Heart sound and breath sound data processing software and method based on artificial intelligence algorithm

Abstract

The invention provides heart sound and breath sound data processing software and method based on an artificial intelligence algorithm, and relates to the technical field of medical data processing. The system can be in Bluetooth connection with an electronic stethoscope to receive data and upload the data to a cloud server, supports user information input, realizes historical record management, and can upgrade a terminal algorithm. During auscultation of a doctor, heart sound and breath audio is automatically transmitted to the mobile terminal, heart noise and dry and wet rale are automatically recognized through an artificial intelligence algorithm, an analysis report is provided, and primary medical treatment can play a greater role in child health care, disease screening and preliminary diagnosis. The problem that efficient heart sound and breath sound data processing software combined with an artificial intelligence algorithm is lacked in the current market, especially application in grading diagnosis and treatment and primary medical treatment is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical data processing, and in particular to heart sound and respiratory sound data processing software and method based on artificial intelligence algorithms. Background Art

[0002] Traditional diagnosis of cardiopulmonary diseases mostly relies on doctors using a stethoscope to perform manual auscultation and subjectively judge the characteristics of heart and respiratory sounds, which is prone to missed diagnoses and misdiagnoses. In recent years, with the rapid development of artificial intelligence technology, the use of machine learning algorithms to perform intelligent analysis of heart and respiratory sounds can improve the accuracy and efficiency of diagnosis.

[0003] However, there is currently a lack of efficient heart and respiratory sound data processing software that combines artificial intelligence algorithms on the market, especially for applications in tiered diagnosis and treatment and primary healthcare.

[0004] Therefore, in view of this, the existing structure and deficiencies are studied and improved, and a heart sound and respiratory sound data processing software and method based on artificial intelligence algorithm is provided, in order to achieve a more practical purpose. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention provides a heart sound and respiratory sound data processing software and method based on artificial intelligence algorithm to solve the problem that there is a lack of efficient heart sound and respiratory sound data processing software combined with artificial intelligence algorithm in the current market, especially in the application of hierarchical diagnosis and treatment and primary medical care.

[0006] The present invention provides a heart sound and respiratory sound data processing software system based on artificial intelligence algorithms, specifically comprising: a mobile terminal APP module: installed on a smart terminal device, having a data transmission function, capable of connecting to an electronic stethoscope via Bluetooth to receive data and upload it to a cloud server, supporting user information entry, realizing historical record management, and capable of terminal algorithm upgrades; Online annotation platform module: used for annotation and review of heart and respiratory sound data, including data upload and storage functions, providing annotation tools, and managing annotation results; Artificial intelligence algorithm module: performs intelligent analysis of heart and respiratory sound data, including signal preprocessing unit, feature extraction unit and intelligent recognition unit.

[0007] Furthermore, the dry and wet rales feature recognition unit in the artificial intelligence algorithm module constructs a sound library containing real dry and wet rales (Cracks) and pseudo dry and wet rales (false Cracks), extracts time domain and frequency domain features, uses a support vector machine (SVM) to train the model, and screens potential dry and wet rales through signal preprocessing and dynamic thresholding.

[0008] Furthermore, the heart murmur feature recognition unit in the artificial intelligence algorithm module uses wavelet packet transform to decompose the heart sound signal, extracts time-frequency domain features and HHT marginal spectral entropy, and performs classification through a dual threshold formula and support vector machine (linear kernel).

[0009] A method for processing heart sound and respiratory sound data based on an artificial intelligence algorithm comprises the following steps: Data collection and transmission: An electronic stethoscope is used to collect heart and respiratory sound signals. The mobile app connects to the electronic stethoscope via Bluetooth, receives respiratory sound signals and terminal algorithm calculation results, and uploads the data to the cloud server under network conditions. User information management: The mobile APP supports user registration and login, and records the basic information of users and those being screened; History record management: The mobile app can query all historical records, recall voice data, and record diagnosis and treatment information related to treatment and rehabilitation; Terminal algorithm upgrade: The mobile APP realizes automatic online upgrade of terminal algorithms; Data annotation and management: The online annotation platform is used for annotation and review of heart and respiratory sound data. Doctors upload the collected heart and respiratory sound data to the platform. The platform provides an annotation interface and tools, supports multiple annotation methods, and the annotation results are saved uniformly. Intelligent analysis: The artificial intelligence algorithm module performs intelligent analysis on heart and respiratory sound data, including signal preprocessing, filtering, noise reduction, and normalization of heart and respiratory sound signals; extracts cardiac murmur and dry and wet rales features through time and frequency domain analysis; and uses machine learning algorithms such as support vector machines (SVM) to intelligently identify lesion features.

