Artificial intelligence-based medical data analysis method, device, equipment and medium

Abstract

The application relates to the technical field of digital medical treatment, and discloses a medical data analysis method, device, equipment and medium based on artificial intelligence, which comprises the following steps: obtaining initial symptom description text of a user to be analyzed and inputting the initial symptom description text into a pre-trained language model; the pre-trained language model comprises an inquiry behavior prediction submodel used for performing inquiry behavior prediction and a reply generation submodel used for performing medical data analysis; generating symptom description entity normalized text based on the initial symptom description text through the reply generation submodel; predicting an inquiry behavior label based on the symptom description entity normalized text through the inquiry behavior prediction submodel; and generating a medical data analysis result for the user to be analyzed based on the symptom description entity normalized text and the inquiry behavior label through the reply generation submodel. The scheme in the application can improve the accuracy of medical data analysis performed by a model.

Description

Technical Field

[0001] This invention relates to the field of digital medical technology, and in particular to a method, apparatus, computer equipment, and medium for medical data analysis based on artificial intelligence. Background Technology

[0002] In automated consultation scenarios within digital healthcare, language models are used to intelligently recognize user-inputted symptom descriptions, enabling intelligent medical data analysis. Typically, a new round of inquiry based on a user's symptom description and the subsequent medical data analysis are handled by two separate language models. This approach presents several problems. For instance, an error in one model's prediction might affect the prediction performance of the other, or there may be isolation between the two language models. Isolating these models makes it difficult to uncover hidden relationships between the inquiry and data analysis through learning, leading to relatively low accuracy in current medical data analysis methods. Summary of the Invention

[0003] This invention provides a medical data analysis method, device, equipment, and medium based on artificial intelligence to solve the technical problem of low accuracy in existing medical data analysis methods.

[0004] Firstly, an artificial intelligence-based medical data analysis method is provided, including:

[0005] The initial symptom description text of the user to be analyzed is obtained and input into a pre-trained language model. The pre-trained language model includes an inquiry behavior prediction sub-model for predicting inquiry behavior and a response generation sub-model for predicting medical data analysis results.

[0006] Based on the initial symptom description text, the response generation sub-model generates normalized text for symptom description entities.

[0007] The inquiry behavior prediction sub-model predicts inquiry behavior labels based on the normalized text of the symptom description entity, and the inquiry behavior labels are either a continue inquiry behavior label or an end inquiry behavior label.

[0008] The response generation sub-model generates medical data analysis results for the user to be analyzed based on the normalized text of the symptom description entity and the query behavior label.

[0009] Secondly, an artificial intelligence-based medical data analysis device is provided, comprising:

[0010] The acquisition module is used to acquire the initial symptom description text of the user to be analyzed and input it into the pre-trained language model. The pre-trained language model includes an inquiry behavior prediction sub-model for inquiry behavior prediction and a response generation sub-model for medical data analysis.

[0011] The first generation module is used to generate normalized text of symptom description entities based on the initial symptom description text through the response generation sub-model;

[0012] The prediction module is used to predict inquiry behavior labels based on the normalized text of the symptom description entity through the inquiry behavior prediction sub-model. The inquiry behavior label is one of the continue inquiry behavior label and the end inquiry behavior label.

[0013] The second generation module is used to generate medical data analysis results for the user to be analyzed based on the normalized text of the symptom description entity and the query behavior label through the response generation sub-model.

[0014] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described medical data analysis method.

[0015] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the above-described medical data analysis method.

[0016] The aforementioned AI-based medical data analysis method, device, computer equipment, and storage medium achieve the following: Initial symptom description text of the user to be analyzed is acquired and input into a pre-trained language model. The pre-trained language model includes an inquiry behavior prediction sub-model for predicting inquiry behavior and a response generation sub-model for medical data analysis. The response generation sub-model generates normalized symptom description entity text based on the initial symptom description text. The inquiry behavior prediction sub-model then predicts inquiry behavior labels based on the normalized symptom description entity text. Finally, the response generation sub-model generates medical data analysis results for the user to be analyzed based on the normalized symptom description entity text and the inquiry behavior labels. This allows the two models to fully explore the hidden relationship between inquiries and data analysis, improving the accuracy of the model's medical data analysis. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of an application environment for a medical data analysis and processing method according to an embodiment of the present invention;

[0019] Figure 2 This is a flowchart illustrating a medical data analysis and processing method according to an embodiment of the present invention;

[0020] Figure 3 yes Figure 1 A flowchart illustrating a specific implementation method of step S1;

[0021] Figure 4 This is a schematic diagram of a medical data analysis and processing device according to an embodiment of the present invention;

[0022] Figure 5 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention;

