architecture of the artificial intelligence model includes a target semantic processing model and a target fully connected neural network model. The collaboration of these two models outputs recommended examination information for the user (i.e., the patient seeking medical treatment).
[0036] In this embodiment of the invention, model training is performed on the semantic processing model (i.e., the semantic processing model to be trained) and the fully connected neural network model (i.e., the fully connected neural network model to be trained) before using the target semantic processing model and the target fully connected neural network.
[0037] Specifically, the training process involves acquiring user information and historical medical history. User information includes basic information related to the patient's condition, such as age, gender, and weight. Historical medical history includes, for example, textual descriptions of the patient's condition, such as: "The patient experienced a squeezing pain behind the sternum after exertion one day ago, had a fever for one day, and had a cough." It can be understood that user information and historical medical history are a set of training data that are associated, meaning that user information and historical medical history have a strong correlation.
[0038] Furthermore, the following methods can be used to train the semantic processing model based on the user information and the historical medical information: The first method: using the user information and historical medical information, train the semantic processing model to be trained to obtain the first semantic processing model; Specifically, the semantic processing model to be trained in this embodiment of the invention can be a BERT model. BERT (Bidirectional Encoder Representations from Transformers) achieves bidirectional semantic understanding through a masked language model (MLM). This model can more accurately capture the semantics of words in specific contexts through bidirectional semantic understanding, thereby improving the semantic understanding capability of long texts corresponding to disease description data.
[0039] Furthermore, using the user information and historical medical information, the semantic processing model to be trained is used to obtain a first semantic processing model; wherein, the first semantic processing model can be used as the target semantic processing model.
[0040] The second method involves obtaining domain training data from a medical domain knowledge base and using the domain training data to train the first semantic processing model to obtain a trained second semantic processing model.
[0041] Preferably, after training the first semantic processing model, a second semantic processing model is further trained based on domain training data in a medical domain knowledge base, and this second semantic processing model is used as the target semantic processing model. The second semantic processing model is a pre-trained BioBERT model (BiomedicalBidirectional Encoder Representations from Transformers), which, through the medical domain knowledge base, can more accurately extract clinical semantic association features, improving the domain-specific professionalism of disease descriptions.
[0042] The third method is to fuse the first semantic processing model and / or the second semantic processing model to obtain the target semantic processing model.
[0043] Specifically, a weighted fusion of the outputs of the first semantic processing model and the second semantic processing model can be performed to obtain the target semantic processing model.
[0044] The step of training the semantic processing model to be trained based on the user information and the historical medical condition information includes: using the user information and the historical medical condition information to train the semantic processing model to be trained to obtain a first semantic processing model; and / or, obtaining domain training data from a medical domain knowledge base, and using the domain training data to train the first semantic processing model to obtain a second semantic processing model; the step of obtaining the target semantic processing model includes: fusing the first semantic processing model and / or the second semantic processing model to obtain the target semantic processing model.
[0045] Further, the step of extracting the user's fused disease features using the semantic processing model to be trained includes: extracting user features corresponding to the user information and extracting disease features corresponding to the historical medical condition information using the semantic processing model to be trained; concatenating the user features and the disease features, or fusing the user features and the disease features using a cross-network; and using the concatenated features or the fused features as the fused disease features.
[0046] Specifically, before extracting user features or disease features, Natural Language Processing (NLP) can be used to segment and identify entities (such as symptoms and locations) from historical medical information (text descriptions) and extract medical keywords (such as "cough" and "abdominal pain").
[0047] It is understood that disease features and user features (such as gender and age) are strongly correlated. In this embodiment of the invention, the user features and the disease features are concatenated, or the user features and the disease features are fused using a cross-network; the concatenated or fused features are used as the fused disease features. For example, by concatenating disease symptoms (such as cough and abdominal pain) with user features (age and gender) or fusing them using a cross-network (CrossNet), the user's fused disease features (such as age + gender + disease features) are obtained, improving the accuracy of feature extraction and thus improving the accuracy of subsequent recommended examination items.
