AI intelligent diagnosis method and system

Abstract

The invention discloses an AI intelligent diagnosis system, which comprises a server main body, an information acquisition module and an answer module, the information acquisition module comprises a collection medical specialist document material, an environment radio device and medical expert knowledge, the answer module is a man-machine interaction machine, and the server main body mainly comprises a data processing module and a data storage module. Compared with the prior art, the system has the advantages that 1, through man-machine interaction inquiry, doctors can be helped to improve the working efficiency, and the inquiry time is saved; support can be provided for new doctors with few experiences, and basic knowledge of the doctors can be improved; and 3, the misjudgment rate of inquiry is reduced, and the medical experience and health guarantee of the patient are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical intelligent judgment technology, and specifically refers to an AI intelligent diagnosis method and system. Background Art

[0002] The shortage of medical resources, especially the insufficient number of doctors, nurses and other personnel who directly provide medical services, has always been a major problem that has plagued the improvement of medical services. How to effectively utilize medical resources so that medical workers can release their energy from simple repetitive work has become one of the key tasks of medical information workers. Moreover, a large number of urban populations are now gathered, and no matter whether they have major or minor illnesses, they go to the hospital for consultation, which further increases the hospital's bearing capacity. Therefore, a system and method that can make intelligent judgments on common diseases or underlying causes is necessary. There are corresponding intelligent question-and-answer systems on the market, but the corresponding accuracy is low, the misdiagnosis rate is high, and they cannot really solve the problem. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide an AI intelligent diagnosis method and system that can greatly improve the accuracy of medical consultation in view of the deficiencies raised in the above-mentioned background technology.

[0004] In order to solve the above technical problems, the technical solution provided by the present invention is: an AI intelligent diagnosis system, which includes a server body, an information acquisition module and an answer module. The information acquisition module includes collecting medical specialty literature materials, environmental sound receiving devices and medical expert knowledge, the answer module is a human-computer interaction machine, and the server body is mainly a data processing module and a data storage module.

[0005] Further, take the following steps:

[0006] Step 1: Train the basic content model by performing structured analysis on medical specialty literature materials, historical clinical diagnosis data, and clinical prescription data;

[0007] Step 2: Collect the patient's consultation content through the environmental sound collection device;

[0008] Step 3: Feedback the answer to the medical question through the basic content model. The answer result is A1.

[0009] Step 4: collect the answers of the frontline doctors through the environmental sound collection device, and the answer result is A2;

[0010] Step 5: Organize medical experts to collect responses to the patient's questions in step 2. The response result is A3.

[0011] Step 6, by comparing the results of A1, A2, and A3, the patient is answered by cooperating with the human-computer interaction machine;

[0012] Step 7: With the help of user feedback and the accuracy of the diagnosis results of physical hospitals, the model knowledge system is continuously optimized to achieve the training of the final content model;

[0013] The medical specialty literature materials selected in step 1 need to be medical literature that is recognized as common knowledge.

[0014] The historical clinical diagnosis data and clinical prescription data in step 1 are both medical data that are confirmed to be non-misdiagnosed.

[0015] In the step 2, the patient's questions are keyword filtered to reduce the interference of sensitive words and redundant content on information analysis, and the text content recognized by the patient is classified and identified. According to different problem classifications, relevant classified literature content is selected in steps 3-5 to extract answer A1, extract the answer of the front-line doctor's consultation A2, and extract the answer of the patient's consultation content by the organization of medical experts A3. The overlap of the three answer results A1, A2, and A3 is compared, and the content with a disposal method greater than 95% is recalled to the large model. The large model completes reasoning based on the recalled content and questions and outputs the answer content.

[0016] In step 7, after the user receives the feedback result from the human-computer interaction machine, he / she scores the result. If the score is lower than the threshold, it is considered that the accuracy of this answer is low and is recorded in the data storage module. At regular intervals, such as 3 days, the parameter weights of the answers with low accuracy are lowered and then re-entered into the large model for training. After the training is completed, the large model is automatically deployed and updated, unsupervised learning of the device is realized, and the question-answering system is optimized.

[0017] After adopting the above structure, the present invention has the following advantages: 1. Through human-computer interactive consultation, it can help doctors improve work efficiency and save consultation time;

[0018] 2. It can provide support to new doctors with less experience and improve their basic knowledge.

[0019] 3. Reduce the misdiagnosis rate of medical consultation and greatly improve patients' medical experience and health assurance. DETAILED DESCRIPTION

[0020] The present invention is described in further detail below.

[0021] The AI ​​intelligent diagnosis system includes a server body, an information acquisition module and an answer module. The information acquisition module includes collecting medical specialty literature materials, environmental sound receiving devices and medical expert knowledge. The answer module is a human-computer interaction machine. The server body is mainly a data processing module and a data storage module.

