Diagnosis method and device for respiratory system diseases of elderly patients

Abstract

The invention discloses a method and a device for diagnosing respiratory system diseases of an elderly patient, and the method comprises the steps: indicating the elderly patient to execute a specified action in a first time period under the condition that the risk level of the to-be-diagnosed respiratory system diseases of the elderly patient is higher than a preset level, and enabling the elderly patient to execute the specified action in the first time period, acquiring state data representing the second time period every second time period, acquiring a state feature vector corresponding to the state data for each piece of state data, and inputting the state feature vector corresponding to each second time period into a trained machine learning model based on time sequence analysis according to a time sequence, and determining the disease index of the respiratory system disease of the elderly patient. According to the application, the elderly patients can diagnose by using the elderly patients at home or going to community hospitals by themselves, and early diagnosis is facilitated.

Description

Technical Field

[0001] This application relates to the field of intelligent medical diagnosis, and particularly to a diagnosis method and device for respiratory diseases of elderly patients. Background Art

[0002] Respiratory diseases, especially obstructive pulmonary diseases, are common diseases characterized by cough, dyspnea, and airflow limitation. Such diseases are common in middle-aged and elderly populations, especially in elderly patients with underlying diseases. With the aggravation of the global population aging in recent years, the incidence and overall prevalence of elderly patients have increased very rapidly.

[0003] For the elderly, respiratory diseases do not affect daily life, but if not intervened early, slow development to the later stage will cause irreversible serious consequences. For example, in the early stage of chronic obstructive pulmonary disease, there may only be chest tightness and shortness of breath after exercise or occasional cough, or there may be other symptoms (such as fatigue). At this stage, many people, especially the elderly, will think that they have an allergic condition or a mild cold. However, if obvious symptoms appear, usually lung damage has already occurred. For the elderly with significantly declined physical functions, this will lead to the emergence of various complications, thus seriously affecting the health status and quality of life. Therefore, it is crucial to be able to diagnose and give corresponding treatment plans in a timely manner before lung damage occurs. That is to say, early diagnosis and intervention are the keys to the treatment of respiratory diseases.

[0004] However, traditional respiratory diseases (such as chronic obstructive pulmonary disease) require patients to actively go to the hospital or medical unit to seek a professional doctor's diagnosis when they are conscious. It is very difficult for ordinary patients to realize that they may have such diseases and miss the best treatment time. In addition, respiratory diseases require special equipment and professional operators. Especially due to the uneven distribution of medical resources and different levels of medical personnel, it is difficult to diagnose them accurately and timely in clinical diagnosis. Therefore, the prior art needs a method that allows elderly patients to diagnose or obtain a reference diagnosis result through simple operations. Summary of the Invention

[0005] The embodiments of this application provide a diagnosis method and device for respiratory diseases of elderly patients, which are used to solve at least the above-mentioned technical problems.

[0006] An embodiment of the present application provides a method for diagnosing respiratory diseases for elderly patients, including: when the risk level of an elderly patient to be diagnosed with a respiratory disease is higher than a preset level, instructing the elderly patient to perform a specified action within a first time period and obtaining, within the first time period, status data representing the status within each second time period at intervals of the second time period, where the status data includes conventional detection data detected by a wearable medical detection instrument of the elderly patient, a respiratory curve image detected by a respiratory airflow detection device, and an audio image detected by an audio receiving device, and the first time period includes a plurality of second time periods; For each piece of status data, obtaining a status feature vector corresponding to the status data, including: Constructing a conventional detection feature vector from the conventional detection data according to the diagnosis result of whether each piece of conventional detection data is abnormal; performing feature extraction on the respiratory curve image detected by the respiratory airflow detection device using a first feature extraction network and performing normalization processing using a first normalization parameter to obtain a respiratory feature vector corresponding to the respiratory curve image; performing feature extraction on the audio image using a second feature extraction network and performing normalization processing using a second normalization parameter to obtain an audio feature vector corresponding to the audio image; fusing the conventional detection feature vector, the respiratory feature vector, and the audio feature vector to obtain a status feature vector; Inputting the status feature vectors corresponding to each second time period into a trained machine learning model based on time series analysis in chronological order to determine the disease index of the elderly patient having the respiratory disease, where the recurrent network layer of the machine learning model includes a gated recurrent unit corresponding to each second time period.

[0007] Optionally, the diagnosis method further includes: obtaining a user database, where the user information of each user in the user database includes basic information and the disease result of whether the user has the respiratory disease, and the basic information includes the age, gender, weight, smoking history, living environment, and genetic factors of each user; performing data fitting on the disease result and the basic information of each user to obtain an objective function for having the respiratory disease; inputting the basic information of the elderly patient into the objective function to determine the risk level of the elderly patient having the respiratory disease.

[0008] Optionally, the specified action refers to walking at a specific step frequency, the first time period is 10 minutes, and the second time period is 1 minute.

[0009] Optionally, the conventional detection data includes body temperature data detected by a thermometer, pulse data detected by a pulse detector, blood pressure data detected by a blood pressure monitor, heart rate data detected by a heart rate detector, and blood oxygen data detected by a blood oxygen monitor.

[0010] Optionally, when the specific step frequency walking includes walking at a first step frequency of less than 60 steps per minute, walking at a second step frequency of more than 60 steps per minute and less than 120 steps per minute, and walking at a third step frequency of more than 120 steps per minute and less than 150 steps per minute, the disease index obtained under the first step frequency will be used as the first disease index, the disease index obtained under the second step frequency will be used as the second disease index, and the disease index obtained under the third step frequency will be used as the third disease index, Optionally, the method further includes: calculating a final disease index by using a first weight factor corresponding to the first disease index, a second weight factor corresponding to the second disease index, and a third weight factor corresponding to the third disease index, where the third weight factor is greater than the second weight factor and the second weight factor, the second weight factor is greater than the first weight factor, and the sum of the first weight factor, the second weight factor, and the third weight factor is 1.

[0011] Optionally, the trained machine learning model for time series analysis is trained in the following manner: obtaining the state feature vectors determined by each user in each second time period and the disease results of each user in the user database; constructing the machine learning model for time series analysis, where the machine learning model is set with full network parameters; using the correspondence between the state feature vectors determined by each user in each second time period and the disease results of each user to train the machine learning model and adjusting the full network parameters until the machine learning model meets the preset requirements.

[0012] Optionally, after the respiratory curve image detected by the respiratory airflow detection device is subjected to feature extraction by the first feature extraction network and normalized by using the first normalization parameter, a respiratory feature vector corresponding to the respiratory curve image is obtained, including: The first feature extraction network performs feature extraction on the respiratory curve image to obtain image features of multiple different scales, and after performing normalization processing on the image features of multiple different scales, merges them through inverse Laplace transform to obtain intermediate respiratory features; the intermediate respiratory features are normalized by using the first statistical information as the first normalization parameter to obtain the respiratory feature vector.

