A control system and device for intelligent stethoscope equipment

Abstract

The present invention relates to a control system and device for intelligent stethoscope equipment, which belongs to the field of medical equipment. It solves the problem that the position of the stethoscope head is not accurate and objective enough to be judged by the doctor's experience and observation, and is easily affected by the doctor's subjective factors. The system includes: at least one memory storing a computer program, and at least one processor is configured to implement a control method for the stethoscope equipment when executing the computer program. The specific steps are as follows: regularly enhance the initial intensity of the cardiopulmonary sound adaptation signal, and monitor the signal-to-noise ratio of the cardiopulmonary sound adaptation signal at the same time. When the signal-to-noise ratio of the cardiopulmonary sound adaptation signal drops to a preset value, stop increasing the amplification factor of the cardiopulmonary sound adaptation signal; extract the key features of abnormal sounds, and input the key features into the sound diagnosis classification model that has completed the training, and output the abnormal sound classification and diagnosis results. The present invention has the following effects: accurately locate the stethoscope part, personalize the signal intensity, and improve the accuracy and reliability of diagnosis.

Description

Technical Field

[0001] The present invention relates to the technical field of medical equipment, and in particular to a management and control system and device for intelligent stethoscope equipment. Background Art

[0002] Smart stethoscope devices play an important role in medical diagnosis. They can help doctors detect patients' heart and lung sounds more accurately, thereby evaluating and diagnosing patients' health conditions.

[0003] At present, there are some intelligent stethoscope devices on the market. These devices usually use the traditional stethoscope method, that is, the patient's heart and lung sound signals are collected through the stethoscope head, and then the signals are transmitted to the device for analysis and processing. Some intelligent stethoscope devices are also equipped with some auxiliary functions, such as noise detection, signal amplification, etc. In terms of determining the body part information where the stethoscope head contacts the patient, the existing technology mainly judges through the doctor's experience and observation, or obtains some basic information through simple sensor detection. In terms of analyzing the initial intensity of the heart and lung sound adaptation signal and the increase in the amplification factor, the existing technology usually adopts fixed parameter settings or adjusts based on some simple rules.

[0004] However, the existing technology has some obvious drawbacks. First, relying on the doctor's experience and observation to determine the position of the stethoscope head is not accurate and objective, and is easily affected by the doctor's subjective factors. Second, simple sensor detection can only obtain limited information and cannot fully and accurately determine the position of the stethoscope head and the patient's physical condition. Summary of the invention

[0005] In order to accurately locate the auscultation site, personalize the signal strength, and improve the accuracy and reliability of diagnosis, the present invention provides a control system and device for intelligent stethoscope equipment.

[0006] In a first aspect, the present invention provides a management and control system for an intelligent stethoscope device, which adopts the following technical solution:

[0007] A management and control system for an intelligent stethoscope device, comprising:

[0008] Auscultatory equipment;

[0009] at least one processor;

[0010] At least one memory storing a computer program, and at least one processor configured to implement a method for controlling a stethoscope device when executing the computer program, the specific steps are as follows:

[0011] Obtain information about the body part where the stethoscope head contacts the patient;

[0012] According to the correspondence between the body part and the initial strength of the heart and lung sound adaptation signal, the initial strength of the heart and lung sound adaptation signal of the body part contacted by the stethoscope head is analyzed and determined, and the stethoscope device is instructed to perform abnormal sound detection according to the analyzed and determined initial strength of the heart and lung sound adaptation signal;

[0013] Analyze whether there is abnormal sound within the preset monitoring period according to the preset abnormal sound judgment method;

[0014] If yes, the initial intensity of the abnormal sound is detected and obtained, and the increase in the amplification factor is analyzed and determined according to the corresponding relationship between the initial intensity range recorded for the initial intensity of the abnormal sound and the increase in the amplification factor of the heart and lung sound adaptation signal;

[0015] The initial strength of the heart and lung sound adaptation signal is regularly enhanced according to the increase of the amplification factor determined by the analysis, and the signal-to-noise ratio of the heart and lung sound adaptation signal is monitored and obtained at the same time. When the signal-to-noise ratio of the heart and lung sound adaptation signal drops to a preset value, the increase of the amplification factor of the heart and lung sound adaptation signal is stopped;

[0016] Extract key features of abnormal sounds, input the key features into the trained sound diagnosis classification model, and output abnormal sound classification and diagnosis results;

[0017] If not, it is preliminarily determined that there is no problem with the corresponding body part of the patient.

[0018] By adopting the above technical solution, the initial intensity of the cardiopulmonary sound adaptation signal at the auscultation site can be accurately determined, thereby improving the accuracy of abnormal sound detection. The amplification factor is adjusted according to the intensity of the abnormal sound, and the signal is enhanced while ensuring the signal-to-noise ratio. By extracting key features and using the sound diagnosis classification model, accurate abnormal sound classification and diagnosis results can be quickly output. If no abnormal sound is detected, it can be preliminarily determined that there is no problem with the body part, which improves diagnostic efficiency and provides patients with more timely and accurate diagnostic services, which helps to detect disease hazards early, improve medical quality and safety, reduce the risk of misdiagnosis, provide medical workers with powerful diagnostic tools, and promote the development of intelligent medical care.

[0019] Optionally, obtaining information about the body part where the stethoscope head contacts the patient includes:

[0020] The camera device preset on the stethoscope head is started to capture and obtain the skin area image within the preset distance range of the stethoscope head, and the pressure detection device preset on the stethoscope head is simultaneously started to detect and obtain the pressure distribution information between the stethoscope head and the patient's body, and the space positioning device preset on the stethoscope head is started to detect and obtain the movement-related information of the stethoscope head;

[0021] The acquired skin area image, pressure distribution information and movement related information of the stethoscope head are respectively preprocessed;

[0022] The skin area image, pressure distribution information and movement-related information of the stethoscope head after data preprocessing are taken as input objects and input into the trained stethoscope head position information analysis neural network model, and the position information of the stethoscope head is output.

[0023] By adopting the above technical solution, the camera device, pressure detection device and spatial positioning device are used to collect information in multiple dimensions to improve the accuracy of position judgment. Data preprocessing ensures the quality of information, and after input into the neural network model, the position information of the stethoscope head can be accurately output. This helps to ensure that subsequent operations such as initial intensity analysis of the cardiopulmonary sound adaptation signal are performed according to the correct body part, thereby improving the pertinence and accuracy of the diagnosis. It provides medical personnel with a more reliable basis for diagnosis and reduces misdiagnosis caused by inaccurate position.

[0024] Optionally, obtaining information about the body part where the stethoscope head contacts the patient includes:

[0025] The camera device preset on the stethoscope head is started to capture and obtain the skin area image within the preset distance range of the stethoscope head, and the pressure detection device preset on the stethoscope head is simultaneously started to detect and obtain the pressure distribution information between the stethoscope head and the patient's body, and the space positioning device preset on the stethoscope head is started to detect and obtain the movement-related information of the stethoscope head;

[0026] Extracting image features from the acquired skin area image within a preset distance range of the stethoscope head, performing similarity matching between the extracted image features and a pre-established body part image feature database, and determining the body part where the stethoscope head is located based on the similarity matching result;

[0027] Extracting the characteristics of pressure distribution from the pressure distribution information and comparing it with typical pressure distribution patterns of different body parts to determine the body part where the stethoscope head is located;

[0028] Extract motion features from the motion-related information of the stethoscope head, and analyze the matching degree between the motion features and the motion patterns of different body parts to determine the body part where the stethoscope head is located;

[0029] According to the preset distribution weights of the camera device, the pressure detection device and the space positioning device, a weighted voting method is adopted to select the body part with the highest weighted vote as the final position of the stethoscope head.

[0030] By adopting the above technical solution, different features are extracted through the camera device, pressure detection device and spatial positioning device to determine the position, thereby improving accuracy. Allocating weights and using weighted voting to determine the final position is more reliable. It can ensure accurate acquisition of information on the body parts where the stethoscope head contacts the patient, providing a basis for subsequent accurate diagnosis.

[0031] Optionally, according to the correspondence between the body part and the initial strength of the heart and lung sound adaptation signal, analyzing and determining the initial strength of the heart and lung sound adaptation signal of the body part contacted by the stethoscope head includes:

[0032] Obtain relevant key information of the patient, including age, gender and body shape;

[0033] Standardize the key information of the patients and extract features;

[0034] According to the extracted features, an individualized model matching the patient is obtained from an individualized model database;

[0035] According to the individualized model matched to the patient, the correspondence between the body part and the initial intensity of the adaptation signal of the heart and lung sounds is analyzed and determined, and the initial intensity of the adaptation signal of the heart and lung sounds of the body part contacted by the stethoscope head is determined.

[0036] By adopting the above technical solutions, key information such as the patient's age, gender, and body shape is taken into consideration to improve the accuracy of adaptation. After standardization and feature extraction, the individualized model is matched to make the analysis more in line with the patient's actual situation. Determining the initial intensity based on the individualized model and the corresponding relationship can improve the accuracy and reliability of auscultation, provide doctors with more accurate diagnostic basis, reduce the risk of misdiagnosis, better meet the personalized needs of different patients, promote the application of intelligent stethoscope equipment in medical diagnosis, and improve medical quality and efficiency.

[0037] Optionally, obtaining an individual model matching the patient from an individualized model database includes:

[0038] Analyze whether the extracted features are matched to the personalized model that is completely consistent with the patient from the personalized model database;

[0039] If yes, then according to the extracted features, a personalized model matching the patient is obtained from the personalized model database;

[0040] If not, all candidate personalized models applicable to the patient are extracted from the personalized model database according to the extracted features;

[0041] According to all candidate individualized models applicable to the patient and the extracted features, a preset similarity measurement method is used to analyze and determine the candidate individualized model with the highest similarity as the individualized model used this time.

[0042] By adopting the above technical solution, an individualized model that matches the patient can be effectively obtained with good results. If a completely consistent model can be matched, it can be used directly, which is efficient and accurate. If there is no completely consistent model, the candidate model is extracted and analyzed to determine the one with the highest similarity to ensure that the model fits the patient as closely as possible. This can improve the accuracy of determining the initial intensity of the cardiopulmonary sound adaptation signal, provide doctors with more accurate diagnostic basis, meet the personalized needs of different patients, improve the diagnostic efficiency of intelligent stethoscope equipment, promote more accurate and personalized medical diagnosis, reduce the risk of misdiagnosis, and improve medical quality and efficiency.