[0048] Furthermore, the step of extracting user features corresponding to the user information using the semantic processing model to be trained, and extracting disease features corresponding to the historical medical condition information, further includes: performing a weight allocation operation on the disease features based on the user features using an attention mechanism. Specifically, due to the strong correlation between user features (e.g., age, gender) and disease features, for example, features such as "advanced age" and "female" are strongly correlated with the probability of disease occurrence and the severity of the condition. Therefore, an attention mechanism can be used to assign higher attention weights to these features when calculating their importance. This allows the weighted features to increase their importance in subsequent tasks such as disease classification and recommended examination items, thereby improving the accuracy and reliability of recommended examination items.
[0049] Further preferably, when it is determined that the disease feature belongs to a preset disease category, the disease feature is semantically expanded to generate expanded training data; the user information, the historical medical information, and the expanded training data are then combined to train the semantic processing model to be trained. Specifically, in this embodiment of the invention, when it is determined that the disease feature belongs to a preset disease category (e.g., rare disease category), the disease feature is semantically expanded (e.g., synonym replacement, word supplementation, etc.) to generate expanded training data. This is further combined with the user information, the historical medical information, and the expanded training data to train the semantic processing model to be trained, thereby increasing the richness of the training data and improving the generalization ability of the target semantic processing model.
[0050] Furthermore, the fused disease features are used to train a fully connected neural network model to be trained, so as to use the trained target fully connected neural network model to output multiple recommended examination items information matching the user.
[0051] Furthermore, in each iteration of training, the user's actual examination items are obtained from the historical medical information, and the similarity between the recommended examination item information and the actual examination items is calculated. If the error of the similarity is determined to be lower than a set error threshold, the target semantic processing model and the target fully connected neural network model are obtained.
[0052] Specifically, the method for calculating the similarity between the recommended inspection item information and the actual inspection items can be cosine similarity, Pearson correlation coefficient, etc.
[0053] Optionally, during multiple iterations of training, the cross-entropy loss function and the Adam (Adaptive Moment Estimation) optimizer can be used to train the target fully connected neural network model to improve training performance.
[0054] like Figure 2 As shown, this embodiment of the invention provides a model training process, which may include the following steps: Step S201: Obtain training data from the basic data source (e.g., case data source).
[0055] Step S202: Extract the user's fusion disease features; that is, train the semantic processing model to be trained based on the user information and the historical medical information, so as to extract the user's fusion disease features using the semantic processing model to be trained.
[0056] Step S203: Output recommended examination item information; that is, use the fused disease features to train the fully connected neural network model to be trained, so as to use the fully connected neural network model to be trained to output multiple recommended examination item information matching the user.
[0057] Step S204: Obtain actual inspection item information for multiple users; obtain recommended inspection item information for multiple users output by the fully connected neural network model to be trained; compare the similarity between the actual inspection item information and the recommended inspection item information; determine whether the error in the similarity is less than a set error threshold. If yes, proceed to step S205; otherwise, proceed to step S206. The set error threshold may be, for example, 10% or 5%.
[0058] Step S205: Determine the trained target model and perform the data entry (storage operation) operation for the training results (such as model parameters).
[0059] Step S206: If the similarity error exceeds a set error threshold, trigger an incremental training step for the semantic processing model to be trained and the fully connected neural network model to be trained.
[0060] That is, if the error in the similarity exceeds the set error threshold, an incremental training step is triggered for the semantic processing model to be trained and the fully connected neural network model to be trained.
[0061] Steps S201 to S206 compare the actual examination information prescribed by the doctor with the recommended examination information output by the model. If the error is greater than 10% (a set error threshold), incremental learning of the model is triggered (e.g., using an online learning algorithm). This triggers incremental training steps for the semantic processing model and the fully connected neural network model to be trained. The model parameters can be updated without manual intervention, which improves the automation and optimization efficiency of model training.
[0062] like Figure 3A As shown in the figure, this embodiment of the invention provides a hospital medical information processing method, which may include the following steps: Step S301: Receive user information and medical condition information; Step S302: Extract the user's fused disease features from the user information and the disease information using the target semantic processing model; Step S303: Using the target fully connected neural network model and the fused disease features, output multiple recommended examination items for the user; send each of the recommended examination items to the user.