[0013] After the audio image is subjected to feature extraction by the second feature extraction network and normalized by using the second normalization parameter, an audio feature vector corresponding to the audio image is obtained, including: The second feature extraction network extracts features from the audio image to obtain image features of multiple different scales, and after performing normalization processing on the image features of multiple different scales, merges them through inverse Laplace transform to obtain intermediate audio features; the intermediate audio features are subjected to normalization processing by using the second statistical information as the second normalization parameter to obtain the audio feature vector.

[0014] An embodiment of the present application further provides a respiratory disease diagnosis device for elderly patients, including: a state data acquisition unit, configured to, when the risk level of a respiratory disease of an elderly patient to be diagnosed is higher than a preset level, instruct the elderly patient to perform a specified action within a first time period and acquire state data representing the state within the second time period at intervals of a second time period within the first time period, where the state data includes conventional detection data detected by the elderly patient using a wearable medical detection instrument, a respiratory curve image detected by a respiratory airflow detection device, and an audio image detected by an audio receiving device, and the first time period includes multiple second time periods; A state feature vector acquisition unit, configured to, for each state data, acquire a state feature vector corresponding to the state data, including: forming a conventional detection feature vector from the conventional detection data according to the diagnosis result of whether each conventional detection data is abnormal; extracting the respiratory curve image detected by the respiratory airflow detection device by using a first feature extraction network and performing normalization processing by using a first normalization parameter to obtain a respiratory feature vector corresponding to the respiratory curve image; performing normalization processing on the audio image by using a second feature extraction network and using a second normalization parameter to obtain an audio feature vector corresponding to the audio image; and performing fusion processing on the conventional detection feature vector, the respiratory feature vector, and the audio feature vector to obtain a state feature vector; A disease index determination unit, configured to input multiple state feature vectors into a machine learning model of time series analysis that has been trained in the order of the second time period, where the recurrent network layer of the machine learning model includes a gated recurrent unit corresponding to each second time period, to determine the disease index of the elderly patient suffering from a respiratory disease.

[0015] An embodiment of the present application further provides a respiratory disease diagnosis device for elderly patients, the device including: a processor; and a memory arranged to store computer-executable instructions, the executable instructions, when executed, causing the processor to execute the above diagnosis method.

[0016] An embodiment of the present application further provides a computer-readable storage medium, on which computer instructions are stored, and when the instructions are executed, the above diagnosis method is implemented.

[0017] The above at least one technical solution adopted in the embodiments of the present application can achieve the following beneficial effects: According to the respiratory disease diagnosis method for elderly patients according to the exemplary embodiments of the present application, it can indicate that elderly patients can achieve self-diagnosis through portable equipment, which is convenient for patients to detect and / or prevent diseases in the initial stage, avoiding lung damage. Further, the diagnosis method takes into account the characteristics of respiratory diseases, especially chronic lung diseases, collects different data sources during the movement of elderly patients, and improves the accuracy of diagnosis. The method also determines the risk level of the elderly patient suffering from respiratory diseases based on the basic information of the elderly patient to be diagnosed. And according to different risk levels, the elderly patient is prompted to perform subsequent operations, realizing hierarchical management of the elderly patient. Further, in the case where the elderly patient has a high risk and does not have home detection, convenience places such as community clinics are linked, so that the elderly patient can complete the diagnosis without going to the hospital and performing complex lung examinations, solving the pain points of difficult access to medical examinations for the elderly patient and alleviating the medical pressure on the hospital. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present application, and constitute a part of the present application. The schematic embodiments and descriptions thereof of the present application are used to explain the present application, and do not constitute an improper limitation to the present application. In the drawings:

[0019] Figure 1 is a schematic flowchart showing a diagnosis method for respiratory diseases of elderly patients according to an embodiment of the present application.

[0020] Figure 2 is a scene diagram showing the diagnosis of respiratory diseases by elderly patients according to an embodiment of the present application.

[0021] Figure 3 is a schematic diagram showing a respiratory curve image according to an exemplary embodiment of the present application.

[0022] Figure 4 is a schematic diagram showing an audio image according to an exemplary embodiment of the present application.

[0023] Figure 5 is a diagram showing a machine learning model according to an exemplary embodiment of the present application.

[0024] Figure 6 is a block diagram showing a diagnosis device for respiratory diseases according to an exemplary embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments of this application and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of this application.

[0026] The following will detail the technical solutions provided by each embodiment of this application in conjunction with the drawings.

[0027] The diagnostic method for respiratory diseases of elderly patients according to the exemplary embodiments of this application is applicable to diagnosing various respiratory diseases. The following will be described by taking chronic obstructive pulmonary disease as an example. As mentioned in the background art, in daily life, it is very difficult for elderly patients to realize that they have chronic obstructive pulmonary disease and can accurately go to a professional medical institution for diagnosis and treatment. Moreover, the existing diagnosis of chronic obstructive pulmonary disease mainly relies on conventional pulmonary function tests, including vital capacity measurement, total lung capacity measurement, forced expiratory volume in the first second measurement, inspiratory capacity, expiratory reserve volume, inspiratory reserve volume, tidal volume, functional residual capacity, residual volume, etc. These test indicators require the assistance of professionals during the test process and take a long time to complete. In view of this, the diagnostic method provided by this application can assist elderly patients or medical staff in diagnosing elderly patients through simple operations and / or portable diagnosis and treatment devices.

[0028] As Figure 1 shown, a schematic flowchart of a diagnostic method for respiratory diseases of elderly patients provided by this application. According to one aspect of this application, it includes: step S110, step S120, and step S130.

[0029] Before performing the diagnosis, this application can first determine the possibility that an elderly patient has a respiratory disease. As mentioned above, many elderly patients are not aware of taking the initiative to check whether they have been infected with a respiratory disease. Through a large amount of clinical experience, it has been found that respiratory diseases, especially chronic obstructive pulmonary disease, show obvious common characteristics. For example, the incidence of respiratory diseases increases with age. Another example is that elderly patients with a smoking history or who are exposed to harmful dust and chemical gases in their living environment have a very high probability of suffering from respiratory diseases. Another example is that the lower the body mass index, the higher the prevalence of chronic obstructive pulmonary disease.

[0030] Therefore, this application can obtain the patient's basic information related to respiratory diseases of the elderly patient. The patient's basic information includes but is not limited to the patient's age, gender, weight, smoking history, underlying diseases, living environment, and genetic factors.