[0023] Figure 6 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] The artificial intelligence-based medical data analysis method provided in this invention can be applied to, for example... Figure 1In this application environment, the client communicates with the server via a network. The server can receive the initial symptom description text of the user to be analyzed from the client. The server inputs the initial symptom description text into a pre-trained language model. The pre-trained language model includes an inquiry behavior prediction sub-model for predicting inquiry behavior and a response generation sub-model for medical data analysis. The response generation sub-model generates normalized text of symptom description entities based on the initial symptom description text and the start inquiry status label. The inquiry behavior prediction sub-model predicts inquiry behavior labels based on the normalized text of symptom description entities and the start inquiry status label. The inquiry behavior label can be either a continue inquiry behavior label or an end inquiry behavior label. The response generation sub-model generates medical data analysis results for the user to be analyzed based on the normalized text of symptom description entities and the inquiry behavior label. The server obtains the medical data analysis results for the user to be analyzed and feeds them back to the client. In this invention, the initial symptom description text of the user to be analyzed is obtained and input into a pre-trained language model. The pre-trained language model includes an inquiry behavior prediction sub-model for predicting inquiry behavior and a response generation sub-model for medical data analysis. The response generation sub-model generates normalized text of symptom description entities based on the initial symptom description text. The inquiry behavior prediction sub-model then predicts inquiry behavior labels based on the normalized text of symptom description entities. Finally, the response generation sub-model generates medical data analysis results for the user to be analyzed based on the normalized text of symptom description entities and the inquiry behavior labels. This allows the two models to fully explore the hidden relationship between inquiries and data analysis, improving the accuracy of the model in medical data analysis. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. The invention will be described in detail below through specific embodiments.

[0026] Please see Figure 2 As shown, Figure 2 A flowchart illustrating an artificial intelligence-based medical data analysis method provided in an embodiment of the present invention includes the following steps:

[0027] S1: Obtain the initial symptom description text of the user to be analyzed and input it into the pre-trained language model. The pre-trained language model includes an inquiry behavior prediction sub-model for inquiry behavior prediction and a response generation sub-model for medical data analysis.

[0028] The medical data analysis method provided by this invention can be applied to intelligent medical data analysis engines such as digital medical intelligent consultation customer service or digital medical intelligent consultation assistant in various application scenarios. These medical data analysis engines are typically implemented through a server-side component, which can obtain the initial symptom description text of the user to be analyzed in real time. For example, in the field of digital healthcare, users often raise simple medical data analysis questions via telephone or online chat. It is often necessary to use a medical data analysis engine to answer these questions, thereby reducing the workload and improving the efficiency of senior doctors or specialists.

[0029] In this embodiment, the initial symptom description text is corpus information entered by the user when they need to consult a doctor to describe the symptoms of a disease. In one embodiment, the initial symptom description text can be online corpus text entered by the user to describe the symptoms of a disease. For example, the initial symptom description text could be: "I have a headache and body aches recently. Do I have a cold or some other illness?"

[0030] It is understandable that the initial symptom description text can also be text corpus information obtained by performing speech recognition on the voice information input by the user to be analyzed, which describes the symptoms of the disease.

[0031] In this embodiment, the pre-trained language model is a language model trained using a large corpus of real patient-doctor dialogues as training samples. The trained language model can then be used for medical data analysis. Internally, this language model includes an inquiry behavior prediction sub-model for predicting inquiry behavior and a response generation sub-model for medical data analysis.

[0032] In this embodiment, the inquiry behavior prediction sub-model outputs inquiry behavior labels, including a "continue inquiry" label to indicate continued medical data inquiry and a "end inquiry" label to indicate continued output of medical data analysis results or termination of the inquiry behavior. The response generation sub-model generates medical data analysis response text for medical data analysis. It is understood that the response generation sub-model can also output question response text guiding the user to conduct a new round of symptom description.

[0033] In this embodiment, the query behavior prediction sub-model and the response generation sub-model can be trained using a Transformer model, or they can be trained using a Convolutional Neural Network (CNN) or a Recurrent Neural Network (RNN), and there is no limitation on this.

[0034] In one embodiment, such as Figure 3 As shown, a specific scheme for obtaining the initial symptom description text of the user to be analyzed is provided. Step S1, that is, the step of obtaining the initial symptom description text of the user to be analyzed, specifically includes the following steps S11-S12:

[0035] S11: Obtain the text of the symptoms described by the user to be analyzed.

[0036] In this embodiment, the corpus text to be analyzed when the user describes symptoms can be online corpus text input by the user to describe disease symptoms, or online voice information input by the user to describe disease symptoms. The corpus text is obtained by converting the online voice information based on speech recognition technology. The speech recognition technology can be a speech recognition model based on recurrent neural network (RNN) or neural network language model (NLM), and no specific limitation is made here.

[0037] S12: Extract feature words describing symptoms from the corpus text to obtain the initial symptom description text.

[0038] In this embodiment of the invention, in order to more accurately obtain specific keywords in the corpus text used to characterize disease symptoms, feature words describing symptoms in the corpus text can be extracted to obtain an initial symptom description text. For example, when the corpus text is "I have a headache and body aches recently, is it a cold or something else?", specific feature words describing symptoms such as "headache", "ache", "body aches" and "cold" can be extracted to obtain an initial symptom description text containing feature words such as "headache", "ache", "body aches" and "cold".

