Artificial intelligence-based pre-consultation method, device, equipment and medium
By using AI-powered pre-consultation methods, and employing predictive models for consultation steps and response generation models, the system automatically generates doctor's response text and displays it through a virtual avatar. This solves the problems of long consultation times and missed or misdiagnosed cases caused by doctors asking questions step by step, achieving efficient and accurate pre-consultation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- PING AN TECH (SHENZHEN) CO LTD
- Filing Date
- 2023-06-20
- Publication Date
- 2026-05-12
Smart Images
Figure CN116775832B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of artificial intelligence and smart healthcare, and in particular to a pre-diagnosis method, device, equipment and medium based on artificial intelligence. Background Technology
[0002] With the increasing population, medical resources are becoming increasingly scarce. Currently, when visiting a hospital, doctors can only obtain information from patients through step-by-step questioning, which is a lengthy process and makes it difficult to improve the efficiency of doctors' consultations. Moreover, patients sometimes cannot clearly express their symptoms, which further prolongs the consultation process and may even lead to missed diagnoses or misdiagnoses. Summary of the Invention
[0003] Based on this, it is necessary to address the technical problems of existing technologies that require doctors to gradually ask patients for information, which is a lengthy consultation process and may even lead to missed diagnoses or misdiagnoses. Therefore, an artificial intelligence-based pre-consultation method, device, equipment, and medium are proposed.
[0004] Firstly, an artificial intelligence-based pre-diagnosis method is provided, the method comprising:
[0005] Obtain the summary data of the pre-consultation dialogue in the i-th round, wherein the summary data of the pre-consultation dialogue in the i-th round includes the pre-consultation dialogue data of each single round from the 1st round to the i-th round, and the single-round pre-consultation dialogue data is the dialogue data of a pre-consultation of the target patient in one round.
[0006] The consultation steps of the (i-1)th round and the summary data of the pre-consultation dialogue of the ith round are input into the consultation step prediction model corresponding to the target doctor to predict the consultation steps of the ith round, wherein the target doctor is the doctor that the target patient wants to see.
[0007] The consultation steps of the i-th round and the summary data of the pre-consultation dialogue are input into the pre-consultation response generation model corresponding to the target doctor to generate the response text and obtain the target response text;
[0008] The target doctor's response text is displayed based on the virtual avatar of the target doctor.
[0009] Secondly, an artificial intelligence-based pre-diagnosis device is provided, the device comprising:
[0010] The data acquisition module is used to acquire the summary data of the pre-consultation dialogue in the i-th round, wherein the summary data of the pre-consultation dialogue in the i-th round includes the pre-consultation dialogue data of each single round from the 1st round to the i-th round, and the single-round pre-consultation dialogue data is the dialogue data of a pre-consultation of the target patient in one round.
[0011] The consultation step prediction module is used to input the consultation steps of the (i-1)th round and the summary data of the pre-consultation dialogue of the ith round into the consultation step prediction model corresponding to the target doctor to predict the consultation steps of the ith round, wherein the target doctor is the doctor that the target patient wants to see.
[0012] The response text generation module is used to input the consultation steps of the i-th round and the summary data of the pre-consultation dialogue into the pre-consultation response generation model corresponding to the target doctor to generate the response text and obtain the target response text;
[0013] The display module is used to display the target doctor's reply text based on the virtual image corresponding to the target doctor.
[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 aforementioned artificial intelligence-based pre-diagnosis method.
[0015] Fourthly, a computer-readable storage medium is provided, the computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described artificial intelligence-based pre-diagnosis method.
[0016] The AI-based pre-consultation method of this application inputs the consultation steps of the (i-1)th round and the summary data of the pre-consultation dialogue of the i-th round into the consultation step prediction model corresponding to the target doctor to predict the consultation steps of the i-th round, wherein the target doctor is the doctor that the target patient wants to see; inputs the consultation steps of the i-th round and the summary data of the pre-consultation dialogue into the pre-consultation response generation model corresponding to the target doctor to generate response text, thereby obtaining the target response text; and displays the target response text based on the virtual image corresponding to the target doctor. This enables automated pre-consultation, which helps shorten the time doctors spend seeing patients. Through a consultation step prediction model and a pre-consultation response generation model, AI-based pre-consultation is achieved, improving accuracy and avoiding missed or misdiagnosed cases. By inputting the consultation steps of the i-th round and the summarized pre-consultation dialogue data into the pre-consultation response generation model corresponding to the target doctor to generate response text, the current round's consultation steps and the current round's summarized pre-consultation dialogue data are combined, ensuring the generated response text conforms to the consultation steps. This improves the intelligence and accuracy of pre-consultation. Attached Figure Description
[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] in:
[0019] Figure 1 This is a diagram illustrating the application environment of an AI-based pre-diagnosis method in one embodiment.