[0043] Optionally, a preset similarity measurement method is used to analyze and determine the candidate individualization models with the highest similarity, including:

[0044] Using multiple similarity measurement methods, analyze and obtain the measurement values ​​of the same candidate individualized model under different similarity measurement methods, and analyze one by one to obtain the measurement values ​​of all candidate individualized models under different candidate individualized models;

[0045] Analyze whether there is a candidate individualized model whose measurement value reaches the preset measurement value;

[0046] If yes, the corresponding candidate individualized model is used as the individualized model for this application;

[0047] If not, the individualized model that meets all the screening conditions is screened out from the candidate individualized models according to the preset individualized model screening conditions as the individualized model for this application, wherein the screening conditions include the measurement thresholds of the same candidate individualized model under different similarity measurement methods.

[0048] By adopting the above technical solutions, it is possible to ensure that the most suitable individualized model for the patient is selected, which is of positive significance. Multiple similarity measurement methods improve accuracy. If there is a measurement value that meets the standard, the model is directly determined, which is efficient and fast. If not, the model is determined according to the screening conditions to ensure the reliability of the model. It can provide an accurate basis for determining the initial strength of the cardiopulmonary sound adaptation signal, improve the accuracy and personalization of diagnosis, and reduce the risk of misdiagnosis.

[0049] Optionally, the magnification increase range is obtained by:

[0050] Analyze whether to obtain the patient's age, physical condition information and medical history;

[0051] If yes, then according to the preset weights matching different age ranges, weights matching different physical conditions and weights matching different medical histories, the average weight of the patient in terms of age, physical condition and medical history is analyzed and obtained as the weight combination result of the patient;

[0052] The weight combination result of the patient and the preliminary intensity range recorded by the preliminary intensity of the abnormal sound are used as input features, input into the analysis model of the amplification adjustment coefficient that has completed the training, and the amplification adjustment coefficient is output;

[0053] If not, the noise detection device preset on the intelligent stethoscope device is activated to detect and obtain the noise intensity within the preset distance range;

[0054] According to the correspondence between the noise intensity interval range that the noise intensity falls into, the preliminary intensity range recorded by the preliminary intensity of the abnormal sound and the increase in the amplification factor of the heart and lung sound adaptation signal, the increase in the amplification factor of the heart and lung sound adaptation signal is analyzed and determined.

[0055] By adopting the above technical solution, if the patient's age, physical condition and other information can be obtained, a more personalized amplification adjustment coefficient can be obtained by combining weight analysis. If not, the coefficient is determined by detecting the noise intensity and combining the corresponding relationship to ensure accuracy. In this way, the appropriate amplification factor can be provided for the cardiopulmonary sound adaptation signal adjustment according to different situations, improving the accuracy and reliability of diagnosis, meeting the needs of different scenarios, and enabling smart stethoscope equipment to better adapt to various situations, providing strong support for medical diagnosis, improving medical quality and efficiency, and reducing the risk of misdiagnosis.

[0056] In another aspect, the present invention provides a control device for an intelligent stethoscope device, comprising:

[0057] An acquisition module, used for acquiring information of the body part where the stethoscope head contacts the patient;

[0058] The analysis module is used to analyze and determine the initial strength of the heart and lung sound adaptation signal of the body part contacted by the auscultation head according to the corresponding relationship between the body part and the initial strength of the heart and lung sound adaptation signal, and instruct the auscultation device to perform abnormal sound detection according to the analyzed and determined initial strength of the heart and lung sound adaptation signal; and analyze whether there is an abnormal sound within a preset monitoring period according to a preset abnormal sound judgment method;

[0059] The execution module detects and obtains the initial intensity of the abnormal sound if there is any abnormal sound, and analyzes and determines the increase in the amplification factor based on the correspondence between the initial intensity range recorded for the initial intensity of the abnormal sound and the increase in the amplification factor of the cardiopulmonary sound adaptation signal; regularly enhances the initial intensity of the cardiopulmonary sound adaptation signal according to the analyzed and determined increase in the amplification factor, and monitors and obtains the signal-to-noise ratio of the cardiopulmonary sound adaptation signal at the same time. When the signal-to-noise ratio of the cardiopulmonary sound adaptation signal drops to a preset value, the increase in the amplification factor of the cardiopulmonary sound adaptation signal is stopped; the key features of the abnormal sound are extracted, and the key features are input into the trained sound diagnosis classification model, and the abnormal sound classification and diagnosis results are output. If there is no abnormal sound, it is preliminarily determined that there is no problem with the corresponding body part of the patient. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 It is a schematic diagram of the overall process of a method for controlling a stethoscope device in a control system of an intelligent stethoscope device according to an embodiment of the present invention.

[0061] Figure 2 It is a schematic diagram of a process of obtaining information about the body part in contact between the stethoscope head and the patient according to another embodiment of the present invention.

[0062] Figure 3 It is a schematic diagram of a process of obtaining information about the body part in contact between the stethoscope head and the patient according to another embodiment of the present invention.

[0063] Figure 4 It is a flowchart of another embodiment of the present invention for analyzing and determining the initial strength of the heart and lung sound adaptation signal of the body part contacted by the stethoscope head according to the corresponding relationship between the body part and the initial strength of the heart and lung sound adaptation signal. DETAILED DESCRIPTION

[0064] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0065] Reference Figure 1 , a management and control system of an intelligent stethoscope device disclosed in the present invention, including a stethoscope device, at least one processor and at least one memory storing a computer program, at least one processor is configured to implement a management and control method of the stethoscope device when executing the computer program, and the specific steps are as follows:

[0066] Step S100, obtaining information about the body part where the stethoscope head contacts the patient.

[0067] Specifically, the "stethoscope head" is the part of the smart stethoscope device that contacts the patient's body to collect sound signals. "Body part information" refers to the specific location of the patient's body that the stethoscope head contacts, such as the chest, back, and other different parts of the body. The characteristics of heart and lung sounds in different parts of the body may be different.

[0068] Acquiring information about the body part where the stethoscope head contacts the patient can be achieved in the following ways:

[0069] 1. Sensor detection: A position sensor is installed on the stethoscope head. When the stethoscope head contacts the patient's body, the sensor can detect the contact position and transmit the position information to the control system of the intelligent stethoscope device. The position sensor can be a pressure sensor, a capacitive sensor, etc., which determines the contacted body part by detecting the pressure change or capacitance change at different positions.

[0070] 2. Image recognition: Use a camera or other image acquisition equipment to capture images of the part where the stethoscope head contacts the patient's body. Then use image recognition technology to analyze the features in the image and determine the part of the body that the stethoscope head contacts. Image recognition can use deep learning algorithms to train a large number of labeled images to improve recognition accuracy.

[0071] Step S200, based on the correspondence between the body part and the initial strength of the heart and lung sound adaptation signal, analyze and determine the initial strength of the heart and lung sound adaptation signal of the body part that the stethoscope head contacts, and instruct the stethoscope device to perform abnormal sound detection according to the analyzed and determined initial strength of the heart and lung sound adaptation signal.

[0072] "Initial strength of the heart and lung sound adaptation signal" refers to the strength standard that the smart stethoscope device should set in the initial state for receiving heart and lung sound signals for different body parts. This strength standard is determined based on the characteristics of different body parts and past medical experience and data.

[0073] The process of establishing the correspondence between body parts and the initial intensity of the heart and lung sound adaptation signal is as follows: First, through a large number of medical studies and clinical experiments, normal heart and lung sound data of different body parts are collected. These data are analyzed to determine the typical heart and lung sound characteristics of each body part, including frequency range, intensity range, etc. Then, based on these characteristics, the corresponding initial intensity of the heart and lung sound adaptation signal is set for each body part. For example, some areas of the chest may require a higher initial intensity, while some parts of the back may require a lower initial intensity.

[0074] The process of analyzing and determining the initial strength of the heart and lung sound adaptation signal of the body part in contact with the stethoscope head is as follows: using the body part information obtained in step S100 that the stethoscope head contacts with the patient, the initial strength of the heart and lung sound adaptation signal corresponding to the body part is searched in the established correspondence.

[0075] The process of instructing the stethoscope device to detect abnormal sounds according to the initial intensity of the heart and lung sound adaptation signal determined by the analysis is as follows: First, after receiving the determined initial intensity, the intelligent stethoscope device adjusts its own signal reception and processing parameters, and receives the heart and lung sound signal based on the initial intensity. Then, using specific algorithms and signal processing techniques, the received heart and lung sound signals are analyzed in real time to detect whether there are abnormal sounds. Abnormal sounds may include murmurs, wheezing, friction sounds, etc. These sounds may be important indicators of disease.

[0076] The abnormal sound detection algorithm used above is an abnormal sound detection algorithm based on wavelet transform and adaptive threshold. This algorithm combines the advantages of wavelet transform in time-frequency analysis and the flexibility of adaptive threshold, and can effectively detect abnormal sounds from cardiopulmonary sound signals.

[0077] The specific steps of the corresponding abnormal sound detection algorithm are as follows: 1. Signal preprocessing: filter the heart and lung sound signals received by the stethoscope to remove high-frequency noise and low-frequency interference. A bandpass filter can be used to filter according to the frequency range of the heart and lung sounds (usually 20Hz-2000Hz).

[0078] Normalize the filtered signal and adjust the signal amplitude to a suitable range for subsequent analysis and processing. 2. Wavelet transform: Perform wavelet transform on the preprocessed heart and lung sound signal. Select a suitable wavelet basis function, such as Daubechies wavelet or Symlet wavelet. Through wavelet transform, the heart and lung sound signal is decomposed into wavelet coefficients of different scales. These wavelet coefficients contain information about the signal at different frequencies and times. 3. Feature extraction: Extract parameters that can reflect the characteristics of the signal from the wavelet coefficients, such as energy, entropy, standard deviation, etc. Calculate the energy distribution of wavelet coefficients at different scales and analyze the energy changes of the signal in different frequency ranges. 4. Adaptive threshold determination: Determine an adaptive threshold based on the characteristic statistics of normal heart and lung sound signals. The mean and standard deviation of the characteristic parameters can be calculated by analyzing a large number of normal heart and lung sound signals, and then the threshold can be determined according to certain rules. For example, a threshold determination method based on the mean and standard deviation can be adopted, such as threshold = mean + k* standard deviation, where k is an adjustment coefficient that can be adjusted according to actual conditions. 5. Abnormal sound detection: The extracted feature parameters are compared with the adaptive threshold. If the feature parameters exceed the threshold, it is considered that there may be abnormal sound. The time period where the abnormal sound may exist is marked, and the signal characteristics in this time period are further analyzed to determine the type and severity of the abnormal sound.

[0079] Step S300, analyzing whether there is abnormal sound in the preset monitoring period according to the preset abnormal sound judgment method. If yes, execute step S400; if no, execute step S700.