[0031] The elderly patient can provide their own basic information for the patient basic information to be provided. For example Figure 2 As shown, in daily life, the elderly patient can regularly input their own basic information on the relevant medical application through a mobile terminal (such as a mobile phone, portable tablet, etc.) according to the prompt information. Then, the server of the medical application can calculate the risk point level of the elderly patient suffering from respiratory diseases based on the collected basic information and provide it to the elderly user. For example, it can directly display on the screen that the risk of the elderly patient suffering from respiratory diseases is high risk, or it can directly display on the screen that the risk of the elderly patient suffering from respiratory diseases is medium risk or low risk. As shown Figure 2 As shown, it can display on the screen "Your disease risk is relatively high. It is recommended to go to the convenience center for testing" and can also give the address of the convenience center closest to the elderly patient. As an embodiment, the server of the medical application can synchronize the preliminary diagnosis result of the elderly patient to a professional doctor

[0032] As a usage scenario, the elderly patient can also perform the above operations through fixed devices set in convenience places, such as elderly activity centers, community clinics, etc. The elderly patient, out of concern for their own physical health or due to external reminders, goes to the convenience place and inputs the patient basic information according to the reminder of the medical application on the fixed device. The background server of the fixed device can calculate the risk point level of the elderly patient suffering from respiratory diseases based on the collected basic information and provide it to the elderly user. For example, it can directly display on the screen that the risk of the elderly patient suffering from respiratory diseases is high risk, or it can directly display on the screen that the risk of the elderly patient suffering from respiratory diseases is medium risk or low risk

[0033] After the background server of the medical application obtains the above patient basic information, it uses the already calculated objective function to calculate the risk level of the elderly patient suffering from respiratory diseases. As shown in the following table, in practice, the background server can determine the specific risk level according to the interval where the function value calculated by the objective function is located. The interval values corresponding to each risk level are set by those skilled in the art based on empirical values

[0034] Numerical range Risk level [0.00,0.25] Low risk [0.26,0.50] Medium risk [0.50,1.00] High risk Table 1 The objective function refers to fitting the result of whether each user has respiratory diseases with the corresponding basic information of the user to obtain the objective function of having respiratory diseases. The basic information includes the age, gender, weight, smoking history, living environment, and genetic factors of each user. The objective function shown in Formula 1

[0035] (1) In Formula 1, F(i) represents the objective function for patients with respiratory diseases, i indicates the user or patient, and x n indicates the basic information corresponding to each user or patient. Here, n = 1, 2... m. In this embodiment, m = 6, and α i indicates the weight factor corresponding to the i-th basic information, and β i is the fluctuation parameter corresponding to each weight factor. This fluctuation parameter can be a parameter set by those skilled in the art based on research results. In practice, the fluctuation parameter can be a fixed parameter set based on experience.

[0036] To determine the weight factor corresponding to each basic information, the background service first obtains the user database and uses the user data in the user database to determine each weight value in the objective function. In practice, the users in the user database know whether they have respiratory diseases. Based on this, the user database stores the disease results and basic information corresponding to each user. Then, data fitting is performed using the disease results and corresponding basic information of each user to determine the weight factor in the objective function.

[0037] Then, the basic information of the elderly patient to be diagnosed is input into the objective function to determine the risk level of the elderly patient having respiratory diseases. As an embodiment, when it is determined through the above processing that the elderly patient has a high risk in this application, subsequent operations are performed, that is, steps S110, S120, and S130 are executed.

[0038] In step S110, when the risk level of the elderly patient to be diagnosed having respiratory diseases is higher than the preset level, it is indicated that the elderly patient performs a specified action within the first time period and, within the first time period, obtains the status data representing the second time period every second time period. Among them, the status data includes the routine detection data detected by the elderly patient using a wearable medical detection instrument, the respiratory curve image detected by a respiratory airflow detection device, and the audio image detected by an audio receiving device. The first time period includes multiple second time periods.

[0039] Preferably, the preset level can be a medium risk level. In practice, when the elderly patient to be diagnosed is diagnosed as a high risk level through the above-mentioned method, at this time, the device used by the elderly patient (for example, the mobile terminal used by the elderly patient or the fixed device used by the elderly patient) will prompt the elderly patient to perform a series of subsequent operations and conduct specific analysis and diagnosis based on this series of operations to determine the disease index of the elderly patient having respiratory diseases.

[0040] The purpose of the method for diagnosing respiratory diseases in elderly patients of the present application is to enable elderly patients to be diagnosed at an early stage of respiratory diseases, especially chronic obstructive pulmonary disease, so as to avoid lung damage. The early symptoms of respiratory diseases, especially chronic obstructive pulmonary disease, are very mild, but they will be manifested after exercise. Based on this, when the risk level of the elderly patient to be diagnosed with a respiratory disease is higher than the preset level, the present application will prompt the elderly patient to perform a specific exercise within a certain period of time, and obtain data of the elderly patient from different data sources during this process, and comprehensively analyze the data from these data sources to determine the disease index of the elderly patient. Therefore, the elderly patient needs to be equipped with instruments capable of obtaining these data sources and these instruments can be transmitted to the background server of the medical application. As Figure 2 shown, the elderly patient can go to a place equipped with relevant instruments and capable of transmitting the data obtained by these instruments to the server.

[0041] As another embodiment, the mobile terminal of the elderly patient can synchronously obtain these data and authorize the transmission of these data to the background server of the medical application. Therefore, the wearable medical detection instruments of the elderly patient include a thermometer, a pulse detector, a blood pressure monitor, a heart rate detector, and a pulse oximeter. In this way, the conventional detection data includes the body temperature data detected by the thermometer, the pulse data detected by the pulse detector, the blood pressure data detected by the blood pressure monitor, the heart rate data detected by the heart rate detector, and the blood oxygen data detected by the pulse oximeter.

[0042] The exemplary specified actions according to the present application may include walking at a first step frequency of less than 60 steps per minute, walking at a second step frequency of more than 60 steps and less than 120 steps per minute, or walking at a third step frequency of more than 120 steps and less than 150 steps per minute. Simply understood, the specified actions can be slow walking, normal walking, and fast walking. As shown above, the early symptoms of respiratory diseases, especially chronic obstructive pulmonary disease, will be manifested after exercise. Considering the exercise ability of the elderly patient, walking at the first step frequency is preferably implemented. In order to improve the detection accuracy, a step frequency meter can be embedded in the medical detection instrument worn by the elderly patient and can remind the elderly user of the current step frequency.

[0043] In addition, in order to obtain a respiratory curve image and an audio image, the elderly patient needs to wear a respiratory airflow detection device and an audio receiving device. The respiratory airflow detection device can be an expiratory flow meter. The expiratory flow meter can measure the gas flow rate at different time points during the patient's exhalation process to form an expiratory flow rate curve graph. The audio receiving device refers to a device that can receive a voice signal and convert it into an audio image (for example, a spectrogram).

