Pneumonia risk warning system based on artificial intelligence

Abstract

The present invention discloses a pneumonia risk warning system based on artificial intelligence, the system comprising an image acquisition module, an audio acquisition module, a text acquisition module and an analysis module, the analysis module comprising a model training unit, an audio recognition unit, a text detection unit, a real-time detection unit and a prompt unit; the pneumonia risk warning system based on artificial intelligence proposed by the present invention performs image analysis on the lung image of the subject to obtain real-time detection results, and combines the conversation audio between the subject and the doctor, as well as the blood test report text of the subject, so as to quickly determine the infection level of the viral pneumonia of the subject; there is no need to spend a long time waiting for the final pneumonia examination result, thereby greatly speeding up the determination of the pneumonia infection level, and is more conducive to early screening and dynamic monitoring of viral pneumonia.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based pneumonia risk warning system. Background Art

[0002] Pneumonia includes bacterial pneumonia, fungal pneumonia and viral pneumonia; among them, viral pneumonia is more contagious, extremely infectious and progresses rapidly; therefore, there is an urgent need for simple and practical methods to conduct early screening and dynamic monitoring of viral pneumonia.

[0003] Currently, pneumonia patients need to undergo step-by-step examinations under the guidance of doctors to get the test results. This process increases the workload and infection risk of doctors, and also occupies a large amount of medical resources. At the same time, it takes a long time to get the test results, which brings great inconvenience to people who do not actually have pneumonia. Therefore, there is an urgent need for a solution that can quickly provide risk warnings for the infectious level of viral pneumonia. Summary of the invention

[0004] The main purpose of the present invention is to provide a pneumonia risk warning system based on artificial intelligence, aiming to solve the problem that there is an urgent need for a solution that can quickly provide risk warnings for the infectious level of viral pneumonia.

[0005] The technical solution proposed by the present invention is:

[0006] A pneumonia risk warning system based on artificial intelligence, the system comprising an image acquisition module, an audio acquisition module, a text acquisition module and an analysis module, the analysis module comprising a model training unit, an audio recognition unit, a text detection unit, a real-time detection unit and a prompt unit;

[0007] The image acquisition module is used to acquire a real-time lung image of the subject; the audio acquisition module is used to acquire the audio of the conversation between the doctor and the subject; the text acquisition module is used to acquire the text of the blood test report of the subject;

[0008] The model training unit is used to establish an image analysis model and train the image analysis model based on historical lung images and historical detection results; the audio recognition unit is used to obtain an audio recognition result based on the conversation audio; the text detection unit is used to obtain a text detection result based on the blood test report text; the real-time detection unit is used to input the real-time lung image into the image analysis model to obtain a real-time detection result; the prompt unit is used to generate a risk level based on the real-time detection result, the audio recognition result and the text detection result.

[0009] Preferably, the historical test results include high-risk results and low-risk results, the high-risk result is the presence of abnormal B-line or abnormal lung consolidation in the lungs, and the low-risk result is the absence of abnormal B-line and abnormal lung consolidation in the lungs; the real-time detection unit is specifically used for:

[0010] The real-time lung image is input into the image analysis model to obtain a real-time detection result, wherein the real-time detection result includes a high-risk result and a low-risk result. The high-risk result is that there is a B-line abnormality or a lung consolidation abnormality in the lungs, and the low-risk result is that there is no B-line abnormality and no lung consolidation abnormality in the lungs.

[0011] Preferably, the audio recognition unit also includes a first subunit, a second subunit and a third subunit; the first subunit is used to obtain preset keywords; the second subunit is used to convert the conversation audio into conversation text, input the conversation text and the preset keywords into a pre-trained NLP semantic matching model for semantic matching, and obtain the semantic matching degree between the conversation text and the preset keywords; the third subunit is used to generate the audio recognition result based on the semantic matching degree.

[0012] Preferably, the third subunit is specifically used to: compare the semantic matching degree with a preset matching threshold, and when the semantic matching degree is greater than the matching threshold, generate the high-risk audio recognition result; when the semantic matching degree is not greater than the matching threshold, generate the low-risk audio recognition result.