[0080] The "monitoring cycle" refers to a specific period of time during which heart and lung sounds are continuously monitored to determine whether there are any abnormal sounds. The monitoring cycle can be determined based on the patient's condition. For patients with serious conditions, such as those with severe heart disease or lung disease, the monitoring cycle can be set to be shorter, such as every 5 minutes. For healthy people who are undergoing general physical examinations, a longer monitoring cycle can be set, such as every 30 minutes.

[0081] "Abnormal sounds" refer to sounds that are different from normal heart and lung sounds, which may indicate that there are some problems with the patient's body. During the preset monitoring period, the smart stethoscope device continuously collects the patient's heart and lung sound signals.

[0082] The collected cardiopulmonary sound signals are compared with the standard cardiopulmonary sound library and analyzed using the preset judgment method. If the signal characteristics exceed the normal range, it is considered that there is an abnormal sound. For example, if the detected cardiopulmonary sound frequency suddenly increases and lasts for a period of time, exceeding the normal frequency range by 20%, it is judged that there may be an abnormal sound.

[0083] Step S400, detecting and obtaining the initial intensity of the abnormal sound, and analyzing and determining the increase in the amplification factor according to the correspondence between the initial intensity range recorded for the initial intensity of the abnormal sound and the increase in the amplification factor of the cardiopulmonary sound adaptation signal.

[0084] Among them, "abnormal sound initial intensity" refers to the value obtained by preliminary measurement of its intensity after the abnormal sound is detected. "Preliminary intensity range" divides the intensity of abnormal sound into different intervals. "Increase in amplification factor of the cardiopulmonary sound adaptation signal" refers to the degree of amplification of the cardiopulmonary sound adaptation signal determined according to the intensity of the abnormal sound.

[0085] The method of detecting and obtaining the preliminary intensity of abnormal sound is as follows: using sensor technology, when the auscultation device detects abnormal sound, the intensity detection function is immediately started. For example, the sound signal can be received by a microphone sensor, and the intensity of the sound can be calculated by a signal processing algorithm. The detected abnormal sound is analyzed in real time to determine its preliminary intensity. For example, after converting the sound signal into a digital signal, its parameters such as amplitude or energy are calculated as an indicator of the preliminary intensity.

[0086] The process of establishing the correspondence between the preliminary intensity range and the increase in amplification factor is as follows: through a large number of experiments and data analysis, the increase in amplification factor of the cardiopulmonary sound adaptation signal corresponding to different preliminary intensity ranges is determined. For example, the preliminary intensity of the abnormal sound is divided into three ranges: low, medium, and high, corresponding to different increases in amplification factors. This correspondence can also be expressed in the form of a table or function. For example, a table is established to list different preliminary intensity ranges and corresponding increases in amplification factors; or a mathematical function is established to input the preliminary intensity of the abnormal sound and output the corresponding increase in amplification factor.

[0087] The process of analyzing and determining the increase in amplification factor is as follows: according to the initial intensity of the abnormal sound detected, the corresponding increase in amplification factor is searched in the established correspondence relationship. For example, if the initial intensity of the abnormal sound belongs to the medium range, then the corresponding medium increase in amplification factor is determined.

[0088] Assume that the initial intensity range of abnormal sounds is divided into three intervals: low intensity interval (0-30 decibels), medium intensity interval (31-60 decibels), and high intensity interval (61 decibels and above). The corresponding increases in the amplification of the cardiopulmonary sound adaptation signal are 2 times, 4 times, and 6 times, respectively.

[0089] If the initial intensity of the abnormal sound detected is 45 decibels, which belongs to the medium intensity range, then according to the corresponding relationship, the increase in the amplification factor is determined to be 4 times. At this time, the stethoscope will amplify the cardiopulmonary sound adaptation signal according to this increase in the amplification factor, so as to detect and analyze the abnormal sound more clearly.

[0090] Step S500, periodically enhance the initial strength of the heart and lung sound adaptation signal according to the increase in the amplification factor determined by the analysis, and monitor the signal-to-noise ratio of the heart and lung sound adaptation signal at the same time. When the signal-to-noise ratio of the heart and lung sound adaptation signal drops to a preset value, stop increasing the amplification factor of the heart and lung sound adaptation signal.

[0091] "Amplification increase" is the degree to which the adaptation signal of the heart and lung sounds is enhanced according to the initial strength of the abnormal sound. "Initial strength of the adaptation signal of the heart and lung sounds" is the starting signal strength for a specific body part determined in step S200. "Signal-to-noise ratio" refers to the ratio of signal to noise, which is used to measure the quality of the adaptation signal of the heart and lung sounds. "Preset value" is the preset lower limit of the signal-to-noise ratio. When the signal-to-noise ratio drops to this value, the signal amplification is stopped.

[0092] The process of enhancing the initial strength of the heart and lung sound adaptation signal according to the amplification increase amplitude is as follows: using an amplification circuit or a digital signal processing algorithm, the initial strength of the heart and lung sound adaptation signal is periodically enhanced according to the determined amplification increase amplitude. For example, the signal can be amplified once according to the amplification increase amplitude at regular intervals.

[0093] The process of monitoring and obtaining the signal-to-noise ratio of the heart and lung sound adaptation signal is as follows: The signal-to-noise ratio of the heart and lung sound adaptation signal is monitored in real time through signal analysis technology. The ratio of the signal power to the noise power can be calculated as an indicator of the signal-to-noise ratio. The noise component in the signal is separated by using sensors and signal processing algorithms, and the noise power is calculated. At the same time, the power of the signal is determined to obtain the signal-to-noise ratio.

[0094] The process of stopping the increase of the amplification factor when the signal-to-noise ratio drops to a preset value is as follows: a reasonable preset value of the signal-to-noise ratio is set, which can be determined based on clinical experience, experimental data or industry standards. For example, after a large number of experiments, it is found that when the signal-to-noise ratio is lower than a certain value, the detection accuracy of abnormal sounds will significantly decrease, and this value can be used as a preset value. When the monitored signal-to-noise ratio drops to the preset value, the increase of the amplification factor of the cardiopulmonary sound adaptation signal is immediately stopped. The operation of stopping amplification can be achieved by controlling the amplification circuit or adjusting the signal processing algorithm.

[0095] Assume that the amplification factor increases by 3 times and the initial strength of the cardiopulmonary sound adaptation signal is 10 millivolts. At the beginning, the signal is enhanced according to the increase in the amplification factor, once every 1 second. At the same time, the signal-to-noise ratio is monitored in real time. If the initial signal-to-noise ratio is 50dB, as the signal is amplified, the noise may also increase accordingly, resulting in a decrease in the signal-to-noise ratio. Assume that the preset value is 30dB. When the signal-to-noise ratio drops to 30dB, stop increasing the signal amplification factor. For example, after several amplifications, the signal-to-noise ratio drops to 32dB. At this time, continuing to amplify may cause the signal-to-noise ratio to drop further, affecting the detection of abnormal sounds. When the signal-to-noise ratio drops to 30dB, stop amplifying immediately to ensure the accuracy and reliability of the detection.

[0096] Step S600, extracting key features of abnormal sounds, and inputting the key features into the trained sound diagnosis classification model, and outputting abnormal sound classification and diagnosis results.

[0097] Among them, "key features of abnormal sounds" refer to unique attributes that can distinguish different types of abnormal sounds, such as frequency characteristics, duration, amplitude changes, etc. The "sound diagnosis classification model" is a trained model for classifying and diagnosing abnormal sounds, which can be an algorithm model based on machine learning or deep learning. "Abnormal sound classification" is to divide the detected abnormal sounds into different categories, such as different types of heart murmurs, types of abnormal lung sounds, etc. The "diagnosis result" is a judgment on the patient's possible diseases or health problems based on the classification of abnormal sounds.

[0098] Extracting the key features of abnormal sounds involves the following: 1. Signal processing technology: Processing the detected abnormal sound signals, such as Fourier transform, wavelet transform, etc., to extract features in frequency, time domain, and frequency domain. For example, the spectrum characteristics of abnormal sounds can be obtained through Fourier transform, including the main frequency components and frequency distribution range. 2. Feature engineering: Design and select appropriate features to describe abnormal sounds. Features such as the intensity, duration, and periodicity of the sound can be considered. For example, for heart murmurs, features such as the intensity of the murmur, the time of occurrence, and the relationship with the heartbeat cycle can be extracted. 3. Feature dimensionality reduction: If the number of extracted features is large, feature dimensionality reduction techniques such as principal component analysis (PCA) can be used to reduce the dimension of the features and improve computational efficiency and model accuracy.

[0099] Inputting key features into the trained sound diagnosis classification model involves the following: 1. Model training: Use a large amount of labeled abnormal sound data to train the sound diagnosis classification model. These data can come from clinical cases, medical databases, or special sound collection experiments. Through training, the model learns the characteristic patterns of different types of abnormal sounds, so that it can classify and diagnose new abnormal sounds. 2. Model selection: You can choose a suitable machine learning or deep learning model, such as support vector machine (SVM), random forest, convolutional neural network (CNN), etc. Different models have different characteristics and advantages, and can be selected according to specific application scenarios and data characteristics. 3. Model evaluation and optimization: Evaluate the trained model and use the test set data to verify the accuracy and generalization ability of the model. If the model performance is not ideal, it can be optimized by adjusting model parameters, increasing training data, improving feature extraction methods, etc.

[0100] Outputting abnormal sound classification and diagnosis results involves the following: 1. Classification results: The model classifies the input abnormal sound into different abnormal sound categories based on its key features. For example, heart murmurs are classified into different types such as systolic murmurs and diastolic murmurs. 2. Diagnosis results: Based on the abnormal sound classification, combined with medical knowledge and clinical experience, possible diagnosis results are given. For example, if a specific type of heart murmur is detected, it may indicate that the patient has problems such as heart valve disease.

[0101] Step S700, preliminarily determine that there is no problem with the corresponding body part of the patient.

[0102] Reference Figure 2 , to obtain information about the body parts where the stethoscope head contacts the patient, including:

[0103] Step S110, start the camera device preset on the stethoscope head to capture and obtain the skin area image within the preset distance range of the stethoscope head, and simultaneously start the pressure detection device preset on the stethoscope head to detect and obtain the pressure distribution information of the contact between the stethoscope head and the patient's body, and start the space positioning device preset on the stethoscope head to detect and obtain the movement-related information of the stethoscope head.

[0104] Specifically, the "camera device" is a device used to capture images of the skin area near the stethoscope head, which can provide visual information to assist in determining the position of the stethoscope head. The "skin area image" is an image of the patient's skin surface within a certain distance around the stethoscope head captured by the camera. The "pressure detection device" is used to detect the pressure distribution when the stethoscope head contacts the patient's body, which can help determine the degree of contact and position between the stethoscope head and the body. "Pressure distribution information" refers to the pressure values ​​and distribution at different positions detected by the pressure detection device. The "spatial positioning device" can detect motion-related information of the stethoscope head, including position, direction, acceleration, etc., which helps to more accurately determine the position change of the stethoscope head. "Motion-related information" is data detected by the spatial positioning device about the motion state of the stethoscope head in space.