[0044] In the case of elderly patients suffering from respiratory diseases, especially chronic obstructive pulmonary disease, they may experience shortness of breath and coughing after exercise, and these symptoms will become more and more obvious as the exercise time increases. In addition, depending on the degree or stage of the respiratory disease, the manifestations of shortness of breath and the frequency of coughing will also vary, and these characteristics will be reflected in the expiratory flow velocity curve graph and the sound spectrogram. Therefore, this application needs to regularly collect and process these data when performing specified actions.

[0045] As an example, the first time period is 10 minutes and the second time period is 1 minute. That is to say, when the method determines that the risk level of an elderly patient to be diagnosed with a respiratory disease is higher than the preset level, it will notify the elderly patient that they need to complete a test of specific actions within 10 minutes. During the test, the elderly patient needs to wear relevant portable instruments and complete the above operations according to the instructions. In one embodiment, after the elderly patient inputs relevant information through the medical application of the mobile terminal at home and determines that the risk level is relatively high (for example, high risk), the medical application can synchronize the situation of the elderly patient to relevant communities / community service centers or locations where subsequent tests can be carried out. The elderly patient can pick up and wear the above portable instruments at these locations and complete the above tests, and the acquisition and processing of relevant data during the test can be carried out in the background server. In another embodiment, the elderly patient to be diagnosed can directly go to a place equipped with portable instruments and a respiratory disease diagnosis device for elderly patients according to the embodiments of the present application. The patient can wear these portable instruments and complete the test of specific actions under the instructions of the device.

[0046] As an example, after wearing the portable device, the elderly patient walks slowly within 10 minutes according to the instructions. According to the respiratory disease diagnosis method for elderly patients according to the embodiments of the present application, the status data of the elderly patient within this one minute is obtained every minute, and then subsequent operations are performed using this status data with time information. The status data for each minute includes the routine test data detected by the wearable medical detection instrument of the elderly patient, the respiratory curve image detected by the respiratory airflow detection device, and the audio image of the elderly patient within the predetermined time interval detected by the audio receiving device.

[0047] As Figure 3 The shown respiratory curve image is a regular curve with each exhalation to inhalation as a cycle. This curve represents the breathing condition of the elderly patient within one minute of performing a specific exercise (for example, slow walking). As Figure 3 In the shown image, the horizontal axis represents the time axis with the unit of seconds, and the vertical axis represents the respiratory flow rate with the unit of liters. The respiratory curve image shows regular periodic fluctuations. As Figure 3As shown, during the inhalation phase of the user, the curve rises rapidly, indicating that air flows quickly into the respiratory tract and lungs. Subsequently, during the exhalation phase, the curve descends smoothly and relatively slowly, meaning that the gas is discharged from the lungs steadily. This cycle repeats, and the intervals between adjacent inhalation and exhalation peaks and valleys are evenly spaced, reflecting a relatively stable respiratory cycle. However, for elderly patients with respiratory diseases, as the exercise continues, the interval between each peak becomes shorter and the peak becomes smaller. Therefore, by learning the characteristic information contained in the respiratory curve image, the disease condition of the elderly patient can be effectively judged.

[0048] As Figure 4 The audio image shown shows the spectrogram of the patient during coughing. The spectrogram is a tool that displays sound data in the form of a two-dimensional image, which can intuitively present characteristics such as the frequency and amplitude of the sound. In implementation, the spectrogram processes the sound signal, such as by using methods like Fourier transform, to convert the time-domain signal into a frequency-domain signal, and uses the abscissa to represent the time series and the ordinate to represent the sound frequency to show the energy distribution of the sound at different times and frequencies.

[0049] For elderly patients with respiratory diseases, as the exercise continues, the wheezing sound of the elderly patient will become louder and the frequency of coughing will increase. This is manifested in the spectrogram as the interval becoming shorter and the duration of each plosive sound becoming longer. And because the wheezing sound becomes louder, the energy of the spectrogram is concentrated in the low-frequency region. Therefore, by analyzing the characteristic information contained in the spectrogram, the disease condition of the elderly patient can be effectively judged.

[0050] Subsequently, as Figure 1 shown, the respiratory disease diagnosis device for elderly patients in the embodiment of the present application executes step S120, and for each state data, obtains a state feature vector corresponding to the state data. As described above, the method has obtained data from multiple data sources according to step S110, and uses different methods for different data sources to obtain feature vectors that can represent the data source, and fuses these feature vectors to finally obtain a state feature vector corresponding to this time period.

[0051] Specifically, the conventional detection data is formed into a conventional detection feature vector according to the diagnosis result of whether each conventional detection data is abnormal. The conventional detection data may include body temperature data detected by a thermometer, pulse data detected by a pulse detector, blood pressure data detected by a blood pressure monitor, heart rate data detected by a heart rate detector, and blood oxygen data detected by a blood oxygen monitor.

[0052] Specifically, the normal range of body temperature data is usually 36.1°C to 37.2°C (this range may vary due to individual differences). The normal range of pulse data is usually 60 to 100 beats per minute. The normal range of blood pressure data is usually 90 - 139 mmHg for systolic blood pressure and 60 - 89 mmHg for diastolic blood pressure. The heart rate data is similar to the pulse data, and the normal range is usually 60 to 100 beats per minute. The normal range of blood oxygen data is usually 95% to 100%. For each test data, check whether it falls within the normal range defined above. If the data is within the normal range, the corresponding feature value is 0 (indicating normal). If the data exceeds the normal range, the corresponding feature value is 1 (indicating abnormal). To detect the accuracy of the data, the elderly patient can input the normal range under normal conditions by themselves. For example, when the elderly patient is in good health, the body temperature is 36.7°C to 37.3°C.

[0053] For example, in the first minute, the elderly patient detects an average body temperature of 36.5°C, a pulse of 85 beats per minute, a blood pressure of 140 / 90 mmHg, a heart rate of 89 beats per minute, and a blood oxygen of 98%. It can be seen that the body temperature is normal, and the corresponding feature value is 0; the pulse is normal, and the corresponding feature value is 0; the systolic blood pressure is normal but the diastolic blood pressure is abnormal, corresponding to two feature values (1, 0); the heart rate is normal, and the corresponding feature value is 0; the blood oxygen is normal, and the corresponding feature value is 0. Then the feature vector is [1, 0, 1, 0, 0, 0]. When the first time period is 10 minutes, the corresponding feature vectors will be output every minute according to the above processing method.