[0020] Figure 2 This is a flowchart of an AI-based pre-diagnosis method in one embodiment;
[0021] Figure 3 This is a structural block diagram of an AI-based pre-diagnosis device in one embodiment;
[0022] Figure 4 This is a structural block diagram of a computer device in one embodiment;
[0023] Figure 5 This is another structural block diagram of a computer device in one embodiment. 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 AI-based pre-diagnosis method provided in this invention can be applied to, for example... Figure 1In this application environment, client 110 communicates with server 120 via a network. Server 120 can receive and obtain the summary data of the pre-consultation dialogue in the i-th round from client 110. This summary data includes the pre-consultation dialogue data for each single round from round 1 to round i, where each single round is the dialogue data for a pre-consultation with the target patient. Server 120 inputs the consultation steps of round (i-1) and the summary data of the pre-consultation dialogue in round i into a consultation step prediction model corresponding to the target doctor to predict the consultation steps for round i, where the target doctor is the doctor the target patient wants to see. Server 120 inputs the consultation steps of round i and the summary data of the pre-consultation dialogue into a pre-consultation response generation model corresponding to the target doctor to generate response text, obtaining the target response text. Based on the virtual avatar of the target doctor, server 120 controls client 110 to display the target response text. This enables automated pre-consultation, which helps shorten the time doctors spend seeing patients. Through a consultation step prediction model and a pre-consultation response generation model, AI-based pre-consultation is achieved, improving accuracy and avoiding missed or misdiagnosed cases. By inputting the consultation steps of the i-th round and the summarized pre-consultation dialogue data into the pre-consultation response generation model corresponding to the target doctor to generate response text, the current round's consultation steps and the current round's summarized pre-consultation dialogue data are combined, ensuring the generated response text conforms to the consultation steps. This improves the intelligence and accuracy of pre-consultation.
[0026] In another embodiment, the client 110 is used to receive and acquire the summary data of the pre-consultation dialogue in the i-th round, wherein the summary data of the pre-consultation dialogue in the i-th round includes the pre-consultation dialogue data of each single round from the 1st round to the i-th round, and the single-round pre-consultation dialogue data is the dialogue data of a pre-consultation of the target patient; the consultation steps of the (i-1)-th round and the summary data of the pre-consultation dialogue in the i-th round are input into the consultation step prediction model corresponding to the target doctor to predict the consultation steps of the i-th round, wherein the target doctor is the doctor that the target patient wants to see; the consultation steps of the i-th round and the summary data of the pre-consultation dialogue are input into the pre-consultation response generation model corresponding to the target doctor to generate response text, thereby obtaining the target response text; and the target response text is displayed based on the virtual image corresponding to the target doctor.
[0027] There are multiple client 110s, including client 110s for patients and client 110s for doctors.
[0028] The client 110 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server 120 can be implemented using a standalone server or a server cluster consisting of multiple servers. The invention will now be described in detail through specific embodiments.
[0029] Please see Figure 2 As shown, Figure 2 A flowchart illustrating an AI-based pre-diagnosis method provided in an embodiment of the present invention includes the following steps:
[0030] S1: Obtain the summary data of the pre-consultation dialogue in the i-th round, wherein the summary data of the pre-consultation dialogue in the i-th round includes the pre-consultation dialogue data of each single round from the 1st round to the i-th round, and the single-round pre-consultation dialogue data is the dialogue data of a pre-consultation of the target patient in one round.
[0031] The pre-consultation dialogue summary data for the i-th round is a summary of all single-round pre-consultation dialogue data from the 1st round to the i-th round. The dialogue data in the pre-consultation dialogue summary data for the i-th round are sorted according to the dialogue order.
[0032] The single-round pre-consultation dialogue data refers to the dialogue data generated during a single round of pre-consultation with the target patient by the program implementing this application. In other words, the single-round pre-consultation dialogue data sequentially includes: the single-round dialogue data of the program in this application and the single-round dialogue data of the target patient. The data in the single-round pre-consultation dialogue data is text.
[0033] Specifically, it can obtain the summary data of the i-th round of pre-consultation dialogue input by the user, or obtain the summary data of the i-th round of pre-consultation dialogue from the storage space, or obtain the summary data of the i-th round of pre-consultation dialogue from a third-party application.
[0034] S2: Input the consultation steps of the (i-1)th round and the summary data of the pre-consultation dialogue of the ith round into the consultation step prediction model corresponding to the target doctor to predict the consultation steps of the ith round, wherein the target doctor is the doctor that the target patient wants to see.
[0035] Specifically, the consultation steps of the (i-1)th round and the summary data of the pre-consultation dialogue of the ith round are input into the consultation step prediction model corresponding to the target doctor to classify and predict the consultation steps. The step corresponding to the vector element with the largest value in the classification prediction vector is taken as the consultation step of the ith round.
[0036] Because the symptoms vary in different departments, the steps for consultation also differ. Therefore, this application adopts a consultation step prediction model corresponding to the target doctor, where the target doctor is the doctor the target patient wants to see, thereby improving the accuracy of the predicted consultation steps.
[0037] The consultation step prediction model is a classification model obtained based on the BERT (Bidirectional Encoder Representation from Transformers) model and a classification layer. The classification layer is a fully connected layer using the softmax activation function. The softmax activation function maps the outputs of multiple neurons to the (0,1) interval.