[0105] Starting the camera device preset on the stethoscope head to capture and obtain the image of the skin area within the preset distance range of the stethoscope head involves the following contents: Installation position: Install the camera device at the appropriate position of the stethoscope head to ensure that the skin area within the preset distance range around the stethoscope head can be captured. For example, a small camera can be installed on the side or top of the stethoscope head. Shooting parameter setting: Set the shooting parameters of the camera device according to actual needs, such as resolution, frame rate, exposure, etc. For example, in order to obtain a clear image, a higher resolution and appropriate exposure can be set. Image acquisition: When the stethoscope head is in contact with the patient's body, start the camera device to shoot and obtain the image of the skin area. Multiple images can be taken continuously for subsequent analysis.

[0106] Synchronously starting the pressure detection device preset on the stethoscope head to detect and obtain the pressure distribution information of the contact between the stethoscope head and the patient's body involves the following contents: Type of pressure detection device: A pressure sensor, such as a thin film pressure sensor, a piezoelectric pressure sensor, etc., can be used, which is installed at the contact part between the stethoscope head and the patient's body. Detection principle: When the stethoscope head is in contact with the patient's body, the pressure detection device is subjected to pressure and generates a corresponding electrical signal. By processing and analyzing these electrical signals, the pressure distribution information can be obtained. Data acquisition: The output signal of the pressure detection device is collected in real time, and the pressure values ​​and changes at different positions are recorded.

[0107] The following contents are involved in obtaining the movement-related information of the stethoscope head by starting the spatial positioning device preset on the stethoscope head: Selection of spatial positioning device: A combination of sensors such as accelerometers, gyroscopes, and magnetometers can be used, or positioning technology based on wireless signals, such as Bluetooth positioning, ultra-wideband positioning, etc. Detection method: The spatial positioning device determines the position, direction, and movement state of the stethoscope head by measuring the acceleration, angular velocity, magnetic field strength and other parameters of the stethoscope head, combined with the positioning algorithm. Data transmission: The detected movement-related information is transmitted to the control system of the intelligent stethoscope device in real time for subsequent processing.

[0108] For example, in an intelligent stethoscope device, a miniature camera is installed on the side of the stethoscope head, which can capture images of skin areas within a diameter of 5 cm. When the stethoscope head contacts the patient's chest, the camera starts shooting to obtain clear skin images. At the same time, a thin film pressure sensor is installed at the part where the stethoscope head contacts the patient's body. When the stethoscope head contacts the patient's body, the pressure sensor detects the pressure distribution at different positions. For example, areas with greater pressure may indicate parts where the stethoscope head is in close contact with the body. In addition, an accelerometer, gyroscope, and magnetometer are installed inside the stethoscope head. These sensors can detect information related to the movement of the stethoscope head, such as acceleration, angular velocity, and magnetic field strength. When the doctor moves the stethoscope head, the spatial positioning device can track the position and direction changes of the stethoscope head in real time.

[0109] Through the coordinated work of the above three devices, skin area images, pressure distribution information and movement-related information within a preset distance range of the stethoscope head can be obtained, providing multi-faceted data support for the subsequent determination of the position of the stethoscope head.

[0110] Step S120, pre-processing the acquired skin area image, pressure distribution information and movement-related information of the stethoscope head.

[0111] Data preprocessing of skin area images involves the following: Image denoising: Use image denoising algorithms to remove noise caused by shooting environment or equipment to improve image quality. For example, median filtering, Gaussian filtering and other methods can be used to remove noise. For example, for a skin area image polluted by salt and pepper noise, using median filtering can effectively remove noise points and make the image clearer. Image enhancement: By enhancing the contrast, brightness and other features of the image, the key information in the image is highlighted. Image enhancement can be performed using methods such as histogram equalization and gamma correction. For example, for a skin area image with low contrast, using histogram equalization can make the grayscale distribution of the image more uniform and enhance the contrast of the image. Image resizing: Adjust the image to a size suitable for neural network input. Interpolation algorithms can be used for image resizing, such as bilinear interpolation, bicubic interpolation, etc. For example, skin area images of different resolutions are uniformly adjusted to a size of 224×224 so that they can be input into the neural network model for processing.

[0112] The data preprocessing of pressure distribution information involves the following: Data normalization: Normalize the pressure values ​​in the pressure distribution information so that they are in the range of [0,1] to facilitate the processing of the neural network model. For example, for pressure distribution information with a pressure value range of [0,100], it can be normalized by dividing it by 100. Data smoothing: Smooth the pressure distribution information to remove noise caused by measurement errors or interference. Data smoothing can be performed using methods such as moving average and low-pass filtering. For example, for a set of pressure distribution data with large fluctuations, the moving average method can be used to smooth the data curve and reduce the impact of noise.

[0113] Data preprocessing of motion-related information involves the following: Data calibration: Calibrate the motion-related information detected by the spatial positioning device to remove deviations caused by equipment errors or environmental factors. For example, the error parameters of the equipment can be determined by conducting a calibration experiment at a known location, and the detected motion-related information can be corrected. Data standardization: Standardize the values ​​in the motion-related information so that they have zero mean and unit variance. The Z-score standardization method can be used for data standardization. For example, for a set of acceleration data, calculate its mean and variance, then subtract the mean from each data point and divide by the variance to obtain the standardized acceleration data.

[0114] Step S130, the skin area image, pressure distribution information and movement-related information of the stethoscope head after data preprocessing are input into the trained stethoscope head position information analysis neural network model, and the position information of the stethoscope head is output.

[0115] "Information after data preprocessing" refers to the skin area image, pressure distribution information and movement-related information of the stethoscope head after processing in step S120. "Stethoscope head position information analysis neural network model" is a model that has been trained to accurately infer the position of the stethoscope head based on multiple input information. "Input object" refers to the three types of preprocessed information, which serve as input data for the model. "Position information output" is a description of the specific position of the stethoscope head on the patient's body given by the model after analysis and calculation.

[0116] Constructing a neural network model for analyzing the position information of the stethoscope head involves the following: 1. Model selection: You can choose a neural network architecture suitable for multimodal data input, such as a combination of convolutional neural network (CNN) and recurrent neural network (RNN), or a deep learning model based on the attention mechanism. For example, a CNN composed of a convolutional layer, a pooling layer, and a fully connected layer can be used to process skin area images, and an LSTM (long short-term memory network) can be used to process time series data of motion-related information and pressure distribution information. 2. Training data collection: Collect a large amount of annotated stethoscope head position information data, including skin area images, pressure distribution information, and motion-related information under different patient body parts, different contact angles and pressure conditions. For example, collect data from 1,000 different patients when they are auscultated in different body parts such as the chest and back. Each data sample includes the corresponding skin area image, pressure distribution information, motion-related information, and accurate stethoscope head position annotation. 3. Model training: Use the collected training data to train the neural network model. The parameters of the model are continuously adjusted through the optimization algorithm so that the model can accurately predict the position of the stethoscope head based on the input information. For example, using the stochastic gradient descent algorithm, the model is trained for multiple iterations with the goal of minimizing the error between the predicted position and the actual position.

[0117] The process of inputting data into the model is as follows: the pre-processed skin area image, pressure distribution information, and motion-related information of the stethoscope head are combined and input in the format required by the model. For example, if the model requires the image to be input in the form of a three-dimensional array, the pressure distribution information to be input in the form of a one-dimensional vector, and the motion-related information to be input in the form of time series data, then the three types of pre-processed information are converted into the corresponding formats and input into the model.

[0118] Output the position information of the stethoscope head as follows: After the model is calculated, it outputs the position information of the stethoscope head. The position information can be expressed in many forms, such as the name of a specific body part, coordinate position, etc. For example, the model output may be "the third rib between the left side of the chest" or a set of three-dimensional coordinate values, indicating the specific position of the stethoscope head on the patient's body.

[0119] Assume that the preprocessed skin area image shows a part of the patient's chest, the pressure distribution information shows that the pressure at the contact point between the stethoscope head and the body is relatively uniform and moderate, and the motion-related information shows that the stethoscope head is relatively stable. Input this information into the trained neural network model. After analysis, the model may output the location information of "patient's chest, near the heart area". This means that based on the various input data, the model determines that the stethoscope head is currently located near the heart area of ​​the patient's chest. Doctors can use this location information to better understand the specific location of the stethoscope and make a more accurate diagnosis.

[0120] Reference Figure 3 , to obtain information about the body parts where the stethoscope head contacts the patient, including:

[0121] Step S1a0, start the camera device preset on the stethoscope head to capture and obtain the skin area image within the preset distance range of the stethoscope head, and simultaneously start the pressure detection device preset on the stethoscope head to detect and obtain the pressure distribution information of the contact between the stethoscope head and the patient's body, and start the space positioning device preset on the stethoscope head to detect and obtain the movement-related information of the stethoscope head.

[0122] This step is the same as the example description of step S110 and will not be described in detail here.

[0123] Step S1b0, extracting image features from the acquired skin area image within a preset distance range of the stethoscope head, performing similarity matching between the extracted image features and a pre-established body part image feature database, and determining the body part where the stethoscope head is located based on the similarity matching result.

[0124] "Image features" are attributes extracted from skin area images that can represent the content and characteristics of the image, such as color, texture, shape, etc. "Body part image feature database" is a pre-established database containing skin image features of different body parts, which is used to compare with the captured skin area images. "Similarity matching" is to determine the matching relationship between the captured skin area image features and the image features of different body parts in the database by calculating the degree of similarity between them. "Body part judgment" is to infer the patient's body part where the stethoscope head is located based on the results of similarity matching.

[0125] Extracting image features from skin area images involves the following: 1. Color feature extraction: You can use color histograms, color moments and other methods to extract the color features of the image. For example, calculate the color histogram of the skin area image on different color channels to obtain the color distribution of the image. 2. Texture feature extraction: Use texture feature extraction algorithms, such as grayscale co-occurrence matrix, local binary pattern, etc. to extract the texture features of the image. For example, by calculating the grayscale co-occurrence matrix, you can get the texture roughness, contrast and other features of the image. 3. Shape feature extraction: Use shape feature extraction methods, such as edge detection, Hough transform, etc. to extract shape features in the image. For example, the shape features of the skin contour in the skin area image can be extracted through the edge detection algorithm.