[0063] Step S304: Based on the recommended examination items and the medical knowledge base, recommend medical information to the user.
[0064] Specifically, steps S301 to S304 are processes of using the target semantic processing model and the target fully connected neural network model trained through steps S101 to S103 to recommend examination items to the user, and recommending medical information to the user based on the recommended examination items and the medical knowledge base.
[0065] Further, the medical information includes information about the medical department and / or doctor; the step of recommending medical information to the user based on the recommended examination items and the medical knowledge base includes: obtaining the mapping relationship between the recommended examination items and the medical information from the medical knowledge base, determining the medical department and / or doctor information corresponding to the recommended examination items according to the mapping relationship; and / or, inputting the recommended examination items into a trained target medical network model, and using the target medical network model to output the medical information; wherein, the target medical network model is obtained by training a fully connected network based on the set of examination items and the medical knowledge base; and recommending the medical information to the user.
[0066] The medical knowledge base can contain detailed information about each department of the hospital, as well as information about the doctors in each department (including doctor profiles, available appointment times, etc.), and information about one or more examination items corresponding to each department (i.e., a set of examination items).
[0067] By recommending specific examination items and combining them with a medical knowledge base, the system further recommends medical departments and doctors to users, thereby improving their medical experience. Preferably, the medical knowledge base can also be used to help users find information such as available appointment times for doctors.
[0068] like Figure 3B As shown, this embodiment of the invention provides a processing flow for an appointment-based medical examination system, which may include the following steps: Step S301A: The user (patient) accesses the appointment examination system. This appointment examination system can be deployed on a medical client or on a device client within the hospital, providing an interface for user interaction through one or more clients.
[0069] Step S302A: Determine whether the user already has an item to be inspected in the appointment inspection system (i.e., the inspection item is known). If yes, proceed to step S304A; otherwise, proceed to step S303A.
[0070] Step S303A: Utilize the target model to output recommended examination item information, that is, call the target semantic processing model trained in this embodiment of the invention to extract the user's fused disease features from the user information and the disease information; and utilize the target fully connected neural network model and the fused disease features to output multiple recommended examination item information for the user; and execute step S305A.
[0071] Step S304A: Obtain existing inspection items.
[0072] Step S305A: Submit and schedule examination items for the user to the corresponding medical system. The medical system can be a hospital system with access to the examination scheduling system.
[0073] Step S306A: Obtain the user's examination results from the medical system.
[0074] Step S307A: Send a medical appointment reminder to the user. The medical appointment reminder may include information such as the test results being ready, as well as recommended medical information (medical department, doctor information, etc.).
[0075] The system through steps S301A to S307A can provide users with automated recommendations for examination items and guidance on medical appointment registration, which greatly improves the user's medical experience.
[0076] like Figure 4A As shown, this embodiment of the invention provides a model training device 400A, including: a training data acquisition module 401A, a semantic model training module 402A, a network model training module 403A, and an evaluation module 404A, wherein: The training data acquisition module 401A is used to acquire user information, historical medical condition information, and historical medical visit information for training. The training semantic model module 402A is used to train a semantic processing model to be trained based on the user information and the historical medical information, so as to extract the user's fused disease features using the semantic processing model to be trained. The training network model module 403A is used to train a fully connected neural network model to be trained using the fused disease features, so as to output multiple recommended examination items information matching the user using the fully connected neural network model to be trained. The evaluation module 404A is used to obtain the user's actual examination items from the historical medical information during iterative training, calculate the similarity between the recommended examination item information and the actual examination items, and obtain the target semantic processing model and the target fully connected neural network model when the error of the similarity is determined to be lower than a set error threshold.
[0077] like Figure 4B As shown, this embodiment of the invention provides a hospital medical information processing device 400B, including: an information receiving module 401B, an item determination module 402B, and a recommendation module 403B; wherein, The information receiving module 401B is used to receive user information and medical condition information of the user. The project determination module 402B is used to extract the user's fused disease features from the user information and the disease information using a target semantic processing model; and to output multiple recommended examination items for the user using a target fully connected neural network and the fused disease features. The recommendation module 403B is used to send information on each of the recommended examination items to the user; and to recommend medical information to the user based on the recommended examination items and the medical knowledge base.