[0039] Understandably, when extracting feature words for symptom description, a feature word library for describing symptoms can be pre-established, the words in the corpus text can be matched with each feature word in the feature word library, and the initial symptom description text can be generated based on the successfully matched feature words.

[0040] Figure 3 In the embodiment shown, by extracting feature words describing symptoms from the corpus text, an initial symptom description text is obtained. Invalid words in the corpus text that cannot reflect symptom description can be filtered out, so as to obtain keywords that can be directly used for medical data analysis and are convenient for symptom description. This can effectively reduce the amount of data processing of the language model and thus improve the efficiency of the language model in medical data analysis.

[0041] Step S2: Based on the initial symptom description text, the response generation sub-model generates normalized text for symptom description entities.

[0042] In this embodiment, the response generation sub-model performs standard entity normalization on the initial symptom description text to generate symptom description entity-normalized text. It can be understood that standard entity normalization refers to the process of transforming keywords in the initial symptom description text into keywords of standard entities, thereby facilitating subsequent recognition processing.

[0043] Optionally, in one embodiment, a mapping relationship between keywords of some non-standard entities and keywords of standard entities can be established, such as the mapping relationship between the non-standard entity "whole body" and the standard entity "limbs", and the mapping relationship between the non-standard entity "head" and the standard entity "head", thereby realizing the rapid standard entity normalization processing of keywords in the initial symptom description text.

[0044] In some embodiments, the response generation sub-model can also convert the initial symptom description text into a feature vector, and perform standard entity normalization processing on the initial symptom description text based on the feature vector to obtain symptom description entity-normalized text. In this embodiment, the response generation sub-model needs to be trained using a large number of text pairs consisting of symptom description texts of non-standard entities and symptom description texts of standard entities. For example, the initial symptom description text of non-standard entities can be "headache, cold, body aches," while the symptom description text of standard entities can be "headache, cold, body aches." By training the response generation sub-model using the above text pairs, the trained response generation sub-model can perform standard entity normalization processing on the keywords of the initial symptom description text to generate symptom description entity-normalized text. In this embodiment, compared to the scheme of performing standard entity normalization processing on the keywords in the initial symptom description text through the aforementioned mapping relationship, the generalization ability of standard entity normalization processing can be improved through model training.

[0045] Step S3: Based on the normalized text of the symptom description entity, the inquiry behavior prediction sub-model predicts the inquiry behavior label, which is either a continue inquiry behavior label or an end inquiry behavior label.

[0046] In this embodiment, the inquiry behavior prediction sub-model predicts inquiry behavior labels based on the normalized text of symptom description entities. These labels are either "continue inquiry" or "end inquiry." The inquiry behavior prediction sub-model outputs these labels, which include a "continue inquiry" label indicating continued inquiry about the disease and an "end inquiry" label indicating continued output of medical data analysis or termination of the inquiry. The inquiry behavior prediction sub-model primarily uses the normalized text of symptom description entities to predict whether to stop the inquiry behavior and outputs both a "continue inquiry" label indicating continued inquiry about the disease and an "end inquiry" label indicating continued output of medical data analysis results or termination of the inquiry to the response generation sub-model.

[0047] Step S4: Based on the normalized text of the symptom description entity and the query behavior label, the response generation sub-model generates medical data analysis results for the user to be analyzed.

[0048] In this embodiment, a response generation sub-model is used to transform the normalized text of the symptom description entity and the query behavior label to obtain a feature vector. Based on the feature vector, a medical data analysis result for the user to be analyzed is generated. This medical data analysis result is the result obtained by analyzing the symptom description of the user to be analyzed. For example, the medical data analysis result for the user's symptom description "fever of 38 degrees Celsius, runny nose" is "cold".

[0049] In one embodiment, step S4 in a scheme for generating medical data analysis results for a user to be analyzed specifically includes steps S41-S45.

[0050] Step S41: If the inquiry behavior label is an end inquiry behavior label, then the response generation sub-model generates medical data analysis results for the user to be analyzed based on the normalized text of the symptom description entity and the end inquiry behavior label.

[0051] When the inquiry behavior label is the end inquiry behavior label, the response generation sub-model will directly generate medical data analysis results for the user to be analyzed based on the normalized text of the symptom description entity and the end inquiry behavior label.

[0052] In one embodiment, a scheme is provided to directly generate medical data analysis results for the user to be analyzed by using a response generation sub-model based on the normalized text of the symptom description entity and the end-of-question behavior label. Step S41 specifically includes the following steps S411-S414.

[0053] Step S411: If the inquiry behavior label is the end inquiry behavior label, then the normalized text of the symptom description entity is used to extract features through the response generation sub-model to obtain the first symptom description features, and the end inquiry behavior label is used to extract features to obtain the first inquiry behavior features.