[0126] Establishing a database of body part image features involves the following: Data collection: Collect a large number of skin images of different body parts, covering a variety of patient ages, skin colors, body shapes, etc. For example, collect 1,000 skin images of different body parts, including the head, neck, chest, abdomen, arms, legs, etc. Feature extraction: Perform feature extraction on the collected skin images to obtain a set of image features for each body part. For example, extract color, texture, shape and other features from the image of each body part. Database construction: Store the extracted image features of different body parts in the database for subsequent similarity matching. For example, establish a relational database, use body parts as the primary key of the table, image features as the fields of the table, and store image feature data of different body parts.

[0127] Performing similarity matching and judging body parts involves the following: Similarity calculation: Using similarity measurement methods, such as Euclidean distance, cosine similarity, etc., calculate the similarity between the captured skin area image features and the image features of different body parts in the database. For example, calculate the cosine similarity between the color features of the skin area image and the color features of each body part in the database. Similarity sorting: Sort the calculated similarity values ​​to find the body part image features that are most similar to the captured skin area image. For example, sort the similarity values ​​of different body parts in order of cosine similarity from large to small. Body part judgment: Determine the body part corresponding to the captured skin area image based on the similarity sorting results. For example, if the similarity with the chest skin image features is the highest, then the body part where the stethoscope head is located is judged to be the chest.

[0128] Step S1c0, extracting the characteristics of the pressure distribution from the pressure distribution information, and comparing it with the typical pressure distribution patterns of different body parts to determine the body part where the stethoscope head is located.

[0129] "Pressure distribution information" refers to the pressure value and distribution at the contact point between the stethoscope head and the patient's body detected by the pressure detection device on the stethoscope head. "Pressure distribution characteristics" are attributes extracted from the pressure distribution information that can describe the characteristics of pressure distribution, such as pressure peak value, pressure mean, shape of pressure distribution, etc. "Typical pressure distribution patterns in different body parts" are the predetermined pressure distribution patterns that different body parts usually present during auscultation.

[0130] Extracting the characteristics of pressure distribution from pressure distribution information involves the following: 1. Calculating the mean pressure: averaging all pressure values ​​in the pressure distribution information to obtain the mean pressure. For example, if the pressure detection device detects pressure values ​​at multiple positions as P1, P2, P3...Pn, the mean pressure is (P1+P2+P3+...+Pn) / n. 2. Determining the peak pressure: Find the maximum pressure value in the pressure distribution information as the peak pressure. For example, in a set of pressure values, the maximum value is Pmax, then Pmax is the peak pressure. 3. Analyzing the shape of the pressure distribution: By observing the graph or data distribution of the pressure distribution information, determine the shape characteristics of the pressure distribution, such as whether it is symmetrical, whether it is concentrated in a specific area, etc. For example, if the pressure is mainly concentrated in a small area, it can be described as a peaked pressure distribution; if the pressure distribution is relatively uniform, it can be described as a relatively flat pressure distribution.

[0131] The typical pressure distribution patterns of different body parts are established as follows: 1. Data collection: Collect a large amount of pressure distribution data during auscultation of different body parts, covering factors such as the body shape, age, and health status of different patients. For example, collect pressure distribution information of 1,000 different patients during auscultation of different body parts such as the chest, abdomen, and arms. 2. Feature analysis: Analyze the collected pressure distribution data of different body parts and summarize the typical pressure distribution characteristics of each body part. For example, for chest auscultation, it may be found that the pressure distribution is relatively uniform, and there is a slightly higher pressure near the heart area; for arm auscultation, the pressure may be mainly concentrated near the blood vessels, etc. 3. Pattern establishment: Based on the analysis results, establish typical pressure distribution patterns of different body parts. These patterns can be represented by mathematical models, charts, or descriptive languages. For example, a function can be used to describe the pressure distribution pattern of the chest, or a graph can be used to show the pressure distribution characteristics of different body parts.

[0132] Step S1d0, extracting motion features from the motion-related information of the stethoscope head, and analyzing the matching degree between the motion features and the motion patterns of different body parts, to determine the body part where the stethoscope head is located.

[0133] Among them, "motion-related information" refers to the motion data of the stethoscope head, such as position, direction, acceleration, etc., detected by the spatial positioning device on the stethoscope head. "Motion features" are attributes extracted from the motion-related information that can describe the motion characteristics of the stethoscope head, such as motion speed, change in motion direction, motion trajectory, etc. "Motion patterns of different body parts" are the pre-determined motion patterns that the stethoscope head may exhibit during the auscultation process for different body parts.

[0134] The process of extracting motion features from motion-related information is as follows: Calculate the motion speed: Calculate the motion speed of the stethoscope head based on the position information of the stethoscope head detected by the spatial positioning device at different time points. For example, the motion speed is obtained by calculating the displacement of the stethoscope head between two time points divided by the time interval. Analyze the change in motion direction: Analyze the change in the motion direction of the stethoscope head based on the direction information of the stethoscope head detected by the spatial positioning device. For example, the change in the angle of the stethoscope head direction between adjacent time points is calculated to determine the stability and degree of change of the motion direction. Determine the motion trajectory: Draw the motion trajectory of the stethoscope head by continuously recording the position information of the stethoscope head. The shape, length, complexity and other characteristics of the motion trajectory can be analyzed. For example, if the motion trajectory presents a relatively regular curve, it may represent a typical motion pattern in a specific body part.

[0135] Analyzing the matching degree between motion features and motion patterns of different body parts involves the following: Similarity calculation: Use appropriate similarity measurement methods, such as dynamic time warping (DTW) algorithm, Euclidean distance, etc., to calculate the similarity between the extracted motion features and the motion patterns of different body parts. For example, use the DTW algorithm to calculate the similarity between the motion trajectory and the typical trajectory of different body parts. Pattern matching: Based on the similarity calculation results, match the extracted motion features with the most similar motion pattern. For example, if the calculated similarity with the chest motion pattern is the highest, then the current motion feature is considered to match the chest pattern.

[0136] Step S1e0, according to the preset distribution weights of the camera device, the pressure detection device and the space positioning device, a weighted voting method is used to select the body part with the highest weighted vote as the final position of the stethoscope head.

[0137] Among them, "the weights assigned to the camera device, pressure detection device and spatial positioning device" are the importance coefficients assigned to the three different information sources, which are used for weighted voting. "Weighted voting method" is a method of voting on the body parts judged by the three types of information according to different weights and calculating the comprehensive results. "The highest number of weighted votes" refers to the body part with the most votes after weighted calculation.

[0138] Determining the allocation weights involves the following: 1. Data collection and analysis: Collect a large number of data from actual auscultation cases, including skin area images, pressure distribution information, and motion-related information of different body parts, as well as the body parts that are ultimately accurately judged. Analyze the contribution of the three types of information to the correct judgment of body parts in different cases. For example, count the number of times in 1,000 cases that the body parts are correctly judged based on skin area images alone, the number of times that the body parts are correctly judged based on pressure distribution information alone, and the number of times that the body parts are correctly judged based on motion-related information alone. 2. Expert evaluation: Invite medical experts to evaluate the importance of the three types of information based on experience. For example, experts may believe that skin area images can provide more intuitive location information in some cases, but may be affected by factors such as lighting; pressure distribution information is relatively stable, but may be inaccurate for patients with certain special body shapes; motion-related information can assist in judgment, but may be affected by the doctor's operating habits. 3. Weight determination: Based on the results of data collection and analysis and expert evaluation, determine the allocation weights of the camera device, pressure detection device, and spatial positioning device. For example, after analysis and evaluation, it is determined that the weight of the skin area image is 0.3, the weight of the pressure distribution information is 0.4, and the weight of the motion-related information is 0.3.

[0139] The process of using the weighted voting method is as follows: First, the body part is determined according to the three types of information: according to step S1b0 (determining the body part from the skin area image), step S1c0 (determining the body part from the pressure distribution information) and step S1d0 (determining the body part from the motion-related information), the body part results determined by the three types of information are obtained. For example, the skin area image is determined to be the chest, the pressure distribution information is determined to be the abdomen, and the motion-related information is determined to be the chest.

[0140] Then, according to the determined weights, weighted voting is performed on the body parts judged by the three types of information. For example, if the skin area image is judged to be the chest and its weight is 0.3, the chest will receive 0.3 votes; if the pressure distribution information is judged to be the abdomen and its weight is 0.4, the abdomen will receive 0.4 votes; if the movement-related information is judged to be the chest and its weight is 0.3, the chest will receive another 0.3 votes.

[0141] Finally, determine the body part with the highest weighted votes: compare the weighted votes of different body parts and determine the body part with the highest votes as the final judgment result. For example, in the above example, the total votes for the chest are 0.6, and the total votes for the abdomen are 0.4, so the body part where the stethoscope head is located is determined to be the chest.

[0142] Reference Figure 4 According to the corresponding relationship between the body part and the initial strength of the heart and lung sound adaptation signal, the initial strength of the heart and lung sound adaptation signal of the body part contacted by the stethoscope head is analyzed and determined, including:

[0143] Step S210, obtaining relevant key information of the patient.

[0144] Among them, relevant key information includes age, gender and body shape.

[0145] Among them, "body shape" includes different types such as obesity, thinness, and normal. Body shape will affect the propagation and reception of sound, and will affect the intensity and characteristics of heart and lung sounds.

[0146] The analysis of body shape is as follows: Professional measuring equipment, such as a height and weight measuring instrument, can be used to measure the patient's height and weight. Then, the body mass index (BMI) is calculated based on the height and weight to determine the patient's body shape. For example, BMI = weight (kg) ÷ height² (m). If the BMI is below 18.5, it can be judged as thin; between 18.5-23.9 is normal; 24 and above is obese.

[0147] Step S220, standardize the acquired patient-related key information and perform feature extraction.

[0148] "Standardization" is to convert the acquired patient-related key information into a unified and comparable format for subsequent analysis and processing. "Feature extraction" is to extract key attributes that can reflect the individual characteristics of patients from the standardized patient-related key information for matching the individualized model.

[0149] Step S230: According to the extracted features, an individualized model matching the patient is obtained from an individualized model database.

[0150] The establishment of the personalized model database is as follows: 1. Data collection: collect relevant key information of a large number of patients, including age, gender, body shape, and the initial strength of the corresponding cardiopulmonary sound adaptation signal. For example, collect information on 1,000 different patients, covering all age groups, different genders, and different body shapes. 2. Model training: Use the collected data to train the model and establish personalized models corresponding to different feature combinations. Machine learning algorithms such as decision trees, random forests, and neural networks can be used. For example, using the random forest algorithm, the initial strength of the corresponding cardiopulmonary sound adaptation signal is predicted based on the patient's age, gender, body shape, and other characteristics, and multiple personalized models are trained and stored in the database. 3. Database construction: The trained personalized models are classified and stored according to the characteristics to build a personalized model database. For example, according to the age range, gender, and body shape category, the personalized model is stored in the corresponding database table for fast retrieval and matching.