[0054] After extracting the respiratory curve image detected by the respiratory airflow detection device using the first feature extraction network and performing normalization processing using the first normalization parameter, a respiratory feature vector corresponding to the respiratory curve image is obtained. Specifically, the first feature extraction network can extract features from the respiratory curve image to obtain image features of multiple different scales, and perform normalizing flow processing on the obtained image features of multiple different scales. The normalizing flow processing can include mapping the image features of multiple different scales to the target space. The multiple image features mapped to the target space can be combined through inverse Laplace transform after normalization processing to obtain the respiratory feature vector. More specifically, the first feature extraction network can be a neural network based on normalizing flow, such as a neural network based on PyramidFlow, a neural network based on FastFlow, a neural network based on MSFlow, etc.

[0055] An exemplary first normalization parameter according to the present application is statistical information obtained by statistically analyzing the distribution of the feature vector values of multiple image features. The feature vector values of the multiple image features are the image features obtained by inputting the respiratory curve images corresponding to multiple elderly patients without diseases into a trained first feature extraction network. The statistical information obtained by statistically analyzing the distribution of the feature vector values of the multiple image features can be used as the first normalization parameter. Among them, the statistical information can be, for example, any one or more parameters that can describe the data distribution, such as mean, variance, standard deviation, quantile, skewness, etc. Taking the mean and standard deviation of the feature vector values of the multiple image features as the first normalization parameter as an example, if the feature vector value of the normalized feature is denoted as x', then the normalized feature can be optionally determined by the formula x' = (x - μ) / σ, where x is the feature vector value of the image feature, and μ and σ are the mean and standard deviation of the feature vector values of the multiple image features, respectively. It can be understood that the larger the absolute value of the difference between the feature vector value of a single image feature and the mean of the feature vector values of the multiple image features, the larger the feature vector value of the obtained normalized feature. Therefore, the feature vector value of the normalized feature can reflect to a certain extent. It should be noted that the first normalization parameter used in the normalization process in the embodiments of the present invention is not limited to the mean and standard deviation, and other parameters such as variance, quantile, skewness, etc. can also be used to obtain the normalized feature.

[0056] Subsequently, after performing normalization processing on the audio image by using a second feature extraction network and a second normalization parameter, an audio feature vector corresponding to the audio image is obtained.

[0057] After extracting the audio image by using a second feature extraction network and performing normalization processing by using a second normalization parameter, an audio feature vector corresponding to the audio image is obtained. Specifically, the second feature extraction network can perform feature extraction on the audio image to obtain multiple image features of different scales, and perform a normalizing flow process on the obtained multiple image features of different scales. The normalizing flow process can include mapping the multiple image features of different scales to a target space. The multiple image features mapped to the target space can be combined through an inverse Laplace transform after normalization processing to obtain an audio feature vector. More specifically, the second feature extraction network can be a neural network based on normalizing flow, such as a neural network based on PyramidFlow, a neural network based on FastFlow, a neural network based on MSFlow, etc.

[0058] An exemplary second normalization parameter according to the present application is statistical information obtained by statistically analyzing the distribution of the eigenvector values of multiple audio-image features. The eigenvector values of the multiple audio-image features are image features obtained by inputting the audio images corresponding to multiple elderly patients without diseases into a trained first feature extraction network. The statistical information obtained by statistically analyzing the distribution of the eigenvector values of the multiple audio-image features can be used as the first normalization parameter. Among them, the statistical information can be, for example, any one or more parameters that can describe the data distribution, such as mean, variance, standard deviation, quantile, skewness, etc. It should be noted that the second normalization parameter used in the normalization process in the embodiments of the present invention is not limited to the mean and standard deviation, and other parameters such as variance, quantile, skewness, etc. can also be used to obtain normalized features.

[0059] In addition, the above-mentioned first feature extraction network and second feature extraction network can be networks with the same network structure and trained using different sample images (for example, respiratory curve images or audio images), or networks with different network structures. For example, the first feature extraction network can adopt a neural network based on Fast Normalization Flow (FastFlow), and the second feature extraction network can adopt a neural network based on Multi-Scale Normalization Flow (MSFlow).

[0060] It should be noted that the respiratory feature vector and the audio feature vector obtained according to the above method can be feature vectors of the same size, which is more conducive to fusion to obtain the state feature vector.

[0061] Subsequently, the conventional detection feature vector, the respiratory feature vector, and the audio feature vector are fused to obtain a state feature vector. According to the embodiments of the present application, before inputting the respiratory curve image and the audio image into the corresponding feature extraction network, the number of channels of the feature vector can be determined first to ensure the subsequent feature vector fusion. In an exemplary embodiment of the present application, the scales of the feature vectors corresponding to the respiratory curve image and the audio image can be set to be the same as the sizes of the respiratory curve image and the audio image. For example, if the scales of the respiratory curve image and the audio image are A*B, then the corresponding feature vector size is A*B. In this case, the state feature vector can be a three-channel feature vector, and each channel represents a data source, that is, [conventional detection feature vector, respiratory feature vector, audio feature vector].

[0062] According to another embodiment of the present application, the current state of the elderly patient can be determined by using conventional detection feature vectors, and then the determined current state is numerically represented and used as a single channel. For example, if the current state of the elderly patient is 1, the state feature vector can be represented as [1, respiratory feature vector, audio feature vector]. Another example is that if the current state of the elderly patient is 0, the state feature vector can be represented as [0, respiratory feature vector, audio feature vector]. For the convenience of calculation, in practice, when all the conventional detection data are 1, the channel value of the state feature vector is represented as 1; as long as there is a 0, the channel value of the state feature vector is represented as 1. For example, for the above-mentioned state feature vector [1, 0, 1, 0, 0, 0], the channel value is represented as 0.

[0063] According to another embodiment of the present application, when the dimensions of the respiratory feature vector and the audio feature vector are the same, the present application can first perform feature reduction on the respiratory feature vector and the audio feature vector by using a feature reduction method, and then use a feature fusion method to perform feature fusion on the reduced respiratory feature vector, audio feature vector and conventional detection feature vector to obtain a state feature vector, where the feature reduction method and feature fusion can both adopt technical means known to those skilled in the art and will not be elaborated herein.

[0064] In step S130, a plurality of state feature vectors are input into a machine learning model for time series analysis that has been trained, and the recurrent network layer of the machine learning model includes gated recurrent units corresponding to each second time period to determine the disease index of the elderly patient suffering from a respiratory system disease.

[0065] As Figure 5 shown, the machine learning model includes a recurrent network layer, a fully connected layer, a logic layer, and an aggregation layer, where the recurrent network layer includes gated recurrent units (GRUs) corresponding to time points at each time interval. As an example, when the first time period is 10 minutes and the second time period is 1 minute, each time point corresponds to each minute, and the number of GRUs can be 10.