[0054] Step S412: Perform feature fusion on the first symptom description feature and the first inquiry behavior feature to obtain the first medical data analysis feature.

[0055] Step S413: Based on the first medical data analysis features, calculate multiple medical data prediction results for the user to be analyzed and the confidence levels corresponding to the multiple medical data prediction results.

[0056] Step S414: Among the multiple medical data prediction results, generate medical data analysis results for the user to be analyzed based on the medical data prediction result with the highest confidence level.

[0057] In steps S411-S414, when the inquiry behavior label is the end inquiry behavior label, the response generation sub-model extracts features from the normalized text of the symptom description entity to obtain the first symptom description feature, and extracts features from the end inquiry behavior label to obtain the first inquiry behavior feature. It can be understood that when the response generation sub-model extracts features from the normalized text of the symptom description entity and the end inquiry behavior label, it specifically converts the words in both the normalized text of the symptom description entity and the end inquiry behavior label into word vectors, and generates a positional feature vector for each word vector based on the positional relationship of the words, thereby obtaining the first symptom description feature and the first inquiry behavior feature.

[0058] After obtaining the first symptom description features and the first inquiry behavior features, feature fusion can be performed on these two features to obtain the first medical data analysis features. During feature fusion, the two features can be directly concatenated to obtain a feature vector with a higher dimension. Alternatively, other feature fusion methods can be used, which are not limited here.

[0059] Once the first medical data analysis feature is obtained, multiple medical data prediction results for the user to be analyzed and their corresponding confidence levels can be calculated based on this feature. Then, based on the prediction result with the highest confidence level among the multiple medical data prediction results, a medical data analysis result for the user to be analyzed is generated. This allows for the prediction of the user's symptom description and the resulting medical data analysis result.

[0060] Step S42: If the inquiry behavior label is a continue inquiry behavior label, the response generation sub-model generates symptom question text for the user to be analyzed based on the symptom description entity normalized text and the continue inquiry behavior label.

[0061] Step S43: Obtain new symptom description text in response to the user's question about the symptoms.

[0062] Step S44: The inquiry behavior prediction sub-model predicts an updated inquiry behavior label based on the normalized text of the symptom description entity, the new symptom description text, and the continued inquiry behavior label. The updated inquiry behavior label is one of the continued inquiry behavior label and the ended inquiry behavior label.

[0063] Step S45: If the updated inquiry behavior label is a continue inquiry behavior label, then repeat steps S43 and S44 until the updated inquiry behavior label output by the inquiry behavior prediction sub-model is an end inquiry behavior label. The response generation sub-model generates medical data analysis results for the user to be analyzed based on the normalized text of the symptom description entity, all new symptom description texts, and the end inquiry behavior label.

[0064] In step S42, when the inquiry behavior label is the end inquiry behavior label, it means that the medical data analysis results cannot be directly generated at present. Therefore, the response generation sub-model will first generate symptom question text for the user to be analyzed to conduct further symptom inquiries.

[0065] In one embodiment, a specific scheme for generating symptom question text using a response generation sub-model is provided, wherein step S42 specifically includes the following steps S421-S424;

[0066] Step S421: If the inquiry behavior label is a continued inquiry behavior label, then the normalized text of the symptom description entity is used to extract features through the response generation sub-model to obtain the second symptom description features, and the continued inquiry behavior label is used to extract features to obtain the second inquiry behavior features.

[0067] Step S422: Perform feature fusion on the second symptom description features and the second inquiry behavior features to obtain medical data analysis features.

[0068] Step S423: Based on the medical data analysis features, calculate multiple recommended question texts for the user to be analyzed and the confidence levels corresponding to the multiple recommended question texts.

[0069] Step S424: Among the multiple recommended question texts, generate a symptom question text for the user to be analyzed based on the recommended question text with the highest confidence.

[0070] In steps S421-S414, when the inquiry behavior label is the "continue inquiry behavior label," the response generation sub-model extracts features from the normalized text of the symptom description entity to obtain the second symptom description features, and extracts features from the "continue inquiry behavior label" to obtain the second inquiry behavior features. It can be understood that when the response generation sub-model extracts features from the normalized text of the symptom description entity and the "continue inquiry behavior label," it specifically converts the words in both the normalized text and the "continue inquiry behavior label" into word vectors, and generates positional feature vectors for each word vector based on the positional relationship of the words, thereby obtaining the second symptom description features and the second inquiry behavior features.

[0071] After obtaining the second symptom description features and the second inquiry behavior features, feature fusion can be performed on the second symptom description features and the second inquiry behavior features to obtain medical data analysis features. During feature fusion, the two features can be directly concatenated to obtain a feature vector with a higher dimension. Of course, other feature fusion methods can also be used, and no limitation is made here.