[0038] S3: Input the consultation steps of the i-th round and the summary data of the pre-consultation dialogue into the pre-consultation response generation model corresponding to the target doctor to generate the response text and obtain the target response text;
[0039] Specifically, the consultation steps of the i-th round and the summary data of the pre-consultation dialogue are input into the GPT model of the pre-consultation response generation model corresponding to the target doctor for feature extraction. The vector output by each prediction bit of the pre-consultation response generation model corresponding to the target doctor is input into the classification unit of the pre-consultation response generation model corresponding to the target doctor for classification prediction. The word corresponding to the vector element with the largest value in the classification prediction vector is taken as the hit word. The hit words are sorted in order according to the prediction bit to obtain the target response text.
[0040] It is understood that the target response text is the dialogue data that the program of this application presents to the target patient in the (i+1)th round (if the i-th round is the current round, then the (i+1)th round is the next round).
[0041] The pre-diagnosis response generation model is a model trained based on GPT (Generative Pre-trained Transformer) and multiple classification units.
[0042] The classification unit consists of a fully connected layer and a softmax activation layer. Each classification unit corresponds one-to-one with a prediction bit.
[0043] S4: Display the target reply text based on the virtual image corresponding to the target doctor.
[0044] Optionally, videos or images of the target doctor can be pre-captured, and then input into a pre-trained virtual avatar generation model to generate a virtual avatar. The virtual avatar generation model is based on a neural network trained on a model.
[0045] Optionally, the target doctor can be photographed in three dimensions beforehand, and a three-dimensional human body model can be constructed based on the photographed data. This three-dimensional human body model can then be used as the virtual image of the target doctor.
[0046] Specifically, based on the virtual image corresponding to the target doctor, the target reply text is directly displayed in text and / or voice.
[0047] Optionally, based on the virtual image corresponding to the target doctor, the client can be invoked to display the target's reply text in text and / or voice.
[0048] It is understood that after displaying the target reply text based on the virtual image corresponding to the target doctor, the dialogue data input by the target patient in response to the target reply text is obtained, and the target reply text and the dialogue data input by the target patient in response to the target reply text are used as the single-round pre-consultation dialogue data for the (i+1)th round; first, i is incremented by 1, and then the process jumps to step S1 to re-execute until the pre-consultation end signal is obtained.
[0049] This embodiment automates pre-consultation, which helps shorten the time doctors spend seeing patients. Through a consultation step prediction model and a pre-consultation response generation model, it achieves AI-based pre-consultation, improving accuracy and avoiding missed or misdiagnosed cases. By inputting the consultation steps of the i-th round and the summarized pre-consultation dialogue data into the pre-consultation response generation model corresponding to the target doctor to generate response text, it combines the consultation steps of the current round with the summarized pre-consultation dialogue data of the current round, ensuring the generated response text conforms to the consultation steps, thus improving the intelligence and accuracy of pre-consultation.
[0050] In one embodiment, the step of inputting the consultation steps of the (i-1)th round and the pre-consultation dialogue summary data of the i-th round into the consultation step prediction model corresponding to the target doctor to predict the consultation steps of the i-th round includes:
[0051] S21: Using a preset first splicing symbol, the consultation steps of the (i-1)th round and the pre-consultation dialogue summary data of the ith round are spliced together to obtain the first spliced data;
[0052] Optionally, the default first concatenation symbol is '&'.
[0053] Specifically, a preset first splicing symbol is used to splice the consultation steps of the (i-1)th round and the pre-consultation dialogue summary data of the ith round, and the spliced data is used as the first spliced data.
[0054] S22: Input the first spliced data into the consultation step prediction model corresponding to the target doctor to predict the consultation step in the i-th round, wherein the consultation step prediction model is a classification model trained based on the BERT model and a classification layer.
[0055] Specifically, the first concatenated data is input into the consultation step prediction model corresponding to the target doctor to classify and predict the consultation steps. The step corresponding to the vector element with the largest value in the classification prediction vector is taken as the consultation step in the i-th round.
[0056] Because the symptoms vary in different departments, the steps of the consultation process also differ. In this embodiment, the consultation steps in the i-th round are predicted by the consultation step prediction model corresponding to the target doctor. The target doctor is the doctor that the target patient wants to see, thereby improving the accuracy of the predicted consultation steps.
[0057] In one embodiment, the step of inputting the consultation steps of the i-th round and the pre-consultation dialogue summary data into the pre-consultation response generation model corresponding to the target doctor to generate response text and obtain the target response text includes:
[0058] S311: Input the consultation steps of the i-th round and the summary data of the pre-consultation dialogue into the pre-consultation response generation model corresponding to the target doctor to generate the response text and obtain the initial response text;
[0059] Specifically, the consultation steps of the i-th round and the summary data of the pre-consultation dialogue are input into the pre-consultation response generation model corresponding to the target doctor for feature extraction. The vector output by each prediction bit of the pre-consultation response generation model corresponding to the target doctor is input into the classification unit for classification prediction. The character corresponding to the vector element with the largest value in the classification prediction vector is taken as the hit character. The hit characters are sorted in order according to the prediction bit to obtain the initial response text.