[0013] Preferably, the first subunit is specifically used to: obtain search data from an Internet search platform within a preset time period in the past; filter out hot keywords including "pneumonia" and ranked in the top preset digits in search volume based on the search data; obtain judgment keywords input by medical personnel; and use the hot keywords and the judgment keywords as the preset keywords.

[0014] Preferably, the third subunit is further specifically used to: obtain the audio duration of the conversation audio and obtain a preset duration; compare the semantic matching degree with a preset matching threshold, and when the semantic matching degree is greater than the matching threshold and the audio duration is greater than the preset duration, generate a high-risk audio recognition result; when the semantic matching degree is not greater than the matching threshold, generate a low-risk audio recognition result.

[0015] Preferably, the text detection unit is specifically used for:

[0016] Obtain key parameter names and corresponding preset parameter intervals; analyze the blood test report text to obtain blood test parameter values ​​corresponding to the key parameter names; when the blood test parameter values ​​all fall into the corresponding preset parameter intervals, generate the text detection results with low risk; when the blood test parameter values ​​do not all fall into the corresponding preset parameter intervals, generate the text detection results with high risk.

[0017] Preferably, the prompt unit is specifically used for:

[0018] When the real-time detection result, the audio detection result and the text detection result are all high risk, the highest risk level is generated; when there are two high risks among the image detection result, the audio detection result and the text detection result, the high risk level is generated; when only one of the image detection result, the audio detection result and the text detection result is high risk, the medium risk level is generated; when the image detection result, the audio detection result and the text detection result are not high risk, the low risk level is generated.

[0019] Preferably, it also includes a display terminal that is communicatively connected to the prompt module; the display terminal is used to display the risk level.

[0020] Preferably, the image analysis model is any one of a convolutional neural network model and a generative adversarial network model.

[0021] Through the above technical solution, the following beneficial effects can be achieved:

[0022] The artificial intelligence-based pneumonia risk warning system proposed in the present invention performs image analysis on the lung images of the person being tested to obtain real-time test results, and combines the audio of the conversation between the person being tested and the doctor, as well as the text of the blood test report of the person being tested, so as to quickly determine the infection level of the viral pneumonia of the person being tested, that is, the infection level of pneumonia can be determined after the person being tested completes the lung image detection, talks with the doctor, and generates the text of the blood test report; there is no need to spend a long time waiting for the final pneumonia test results, thereby greatly speeding up the determination of the pneumonia infection level, and being more conducive to early screening and dynamic monitoring of viral pneumonia. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying creative work.

[0024] Figure 1 This is a structural diagram of an embodiment of the pneumonia risk warning system based on artificial intelligence proposed by the present invention. DETAILED DESCRIPTION

[0025] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0026] The present invention proposes a pneumonia risk warning system based on artificial intelligence.

[0027] As attached Figure 1 As shown, in one embodiment of an artificial intelligence-based pneumonia risk warning system proposed by the present invention, the system includes an image acquisition module, an audio acquisition module, a text acquisition module and an analysis module; the image acquisition module, the audio acquisition module and the text acquisition module are all communicatively connected to the analysis module; the analysis module here is preferably a processor; the analysis module includes a model training unit, an audio recognition unit, a text detection unit, a real-time detection unit and a prompt unit.

[0028] The image acquisition module (such as a camera) is used to acquire a real-time lung image of the examinee, that is, a real-time lung image of the examinee (such as a lung X-ray image, a CT image), that is, the real-time lung image is taken by a camera; the audio acquisition module is used to acquire the audio of the conversation between the doctor and the examinee, and the hard disk acquisition module here is preferably a microphone; the text acquisition module is used to acquire the blood test report text of the examinee, and the text acquisition module here is directly connected to the result database of the blood testing department for communication, and retrieves the blood test report text of the examinee from the result database.