[0066] GRU belongs to the Recurrent Neural Network (RNN) and is used to solve problems such as long-term memory and gradients in backpropagation. GRU includes an update gate and a reset gate. Among them, the update gate is used to control the degree to which the state information of the previous moment is brought into the current state. The larger the value of the update gate, the more state information of the previous moment is brought in. The reset gate controls how much information of the previous state is written into the current candidate set. The smaller the reset gate, the less information of the previous state is written.

[0067] In this application, after the state feature vector is input into the machine learning model through the recurrent network layer, the recurrent network layer can extract feature data with temporal information. Subsequently, these feature data can be fully connected through a fully connected layer. That is to say, the features of these feature data are retained and mapped through the fully connected layer. In order to extract more features, the fully connected layer usually increases the dimension of the feature data. As an example, the fully connected layer can be composed of linear neurons several times the number of GRUs. Among them, the activation function of the linear neuron can be Tanh(), that is to say, the fully connected layer can be composed of linear neurons with Tanh() as the activation function several times the number of GRUs.

[0068] Tanh() belongs to hyperbolic functions. In mathematics, the hyperbolic tangent is derived from the basic hyperbolic functions hyperbolic sine and hyperbolic cosine. Tanh() can expand and display the features represented in the data generated in the recurrent network layer.

[0069] The logic layer can convert the data output by the fully connected layer into the abnormal probability value at each time point. As an example, the logic layer can be composed of linear neurons with Sigmoid() as the activation function. Sigmoid() is a common S-shaped function and is usually used in the output layer for binary classification. The output of the logic layer can be input into the aggregation layer ( Figure 5 the MAX in). The aggregation layer can be implemented using the function MAX(). That is to say, taking the abnormal probability value p at each time point output by the logic layer max as the input and processing it using MAX(), the maximum abnormal probability value is selected as the disease index for the elderly patient suffering from respiratory diseases. In addition, as an example, a threshold can be set in advance. If p max is greater than the threshold, it is determined that the elderly patient has respiratory diseases. If it is lower than the threshold, it is determined that the elderly patient does not have respiratory diseases.

[0070] The construction of the machine learning model has been completed above. In order to use the machine learning model to diagnose the elderly patients to be diagnosed, it is necessary to train the machine learning model using training data, so as to continuously adjust the full network parameters in the machine learning model.

[0071] Specifically, to train the machine learning model, a large amount of user data needs to be collected first to form a user database, which contains user information of users of all ages. In the implementation process, the medical system can recruit volunteers of all ages, and then conduct lung examinations on these volunteers to confirm whether they have respiratory diseases. Subsequently, each volunteer is instructed to perform a specified action within a first time period, and within the first time period, state data representing the state within each second time period is obtained at intervals of the second time period. According to an exemplary embodiment, after each volunteer wears the portable device, the state data of the volunteer within each minute can be obtained every minute. The state data of each minute includes the conventional detection data detected by the volunteer using the wearable medical detection instrument, the respiratory curve image detected by the respiratory airflow detection device, and the audio image of the volunteer within the predetermined time interval detected by the audio receiving device. Finally, each volunteer corresponds to the information on whether they have chronic lung diseases and the state data of each minute within these 10 minutes.

[0072] Using the correspondence between the state feature vectors determined for each user (e.g., volunteer) in each user database under each second time period and the result of whether each user has a respiratory disease to train the machine learning model, and adjusting the parameters of the entire network until the machine learning model meets the preset requirements. For example, the accuracy rate reaches more than 90%.

[0073] In addition, to diagnose elderly patients more accurately, the method can also let elderly patients perform walks with three different step frequencies respectively, and then confirm the morbidity indices under each step frequency of walking. That is, when the risk level of a to-be-diagnosed elderly patient having a respiratory disease is higher than the preset level, the method can instruct the elderly patient to perform different specified actions within multiple first time periods, obtain the state data within multiple second time periods within each first time period, and then perform subsequent operations according to each state data to determine the morbidity index.

[0074] For example, when the specified actions include walking at a first step frequency of less than 60 steps per minute, a second step frequency of more than 60 steps and less than 120 steps per minute, and a third step frequency of more than 120 steps and less than 150 steps per minute, the diagnostic method according to the embodiment of the present application respectively executes steps S110 to S130 to obtain a first morbidity index, a second morbidity index, and a third morbidity index.

[0075] In other words, the elderly patient to be diagnosed can perform the first walking frequency of less than 60 steps per minute within the first 10 minutes. After the diagnostic method performs the above processing, the first disease index is determined. Then, within the second 10 minutes, the patient performs the second walking frequency of more than 60 steps per minute and less than 120 steps per minute. After the diagnostic method performs the above processing, the second disease index is determined. And within the third 10 minutes, after the patient performs the third walking frequency of more than 120 steps per minute and less than 150 steps per minute, the diagnostic method performs the above processing to determine the third disease index.

[0076] According to an exemplary embodiment of the present application, after obtaining the first disease index, the second disease index, and the third disease index, the diagnostic method can calculate the final disease index by using the first weight factor corresponding to the first disease index, the second weight factor corresponding to the second disease index, and the third weight factor corresponding to the third disease index. Among them, the third weight factor is greater than the first weight factor and the second weight factor, the second weight factor is greater than the first weight factor, and the sum of the first weight factor, the second weight factor, and the third weight factor is 1. This is because the early symptoms of respiratory diseases, especially chronic obstructive pulmonary disease, are very mild, but they will be manifested after exercise, and the more intense the exercise, the easier it is to detect. And the application object of the diagnostic method of the present application is middle-aged and elderly patients without obvious manifestations. Therefore, those skilled in the art have found through multiple experiments that the accuracy rate of the third disease index is the highest, followed by the second disease index, and finally the first disease index. Therefore, the above three weight factors are set based on experience. Preferably, the first weight factor can be set to 0.5, the second weight factor can be set to 0.3, and the third weight factor can be set to 0.2.

[0077] In summary, according to the exemplary embodiment of the present application, the diagnostic method for respiratory diseases of elderly patients can indicate that elderly patients can achieve self-diagnosis through portable equipment, which is convenient for patients to detect and / or prevent in the early stage of the disease and avoid lung damage. Furthermore, the diagnostic method takes into account the characteristics of chronic lung diseases, collects different data sources during the exercise of elderly patients, and improves the accuracy of diagnosis. The method also determines the risk level of the elderly patient suffering from respiratory diseases based on the basic information of the elderly patient to be diagnosed. And according to different risk levels, the elderly patient is prompted to perform subsequent operations, realizing the hierarchical management of elderly patients. Furthermore, when the elderly patient has a high risk and does not have home detection conditions, community clinics and other convenient places are linked, so that the elderly patient can complete the diagnosis without going to the hospital and performing complex lung examinations, solving the pain points of difficult access to medical examinations for the elderly and alleviating the medical pressure on hospitals.