[0151] Matching is performed based on the feature extraction results as follows: 1. Feature input: The patient features extracted in step S220 are used as input, such as middle-aged age, male gender, normal body shape, etc. 2. Database search: Search in the personalized model database to find the personalized model that is most similar to the input features. Similarity metrics such as Euclidean distance, cosine similarity, etc. can be used. For example, the cosine similarity between the input features and the features of each personalized model in the database is calculated to find the model with the highest similarity. 3. Model matching: Based on the similarity results, the most matching personalized model is selected as the personalized model that matches the patient. For example, if the similarity with a certain personalized model exceeds a certain threshold (such as 0.8), the model is considered to match the patient.

[0152] Step S240, analyzing and determining the initial intensity of the heart and lung sound adaptation signal of the body part contacted by the stethoscope head according to the individualized model matched with the patient, the corresponding relationship between the body part and the initial intensity of the heart and lung sound adaptation signal.

[0153] The process of using the personalized model is as follows: extract information related to the initial strength of the heart and lung sound adaptation signal from the personalized model, such as a specific adjustment coefficient or trend indication. For example, for a middle-aged male patient with normal body shape, the model may give an adjustment coefficient of 1.2, indicating that the initial strength of the heart and lung sound adaptation signal of the patient should be increased by 20% relative to the general situation.

[0154] Assume that the matched individualized model indicates that the patient is a 50-year-old female with an obese body type, and the adjustment coefficient is 1.1. The body part that the stethoscope head contacts is the right side of the chest. According to the correspondence between the body part and the initial intensity of the heart and lung sound adaptation signal, the initial intensity range of the right side of the chest is [80,100]. First, according to the adjustment coefficient of the individualized model, the adjusted intensity range is calculated as 1.1*[80,100]=[88,110]. Then, the middle value is taken as the initial intensity of the heart and lung sound adaptation signal determined initially, that is, (88+110) / 2=99.

[0155] According to the extracted features, the individual model matching the patient is obtained from the individualized model database, including:

[0156] Step S231, analyzing whether the extracted features are matched to an individualized model that is completely consistent with the patient from the individualized model database. If yes, execute step S232; if no, execute step S233.

[0157] Matching with the personalized model database involves the following: 1. Database query: query the personalized model database and try to find a personalized model that is completely consistent with the extracted features. You can use a database query language or tool to perform an exact match query based on the feature representation. For example, if the extracted feature representation is "03 (adult), 1 (male), 02 (normal body shape)", query the database to see if there is a personalized model with the same feature representation. 2. Compare the query results with the extracted features one by one to determine whether they are completely consistent. The comparison includes various feature dimensions such as age, gender, and body shape. If all features are completely matched, it is considered that a personalized model that is completely consistent with the patient has been found.

[0158] Step S232: According to the extracted features, an individualized model matching the patient is obtained from an individualized model database.

[0159] Step S233: extract all candidate individualized models suitable for the patient from the individualized model database according to the extracted features.

[0160] “Candidate personalized models” are a collection of personalized models that may be suitable for the current patient and are screened from the database.

[0161] Determining the candidate personalized model set involves the following: 1. Based on the extracted features, query the personalized model database to screen out all candidate personalized models that may be applicable to the patient. The query conditions can be partial matches or conditions that meet a specific range. For example, if the patient is a middle-aged male with a normal body shape, then query the database for all personalized models with an age range (such as 35-55 years old), male gender, and normal or near-normal body shape. 2. Model screening: Further screen the query results to remove models that are obviously irrelevant or inapplicable. For example, if the patient characteristics corresponding to a certain model are significantly different from those of the current patient, such as an age difference of more than 20 years or a completely different body shape, it can be excluded from the candidate set. 3. Set construction: The screened personalized models are combined into a candidate personalized model set. The models in this set all match the characteristics of the current patient to a certain extent, which may be helpful in determining the final personalized model.

[0162] Step S234, based on all candidate individualized models applicable to the patient and the extracted features, a preset similarity measurement method is used to analyze and determine the candidate individualized model with the highest similarity as the individualized model used this time.

[0163] The preset similarity measurement method can choose a variety of similarity measurement methods, such as Euclidean distance, cosine similarity, Pearson correlation coefficient, etc.

[0164] The process of analyzing and determining the candidate individualized model with the highest similarity is as follows: 1. Feature comparison: Compare the features of each candidate individualized model with the features extracted from the patient. For example, if the features of the candidate model include age, gender, body shape, etc., compare these features with the corresponding features of the patient one by one. 2. Similarity calculation: Use the preset similarity measurement method to calculate the similarity value between each candidate individualized model and the patient's features. Perform corresponding calculations according to the specific method to obtain a quantitative similarity result. For example, use the Euclidean distance calculation to obtain the distance value between each candidate model and the patient's features. The smaller the distance, the higher the similarity. 3. Determine the highest similarity model: Compare the similarity values ​​of all candidate individualized models to find the model with the highest similarity. This can be determined by sorting or direct comparison. For example, sort the similarity values ​​of all candidate models and select the model with the smallest similarity value (indicating the most similar) as the individualized model used this time.

[0165] Using the preset similarity measurement method, the candidate individualization models with the highest similarity are analyzed and determined to include:

[0166] Step S234.1, using multiple similarity measurement methods, analyzing and obtaining the measurement values ​​of the same candidate individualized model under different similarity measurement methods, and analyzing one by one to obtain the measurement values ​​of all candidate individualized models under different candidate individualized models.

[0167] Calculating the metric values ​​of the same candidate individualized model under different similarity measurement methods involves the following: 1. Data preparation: Organize and prepare the features of the candidate individualized model and the patient's features for similarity calculation. For example, if the Euclidean distance is used, the features need to be represented as numerical vectors. 2. Calculate one by one: For the same candidate individualized model, use the selected multiple similarity measurement methods to calculate respectively. According to the calculation formula and steps of each method, compare the features of the candidate model with the patient's characteristics to obtain the corresponding metric values. 3. Record the results: Record the different metric values ​​calculated for subsequent analysis. For example, the metric values ​​of each candidate individualized model under different similarity measurement methods can be stored in a table.

[0168] Calculating the metric values ​​of all candidate individualized models under different similarity measurement methods involves the following: Repeating steps: Repeat the above calculation process for all candidate individualized models to ensure that each model has metric values ​​under different similarity measurement methods. Comprehensive analysis: By calculating the metric values ​​of all candidate models, we can have a comprehensive understanding of the similarity of different models under different methods, providing a basis for subsequent screening and determination.

[0169] Step S234.2, analyzing whether there is a candidate individualized model whose metric value reaches the preset metric value. If yes, execute step S234.3; if no, execute step S234.4.

[0170] Assume that we use two similarity measurement methods, Euclidean distance and cosine similarity, with the preset measurement values ​​being Euclidean distance less than or equal to 10 and cosine similarity greater than or equal to 0.8. There are three candidate individualized models A, B, and C. For model A, the calculated Euclidean distance is 8 and the cosine similarity is 0.85. For model B, the Euclidean distance is 12 and the cosine similarity is 0.7. For model C, the Euclidean distance is 15 and the cosine similarity is 0.6.

[0171] In this example, the Euclidean distance of model A is less than or equal to 10, and the cosine similarity is greater than or equal to 0.8, so the metric value of model A reaches the preset metric value. The metric values ​​of models B and C do not reach the preset metric value.

[0172] Step S234.3, taking the corresponding candidate individualized model as the individualized model for this application.

[0173] Step S234.4, according to the preset individualized model screening conditions, the individualized model that meets all the screening conditions is screened out from the candidate individualized models as the individualized model for this application, wherein the screening conditions include the measurement threshold of the same candidate individualized model under different similarity measurement methods.

[0174] Among them, the personalized model screening conditions are a series of standards and requirements used to determine the most suitable model for the current patient from the candidate personalized models. "Candidate personalized models" are multiple models that may be suitable for patients extracted from the personalized model database. "Measurement threshold" is the numerical range set under different similarity measurement methods to determine whether the candidate personalized model meets the screening conditions.

[0175] Determining the screening conditions for individualized models involves the following: Considering multiple factors: The screening conditions may include the measurement thresholds of the same candidate individualized model under different similarity measurement methods, the patient's specific clinical needs, the accuracy and reliability of the model, etc. For example, for patients with certain specific diseases, it may be necessary to select a model that has a higher match with the disease-related characteristics under a specific similarity measurement method. Setting measurement thresholds: Based on historical data, expert experience or experimental results, set reasonable measurement thresholds for different similarity measurement methods. These thresholds can be adjusted according to different patient groups and application scenarios. For example, for Euclidean distance, setting a smaller threshold means that the model is closer to the patient's characteristics; for cosine similarity, setting a higher threshold means that the direction of the model and the patient's characteristics is more consistent.

[0176] Screening from candidate individualized models involves the following: Evaluate one by one: Evaluate each candidate individualized model to determine whether it meets all the screening conditions. For example, for a candidate model, check whether its metric values ​​under different similarity measurement methods are within the corresponding metric threshold range, and whether it meets other screening conditions. Comprehensive comparison: If multiple candidate individualized models meet some of the screening conditions, a comprehensive comparison is required to determine the model that best meets all conditions. You can assign corresponding weights according to the importance of different conditions, and then calculate the comprehensive score of each model. For example, assign different weights to the similarity measurement value and other conditions, and calculate the weighted total score of each model. Determine the final model: Based on the results of the comprehensive comparison, select the candidate individualized model with the highest score or the one that best meets all the screening conditions as the individualized model for this application.

[0177] Assume that we have three candidate individual models A, B, and C. We use two similarity measurement methods, Euclidean distance and cosine similarity, with the following screening conditions: Euclidean distance is less than or equal to 10. Cosine similarity is greater than or equal to 0.8. Age deviation is no more than 5 years.

[0178] For model A: Euclidean distance is 8. Cosine similarity is 0.85. The age difference with the patient is 3 years. For model B: Euclidean distance is 12. Cosine similarity is 0.7. The age difference with the patient is 4 years. For model C: Euclidean distance is 15. Cosine similarity is 0.6. The age difference with the patient is 6 years.

[0179] In this example, model A meets the screening criteria in terms of Euclidean distance, cosine similarity, and age deviation. Models B and C do not meet the screening criteria in different aspects and are excluded.

[0180] The magnification increase range is obtained by:

[0181] Step S410, analyzing whether the patient's age, physical condition information and medical history are obtained. If yes, execute step S420; if no, execute step S440.