[0010] Furthermore, the identification of dry and wet rales features in the intelligent analysis specifically includes: constructing a sound library containing real dry and wet rales (Cracks) and a sound library containing pseudo dry and wet rales (false Cracks); extracting time domain features, including waveform zero-crossing rate, short-time energy, peak factor, etc., extracting frequency domain features, and using FFT to calculate the spectrum center of mass and frequency band energy ratio; using support vector machine (SVM) for binary classification training, optimizing kernel function and regularization parameters through grid search, and introducing cross-validation to improve model generalization ability; using resampling and bandpass filtering for signal preprocessing, dynamically calculating thresholds based on Gaussian mixture model (GMM), and screening potential dry and wet rales candidate segments.

[0011] Furthermore, the recognition of cardiac murmur features in the intelligent analysis specifically includes: performing multi-layer decomposition of cardiac sound signals using wavelet packet transform to obtain sub-signals of different frequency bands; extracting the energy proportion, entropy value, kurtosis, and skewness of each frequency band sub-signal, calculating the marginal spectral entropy through Hilbert-Huang transform (HHT), and analyzing the signal complexity using a sliding window; designing a dual-threshold formula, balancing sensitivity and specificity by adjusting the parameters in the formula, and using a support vector machine (linear kernel) for classification.

[0012] Compared with the prior art, the present invention has the following beneficial effects: The heart sound and respiratory sound data processing system of the present invention combines intelligent auscultation application software and artificial intelligence-assisted auscultation algorithm. The electronic stethoscope is used in the same way as a traditional stethoscope. When the doctor listens to the stethoscope, the heart sound and respiratory audio are automatically transmitted to the mobile terminal. The artificial intelligence algorithm automatically identifies heart murmurs, dry and wet rales and provides an analysis report, which helps primary medical care play a greater role in child health care, disease screening, and preliminary diagnosis, improves the efficiency and accuracy of triage in primary medical units in tiered diagnosis and treatment, and enhances pediatric service capabilities.

[0013] It can realize remote intelligent diagnosis and clinical efficacy evaluation of pediatric respiratory diseases and major birth defects (congenital heart disease), greatly improve the clinical diagnosis level of these diseases in grassroots and remote areas, and reduce the human, material and financial resources consumed by patients when seeking medical treatment.

[0014] It can conduct remote diagnosis and intelligent screening for grassroots hospitals, provide off-site follow-up services for discharged patients, break through the national bed capacity limit, expand the scale of hospital services through remote technology and intelligent equipment, and increase the direct and indirect benefits of the hospital. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings of the embodiments are briefly introduced below.

[0016] In the attached figure: Figure 1 A schematic diagram of a system according to an embodiment of the present invention is shown; Figure 2 A schematic diagram of an APP software interface according to an embodiment of the present invention is shown; Figure 3 A schematic diagram of displaying a heart sound and respiratory sound spectrum according to an embodiment of the present invention is shown; Figure 4 A schematic diagram of an interface of a third-party online light consultation platform according to an embodiment of the present invention is shown; Figure 5 A schematic diagram of an online annotation platform according to an embodiment of the present invention is shown; Figure 6 A schematic diagram of a labeling process according to an embodiment of the present invention is shown; Figure 7 A schematic diagram of a labeling result according to an embodiment of the present invention is shown; Figure 8 A schematic diagram of the software interface for the automatic identification algorithm of dry and wet rales according to an embodiment of the present invention is shown; Figure 9 A schematic diagram of a feature extraction technology route according to an embodiment of the present invention is shown; Figure 10 A schematic diagram of a technical roadmap of an automatic dry and wet rales recognition algorithm according to an embodiment of the present invention is shown; Figure 11 A schematic diagram showing the result of vector machine processing according to an embodiment of the present invention is shown; Figure 12 A schematic diagram showing a comparison of different thresholds for abnormal heart sound signals according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0017] The following will be combined with the accompanying drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. Obviously, the embodiments described are only part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present disclosure.

[0018] Unless otherwise defined, all terms (including technical and scientific terms) used in the embodiments of the present disclosure have the same meaning as commonly understood by those skilled in the art to which the present disclosure belongs. It should also be understood that terms such as those defined in common dictionaries should be interpreted as having meanings consistent with their meanings in the context of the relevant technology, and should not be interpreted in an idealized or highly formal sense, unless explicitly defined in this manner in the embodiments of the present disclosure.

[0019] Example: As attached Figure 1 To the attached Figure 12 As shown: The present invention provides a heart sound and respiratory sound data processing software system based on artificial intelligence algorithms, including: a mobile terminal APP module: installed on a smart terminal device, with a data transmission function, capable of connecting to an electronic stethoscope via Bluetooth to receive data and upload it to a cloud server, supporting user information entry, realizing historical record management, and capable of terminal algorithm upgrades; Online annotation platform module: used for annotation and review of heart and respiratory sound data, including data upload and storage functions, providing annotation tools, and managing annotation results; Artificial intelligence algorithm module: performs intelligent analysis of heart and respiratory sound data, including signal preprocessing unit, feature extraction unit and intelligent recognition unit.