[0078] Figure 6 The block diagram of the diagnostic device for respiratory diseases of elderly patients according to an exemplary embodiment of the present application is shown. Refer toFigure 6 At the hardware level, the device includes a processor, an internal bus, and a computer-readable storage medium. Among them, the computer-readable storage medium includes a volatile memory and a non-volatile memory. The processor reads the corresponding computer program from the non-volatile memory and then runs it. Of course, in addition to the software implementation, this application does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, and can also be hardware or logic devices.

[0079] Specifically, the processor performs the following operations: When the risk level of a geriatric patient to be diagnosed with a respiratory disease is higher than a preset level, instruct the geriatric patient to perform a specified action within a first time period and, within the first time period, obtain status data representing the status within each second time period at intervals of the second time period. Among them, the status data includes routine test data detected by a wearable medical detection instrument of the geriatric patient, a respiratory curve image detected by a respiratory airflow detection device, and an audio image detected by an audio receiving device. The first time period includes a plurality of second time periods. For each status data, obtain a status feature vector corresponding to the status data, including: Construct a routine test feature vector from the routine test data according to the diagnosis result of whether each routine test data is abnormal; after extracting the respiratory curve image detected by the respiratory airflow detection device using a first feature extraction network and performing normalization processing using a first normalization parameter, obtain a respiratory feature vector corresponding to the respiratory curve image; after performing normalization processing on the audio image using a second feature extraction network and a second normalization parameter, obtain an audio feature vector corresponding to the audio image; perform fusion processing on the routine test feature vector, the respiratory feature vector, and the audio feature vector to obtain a status feature vector. Input multiple status feature vectors into a trained machine learning model for time series analysis in the order of the second time period. The recurrent network layer of the machine learning model includes a gated recurrent unit corresponding to each second time period, and determine the disease index of the geriatric patient suffering from a respiratory disease.

[0080] Optionally, the processor may also perform the following operations: Obtain a user database for determining whether a user has a respiratory disease. Among them, each user corresponds to basic information and the result of whether they have a respiratory disease. Among them, the basic information includes the age, gender, weight, smoking history, living environment, and genetic factors of each user. Perform data fitting on the result of whether each user has a respiratory disease and the basic information corresponding to the user to obtain an objective function for having a respiratory disease. Input the basic information of the elderly patient to be diagnosed into the objective function to determine the risk level of the elderly patient suffering from respiratory diseases.

[0081] Optionally, when the specified action is walking at a specific step frequency, the first time period is 10 minutes and the second time period is 1 minute.

[0082] Optionally, the routine detection data includes body temperature data detected by a thermometer, pulse data detected by a pulse detector, blood pressure data detected by a sphygmomanometer, heart rate data detected by a heart rate detector, and blood oxygen data detected by a pulse oximeter.

[0083] Optionally, when the specific step frequency walking includes walking at a first step frequency lower than 60 per minute, a second step frequency higher than 60 and lower than 120 per minute, and a third step frequency higher than 120 and lower than 150 per minute, the disease index obtained under the first step frequency will be used as the first disease index, the disease index obtained under the second step frequency will be used as the second disease index, and the disease index obtained under the third step frequency will be used as the third disease index. Optionally, the processor can also perform the following operations: calculate the final disease index using the first weight factor corresponding to the first disease index, the second weight factor corresponding to the second disease index, and the third weight factor corresponding to the third disease index, where the third weight factor is greater than the first weight factor and the second weight factor, the second weight factor is greater than the first weight factor, and the sum of the first weight factor, the second weight factor, and the third weight factor is 1.

[0084] Optionally, the trained machine learning model for time series analysis is trained in the following manner: Obtain the state feature vectors determined for each user in each second time period in the user database and the results of whether each user suffers from respiratory diseases; Construct the machine learning model for time series analysis, and the machine learning model is set with all network parameters; Use the correspondence between the state feature vectors determined for each user in each second time period and the results of whether each user suffers from respiratory diseases to train the machine learning model and adjust the all network parameters until the machine learning model meets the preset requirements.

[0085] Optionally, the processor can also perform the following operations: The first feature extraction network extracts features from the respiratory curve image to obtain image features of multiple different scales, performs normalization processing on the multiple different scales, and then merges them through inverse Laplace transform to obtain intermediate respiratory features; the intermediate respiratory features are subjected to normalization processing by using the first statistical information as the first normalization parameter to obtain the respiratory feature vector.

[0086] After the audio image is subjected to normalization processing by using the second feature extraction network and the second normalization parameter, an audio feature vector corresponding to the audio image is obtained, including: The second feature extraction network extracts features from the audio image to obtain image features of multiple different scales, performs normalization processing on the multiple different scales, and then merges them through inverse Laplace transform to obtain intermediate audio features; the intermediate audio features are subjected to normalization processing by using the second statistical information as the second normalization parameter to obtain the audio feature vector.

[0087] In summary, the respiratory system disease diagnosis device for elderly patients according to the exemplary embodiment of the present application can indicate that elderly patients can achieve self-diagnosis through portable equipment, which is convenient for patients to detect and / or prevent in the early stage of the disease in a timely manner and avoid causing lung damage. Further, considering the characteristic manifestations of respiratory system diseases, the diagnosis device collects different data sources during the movement of elderly patients, improving the accuracy of diagnosis. The diagnosis device also determines the risk level of the elderly patient suffering from respiratory system diseases based on the basic information of the elderly patient to be diagnosed. And according to different risk levels, the elderly patient is prompted to perform subsequent operations, realizing hierarchical management of elderly patients. Further, in the case where the elderly patient has a high risk and does not have home detection, convenient places such as community clinics are linked, so that the elderly patient can complete the diagnosis without going to the hospital and performing complex lung examinations, solving the pain points of difficult going out and difficult examination for elderly patients, and also alleviating the medical pressure on hospitals.

[0088] It should be noted that the execution subject of each step of the method provided in Embodiment 1 can be the same device, or the method can also be executed by different devices as the execution subject. For example, the execution subject of steps 21 and 22 can be device 1, and the execution subject of step 23 can be device 2; or, the execution subject of step 21 can be device 1, and the execution subjects of steps 22 and 23 can be device 2; and so on.

[0089] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0090] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0091] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0092] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0093] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0094] The memory may include non-permanent memory in the computer-readable medium, random access memory (RAM), and / or non-volatile memory in the form of, for example, read-only memory (ROM) or flash memory (flash RAM). The memory is an example of a computer-readable medium.

[0095] A computer-readable medium includes permanent and non-permanent, removable and non-removable media and can implement information storage by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transitory medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer-readable medium does not include transitory computer-readable media, such as modulated data signals and carrier waves.