[0182] Step S420, based on the preset weights matching different age ranges, the weights matching different physical conditions, and the weights matching different medical histories, analyze and obtain the average weights of the patient's age, physical condition, and medical history as the patient's weight combination result.

[0183] Among them, "weights matched by different age ranges" are the importance coefficients used to adjust the magnification factor according to different age ranges. "Weights matched by different physical conditions" are coefficients that affect the magnification factor based on the patient's different overall physical health status. "Weights matched by different medical histories" are coefficients that affect the magnification factor based on the diseases the patient has suffered in the past. "Weight combination result" is the weight value obtained after comprehensively considering multiple factors of the patient for subsequent analysis.

[0184] Step S430, taking the weight combination result of the patient and the preliminary intensity range recorded by the preliminary intensity of the abnormal sound as input features, inputting them into the analysis model of the trained amplification adjustment coefficient, and outputting the amplification adjustment coefficient.

[0185] Among them, the "preliminary intensity range of abnormal sound" is the intensity range determined after the preliminary judgment of the abnormal sound detected by auscultation. The "analysis model of the amplification adjustment coefficient" is a model that has been trained to output the appropriate amplification adjustment coefficient according to the input features.

[0186] Establishing an analytical model for the magnification adjustment coefficient involves the following processes: 1. Data collection: Collect a large number of patients' age, physical condition, medical history, abnormal sound intensity range, and corresponding appropriate magnification adjustment coefficients. For example, collect cases of 1,000 different patients, record their various characteristics and the magnification adjustment actually used. 2. Feature extraction: Extract useful features from the collected data, such as age range, physical condition category, medical history type, abnormal sound intensity range, etc. Encode or digitize these features for model input. 3. Model selection and training: You can select a suitable machine learning algorithm, such as a neural network, decision tree, etc., to establish an analytical model. Use the collected data to train the model, and continuously adjust the model's parameters through the optimization algorithm so that the model can accurately predict the magnification adjustment coefficient based on the input features. For example, use the back propagation algorithm to train the neural network model to minimize the error between the predicted magnification adjustment coefficient and the actual value.

[0187] Inputting features into the model involves the following: Arranging input data: Arranging the patient's weighted combination results and the preliminary intensity range of abnormal sounds in the format required by the model. For example, if the model requires the input to be in vector form, convert the values ​​of the weighted combination results and the intensity range of abnormal sounds into vectors. Inputting into the model: Inputting the arranged input features into the trained analysis model of the amplification adjustment coefficient. For example, by calling the prediction function of the model, passing the input features to the model for calculation.

[0188] Output amplification adjustment coefficient: The analysis model calculates based on the input features and outputs an amplification adjustment coefficient. This coefficient reflects the degree to which the cardiopulmonary sound adaptation signal should be amplified according to the patient's specific situation and the intensity of the abnormal sound.

[0189] Assume that the weight combination result of a patient is 0.6, the initial intensity of the abnormal sound is judged to be medium intensity, and the corresponding initial intensity range is [50,70]. This weight combination result and the abnormal sound intensity range are input into the trained analysis model of the amplification adjustment factor. After calculation, the model outputs the amplification adjustment factor of 1.4. This means that according to the patient's condition and the intensity of the abnormal sound, the amplification of the cardiopulmonary sound adaptation signal should be increased to 1.4 times the original, so that the abnormal sound can be heard more clearly and further diagnosis can be made.

[0190] Step S440, starting the noise detection device preset on the intelligent stethoscope device to detect and obtain the noise intensity within a preset distance range.

[0191] "Noise detection device" is a device used to detect ambient noise, usually installed on a smart stethoscope. "Preset distance range" is the distance range within which the noise detection device can effectively detect noise. "Noise intensity" refers to the magnitude of the detected ambient noise, usually expressed in units such as decibels (dB).

[0192] Detecting the noise intensity within a preset distance range involves the following: Starting the device: When the noise intensity needs to be detected, start the noise detection device. For example, when the intelligent stethoscope starts working, the noise detection device is started at the same time. Detection range setting: Determine the detection distance range of the noise detection device, that is, the preset distance range. This range can be adjusted according to the actual application scenario. For example, set the detection distance to noise within 1 meter. Data acquisition: The noise detection device starts to collect sound signals from the surrounding environment and converts them into digital signals for analysis. By calculating the intensity of the sound signal, the numerical value of the noise intensity can be obtained. For example, use algorithms such as fast Fourier transform (FFT) to analyze the sound signal and calculate the decibel value of the noise.

[0193] Step S450, analyzing and determining the increase in the amplification factor of the cardiopulmonary sound adaptation signal according to the correspondence between the noise intensity interval range into which the noise intensity falls, the preliminary intensity range recorded by the preliminary intensity of the abnormal sound, and the increase in the amplification factor of the cardiopulmonary sound adaptation signal.

[0194] Assuming that the noise intensity range is divided into three intervals: low (0-40 decibels), medium (41-60 decibels), and high (61 decibels and above), the preliminary intensity range of abnormal sounds is divided into three levels: weak, medium, and strong.

[0195] The corresponding relationship is as follows: when the noise level is low and the abnormal sound is weak, the amplification factor increases by 1.2 times; when the noise level is low and the abnormal sound is medium, the amplification factor increases by 1.3 times; when the noise level is low and the abnormal sound is strong, the amplification factor increases by 1.4 times; when the noise level is medium and the abnormal sound is weak, the amplification factor increases by 1.3 times; when the noise level is medium and the abnormal sound is medium, the amplification factor increases by 1.5 times; when the noise level is medium and the abnormal sound is strong, the amplification factor increases by 1.6 times; when the noise level is high and the abnormal sound is weak, the amplification factor increases by 1.4 times; when the noise level is high and the abnormal sound is medium, the amplification factor increases by 1.6 times; when the noise level is high and the abnormal sound is strong, the amplification factor increases by 1.8 times.

[0196] If the detected noise intensity is 50 decibels, it falls into the medium noise range, and the abnormal sound is initially judged to be of medium intensity. According to the corresponding relationship, the analysis determines that the amplification factor of the cardiopulmonary sound adaptation signal is increased by 1.5 times.

[0197] A management and control system of an intelligent stethoscope device also includes steps in parallel with outputting abnormal sound classification and diagnosis results, specifically as follows:

[0198] Obtain heart and lung sound data from the smart stethoscope, obtain electrocardiogram data from the electrocardiograph, and obtain blood pressure data from the sphygmomanometer. The heart and lung sound data include the frequency, intensity, duration and other characteristics of the sound. Filter the heart and lung sounds to remove environmental noise and interference signals; the electrocardiogram data includes characteristics such as heart rate, heart rhythm, P wave, QRS wave, T wave, etc.; perform baseline correction and filtering on the electrocardiogram to improve signal quality, and the blood pressure data includes characteristics such as systolic pressure, diastolic pressure, and mean arterial pressure. Perform stability checks on the blood pressure data to remove abnormal values.

[0199] The three indicators of heart and lung sounds, electrocardiogram, and blood pressure are evaluated for abnormality. Specifically, for heart and lung sounds, different types of abnormal sounds are defined, such as murmurs, wheezes, moist rales, etc., and are divided into three levels of mild, moderate, and severe according to their severity. For example, a mild murmur may be given a lower abnormality score, while a severe murmur is given a higher score. For electrocardiograms, abnormal heart rate (tachycardia, bradycardia), abnormal heart rhythm (premature beats, atrial fibrillation, etc.), and abnormal waveform (ST segment changes, T wave inversion, etc.) are analyzed and scored in the same way as mild, moderate, and severe levels. For blood pressure, according to the standards of hypertension and hypotension, as well as the fluctuation of blood pressure, it is scored in four levels: normal, mild abnormality, moderate abnormality, and severe abnormality.

[0200] The weight distribution can be adjusted according to different disease types and patient groups. For example, for elderly patients or patients with a family history of cardiovascular disease, the weight of the electrocardiogram can be appropriately increased; for patients with respiratory diseases, the weight of heart and lung sounds can be increased.

[0201] The abnormality score of each indicator is multiplied by its corresponding weight to obtain the weighted abnormal value. The weighted abnormal values ​​of each indicator such as heart and lung sounds, electrocardiogram, blood pressure, etc. are added together to obtain a comprehensive disease assessment value. The comprehensive disease assessment value is sent as notification information to the terminal held by the relevant person in charge.

[0202] For example, if the abnormality score of heart and lung sounds is 40 (moderate abnormality), the weight is 0.3; the abnormality score of electrocardiogram is 60 (severe abnormality), the weight is 0.5; the abnormality score of blood pressure is 30 (mild abnormality), the weight is 0.2. Then the comprehensive condition assessment value is: 40×0.3+60×0.5+30×0.2=12+30+6=48.

[0203] The present invention also provides a control device for an intelligent stethoscope device, which specifically includes:

[0204] The acquisition module is used to obtain the body part information where the stethoscope head contacts the patient.

[0205] The analysis module is used to analyze and determine the initial intensity of the cardiopulmonary sound adaptation signal of the body part contacted by the auscultation head according to the corresponding relationship between the body part and the initial intensity of the cardiopulmonary sound adaptation signal, and instruct the auscultation device to perform abnormal sound detection according to the analyzed and determined initial intensity of the cardiopulmonary sound adaptation signal; and analyze whether there is an abnormal sound within a preset monitoring period according to a preset abnormal sound judgment method.

[0206] The execution module detects and obtains the initial intensity of the abnormal sound if there is any abnormal sound, and analyzes and determines the increase in the amplification factor based on the correspondence between the initial intensity range recorded for the initial intensity of the abnormal sound and the increase in the amplification factor of the cardiopulmonary sound adaptation signal; regularly enhances the initial intensity of the cardiopulmonary sound adaptation signal according to the analyzed and determined increase in the amplification factor, and monitors and obtains the signal-to-noise ratio of the cardiopulmonary sound adaptation signal at the same time. When the signal-to-noise ratio of the cardiopulmonary sound adaptation signal drops to a preset value, the increase in the amplification factor of the cardiopulmonary sound adaptation signal is stopped; the key features of the abnormal sound are extracted, and the key features are input into the trained sound diagnosis classification model, and the abnormal sound classification and diagnosis results are output. If there is no abnormal sound, it is preliminarily determined that there is no problem with the corresponding body part of the patient.