[0078] This invention also provides an electronic device for a model training method or a hospital medical information processing method, comprising: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method provided in any of the above embodiments.
[0079] This invention also provides a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the method provided in any of the above embodiments.
[0080] This invention also provides a computer program product, including a computer program that, when executed by a processor, implements any one of the methods described above, such as the model training method or the hospital medical information processing method.
[0081] Figure 5 An exemplary system architecture 500 is shown that can be applied to the model training method, hospital medical information processing method, model training device, or hospital medical information processing device of the present invention.
[0082] like Figure 5As shown, system architecture 500 may include terminal devices 501, 502, and 503, a network 504, and a server 505. Network 504 serves as the medium for providing communication links between terminal devices 501, 502, and 503 and server 505. Network 504 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.
[0083] Users can use terminal devices 501, 502, and 503 to interact with server 505 via network 504 to receive or send messages, etc. Various client applications, such as medical application clients, can be installed on terminal devices 501, 502, and 503.
[0084] Terminal devices 501, 502, and 503 can be various electronic devices with displays and supporting various client applications, including but not limited to smartphones, tablets, laptops, and desktop computers.
[0085] Server 505 can be a server that provides various services, such as a backend management server that supports client applications used by users through terminal devices 501, 502, and 503. The backend management server can process the received user information and medical condition information, and feed back recommended examination information and / or medical information to the terminal devices.
[0086] It should be noted that the model training method or hospital medical information processing method provided in the embodiments of the present invention is generally executed by the server 505, and correspondingly, the model training device or hospital medical information processing device is generally set in the server 505.
[0087] It should be understood that Figure 5 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.
[0088] The following is for reference. Figure 6 It shows a schematic diagram of the structure of a computer system 600 suitable for implementing a terminal device of the present invention. Figure 6 The terminal device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.
[0089] like Figure 6As shown, the computer system 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 602 or programs loaded from storage section 608 into random access memory (RAM) 603. The RAM 603 also stores various programs and data required for the operation of the system 600. The CPU 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.
[0090] The following components are connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. A removable medium 611, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 610 as needed so that computer programs read from it can be installed into storage section 608 as needed.
[0091] In particular, according to the embodiments disclosed in this invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 609, and / or installed from removable medium 611. When the computer program is executed by central processing unit (CPU) 601, it performs the functions defined above in the system of this invention.
[0092] It should be noted that the computer-readable medium shown in this invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.
[0093] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.
[0094] The modules and / or units described in the embodiments of the present invention can be implemented in software or hardware. The described modules and / or units can also be housed in a processor. For example, a processor can be described as including a module for acquiring training data, a module for training a semantic model, and a module for training a network model. The names of these modules do not necessarily limit the module itself; for example, the module for acquiring training data can also be described as "a module for acquiring user information and historical medical information for training." As another aspect, the present invention also provides a computer-readable medium, which can be included in the device described in the above embodiments or exist independently and not assembled into the device. The computer-readable medium carries one or more programs that, when executed by the device, cause the device to include: receiving user information and medical information; extracting fused disease features of the user from the user information and medical information using a target semantic processing model; outputting multiple recommended examination items for the user using a target fully connected neural network model and the fused disease features; sending each recommended examination item to the user; and recommending medical information to the user based on the recommended examination items and a medical knowledge base. This invention uses an artificial intelligence model to recommend examination items to users, shortening the time required for their medical treatment (eliminating the need for registration, obtaining examination orders, and re-registering after the examination), thus improving the efficiency and experience of medical treatment.
[0095] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.
Claims
1. A model training method, characterized in that, include: Obtain user information, historical medical records, and historical medical history for training purposes; Based on the user information and the historical medical information, a semantic processing model is trained to extract the user's fused disease features. The fused disease features are used to train a fully connected neural network model to be trained, so that the fully connected neural network model to be trained can output multiple recommended examination items information that match the user; In each iteration of training, the user's actual examination items are obtained from the historical medical information, and the similarity between the recommended examination item information and the actual examination items is calculated. If the error of the similarity is determined to be lower than a set error threshold, the target semantic processing model and the target fully connected neural network model are obtained.