[0096] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising the element.

[0097] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0098] The above description is only for the embodiments of the present application and is not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A method for diagnosing respiratory diseases in elderly patients, characterized in that: include: In case the risk level of the elderly patient to be diagnosed suffering from a respiratory disease is higher than a preset level, instructing the elderly patient to perform a specified action within a first time period and acquiring state data representing the second time period every second time period within the first time period, wherein the state data includes conventional detection data detected by the elderly patient using a wearable medical detection instrument, a respiratory curve image detected by a respiratory airflow detection device, and an audio image detected by an audio receiving device, and the first time period includes multiple second time periods; For each state data, a state feature vector corresponding to the state data is obtained, including: The conventional detection data are formed into a conventional detection feature vector according to the diagnosis result of whether each conventional detection data has an abnormality; the respiratory curve image detected by the respiratory airflow detection device is subjected to feature extraction using a first feature extraction network and normalization processing using a first normalization parameter to obtain a respiratory feature vector corresponding to the respiratory curve image; the audio image is subjected to feature extraction using a second feature extraction network and normalization processing using a second normalization parameter to obtain an audio feature vector corresponding to the audio image; the conventional detection feature vector, the respiratory feature vector, and the audio feature vector are fused to obtain a state feature vector; The state feature vector corresponding to each second time period is input into a trained machine learning model based on time series analysis in chronological order to determine the morbidity index of the elderly patient suffering from the respiratory disease. The recurrent network layer of the machine learning model includes a gated recurrent unit corresponding to each second time period.

2. The diagnostic method according to claim 1, characterized in that Also includes: Acquire a user database, wherein the user information of each user in the user database includes basic information and a result of whether the user suffers from the respiratory disease, and the basic information includes the age, gender, weight, smoking history, living environment and genetic factors of each user; Performing data fitting on each user's disease result and basic information to obtain a target function for the respiratory disease; The basic information of the elderly patient is input into the objective function to determine the risk level of the elderly patient suffering from the respiratory disease, wherein the basic information of the elderly patient includes the age, gender, weight, smoking history, living environment and genetic factors of the elderly patient.

3. The diagnostic method according to claim 1, characterized in that The designated action is walking at a specific step frequency, the first time period is 10 minutes, and the second time period is 1 minute.

4. The diagnostic method according to claim 3, characterized in that The conventional detection data includes body temperature data detected by a thermometer, pulse data detected by a pulse detector, blood pressure data detected by a blood pressure meter, heart rate data detected by a heart rate detector, and blood oxygen data detected by a blood oximeter.

5. The diagnostic method according to claim 4, characterized in that The specific walking frequency includes walking at a first frequency of less than 60 steps per minute, walking at a second frequency of more than 60 steps per minute and less than 120 steps per minute, and walking at a third frequency of more than 120 steps per minute and less than 150 steps per minute, respectively obtaining the disease index at the first walking frequency as the first disease index, the disease index at the second walking frequency as the second disease index, and the disease index at the third walking frequency as the third disease index, The method further comprises: The final disease index is calculated using the first weight factor corresponding to the first disease index, the second weight factor corresponding to the second disease index, and the third weight factor corresponding to the third disease index, wherein the third weight factor is greater than the first weight factor and the second weight factor, the second weight factor is greater than the first weight factor, and the sum of the first weight factor, the second weight factor and the third weight factor is 1.

6. The diagnostic method according to claim 1, characterized in that The machine learning model for time series analysis that has been trained is trained in the following manner: Obtaining a state feature vector determined for each user in the user database in each second time period and a disease result of each user; Constructing a machine learning model for the time series analysis, wherein the machine learning model is provided with full network parameters; The machine learning model is trained using the correspondence between the state feature vector determined for each user in each second time period and the disease outcome of each user, and the full network parameters are adjusted until the training accuracy of the machine learning model reaches more than 90 percent.

7. The diagnostic method according to claim 1, characterized in that After performing feature extraction on a respiratory curve image detected by a respiratory airflow detection device using a first feature extraction network and performing normalization processing using a first normalization parameter, a respiratory feature vector corresponding to the respiratory curve image is obtained, including: The first feature extraction network performs feature extraction on the respiratory curve image to obtain multiple image features of different scales, and performs normalization processing on the multiple image features of different scales and then merges them through inverse Laplace transform to obtain a respiratory intermediate feature; after using the first statistical information as a first normalization parameter, the respiratory intermediate feature is normalized to obtain the respiratory feature vector; After performing feature extraction on the audio image using a second feature extraction network and performing normalization processing using a second normalization parameter, obtaining an audio feature vector corresponding to the audio image includes: The second feature extraction network performs feature extraction on the audio image to obtain multiple image features of different scales, and performs normalization processing on the multiple image features of different scales and then merges them through inverse Laplace transform to obtain audio intermediate features; and performs normalization processing on the audio intermediate features after using the second statistical information as a second normalization parameter to obtain the audio feature vector.

8. A diagnostic device for respiratory diseases in elderly patients, characterized in that: include: a state data acquisition unit, for instructing the elderly patient to be diagnosed to perform a specified action within a first time period and acquiring state data representing the second time period every second time period within the first time period, when the risk level of the elderly patient suffering from a respiratory disease is higher than a preset level, wherein the state data includes conventional detection data detected by the elderly patient using a wearable medical detection instrument, a respiratory curve image detected by a respiratory airflow detection device, and an audio image detected by an audio receiving device, and the first time period includes a plurality of second time periods; The state feature vector acquisition unit is used to acquire the state feature vector corresponding to the state data, including: The conventional detection data are formed into a conventional detection feature vector according to the diagnosis result of whether each conventional detection data has an abnormality; the respiratory curve image detected by the respiratory airflow detection device is subjected to feature extraction using a first feature extraction network and normalization processing using a first normalization parameter to obtain a respiratory feature vector corresponding to the respiratory curve image; the audio image is subjected to feature extraction using a second feature extraction network and normalization processing using a second normalization parameter to obtain an audio feature vector corresponding to the audio image; the conventional detection feature vector, the respiratory feature vector, and the audio feature vector are fused to obtain a state feature vector; The disease index determination unit is used to input the state feature vector corresponding to each second time period into a trained machine learning model based on time series analysis in chronological order to determine the disease index of the elderly patient suffering from the respiratory disease. The recurrent network layer of the machine learning model includes a gated recurrent unit corresponding to each second time period.

9. A diagnostic device for respiratory diseases in elderly patients, characterized in that: include: processor; as well as A memory arranged to store computer executable instructions, which when executed cause the processor to perform the respiratory disease diagnosis method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the instructions are executed, the respiratory disease diagnosis method according to any one of claims 1 to 7 is implemented.

Images