[0072] Once the medical data analysis features are obtained, multiple recommended question texts for the user to be analyzed and their corresponding confidence levels can be calculated based on these features. Then, based on the recommended question text with the highest confidence level, a symptom question text for the user to be analyzed can be generated. This facilitates further symptom inquiries based on the symptom question text for the user to be analyzed, thereby obtaining a lot of new symptom description information, so as to obtain more accurate medical data analysis results.

[0073] In step S43, the generated symptom query text is provided to the user to be analyzed for review, thereby facilitating the user to provide new symptom descriptions. Based on the new symptom description information provided by the user in response to the symptom query text, a new symptom description text is generated. The specific method for generating the new symptom description text can be referred to the description in the foregoing embodiments, and will not be repeated here.

[0074] In step S44, the inquiry behavior prediction sub-model predicts the updated inquiry behavior label based on the normalized text of the symptom description entity, the new symptom description text, and the continue inquiry behavior label. The updated inquiry behavior label is one of the continue inquiry behavior label and the end inquiry behavior label.

[0075] In one embodiment, a specific scheme is provided to predict the updated inquiry behavior label through an inquiry behavior prediction sub-model, wherein step S44 specifically includes the following steps S441-S443:

[0076] Step S441: The normalized text of the symptom description entity and the new symptom description text are extracted using the response generation sub-model to obtain the third medical data analysis features, and the feature of the continued inquiry behavior label is extracted to obtain the third inquiry behavior features.

[0077] Step S442: Perform feature fusion on the third medical data analysis features and the third inquiry behavior features to obtain inquiry behavior prediction features.

[0078] Step S443: Based on the query behavior prediction features, predict the updated query behavior label.

[0079] In steps S441-S443, the response generation sub-model extracts features from the normalized text of the symptom description entity and the new symptom description text to obtain the third medical data analysis features, and extracts features from the follow-up inquiry behavior label to obtain the third inquiry behavior features. It can be understood that when the response generation sub-model extracts features from the normalized text of the symptom description entity, the new symptom description text, and the follow-up inquiry behavior label, it specifically converts the words in these three texts into word vectors, and generates positional feature vectors for each word vector based on the positional relationships of the words, thereby obtaining the third medical data analysis features and the third inquiry behavior features.

[0080] After obtaining the third-party medical data analysis features and the third-party inquiry behavior features, feature fusion can be performed on these features to obtain inquiry behavior prediction features. During feature fusion, the two features can be directly concatenated to obtain a feature vector with a higher dimension. Alternatively, other feature fusion methods can be used, which are not limited here.

[0081] Once the query behavior prediction features are obtained, the updated query behavior labels can be predicted based on these features.

[0082] In step S45, if the updated inquiry behavior label is the "continue inquiry behavior label", then the contents of steps S43 and S44 are repeated until the updated inquiry behavior label output by the inquiry behavior prediction sub-model is the "end inquiry behavior label". Then, the final response generation sub-model generates medical data analysis results for the user to be analyzed based on the symptom description entity normalized text, all new symptom description text, and the "end inquiry behavior label".

[0083] As can be seen, in the above scheme, the initial symptom description text of the user to be analyzed is obtained and input into a pre-trained language model. The pre-trained language model includes an inquiry behavior prediction sub-model for predicting inquiry behavior and a response generation sub-model for medical data analysis. The response generation sub-model generates normalized text of symptom description entities based on the initial symptom description text. The inquiry behavior prediction sub-model then predicts inquiry behavior labels based on the normalized text of symptom description entities. Finally, the response generation sub-model generates medical data analysis results for the user to be analyzed based on the normalized text of symptom description entities and the inquiry behavior labels. This allows the two models to fully explore the hidden relationship between inquiries and data analysis, thereby improving the accuracy of the model in medical data analysis.

[0084] In one embodiment, after step S4, the AI-based medical data analysis method of the present invention may further include: step S5, determining medical treatment text associated with the medical data analysis results in a preset medical information database based on the medical data analysis results; step S6, generating medical treatment suggestions based on the medical treatment text.

[0085] In this embodiment, after the medical data analysis results are obtained, in order to facilitate the recommendation of reasonable medical treatment suggestions to the user to be analyzed, medical treatment suggestions associated with the medical data analysis results can be determined in a preset medical information database. The preset medical information database contains various diseases and medical treatment texts for each disease. The medical treatment texts include various information such as treatment opinions and treatment drugs. After determining the medical treatment text associated with the medical data analysis results, the medical treatment text can be used as a medical treatment suggestion.

[0086] In this embodiment, by generating medical treatment suggestions for the user to be analyzed based on the medical data analysis results, the richness of the generated medical data analysis results can be improved to a certain extent, and at the same time, the user to be analyzed can have a better medical consultation experience.

[0087] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0088] In one embodiment, an artificial intelligence-based medical data analysis device is provided, which corresponds one-to-one with the artificial intelligence-based medical data analysis method described in the above embodiments. For example... Figure 4As shown, the medical data analysis device includes an acquisition module 101, a first generation module 102, a prediction module 103, a second generation module 104, a matching module 105, and a third generation module 106. Detailed descriptions of each functional module are as follows:

[0089] The acquisition module 101 is used to acquire the initial symptom description text of the user to be analyzed and input it into the pre-trained language model. The pre-trained language model includes an inquiry behavior prediction sub-model for predicting inquiry behavior and a response generation sub-model for predicting medical data analysis results.