[0020] Among them, the dry and wet rales feature recognition unit in the artificial intelligence algorithm module constructs a sound library containing real dry and wet rales (Cracks) and pseudo dry and wet rales (false Cracks), extracts time domain and frequency domain features, uses support vector machine (SVM) to train the model, and screens potential dry and wet rales through signal preprocessing and dynamic thresholding.

[0021] Among them, the heart murmur feature recognition unit in the artificial intelligence algorithm module uses wavelet packet transform to decompose the heart sound signal, extracts time-frequency domain features and HHT marginal spectral entropy, and performs classification through a dual threshold formula and support vector machine (linear kernel).

[0022] A method for processing heart sound and respiratory sound data based on an artificial intelligence algorithm comprises the following steps: Data collection and transmission: An electronic stethoscope is used to collect heart and respiratory sound signals. The mobile app connects to the electronic stethoscope via Bluetooth, receives respiratory sound signals and terminal algorithm calculation results, and uploads the data to the cloud server under network conditions. User information management: The mobile APP supports user registration and login, and records the basic information of users and those being screened; History record management: The mobile app can query all historical records, recall voice data, and record diagnosis and treatment information related to treatment and rehabilitation; Terminal algorithm upgrade: The mobile APP realizes automatic online upgrade of terminal algorithms; Data annotation and management: The online annotation platform is used for annotation and review of heart and respiratory sound data. Doctors upload the collected heart and respiratory sound data to the platform. The platform provides an annotation interface and tools, supports multiple annotation methods, and the annotation results are saved uniformly. Intelligent analysis: The artificial intelligence algorithm module performs intelligent analysis on heart and respiratory sound data, including signal preprocessing, filtering, noise reduction, and normalization of heart and respiratory sound signals; extracts cardiac murmur and dry and wet rales features through time and frequency domain analysis; and uses machine learning algorithms such as support vector machines (SVM) to intelligently identify lesion features.

[0023] Among them, the identification of dry and wet rales features in the intelligent analysis specifically includes: constructing a sound library containing real dry and wet rales (Cracks) and a sound library containing pseudo dry and wet rales (false Cracks); extracting time domain features, including waveform zero-crossing rate, short-time energy, peak factor, etc., extracting frequency domain features, and using FFT to calculate the spectrum center of mass and frequency band energy ratio; using support vector machine (SVM) for binary classification training, optimizing kernel function and regularization parameters through grid search, and introducing cross-validation to improve model generalization ability; using resampling and bandpass filtering for signal preprocessing, dynamically calculating thresholds based on Gaussian mixture model (GMM), and screening potential dry and wet rales candidate segments.

[0024] Among them, the recognition of cardiac murmur features in the intelligent analysis specifically includes: using wavelet packet transform to perform multi-layer decomposition of cardiac sound signals to obtain sub-signals of different frequency bands; extracting the energy proportion, entropy value, kurtosis, and skewness of each frequency band sub-signal, calculating the marginal spectral entropy through Hilbert-Huang transform (HHT), and using a sliding window to analyze signal complexity; designing a dual-threshold formula, balancing sensitivity and specificity by adjusting the parameters in the formula, and using a support vector machine (linear kernel) for classification.

[0025] When using: (1) Mobile APP: Data transmission: The mobile APP communicates with the stethoscope via Bluetooth, receives the respiratory sound signals sent by the stethoscope and the calculation results of the terminal algorithm, and communicates with the cloud server when network conditions are available to upload the collected data to the cloud server to realize data transmission and storage.

[0026] User information entry: supports users to register and log in to their accounts, and can enter detailed information such as the name, date of birth, and place of birth of the screened person to facilitate the establishment of patient files.

[0027] History record management: Ability to query all historical records, recall voice data, and enter and query treatment, rehabilitation and other medical information of the screened person. Important node information can also be pushed to users who follow the screened person to facilitate tracking of the patient's condition.

[0028] Terminal algorithm upgrade: adopts the cloud training and terminal execution mode; when the cloud algorithm is trained with new data and a new terminal algorithm is released, the mobile APP automatically downloads the algorithm module to the mobile phone; after the stethoscope is connected to the mobile phone APP, the algorithm module on the stethoscope side is automatically upgraded to ensure the timeliness and accuracy of the algorithm.