[0060] S312: Input the initial response text into the text style conversion model corresponding to the target doctor to perform text style conversion, and obtain the target response text.
[0061] Specifically, the initial response text is input into the text style conversion model corresponding to the target doctor for text style conversion to obtain a response text with the target doctor's personal style, and this response text is used as the target response text.
[0062] Text style conversion is the process of changing the style of expression in a text while keeping the text's theme unchanged.
[0063] Text style transfer models are used to perform style transfer on the text input to a model, essentially refining the text to improve its quality. For example, a text style transfer model can be a BART model that has been fine-tuned.
[0064] BART (Bidirectional and Auto-Regressive Transformers) models are denoising autoencoders used for pre-training seq-to-seq (Sequence-to-Sequence) models. Fine-tuning refers to tweaking the pre-trained model to avoid retraining, improving training efficiency and reducing resource consumption. Specifically, the BART model used for pre-training can be the open-source Chinese-BART-Large model. After fine-tuning, a text style transfer model is obtained, which is then trained to produce a well-trained text style transfer model.
[0065] This embodiment inputs the initial response text into the text style conversion model corresponding to the target doctor and performs text style conversion to obtain the target response text. This makes the target response text conform to the target doctor's personal expression style, ensuring that the expression style is consistent between the pre-consultation and the target doctor's actual consultation, thereby improving the patient's consultation experience and increasing the patient's acceptance of automated pre-consultation.
[0066] In one embodiment, the step of inputting the consultation steps of the i-th round and the pre-consultation dialogue summary data into the pre-consultation response generation model corresponding to the target doctor to generate response text and obtain the target response text includes:
[0067] S321: Extract symptom information for the i-th round based on the summary data of the pre-diagnosis dialogue in the i-th round;
[0068] Specifically, the summary data of the pre-consultation dialogue in the i-th round is input into the symptom prediction model corresponding to the target doctor to extract and predict symptom information. The symptom label corresponding to each vector element whose median value is greater than a preset probability value is taken as the hit symptom, and each hit symptom is taken as the symptom information of the i-th round.
[0069] The symptom prediction model is a model obtained by training a neural network.
[0070] S322: Using a preset second splicing symbol, the consultation steps of the i-th round, the symptom information of the i-th round, and the pre-consultation dialogue summary data of the i-th round are spliced together to obtain the second splicing data;
[0071] Specifically, a preset second splicing symbol is used to splice the consultation steps of the i-th round, the symptom information of the i-th round, and the pre-consultation dialogue summary data of the i-th round, and the spliced data is used as the second splicing data.
[0072] For example, the preset second concatenation symbol is '&', which concatenates the consultation steps of the i-th round, the symptom information of the i-th round, and the pre-consultation dialogue summary data of the i-th round to obtain A&B&C, where A is the consultation steps of the i-th round, B is the symptom information of the i-th round, and C is the pre-consultation dialogue summary data of the i-th round.
[0073] S323: Input the second spliced data into the pre-consultation response generation model corresponding to the target doctor to generate the response text, thereby obtaining the target response text, wherein the pre-consultation response generation model is a model trained based on GPT.
[0074] Specifically, the second concatenated data is input into the pre-consultation response generation model corresponding to the target doctor to generate response text and extract features. The vector output by each prediction bit of the pre-consultation response generation model corresponding to the target doctor is input into the classification unit for classification prediction. The character corresponding to the vector element with the largest value in the classification prediction vector is taken as the hit character. The hit characters are sorted in order according to the prediction bit to obtain the target response text.
[0075] In this embodiment, the second spliced data is input into the pre-consultation response generation model corresponding to the target doctor to generate response text. This combines the symptom information, consultation steps, and pre-consultation dialogue summary data of the current round, so that the generated response text conforms to the consultation steps and is an extension of the symptom information of the current round. This further improves the intelligence and accuracy of pre-consultation.
[0076] In one embodiment, the step of displaying the target reply text based on the virtual avatar corresponding to the target doctor includes:
[0077] S41: Based on the voice features corresponding to the target doctor, the target reply text is converted into speech to obtain the target speech;
[0078] Specifically, an audio segment of data is pre-recorded for the target doctor, and sound features are extracted based on the recorded audio data; the sound features corresponding to the target doctor and the target reply text are input into a preset sound generation model to generate speech, and the generated speech is used as the target speech.
[0079] Sound generation models, also known as speech generation models, can have their model structures and training methods chosen from existing technologies, which will not be elaborated upon here.
[0080] S42: Based on the virtual image corresponding to the target doctor, the target voice is displayed, wherein, during the display process, the virtual image corresponding to the target doctor is controlled based on the facial expression features and movement features corresponding to the target doctor.
[0081] Specifically, the virtual image corresponding to the target doctor is displayed on the interface, and the target voice is played. During the playback of the target voice, the facial expressions of the virtual image corresponding to the target doctor are controlled based on the facial expression features of the target doctor, and the actions of the virtual image corresponding to the target doctor are controlled based on the action features of the target doctor.