[0029] The model training unit is used to establish an image analysis model and train the image analysis model based on historical lung images and historical test results, where the historical lung images and historical test results are public training data; the audio recognition unit is used to obtain an audio recognition result based on the conversation audio; the text detection unit is used to obtain a text detection result based on the blood test report text; the real-time detection unit is used to input the real-time lung image into the image analysis model to obtain a real-time detection result; the prompt unit is used to generate a risk level based on the real-time detection result, the audio recognition result and the text detection result.

[0030] The artificial intelligence-based pneumonia risk warning system proposed in the present invention performs image analysis on the lung images of the person being tested to obtain real-time test results, and combines the audio of the conversation between the person being tested and the doctor, as well as the text of the blood test report of the person being tested, so as to quickly determine the infection level of the viral pneumonia of the person being tested, that is, the infection level of pneumonia can be determined after the person being tested completes the lung image detection, talks with the doctor, and generates the text of the blood test report; there is no need to spend a long time waiting for the final pneumonia test results, thereby greatly speeding up the determination of the pneumonia infection level, and being more conducive to early screening and dynamic monitoring of viral pneumonia.

[0031] In addition, the historical test results include high-risk results and low-risk results. The high-risk result is the presence of abnormal B lines or abnormal lung consolidation in the lungs, and the low-risk result is the absence of abnormal B lines and abnormal lung consolidation in the lungs.

[0032] Specifically, the division of lung regions includes: using the parasternal line, anterior axillary line, posterior axillary line, and paraspinal line as boundaries, each lung is divided into three regions: anterior, lateral, and posterior, namely the anterior chest region, lateral chest region, and posterior chest region. Each region is then divided into upper and lower parts at the level of the nipple line, so each lung can be divided into the anterior upper region, anterior lower region, lateral upper region, lateral lower region, posterior upper region, and posterior lower region, for a total of 12 lung regions for both lungs.

[0033] Among them, the B line refers to a strip-like strong echo perpendicular to the pleural line (or A line), also called the "rocket" sign, which radiates radially to the deep lung field.

[0034] Among them, pulmonary consolidation (i.e., subpleural pulmonary consolidation) manifests as a subpleural hypoechoic area or a tissue-like echo structure area, which is a change that occurs after the lung tissue loses air. In lung ultrasound, most pulmonary consolidations (intralobar consolidation) show "fragmented" characteristics, that is, the boundary between the consolidated lung tissue and the aerated lung tissue is irregular and appears as torn fragments. Interlobar consolidation shows the echo of liver and spleen tissue. The dynamic bronchial "inflation sign" visible in pulmonary consolidation can be used as an important evidence for the determination of pneumonia and can rule out obstructive atelectasis. Pulmonary consolidation is often accompanied by pleural effusion, and ultrasound has a high sensitivity and accuracy for pleural effusion, and can detect very small amounts of fluid in the pleural cavity.

[0035] Therefore, if the lung image shows abnormal B-line or abnormal lung consolidation, it means that the person being examined has a higher risk of infectious pneumonia, that is, a high-risk real-time test result is obtained. On the contrary, if the lung image does not show abnormal B-line and abnormal lung consolidation, it means that the person being examined has a lower risk of infectious pneumonia, that is, a low-risk real-time test result is obtained.

[0036] The real-time detection unit is specifically used for:

[0037] The real-time lung image is input into the image analysis model to obtain a real-time detection result, wherein the real-time detection result includes a high-risk result and a low-risk result, wherein the high-risk result is the presence of abnormal B-line or abnormal lung consolidation in the lungs, and the low-risk result is the absence of abnormal B-line and abnormal lung consolidation in the lungs. Through the above technical solution, the specific functions of the real-time detection unit are improved.

[0038] At the same time, the audio recognition unit also includes a first subunit, a second subunit and a third subunit; the first subunit is used to obtain preset keywords; the second subunit is used to convert the conversation audio into conversation text, input the conversation text and the preset keywords into a pre-trained NLP semantic matching model for semantic matching, and obtain the semantic matching degree between the conversation text and the preset keywords (the semantic matching degree here is expressed in the form of a percentage); the third subunit is used to generate the audio recognition result based on the semantic matching degree.