[0207] The embodiments of this specific implementation method are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A management and control system for intelligent stethoscope equipment, characterized in that: include: Auscultatory equipment; at least one processor; At least one memory storing a computer program, and at least one processor configured to implement a method for controlling a stethoscope device when executing the computer program, the specific steps are as follows: Obtain information about the body part where the stethoscope head contacts the patient; According to the correspondence between the body part and the initial strength of the heart and lung sound adaptation signal, the initial strength of the heart and lung sound adaptation signal of the body part contacted by the stethoscope head is analyzed and determined, and the stethoscope device is instructed to perform abnormal sound detection according to the analyzed and determined initial strength of the heart and lung sound adaptation signal; Analyze whether there is abnormal sound within the preset monitoring period according to the preset abnormal sound judgment method; If yes, the initial intensity of the abnormal sound is detected and obtained, and the increase in the amplification factor is analyzed and determined according to the corresponding relationship between the initial intensity range recorded for the initial intensity of the abnormal sound and the increase in the amplification factor of the heart and lung sound adaptation signal; The initial strength of the heart and lung sound adaptation signal is regularly enhanced according to the increase of the amplification factor determined by the analysis, and the signal-to-noise ratio of the heart and lung sound adaptation signal is monitored and obtained at the same time. When the signal-to-noise ratio of the heart and lung sound adaptation signal drops to a preset value, the increase of the amplification factor of the heart and lung sound adaptation signal is stopped; Extract key features of abnormal sounds, input the key features into the trained sound diagnosis classification model, and output abnormal sound classification and diagnosis results; If no, it is preliminarily determined that there is no problem with the corresponding body part of the patient; The information about the body parts where the stethoscope head contacts the patient includes: The camera device preset on the stethoscope head is started to capture and obtain the skin area image within the preset distance range of the stethoscope head, and the pressure detection device preset on the stethoscope head is simultaneously started to detect and obtain the pressure distribution information between the stethoscope head and the patient's body, and the space positioning device preset on the stethoscope head is started to detect and obtain the movement-related information of the stethoscope head; The acquired skin area image, pressure distribution information and movement related information of the stethoscope head are respectively preprocessed; The skin area image, pressure distribution information and movement-related information of the stethoscope head after data preprocessing are input into the trained stethoscope head position information analysis neural network model, and the position information of the stethoscope head is output; Alternatively, obtain information about the part of the body that the stethoscope is in contact with the patient including: The camera device preset on the stethoscope head is started to capture and obtain the skin area image within the preset distance range of the stethoscope head, and the pressure detection device preset on the stethoscope head is simultaneously started to detect and obtain the pressure distribution information between the stethoscope head and the patient's body, and the space positioning device preset on the stethoscope head is started to detect and obtain the movement-related information of the stethoscope head; Extracting image features from the acquired skin area image within a preset distance range of the stethoscope head, performing similarity matching between the extracted image features and a pre-established body part image feature database, and determining the body part where the stethoscope head is located based on the similarity matching result; Extracting the characteristics of pressure distribution from the pressure distribution information and comparing it with typical pressure distribution patterns of different body parts to determine the body part where the stethoscope head is located; Extract motion features from the motion-related information of the stethoscope head, and analyze the matching degree between the motion features and the motion patterns of different body parts to determine the body part where the stethoscope head is located; According to the preset distribution weights of the camera device, the pressure detection device and the spatial positioning device, a weighted voting method is used to select the body part with the highest weighted vote as the final position of the stethoscope head; According to the corresponding relationship between the body part and the initial strength of the heart and lung sound adaptation signal, the initial strength of the heart and lung sound adaptation signal of the body part contacted by the stethoscope head is analyzed and determined, including: Obtain relevant key information of the patient, including age, gender and body shape; Standardize the key information of the patients and extract features; According to the extracted features, an individualized model matching the patient is obtained from an individualized model database; According to the individualized model matched to the patient, the corresponding relationship between the body part and the initial strength of the adaptation signal of the heart and lung sounds, the initial strength of the adaptation signal of the heart and lung sounds of the body part contacted by the stethoscope head is analyzed and determined; The magnification increase range is obtained by: Analyze whether to obtain the patient's age, physical condition information and medical history; If yes, then according to the preset weights matching different age ranges, weights matching different physical conditions and weights matching different medical histories, the average weight of the patient in terms of age, physical condition and medical history is analyzed and obtained as the weight combination result of the patient; The weight combination result of the patient and the preliminary intensity range recorded by the preliminary intensity of the abnormal sound are used as input features, input into the analysis model of the amplification adjustment coefficient that has completed the training, and the amplification adjustment coefficient is output; If not, the noise detection device preset on the intelligent stethoscope device is activated to detect and obtain the noise intensity within the preset distance range; According to the correspondence between the noise intensity interval range that the noise intensity falls into, the preliminary intensity range recorded by the preliminary intensity of the abnormal sound and the increase in the amplification factor of the heart and lung sound adaptation signal, the increase in the amplification factor of the heart and lung sound adaptation signal is analyzed and determined.

2. The management and control system of the intelligent stethoscope device according to claim 1, characterized in that: According to the extracted features, the individual model matching the patient is obtained from the individualized model database, including: Analyze whether the extracted features are matched to the personalized model that is completely consistent with the patient from the personalized model database; If yes, then according to the extracted features, a personalized model matching the patient is obtained from the personalized model database; If not, all candidate personalized models applicable to the patient are extracted from the personalized model database according to the extracted features; According to all candidate individualized models applicable to the patient and the extracted features, a preset similarity measurement method is used to analyze and determine the candidate individualized model with the highest similarity as the individualized model used this time.

3. The management and control system of the intelligent stethoscope device according to claim 2, characterized in that: Using the preset similarity measurement method, the candidate individualization models with the highest similarity are analyzed and determined to include: Using multiple similarity measurement methods, analyze and obtain the measurement values ​​of the same candidate individualized model under different similarity measurement methods, and analyze one by one to obtain the measurement values ​​of all candidate individualized models under different candidate individualized models; Analyze whether there is a candidate individualized model whose measurement value reaches the preset measurement value; If yes, the corresponding candidate individualized model is used as the individualized model for this application; If not, the individualized model that meets all the screening conditions is screened out from the candidate individualized models according to the preset individualized model screening conditions as the individualized model for this application, wherein the screening conditions include the measurement thresholds of the same candidate individualized model under different similarity measurement methods.

4. A control device for an intelligent stethoscope device, characterized in that: include: An acquisition module, used for acquiring information of the body part where the stethoscope head contacts the patient; The information about the body parts that the stethoscope head contacts with the patient includes: The camera device preset on the stethoscope head is started to capture and obtain the skin area image within the preset distance range of the stethoscope head, and the pressure detection device preset on the stethoscope head is simultaneously started to detect and obtain the pressure distribution information between the stethoscope head and the patient's body, and the space positioning device preset on the stethoscope head is started to detect and obtain the movement-related information of the stethoscope head; The acquired skin area image, pressure distribution information and movement related information of the stethoscope head are respectively preprocessed; The skin area image, pressure distribution information and movement-related information of the stethoscope head after data preprocessing are input into the trained stethoscope head position information analysis neural network model, and the position information of the stethoscope head is output; Alternatively, obtain information about the part of the body that the stethoscope is in contact with the patient including: The camera device preset on the stethoscope head is started to capture and obtain the skin area image within the preset distance range of the stethoscope head, and the pressure detection device preset on the stethoscope head is simultaneously started to detect and obtain the pressure distribution information between the stethoscope head and the patient's body, and the space positioning device preset on the stethoscope head is started to detect and obtain the movement-related information of the stethoscope head; Extracting image features from the acquired skin area image within a preset distance range of the stethoscope head, performing similarity matching between the extracted image features and a pre-established body part image feature database, and determining the body part where the stethoscope head is located based on the similarity matching result; Extracting the characteristics of pressure distribution from the pressure distribution information and comparing it with typical pressure distribution patterns of different body parts to determine the body part where the stethoscope head is located; Extract motion features from the motion-related information of the stethoscope head, and analyze the matching degree between the motion features and the motion patterns of different body parts to determine the body part where the stethoscope head is located; According to the preset distribution weights of the camera device, the pressure detection device and the spatial positioning device, a weighted voting method is used to select the body part with the highest weighted vote as the final position of the stethoscope head; The analysis module is used to analyze and determine the initial strength of the heart and lung sound adaptation signal of the body part contacted by the auscultation head according to the corresponding relationship between the body part and the initial strength of the heart and lung sound adaptation signal, and instruct the auscultation device to perform abnormal sound detection according to the analyzed and determined initial strength of the heart and lung sound adaptation signal; and analyze whether there is an abnormal sound within a preset monitoring period according to a preset abnormal sound judgment method; Among them, according to the corresponding relationship between the body part and the initial strength of the heart and lung sound adaptation signal, the initial strength of the heart and lung sound adaptation signal of the body part contacted by the stethoscope head is analyzed and determined, including: Obtain relevant key information of the patient, including age, gender and body shape; Standardize the key information of the patients and extract features; According to the extracted features, an individualized model matching the patient is obtained from an individualized model database; According to the individualized model matched to the patient, the corresponding relationship between the body part and the initial strength of the adaptation signal of the heart and lung sounds, the initial strength of the adaptation signal of the heart and lung sounds of the body part contacted by the stethoscope head is analyzed and determined; The execution module detects and obtains the initial intensity of the abnormal sound if there is an abnormal sound, and analyzes and determines the increase in the amplification factor according to the corresponding relationship between the initial intensity range recorded for the initial intensity of the abnormal sound and the increase in the amplification factor of the heart and lung sound adaptation signal; regularly enhances the initial intensity of the heart and lung sound adaptation signal according to the analyzed and determined increase in the amplification factor, and monitors and obtains the signal-to-noise ratio of the heart and lung sound adaptation signal at the same time. When the signal-to-noise ratio of the heart and lung sound adaptation signal drops to a preset value, the increase in the amplification factor of the heart and lung sound adaptation signal is stopped; the key features of the abnormal sound are extracted, and the key features are input into the trained sound diagnosis classification model, and the abnormal sound classification and diagnosis results are output. If there is no abnormal sound, it is preliminarily determined that there is no problem with the corresponding body part of the patient; Among them, the acquisition of the increase in magnification includes: Analyze whether to obtain the patient's age, physical condition information and medical history; If yes, then according to the preset weights matching different age ranges, weights matching different physical conditions and weights matching different medical histories, the average weight of the patient in terms of age, physical condition and medical history is analyzed and obtained as the weight combination result of the patient; The weight combination result of the patient and the preliminary intensity range recorded by the preliminary intensity of the abnormal sound are used as input features, input into the analysis model of the amplification adjustment coefficient that has completed the training, and the amplification adjustment coefficient is output; If not, the noise detection device preset on the intelligent stethoscope device is activated to detect and obtain the noise intensity within the preset distance range; According to the correspondence between the noise intensity interval range that the noise intensity falls into, the preliminary intensity range recorded by the preliminary intensity of the abnormal sound and the increase in the amplification factor of the heart and lung sound adaptation signal, the increase in the amplification factor of the heart and lung sound adaptation signal is analyzed and determined.

Images