[0082] This embodiment performs speech conversion on the target reply text based on the voice characteristics of the target doctor, so that the target speech simulates the voice of the target doctor. When displaying the target speech, the virtual image of the target doctor is controlled based on the facial expression and movement characteristics of the target doctor, so that the virtual image displayed on the interface matches the personal image of the target doctor. This ensures that the expression style of the pre-consultation and the actual consultation of the target doctor is consistent, improving the patient's consultation experience and increasing the patient's acceptance of automated pre-consultation.
[0083] In one embodiment, the method includes:
[0084] S51: Obtain the pre-consultation end signal corresponding to the target patient;
[0085] The pre-consultation ends signal, indicating the completion of the pre-consultation with the target patient.
[0086] Specifically, it could be a pre-consultation end signal triggered by the target patient, or a pre-consultation end signal actively triggered by the program implementing this application based on preset conditions. For example, the pre-consultation end signal is actively triggered when the pre-consultation end conditions are met, where the pre-consultation end condition is that the consultation step is the end step.
[0087] S52: In response to the pre-consultation end signal, acquire the single-round pre-consultation dialogue data corresponding to the target patient as data to be analyzed;
[0088] Specifically, upon receiving the pre-consultation end signal, all single-round pre-consultation dialogue data of the target patient from round 1 to the current round are acquired, and all acquired single-round pre-consultation dialogue data are used as data to be analyzed. The dialogue data in the data to be analyzed is sorted according to the dialogue order.
[0089] S53: Input the data to be analyzed into the pre-consultation conclusion generation model corresponding to the target doctor to generate the pre-consultation conclusion and obtain the initial pre-consultation conclusion.
[0090] Specifically, the data to be analyzed is input into the pre-consultation conclusion generation model corresponding to the target doctor to predict the pre-consultation conclusion, and a consultation conclusion vector is obtained. The pre-consultation conclusion is determined based on the consultation conclusion vector, and the determined pre-consultation conclusion is used as the initial pre-consultation conclusion.
[0091] In the consultation conclusion vector, each vector element corresponds to a conclusion sub-category, and the value of each vector element is a probability value. Each conclusion sub-category corresponds to one conclusion category. That is, each conclusion category corresponds to at least one conclusion sub-category.
[0092] For example, the range of values for the conclusion category includes: gender, age, chief complaint, main symptom, accompanying symptoms, onset time, triggering factor, treatment process, past medical history, and allergy history. The conclusion category for gender corresponds to the subcategories for male and female, and the conclusion category for age includes all age groups. The conclusion category for main symptom corresponds to the subcategories for all symptoms. The conclusion category for accompanying symptoms corresponds to the subcategories for all symptoms.
[0093] The pre-diagnosis conclusion generation model is a classification model obtained by training a neural network.
[0094] The preliminary consultation conclusion is determined based on the consultation conclusion vector. This means selecting vector elements whose values are greater than the probability threshold corresponding to the conclusion category from the vector elements corresponding to the consultation conclusion vector, taking the conclusion sub-category corresponding to the selected vector element as the hit sub-category, and taking all hit sub-categories corresponding to the conclusion category as the conclusions corresponding to that conclusion category in the initial preliminary consultation conclusion.
[0095] This embodiment generates pre-consultation conclusions by inputting the data to be analyzed into the pre-consultation conclusion generation model corresponding to the target doctor, thereby automating the generation of pre-consultation conclusions. Furthermore, artificial intelligence is used to improve the accuracy of the pre-consultation conclusions, thus enhancing the automation level of this application.
[0096] In one embodiment, after the step of inputting the data to be analyzed into the pre-consultation conclusion generation model corresponding to the target doctor to generate a pre-consultation conclusion and obtain an initial pre-consultation conclusion, the method further includes:
[0097] S54: Obtain supplementary consultation signals;
[0098] The supplementary consultation signal is a signal to conduct a supplementary consultation with the target patient manually.
[0099] Specifically, it involves obtaining supplementary consultation signals sent by the target doctor in the program or client implementing this application.
[0100] S55: In response to the supplementary consultation signal, display the initial pre-consultation conclusion;
[0101] Specifically, in response to the supplementary consultation signal, the consultation interface displays the initial pre-consultation conclusion.
[0102] S56: Based on the displayed initial pre-diagnosis conclusion, obtain the supplementary consultation information input by the target doctor;
[0103] Specifically, the target doctor enters supplementary consultation information on the consultation interface, and clicks the "End" button after completing the information. When the "End" button is clicked, a signal indicating that the supplementary information is complete is generated.
[0104] S57: Obtain completion signal;
[0105] S58: In response to the supplementary completion signal, the supplementary consultation information and the data to be analyzed are input into the pre-consultation conclusion generation model corresponding to the target doctor to generate a pre-consultation conclusion and obtain the target pre-consultation conclusion.
[0106] Specifically, upon receiving the supplementary completion signal, the supplementary consultation information on the consultation interface and the data to be analyzed are concatenated. The concatenated data is then input into the pre-consultation conclusion generation model corresponding to the target doctor to generate a pre-consultation conclusion. The generated pre-consultation conclusion is then used as the target pre-consultation conclusion.