[0022] Take the following steps:

[0023] Step 1: Train the basic content model by performing structured analysis on medical specialty literature materials, historical clinical diagnosis data, and clinical prescription data. The selected medical specialty literature materials must be medical literature of recognized common sense, that is, some medical literature that has been authoritatively certified and medical literature with a long history. Literature with a short history and low publication visibility cannot be imported. This can filter out some useless and erroneous literature with the highest probability. The historical clinical diagnosis data and clinical prescription data are all medical data that are confirmed to be non-misdiagnosed. Even for expert consultations, there is a certain misdiagnosis rate. Eliminating misdiagnosis data in clinical data can further reduce the misdiagnosis rate.

[0024] Step 2, through the environmental sound collection device, the patient's consultation content is collected, keyword filtering is performed, the interference of sensitive words and redundant content on information analysis is reduced, and the text content recognized by the patient is classified and identified according to different problem classifications, such as different medical classifications.

[0025] Step 3: Feedback the answer to the medical question through the basic content model. The answer result is A1.

[0026] Step 4: collect the answers of the frontline doctors through the environmental sound collection device, and the answer result is A2;

[0027] Step 5: Organize medical experts to collect responses to the patient's questions in step 2. The response result is A3.

[0028] Step 6, by comparing the three answer results A1, A2, and A3 for overlap, and cooperating with the human-computer interaction machine to answer the patient, the patient's question in step 2, selects the relevant classified literature content in steps 3-5 to extract answer A1, extracts the front-line doctor's consultation answer A2, and extracts the organization's medical experts to answer the patient's consultation content A3. By comparing the overlap of the three answer results A1, A2, and A3, the processing method is greater than 95%, and the content is recalled to the big model. The big model completes the reasoning based on the recalled content and questions, and outputs the answer content.

[0029] Step 7, with the help of user result feedback, in conjunction with the accuracy of the diagnosis result detection of the physical hospital, the model knowledge system is continuously optimized to realize the training of the final content large model. In step 7, after the user receives the feedback result from the human-computer interaction machine, he scores the result. If the score content is lower than the threshold, it is considered that the accuracy of this answer is low, and it is recorded in the data storage module. At regular intervals, such as 3 days, the parameter weights of the answers with low accuracy are lowered, and then they are re-invested in the large model for training. After the training is completed, the large model is automatically deployed and updated, the unsupervised learning of the equipment is realized, and the question-answering system is optimized.

[0030] The present invention and its implementation methods are described above, and such description is not restrictive, and the actual structure is not limited thereto. In short, if a person skilled in the art is inspired by it, and does not deviate from the purpose of the invention, and does not creatively design a structure and implementation method similar to the technical solution, they should all fall within the protection scope of the present invention.

Claims

1. AI intelligent diagnosis system, characterized by: It includes a server body, an information acquisition module and an answer module. The information acquisition module includes collecting medical specialty literature materials, environmental sound receiving devices and medical expert knowledge. The answer module is a human-computer interaction machine. The server body mainly includes a data processing module and a data storage module.

2. The method for implementing the AI ​​intelligent diagnosis system according to claim 1, characterized in that: Take the following steps: Step 1: Train the basic content model by performing structured analysis on medical specialty literature materials, historical clinical diagnosis data, and clinical prescription data; Step 2: Collect the patient's consultation content through the environmental sound collection device; Step 3: Feedback the answer to the medical question through the basic content model. The answer result is A1. Step 4: collect the answers of the frontline doctors through the environmental sound collection device, and the answer result is A2; Step 5: Organize medical experts to collect responses to the patient's questions in step 2. The response result is A3. Step 6, by comparing the results of A1, A2, and A3, the patient is answered by cooperating with the human-computer interaction machine; Step 7: With the help of user feedback and the accuracy of the diagnosis results of physical hospitals, the model knowledge system is continuously optimized to achieve the training of the final content model; 3. The method for implementing the AI ​​intelligent diagnosis system according to claim 2, characterized in that: The medical specialty literature materials selected in step 1 need to be medical literature that is recognized as common knowledge.

4. The method for implementing the AI ​​intelligent diagnosis system according to claim 2, characterized in that: The historical clinical diagnosis data and clinical prescription data in step 1 are both medical data that are confirmed to be non-misdiagnosed.

5. The method for implementing the AI ​​intelligent diagnosis system according to claim 2, characterized in that: In the step 2, the patient's questions are keyword filtered to reduce the interference of sensitive words and redundant content on information analysis, and the text content recognized by the patient is classified and identified. According to different problem classifications, relevant classified literature content is selected in steps 3-5 to extract answer A1, extract the answer of the front-line doctor's consultation A2, and extract the answer of the patient's consultation content by the organization of medical experts A3. The overlap of the three answer results A1, A2, and A3 is compared, and the content with a disposal method greater than 95% is recalled to the large model. The large model completes reasoning based on the recalled content and questions and outputs the answer content.

6. The method for implementing the AI ​​intelligent diagnosis system according to claims 1-2, characterized in that: In step 7, after the user receives the feedback result from the human-computer interaction machine, he / she scores the result. If the score is lower than the threshold, it is considered that the accuracy of this answer is low and is recorded in the data storage module. At regular intervals, such as 3 days, the parameter weights of the answers with low accuracy are lowered and then re-entered into the large model for training. After the training is completed, the large model is automatically deployed and updated, unsupervised learning of the device is realized, and the question-answering system is optimized.