[0090] The first generation module 102 is used to generate normalized text of symptom description entities based on the initial symptom description text through the response generation sub-model;

[0091] Prediction module 103 is used to predict inquiry behavior labels based on the normalized text of the symptom description entity through the inquiry behavior prediction sub-model. The inquiry behavior label is one of a continue inquiry behavior label and an end inquiry behavior label.

[0092] The second generation module 104 is used to generate medical data analysis results for the user to be analyzed based on the normalized text of the symptom description entity and the query behavior label through the response generation sub-model.

[0093] In one embodiment, the second generation module 104 is specifically used to: if the inquiry behavior label is an end inquiry behavior label, then generate medical data analysis results for the user to be analyzed based on the normalized text of the symptom description entity and the end inquiry behavior label through the response generation sub-model.

[0094] In one embodiment, the second generation module 104 is specifically configured to: if the inquiry behavior label is an end-inquiry behavior label, then extract features from the normalized text of the symptom description entity through the response generation sub-model to obtain a first symptom description feature, and extract features from the end-inquiry behavior label to obtain a first inquiry behavior feature; perform feature fusion on the first symptom description feature and the first inquiry behavior feature to obtain a first medical data analysis feature; calculate multiple medical data prediction results for the user to be analyzed and the confidence level corresponding to each medical data prediction result based on the first medical data analysis feature; and generate a medical data analysis result for the user to be analyzed based on the medical data prediction result with the highest confidence level among the multiple medical data prediction results.

[0095] In one embodiment, the second generation module 104 is further configured to: if the inquiry behavior label is a continue inquiry behavior label, generate symptom question text for the user to be analyzed based on the symptom description entity normalized text and the continue inquiry behavior label through the response generation sub-model; obtain new symptom description text fed back by the user to be analyzed in response to the symptom question text; predict an updated inquiry behavior label based on the symptom description entity normalized text, the new symptom description text, and the continue inquiry behavior label, wherein the updated inquiry behavior label is one of a continue inquiry behavior label and an end inquiry behavior label; if the updated inquiry behavior label is a continue inquiry behavior label, re-execute obtaining new symptom description text fed back by the user to be analyzed in response to the symptom question text, until the updated inquiry behavior label output by the inquiry behavior prediction sub-model is an end inquiry behavior label, and generate medical data analysis results for the user to be analyzed based on the symptom description entity normalized text, all new symptom description texts, and the end inquiry behavior label through the response generation sub-model.

[0096] In one embodiment, the second generation module 104 is specifically configured to: if the inquiry behavior label is a continued inquiry behavior label, then extract features from the normalized text of the symptom description entity through the response generation sub-model to obtain second symptom description features, and extract features from the continued inquiry behavior label to obtain second inquiry behavior features; perform feature fusion on the second symptom description features and the second inquiry behavior features to obtain medical data analysis features; calculate multiple recommended question texts for the user to be analyzed and the confidence levels corresponding to the multiple recommended question texts based on the medical data analysis features; and generate symptom question texts for the user to be analyzed based on the recommended question text with the highest confidence level among the multiple recommended question texts.

[0097] The second generation module 104 is further configured to: extract features from the normalized text of the symptom description entity through the response generation sub-model to obtain a third medical data analysis feature, and extract features from the continued inquiry behavior label to obtain a third inquiry behavior feature; fuse the third medical data analysis feature and the third inquiry behavior feature to obtain an inquiry behavior prediction feature; and predict the updated inquiry behavior label based on the inquiry behavior prediction feature.

[0098] In one embodiment, the acquisition module 101 is specifically used to: acquire the corpus text of the user to be analyzed when describing symptoms; extract feature words of symptom description from the corpus text to obtain the initial symptom description text.

[0099] In one embodiment, the matching module 105 is used to determine, based on the medical data analysis results, a medical treatment text associated with the medical data analysis results in a preset medical information database; the third generation module 106 is used to generate a medical treatment suggestion based on the medical treatment text.

[0100] This invention provides a medical data analysis device. It acquires the initial symptom description text of the user to be analyzed and inputs it into a pre-trained language model. The pre-trained language model includes an inquiry behavior prediction sub-model for predicting inquiry behavior and a response generation sub-model for medical data analysis. The response generation sub-model generates normalized text of symptom description entities based on the initial symptom description text. The inquiry behavior prediction sub-model then predicts inquiry behavior tags based on the normalized text of symptom description entities. Finally, the response generation sub-model generates medical data analysis results for the user to be analyzed based on the normalized text of symptom description entities and the inquiry behavior tags. This allows the two models to fully explore the hidden relationship between inquiries and data analysis, improving the accuracy of the model in medical data analysis.