2. The method according to claim 1, characterized in that, The process of training the semantic processing model based on the user information and the historical medical information includes: Using the user information and historical medical information, the semantic processing model to be trained is used to obtain the first semantic processing model; And / or, obtain domain training data from a medical domain knowledge base, and use the domain training data to train the first semantic processing model to obtain the second semantic processing model; The obtained target semantic processing model includes: The target semantic processing model is obtained by fusing the first semantic processing model and / or the second semantic processing model.
3. The method according to claim 1, characterized in that, The step of extracting the user's fused disease features using the semantic processing model to be trained includes: The user features corresponding to the user information and the disease features corresponding to the historical medical condition information are extracted using the semantic processing model to be trained. The user features and the disease features can be concatenated, or the user features and the disease features can be fused using a cross-network. The spliced or fused features are used as the fused disease features.
4. The method according to claim 3, characterized in that, The step of extracting user features corresponding to the user information and extracting disease features corresponding to the historical medical condition information using the semantic processing model to be trained further includes: The disease features are weighted based on the user features using an attention mechanism.
5. The method according to claim 3, characterized in that, Further includes: If the disease feature is determined to belong to a preset disease category, the disease feature is semantically expanded to generate expanded training data. The semantic processing model to be trained is trained by combining the user information, the historical medical information, and the expanded training data.
6. The method according to claim 1, characterized in that, Further includes: If the error in the similarity is determined to exceed the set error threshold, an incremental training step is triggered for the semantic processing model to be trained and the fully connected neural network model to be trained.
7. A hospital medical information processing method based on the model training method according to any one of claims 1 to 6, characterized in that, include: Receive user information and medical condition information from users; The target semantic processing model is used to extract the user's fused disease features from the user information and the disease information; Using a target fully connected neural network model and the fused disease features, output multiple recommended examination items for the user; send each of the recommended examination items to the user; Based on the recommended examination items and the medical knowledge base, medical information is recommended to the user.
8. The method according to claim 7, characterized in that, The medical information includes the department and / or doctor's information; The process of recommending medical information to users based on the recommended examination items and the medical knowledge base includes: Obtain the mapping relationship between the recommended examination items and medical information from the medical knowledge base, and determine the medical department and / or doctor information corresponding to the recommended examination items based on the mapping relationship; And / or, The recommended examination items are input into the target medical treatment network model, and the target medical treatment network model is used to output medical information; wherein, the target medical treatment network model is obtained by training a fully connected network based on the examination item set and the medical treatment knowledge base; The medical information mentioned above will be recommended to the user.
9. A model training device, characterized in that, include: The system includes modules for acquiring training data, training the semantic model, training the network model, and evaluation; among these, The training data acquisition module is used to acquire user information, historical medical condition information, and historical medical visit information for training purposes. The training semantic model module is used to train a semantic processing model to be trained based on the user information and the historical medical information, so as to extract the user's fused disease features using the semantic processing model to be trained. The training network model module is used to train a fully connected neural network model to be trained using the fused disease features, so as to output multiple recommended examination items information matching the user using the fully connected neural network model to be trained; The evaluation module is used to obtain the user's actual examination items from the historical medical information during iterative training, calculate the similarity between the recommended examination item information and the actual examination items, and obtain the target semantic processing model and the target fully connected neural network model when the error of the similarity is determined to be lower than a set error threshold.
10. A hospital medical information processing device, characterized in that, include: The system includes an information receiving module, a project selection module, and a recommendation module; among which, The information receiving module is used to receive user information and medical condition information; The project determination module is used to extract the user's fused disease features from the user information and the disease information using a target semantic processing model; and to output multiple recommended examination items for the user using a target fully connected neural network model and the fused disease features. The recommendation module is used to send information on each of the recommended examination items to the user; and to recommend medical information to the user based on the recommended examination items and the medical knowledge base.
11. An electronic device, characterized in that, include: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-8.
12. A computer-readable medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-8.
13. A computer program product, comprising a computer program, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-8.