[0039] Specifically, the preset keywords here are keywords related to pneumonia symptoms with a high infectious risk level. During the conversation between the examinee and the doctor, if the number of preset keywords that appear is greater, it proves that the examinee has a higher infectious risk level of infectious pneumonia, and a high-risk audio recognition result is obtained.

[0040] In addition, the third subunit is specifically used to: compare the semantic matching degree with a preset matching threshold, and when the semantic matching degree is greater than the matching threshold (for example, 80%), generate a high-risk audio recognition result; when the semantic matching degree is not greater than the matching threshold, generate a low-risk audio recognition result.

[0041] At the same time, the first subunit is specifically used to: obtain search data from the Internet search platform within a preset time period in the past; filter out hot keywords including "pneumonia" and ranked in the top preset positions in search volume based on the search data; obtain judgment keywords input by medical personnel; and use the hot keywords and the judgment keywords as the preset keywords.

[0042] Specifically, the first subunit is used to obtain the search data of the Internet search platform in the past preset time period (for example, the past 30 days), and screen out the hot keywords including the two words "pneumonia" and the top 10 search volumes based on the search data; and the judgment keywords here are the keywords related to high-risk pneumonia summarized by medical staff during daily consultations, and the hot keywords and judgment keywords are used as preset keywords.

[0043] In addition, the third subunit is specifically used to: obtain the audio duration of the conversation audio and obtain a preset duration (for example, 10 minutes); compare the semantic matching degree with a preset matching threshold, and when the semantic matching degree is greater than the matching threshold and the audio duration is greater than the preset duration, generate a high-risk audio recognition result; when the semantic matching degree is not greater than the matching threshold, generate a low-risk audio recognition result.

[0044] Specifically, the preset duration here is used to express the average duration of the conversation between the doctor and the examinee. If it is greater than this duration, it means that the duration of the conversation between the examinee and the doctor exceeds the normal average duration, and the examinee is at greater risk of highly contagious pneumonia; therefore, when generating audio recognition results, the duration of the conversation audio needs to be considered.

[0045] At the same time, the text detection unit is specifically used to:

[0046] Obtain the key parameter name and the corresponding preset parameter interval (for example, the key parameter name is the white blood cell count tree); analyze the blood test report text to obtain the blood test parameter value corresponding to the key parameter name; when the blood test parameter values ​​all fall into the corresponding preset parameter interval, generate the text detection result of low risk; when the blood test parameter values ​​do not all fall into the corresponding preset parameter interval, generate the text detection result of high risk. Through the above technical solution, the specific functions of the text detection unit are improved.

[0047] In addition, the prompt unit is specifically used for:

[0048] When the real-time detection result, the audio detection result, and the text detection result are all high risk, the highest risk level is generated; when two of the image detection result, the audio detection result, and the text detection result are high risk, the high risk level is generated; when only one of the image detection result, the audio detection result, and the text detection result is high risk, the medium risk level is generated; when the image detection result, the audio detection result, and the text detection result are not high risk, the low risk level is generated. Through the above technical solution, the specific functions of the prompt unit are improved.

[0049] At the same time, the pneumonia risk warning system based on artificial intelligence also includes a display terminal connected to the prompt module; the display terminal is used to display the risk level, and the display terminal here is preferably a display screen or a smart phone terminal. The image analysis model is any one of a convolutional neural network model and a generative adversarial network model.