[0101] Specific limitations regarding the medical data analysis device can be found in the limitations of the medical data analysis method described above, and will not be repeated here. Each module in the aforementioned medical data analysis device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0102] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements server-side functions or steps of an artificial intelligence-based medical data analysis method.

[0103] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 6As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements client-side functions or steps of an artificial intelligence-based medical data analysis method.

[0104] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps:

[0105] The initial symptom description text of the user to be analyzed is obtained and input into a pre-trained language model, which includes an inquiry behavior prediction sub-model for inquiry behavior prediction and a response generation sub-model for medical data analysis.

[0106] Based on the initial symptom description text, the response generation sub-model generates normalized text for symptom description entities.

[0107] The inquiry behavior prediction sub-model predicts inquiry behavior labels based on the normalized text of the symptom description entity, and the inquiry behavior labels are either a continue inquiry behavior label or an end inquiry behavior label.

[0108] The response generation sub-model generates medical data analysis results for the user to be analyzed based on the normalized text of the symptom description entity and the query behavior label.

[0109] This invention provides a computer device that acquires initial symptom description text of a user to be analyzed and inputs it into a pre-trained language model. The pre-trained language model includes an inquiry behavior prediction sub-model for predicting inquiry behavior and a response generation sub-model for medical data analysis. The response generation sub-model generates normalized text of symptom description entities based on the initial symptom description text. The inquiry behavior prediction sub-model then predicts inquiry behavior labels based on the normalized text of symptom description entities. Finally, the response generation sub-model generates medical data analysis results for the user to be analyzed based on the normalized text of symptom description entities and the inquiry behavior labels. This allows the two models to fully explore the hidden relationship between inquiries and data analysis, improving the accuracy of the model in medical data analysis.

[0110] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0111] The initial symptom description text of the user to be analyzed is obtained and input into a pre-trained language model, which includes an inquiry behavior prediction sub-model for inquiry behavior prediction and a response generation sub-model for medical data analysis.

[0112] Based on the initial symptom description text, the response generation sub-model generates normalized text for symptom description entities.

[0113] The inquiry behavior prediction sub-model predicts inquiry behavior labels based on the normalized text of the symptom description entity, and the inquiry behavior labels are either a continue inquiry behavior label or an end inquiry behavior label.

[0114] The response generation sub-model generates medical data analysis results for the user to be analyzed based on the normalized text of the symptom description entity and the query behavior label.

[0115] This invention provides a computer-readable storage medium that acquires the initial symptom description text of the user to be analyzed and inputs it into a pre-trained language model. The pre-trained language model includes an inquiry behavior prediction sub-model for predicting inquiry behavior and a response generation sub-model for medical data analysis. The response generation sub-model generates normalized text of symptom description entities based on the initial symptom description text. The inquiry behavior prediction sub-model then predicts inquiry behavior labels based on the normalized text of symptom description entities. Finally, the response generation sub-model generates medical data analysis results for the user to be analyzed based on the normalized text of symptom description entities and the inquiry behavior labels. This allows the two models to fully explore the hidden relationship between inquiries and data analysis, improving the accuracy of the model in medical data analysis.

[0116] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0117] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0118] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0119] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A medical data analysis method based on artificial intelligence, characterized in that, include: The initial symptom description text of the user to be analyzed is obtained and input into a pre-trained language model, which includes an inquiry behavior prediction sub-model for inquiry behavior prediction and a response generation sub-model for medical data analysis. Based on the initial symptom description text, the response generation sub-model generates normalized text for symptom description entities. The inquiry behavior prediction sub-model predicts inquiry behavior labels based on the normalized text of the symptom description entity, and the inquiry behavior labels are either a continue inquiry behavior label or an end inquiry behavior label. The response generation sub-model generates medical data analysis results for the user to be analyzed based on the normalized text of the symptom description entity and the inquiry behavior tags, including: If the inquiry behavior label is an end-inquiry behavior label, then the response generation sub-model extracts features from the normalized text of the symptom description entity to obtain a first symptom description feature, and extracts features from the end-inquiry behavior label to obtain a first inquiry behavior feature; the first symptom description feature and the first inquiry behavior feature are fused to obtain a first medical data analysis feature; based on the first medical data analysis feature, multiple medical data prediction results for the user to be analyzed and the confidence level corresponding to each medical data prediction result are calculated; among the multiple medical data prediction results, the medical data analysis result for the user to be analyzed is generated based on the medical data prediction result with the highest confidence level. If the inquiry behavior label is a continue inquiry behavior label, then the response generation sub-model generates symptom question text for the user to be analyzed based on the normalized text of the symptom description entity and the continue inquiry behavior label; obtains new symptom description text in response to the symptom question text from the user to be analyzed; and predicts an updated inquiry behavior label based on the normalized text of the symptom description entity, the new symptom description text, and the continue inquiry behavior label using the inquiry behavior prediction sub-model. The updated inquiry behavior label is either a continue inquiry behavior label or an end inquiry behavior label.