[0107] This embodiment obtains supplementary consultation information input by the target doctor through a consultation interface that displays the initial pre-consultation conclusion. Then, the supplementary consultation information and the data to be analyzed are input into the pre-consultation conclusion generation model corresponding to the target doctor to generate the pre-consultation conclusion, which satisfies the doctor's supplementary consultation needs and improves the accuracy of the determined target pre-consultation conclusion.
[0108] Please see Figure 3 As shown, in one embodiment, an artificial intelligence-based pre-diagnosis device is provided, the device comprising:
[0109] The data acquisition module 801 is used to acquire the summary data of the pre-consultation dialogue in the i-th round, wherein the summary data of the pre-consultation dialogue in the i-th round includes the pre-consultation dialogue data of each single round from the 1st round to the i-th round, and the single-round pre-consultation dialogue data is the dialogue data of a pre-consultation of the target patient in one round.
[0110] The consultation step prediction module 802 is used to input the consultation steps of the (i-1)th round and the summary data of the pre-consultation dialogue of the ith round into the consultation step prediction model corresponding to the target doctor to predict the consultation steps of the ith round, wherein the target doctor is the doctor that the target patient wants to see.
[0111] The response text generation module 803 is used to input the consultation steps of the i-th round and the summary data of the pre-consultation dialogue into the pre-consultation response generation model corresponding to the target doctor to generate response text and obtain the target response text;
[0112] The display module 804 is used to display the target reply text based on the virtual image corresponding to the target doctor.
[0113] This embodiment automates pre-consultation, which helps shorten the time doctors spend seeing patients. Through a consultation step prediction model and a pre-consultation response generation model, it achieves AI-based pre-consultation, improving accuracy and avoiding missed or misdiagnosed cases. By inputting the consultation steps of the i-th round and the summarized pre-consultation dialogue data into the pre-consultation response generation model corresponding to the target doctor to generate response text, it combines the consultation steps of the current round with the summarized pre-consultation dialogue data of the current round, ensuring the generated response text conforms to the consultation steps, thus improving the intelligence and accuracy of pre-consultation.
[0114] In one embodiment, the step of the consultation step prediction module 802 inputting the consultation steps of the (i-1)th round and the pre-consultation dialogue summary data of the i-th round into the consultation step prediction model corresponding to the target doctor to predict the consultation steps of the i-th round includes:
[0115] Using a preset first splicing symbol, the consultation steps of the (i-1)th round and the pre-consultation dialogue summary data of the ith round are spliced together to obtain the first spliced data;
[0116] The first spliced data is input into the consultation step prediction model corresponding to the target doctor to predict the consultation step in the i-th round, wherein the consultation step prediction model is a classification model trained based on the BERT model and a classification layer.
[0117] In one embodiment, the step of the response text generation module 803 in generating the response text by inputting the consultation steps of the i-th round and the summary data of the pre-consultation dialogue into the pre-consultation response generation model corresponding to the target doctor to generate the response text includes:
[0118] The consultation steps of the i-th round and the summary data of the pre-consultation dialogue are input into the pre-consultation response generation model corresponding to the target doctor to generate the response text and obtain the initial response text;
[0119] The initial response text is input into the text style conversion model corresponding to the target doctor to perform text style conversion, thereby obtaining the target response text.
[0120] In one embodiment, the step of the response text generation module 803 in generating the response text by inputting the consultation steps of the i-th round and the summary data of the pre-consultation dialogue into the pre-consultation response generation model corresponding to the target doctor to generate the response text includes:
[0121] Based on the summary data of the pre-diagnosis dialogue in the i-th round, extract the symptom information for the i-th round;
[0122] Using a preset second splicing symbol, the consultation steps of the i-th round, the symptom information of the i-th round, and the pre-consultation dialogue summary data of the i-th round are spliced together to obtain the second spliced data;
[0123] The second concatenated data is input into the pre-consultation response generation model corresponding to the target doctor to generate the response text, thereby obtaining the target response text. The pre-consultation response generation model is a model trained based on GPT.
[0124] In one embodiment, the step of displaying the target reply text based on the virtual image corresponding to the target doctor in the display module 804 includes:
[0125] Based on the voice features corresponding to the target doctor, the target reply text is converted into speech to obtain the target speech;
[0126] Based on the virtual avatar corresponding to the target doctor, the target voice is displayed. During the display process, the virtual avatar corresponding to the target doctor is controlled based on the facial expression features and movement features of the target doctor.
[0127] In one embodiment, the device includes: a consultation conclusion generation module;
[0128] The consultation conclusion generation module is used to obtain the pre-consultation end signal corresponding to the target patient, respond to the pre-consultation end signal, obtain the single-round pre-consultation dialogue data corresponding to the target patient as the data to be analyzed, input the data to be analyzed into the pre-consultation conclusion generation model corresponding to the target doctor to generate the pre-consultation conclusion, and obtain the initial pre-consultation conclusion.