[0050] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0051] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), including a number of instructions for a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0052] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A pneumonia risk warning system based on artificial intelligence, characterized in that: The system includes an image acquisition module, an audio acquisition module, a text acquisition module and an analysis module, wherein the analysis module includes a model training unit, an audio recognition unit, a text detection unit, a real-time detection unit and a prompt unit; The image acquisition module is used to acquire a real-time lung image of the subject; the audio acquisition module is used to acquire the audio of the conversation between the doctor and the subject; The text acquisition module is used to acquire the blood test report text of the examinee; The model training unit is used to establish an image analysis model and train the image analysis model based on historical lung images and historical detection results; the audio recognition unit is used to obtain an audio recognition result based on the conversation audio; The text detection unit is used to obtain a text detection result based on the blood test report text; The real-time detection unit is used to input the real-time lung image into the image analysis model to obtain a real-time detection result; The prompt unit is used to generate a risk level based on the real-time detection result, the audio recognition result and the text detection result; The audio recognition unit further includes a first subunit, a second subunit and a third subunit; the first subunit is used to obtain preset keywords; the second subunit is used to convert the conversation audio into conversation text, input the conversation text and the preset keywords into a pre-trained NLP semantic matching model for semantic matching, and obtain the semantic matching degree between the conversation text and the preset keywords; The third subunit is used to generate the audio recognition result based on the semantic matching degree.

2. According to claim 1, a pneumonia risk warning system based on artificial intelligence is characterized in that: The historical test results include high-risk results and low-risk results, wherein the high-risk result is the presence of abnormal B-line or abnormal lung consolidation in the lungs, and the low-risk result is the absence of abnormal B-line and abnormal lung consolidation in the lungs; The real-time detection unit is specifically used for: The real-time lung image is input into the image analysis model to obtain a real-time detection result, wherein the real-time detection result includes a high-risk result and a low-risk result. The high-risk result is that there is a B-line abnormality or a lung consolidation abnormality in the lungs, and the low-risk result is that there is no B-line abnormality and no lung consolidation abnormality in the lungs.

3. According to claim 1, a pneumonia risk warning system based on artificial intelligence is characterized in that: The third subunit is specifically used to: compare the semantic matching degree with a preset matching threshold, and when the semantic matching degree is greater than the matching threshold, generate the high-risk audio recognition result; when the semantic matching degree is not greater than the matching threshold, generate the low-risk audio recognition result.

4. According to claim 1, a pneumonia risk warning system based on artificial intelligence is characterized in that: The first subunit is specifically used to: obtain search data from an Internet search platform within a preset time period in the past; filter out hot keywords including "pneumonia" and with a search volume ranking in the top preset digits based on the search data; obtain judgment keywords input by medical personnel; and use the hot keywords and the judgment keywords as the preset keywords.

5. According to the artificial intelligence-based pneumonia risk warning system of claim 1, it is characterized in that: The third subunit is further specifically used to: obtain the audio duration of the conversation audio and obtain a preset duration; compare the semantic matching degree with a preset matching threshold, and when the semantic matching degree is greater than the matching threshold and the audio duration is greater than the preset duration, generate a high-risk audio recognition result; when the semantic matching degree is not greater than the matching threshold, generate a low-risk audio recognition result.

6. According to claim 3, a pneumonia risk warning system based on artificial intelligence is characterized in that: The text detection unit is specifically used for: Obtain key parameter names and corresponding preset parameter intervals; analyze the blood test report text to obtain blood test parameter values ​​corresponding to the key parameter names; when the blood test parameter values ​​all fall into the corresponding preset parameter intervals, generate the text detection results with low risk; when the blood test parameter values ​​do not all fall into the corresponding preset parameter intervals, generate the text detection results with high risk.

7. The artificial intelligence-based pneumonia risk warning system according to claim 6, characterized in that: The prompt unit is specifically used for: When the real-time detection result, the audio recognition result, and the text detection result are all high risk, generating the risk level of the highest risk; When two of the real-time detection result, the audio recognition result, and the text detection result are high risks, generating the risk level of high risk; When only one of the real-time detection result, the audio recognition result, and the text detection result is high risk, generating the risk level of medium risk; When the real-time detection result, the audio recognition result, and the text detection result are all not high risk, the risk level of low risk is generated.

8. The artificial intelligence-based pneumonia risk warning system according to claim 1, characterized in that: It also includes a display terminal that is communicatively connected to the prompt unit; the display terminal is used to display the risk level.

9. The artificial intelligence-based pneumonia risk warning system according to claim 1, characterized in that: The image analysis model is any one of a convolutional neural network model and a generative adversarial network model.

Images