2. The artificial intelligence-based medical data analysis method as described in claim 1, characterized in that, The response generation sub-model generates medical data analysis results for the user to be analyzed based on the normalized text of the symptom description entity and the inquiry behavior tags, including: If the updated inquiry behavior label is a "continue inquiry behavior label", then the process of obtaining the new symptom description text of the user's question text in response to the symptom is repeated until the updated inquiry behavior label output by the inquiry behavior prediction sub-model is an "end inquiry behavior label". Then, the response generation sub-model generates medical data analysis results for the user based on the symptom description entity normalized text, all new symptom description texts, and the end inquiry behavior label.

3. The artificial intelligence-based medical data analysis method as described in claim 2, characterized in that, If the inquiry behavior label is a follow-up inquiry behavior label, then the response generation sub-model generates symptom question text for the user to be analyzed based on the symptom description entity normalized text and the follow-up inquiry behavior label, including: If the inquiry behavior label is a follow-up inquiry behavior label, then the normalized text of the symptom description entity is used to extract features through the response generation sub-model to obtain the second symptom description features, and the follow-up inquiry behavior label is used to extract features to obtain the second inquiry behavior features. The second symptom description features and the second inquiry behavior features are fused to obtain the second medical data analysis features; Based on the medical data analysis features, multiple recommended question texts for the user to be analyzed and the confidence levels corresponding to the multiple recommended question texts are calculated. Among the multiple recommended question texts, a symptom question text for the user to be analyzed is generated based on the recommended question text with the highest confidence. The inquiry behavior prediction sub-model predicts updated inquiry behavior labels based on the symptom description entity normalized text, the new symptom description text, and the continued inquiry behavior label, including: The response generation sub-model extracts features from the normalized text of the symptom description entity and the new symptom description text to obtain the third medical data analysis features, and extracts features from the continued inquiry behavior label to obtain the third inquiry behavior features. The third medical data analysis feature and the third inquiry behavior feature are fused to obtain the inquiry behavior prediction feature; Based on the aforementioned query behavior prediction features, updated query behavior labels are predicted.

4. The artificial intelligence-based medical data analysis method as described in claim 1, characterized in that, The process of obtaining the initial symptom description text of the user to be analyzed includes: Obtain the text of the user's symptom descriptions; Feature words describing symptoms are extracted from the corpus text to obtain the initial symptom description text.

5. The artificial intelligence-based medical data analysis method as described in claim 1, characterized in that, After generating medical data analysis results for the user to be analyzed based on the symptom description entity normalized text and the inquiry behavior tags through the response generation sub-model, the method further includes: Based on the medical data analysis results, medical processing text associated with the medical data analysis results is determined in a preset medical information database; Based on the medical treatment text, a medical treatment suggestion is generated.

6. A medical data analysis device based on artificial intelligence, characterized in that, include: The acquisition module is used to acquire the initial symptom description text of the user to be analyzed and input it into the pre-trained language model. The pre-trained language model includes an inquiry behavior prediction sub-model for predicting inquiry behavior and a response generation sub-model for predicting medical data analysis results. The first generation module is used to generate normalized text of symptom description entities based on the initial symptom description text through the response generation sub-model; The prediction module is used to predict inquiry behavior labels based on the normalized text of the symptom description entity through the inquiry behavior prediction sub-model. The inquiry behavior label is one of the continue inquiry behavior label and the end inquiry behavior label. The second generation module is used to generate medical data analysis results for the user to be analyzed based on the symptom description entity normalized text and the inquiry behavior tags through the response generation sub-model, including: If the inquiry behavior label is an end-inquiry behavior label, then the response generation sub-model extracts features from the normalized text of the symptom description entity to obtain a first symptom description feature, and extracts features from the end-inquiry behavior label to obtain a first inquiry behavior feature; the first symptom description feature and the first inquiry behavior feature are fused to obtain a first medical data analysis feature; based on the first medical data analysis feature, multiple medical data prediction results for the user to be analyzed and the confidence level corresponding to each medical data prediction result are calculated; among the multiple medical data prediction results, the medical data analysis result for the user to be analyzed is generated based on the medical data prediction result with the highest confidence level. If the inquiry behavior label is a continue inquiry behavior label, then the response generation sub-model generates symptom question text for the user to be analyzed based on the normalized text of the symptom description entity and the continue inquiry behavior label; obtains new symptom description text in response to the symptom question text from the user to be analyzed; and predicts an updated inquiry behavior label based on the normalized text of the symptom description entity, the new symptom description text, and the continue inquiry behavior label using the inquiry behavior prediction sub-model. The updated inquiry behavior label is either a continue inquiry behavior label or an end inquiry behavior label.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the artificial intelligence-based medical data analysis method as described in any one of claims 1 to 5.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the artificial intelligence-based medical data analysis method as described in any one of claims 1 to 5.

Images