[0129] In one embodiment, after the step of inputting the data to be analyzed into the pre-consultation conclusion generation model corresponding to the target doctor to generate a pre-consultation conclusion and obtain an initial pre-consultation conclusion, the consultation conclusion generation module further includes:
[0130] Obtain supplementary consultation signals;
[0131] In response to the supplementary consultation signal, the initial pre-consultation conclusion is displayed;
[0132] Based on the initial pre-consultation conclusion shown, obtain the supplementary consultation information input by the target doctor;
[0133] Receive completion signal;
[0134] In response to the completion signal of the supplementary consultation, the supplementary consultation information and the data to be analyzed are input into the pre-consultation conclusion generation model corresponding to the target doctor to generate a pre-consultation conclusion and obtain the target pre-consultation conclusion.
[0135] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 4 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 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 the functions or steps of a pre-diagnosis method based on artificial intelligence on the server side.
[0136] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 5As 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 pre-diagnosis method.
[0137] 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, when executing the computer program, performs the following steps:
[0138] Obtain the summary data of the pre-consultation dialogue in the i-th round, wherein the summary data of the pre-consultation dialogue in the i-th round includes the pre-consultation dialogue data of each single round from the 1st round to the i-th round, and the single-round pre-consultation dialogue data is the dialogue data of a pre-consultation of the target patient in one round.
[0139] The consultation steps of the (i-1)th round and the summary data of the pre-consultation dialogue of the ith round are input into the consultation step prediction model corresponding to the target doctor to predict the consultation steps of the ith round, wherein the target doctor is the doctor that the target patient wants to see.
[0140] The consultation steps of the i-th round and the summary data of the pre-consultation dialogue are input into the pre-consultation response generation model corresponding to the target doctor to generate the response text and obtain the target response text;
[0141] The target doctor's response text is displayed based on the virtual avatar of the target doctor.
[0142] This embodiment automates pre-consultation, which helps shorten the time doctors spend seeing patients. Through a consultation step prediction model and a pre-consultation response generation model, it achieves AI-based pre-consultation, improving accuracy and avoiding missed or misdiagnosed cases. By inputting the consultation steps of the i-th round and the summarized pre-consultation dialogue data into the pre-consultation response generation model corresponding to the target doctor to generate response text, it combines the consultation steps of the current round with the summarized pre-consultation dialogue data of the current round, ensuring the generated response text conforms to the consultation steps, thus improving the intelligence and accuracy of pre-consultation.
[0143] In one embodiment, a computer-readable storage medium is provided that stores a computer program, which, when executed by a processor, performs the following steps:
[0144] Obtain the summary data of the pre-consultation dialogue in the i-th round, wherein the summary data of the pre-consultation dialogue in the i-th round includes the pre-consultation dialogue data of each single round from the 1st round to the i-th round, and the single-round pre-consultation dialogue data is the dialogue data of a pre-consultation of the target patient in one round.
[0145] The consultation steps of the (i-1)th round and the summary data of the pre-consultation dialogue of the ith round are input into the consultation step prediction model corresponding to the target doctor to predict the consultation steps of the ith round, wherein the target doctor is the doctor that the target patient wants to see.
[0146] The consultation steps of the i-th round and the summary data of the pre-consultation dialogue are input into the pre-consultation response generation model corresponding to the target doctor to generate the response text and obtain the target response text;
[0147] The target doctor's response text is displayed based on the virtual avatar of the target doctor.
[0148] This embodiment automates pre-consultation, which helps shorten the time doctors spend seeing patients. Through a consultation step prediction model and a pre-consultation response generation model, it achieves AI-based pre-consultation, improving accuracy and avoiding missed or misdiagnosed cases. By inputting the consultation steps of the i-th round and the summarized pre-consultation dialogue data into the pre-consultation response generation model corresponding to the target doctor to generate response text, it combines the consultation steps of the current round with the summarized pre-consultation dialogue data of the current round, ensuring the generated response text conforms to the consultation steps, thus improving the intelligence and accuracy of pre-consultation.
[0149] 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.
[0150] 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.
[0151] 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.
[0152] 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 pre-diagnosis method based on artificial intelligence, the method comprising: Obtain the summary data of the pre-consultation dialogue in the i-th round, wherein the summary data of the pre-consultation dialogue in the i-th round includes the pre-consultation dialogue data of each single round from the 1st round to the i-th round, and the single-round pre-consultation dialogue data is the dialogue data of a pre-consultation of the target patient in one round. The consultation steps of the (i-1)th round and the summary data of the pre-consultation dialogue of the ith round are input into the consultation step prediction model corresponding to the target doctor to predict the consultation steps of the ith round, wherein the target doctor is the doctor that the target patient wants to see. The consultation steps of the i-th round and the summary data of the pre-consultation dialogue are input into the pre-consultation response generation model corresponding to the target doctor to generate the response text and obtain the target response text; The target doctor's response text is displayed based on the virtual avatar of the target doctor. The step of inputting the consultation steps of the (i-1)th round and the pre-consultation dialogue summary data of the i-th round into the consultation step prediction model corresponding to the target doctor to predict the consultation steps of the i-th round includes: Using a preset first splicing symbol, the consultation steps of the (i-1)th round and the pre-consultation dialogue summary data of the ith round are spliced together to obtain the first spliced data; The first spliced data is input into the consultation step prediction model corresponding to the target doctor to predict the consultation step in the i-th round, wherein the consultation step prediction model is a classification model trained based on the BERT model and a classification layer.
2. The pre-diagnosis method based on artificial intelligence according to claim 1, characterized in that, The step of inputting the consultation steps of the i-th round and the summary data of the pre-consultation dialogue into the pre-consultation response generation model corresponding to the target doctor to generate the response text and obtain the target response text includes: The consultation steps of the i-th round and the summary data of the pre-consultation dialogue are input into the pre-consultation response generation model corresponding to the target doctor to generate the response text and obtain the initial response text; The initial response text is input into the text style conversion model corresponding to the target doctor to perform text style conversion, thereby obtaining the target response text.
3. The pre-diagnosis method based on artificial intelligence according to claim 1, characterized in that, The step of inputting the consultation steps of the i-th round and the summary data of the pre-consultation dialogue into the pre-consultation response generation model corresponding to the target doctor to generate the response text and obtain the target response text includes: Based on the summary data of the pre-diagnosis dialogue in the i-th round, extract the symptom information for the i-th round; Using a preset second splicing symbol, the consultation steps of the i-th round, the symptom information of the i-th round, and the pre-consultation dialogue summary data of the i-th round are spliced together to obtain the second spliced data; The second concatenated data is input into the pre-consultation response generation model corresponding to the target doctor to generate the response text, thereby obtaining the target response text. The pre-consultation response generation model is a model trained based on GPT.
4. The pre-diagnosis method based on artificial intelligence according to claim 1, characterized in that, The step of displaying the target reply text based on the virtual avatar corresponding to the target doctor includes: Based on the voice features corresponding to the target doctor, the target reply text is converted into speech to obtain the target speech; Based on the virtual avatar corresponding to the target doctor, the target voice is displayed. During the display process, the virtual avatar corresponding to the target doctor is controlled based on the facial expression features and movement features of the target doctor.
5. The pre-diagnosis method based on artificial intelligence according to claim 1, characterized in that, The method includes: Obtain the pre-consultation end signal corresponding to the target patient; In response to the pre-consultation end signal, acquire the single-round pre-consultation dialogue data corresponding to the target patient as data to be analyzed. The data to be analyzed is input into the pre-consultation conclusion generation model corresponding to the target doctor to generate the pre-consultation conclusion and obtain the initial pre-consultation conclusion.
6. The pre-diagnosis method based on artificial intelligence according to claim 5, characterized in that, After the step of inputting the data to be analyzed into the pre-consultation conclusion generation model corresponding to the target doctor to generate a pre-consultation conclusion and obtain an initial pre-consultation conclusion, the method further includes: Obtain supplementary consultation signals; In response to the supplementary consultation signal, the initial pre-consultation conclusion is displayed; Based on the initial pre-consultation conclusion shown, obtain the supplementary consultation information input by the target doctor; Receive completion signal; In response to the completion signal of the supplementary consultation, the supplementary consultation information and the data to be analyzed are input into the pre-consultation conclusion generation model corresponding to the target doctor to generate a pre-consultation conclusion and obtain the target pre-consultation conclusion.
7. A pre-diagnosis device based on artificial intelligence, characterized in that, The device includes: The data acquisition module is used to acquire the summary data of the pre-consultation dialogue in the i-th round, wherein the summary data of the pre-consultation dialogue in the i-th round includes the pre-consultation dialogue data of each single round from the 1st round to the i-th round, and the single-round pre-consultation dialogue data is the dialogue data of a pre-consultation of the target patient in one round. The consultation step prediction module is used to input the consultation steps of the (i-1)th round and the summary data of the pre-consultation dialogue of the ith round into the consultation step prediction model corresponding to the target doctor to predict the consultation steps of the ith round, wherein the target doctor is the doctor that the target patient wants to see. The response text generation module is used to input the consultation steps of the i-th round and the summary data of the pre-consultation dialogue into the pre-consultation response generation model corresponding to the target doctor to generate the response text and obtain the target response text; The display module is used to display the target reply text based on the virtual image corresponding to the target doctor; The step of inputting the consultation steps of the (i-1)th round and the pre-consultation dialogue summary data of the i-th round into the consultation step prediction model corresponding to the target doctor to predict the consultation steps of the i-th round includes: Using a preset first splicing symbol, the consultation steps of the (i-1)th round and the pre-consultation dialogue summary data of the ith round are spliced together to obtain the first spliced data; The first spliced data is input into the consultation step prediction model corresponding to the target doctor to predict the consultation step in the i-th round, wherein the consultation step prediction model is a classification model trained based on the BERT model and a classification layer.
8. 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 pre-diagnosis method as described in any one of claims 1 to 6.
9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the artificial intelligence-based pre-diagnosis method as described in any one of claims 1 to 6.