[0029] (2) Online annotation platform: The annotator must have intermediate or above qualifications as a cardiologist or respiratory physician, and use online annotation. The annotator registers an independent account, and the respiratory sounds collected by the electronic stethoscope are automatically uploaded to the annotation platform. When annotating, the audio range is selected by sliding the mouse, and the annotation label is selected after online auscultation. After the annotation is completed, the annotation results are saved uniformly, including key information such as the annotator, annotation time, audio start and end time points, annotation label and corresponding disease.

[0030] (3) Artificial Intelligence Algorithm Module: Feature recognition of dry and wet rales: A dual-sound library containing real dry and wet rales (cracks) and pseudo dry and wet rales (false cracks) is constructed, ensuring sample diversity by combining clinical acquisition with simulation generation. Time domain features (such as waveform zero-crossing rate, short-time energy, and crest factor) and frequency domain features (such as using FFT to calculate the spectral centroid and the energy proportion of the 200-600 Hz frequency band) are extracted. Support vector machines (SVMs) are used for binary classification training. Grid search is used to optimize kernel functions (such as RBF kernels) and regularization parameters, and 5-fold cross-validation is introduced to improve model generalization. Signal preprocessing involves resampling (to enhance weak signals) and bandpass filtering (to retain the 20-2000 Hz frequency band). Intelligent thresholds are dynamically calculated based on the Gaussian mixture model (GMM) to screen for potential candidate segments of dry and wet rales. After testing 218 cases of moist rales data and 428 cases of normal breath sound data, the sensitivity reached 92.25%, the specificity reached 88.32%, and the accuracy reached 90.28%; after testing 264 cases of wheezing data and 264 cases of normal breath sound data, the sensitivity reached 86.74%, the specificity reached 87.50%, and the accuracy reached 87.12%.

[0031] Heart murmur feature recognition: Wavelet packet transform (e.g., db4 wavelet, 5-layer decomposition) is used to perform multi-layer decomposition of heart sound signals to obtain sub-signals in different frequency bands. Time-frequency domain features such as energy proportion, entropy, kurtosis, and skewness of each frequency band sub-signal are extracted, and marginal spectral entropy is calculated using the Hilbert-Huang transform (HHT). Signal complexity is analyzed using a sliding window with a 64-point window length and a 1-point frame shift. Design a dual threshold formula: Threshold 1=min(nfzcc)+k1*(max(nfzcc)-min(nfzcc)); Threshold 2=min(nfzcc)+k2*(max(nfzcc)-min(nfzcc))(k1>k2); By adjusting k1 and k2 (such as k1=0.4, k2=0.225) to balance sensitivity and specificity, SVM (linear kernel) was used for classification. After testing 678 cases of abnormal heart sound data and 324 cases of normal heart sound data, the sensitivity reached 92.27%, the specificity reached 90.22%, and the accuracy reached 91.24%.

[0032] The above are only specific embodiments of the present disclosure, but the protection scope of the present disclosure is not limited thereto. The protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A heart sound and respiratory sound data processing software system based on artificial intelligence algorithm, characterized by: include: Mobile APP module: installed on smart terminal devices, with data transmission function, can connect to the electronic stethoscope via Bluetooth to receive data and upload it to the cloud server, support user information entry, realize historical record management, and can upgrade terminal algorithms; Online annotation platform module: used for annotation and review of heart and respiratory sound data, including data upload and storage functions, providing annotation tools, and managing annotation results; Artificial intelligence algorithm module: performs intelligent analysis of heart and respiratory sound data, including signal preprocessing unit, feature extraction unit and intelligent recognition unit.

2. A heart sound and respiratory sound data processing software system based on an artificial intelligence algorithm as claimed in claim 1, characterized in that: The dry and wet rales feature recognition unit in the artificial intelligence algorithm module constructs a sound library containing real dry and wet rales and pseudo dry and wet rales, extracts time domain and frequency domain features, uses a support vector machine to train the model, and screens potential dry and wet rales through signal preprocessing and dynamic thresholding.

3. The heart sound and respiratory sound data processing software system based on artificial intelligence algorithm as claimed in claim 1, characterized in that: The heart murmur feature recognition unit in the artificial intelligence algorithm module uses wavelet packet transform to decompose the heart sound signal, extracts time-frequency domain features and HHT marginal spectral entropy, and performs classification through a double threshold formula and a support vector machine linear kernel.

4. A heart sound and respiratory sound data processing method based on artificial intelligence algorithm, characterized in that: The following steps are involved: Data collection and transmission: An electronic stethoscope is used to collect heart and respiratory sound signals. The mobile app connects to the electronic stethoscope via Bluetooth, receives respiratory sound signals and terminal algorithm calculation results, and uploads the data to the cloud server under network conditions. User information management: The mobile APP supports user registration and login, and records the basic information of users and those being screened; History record management: The mobile app can query all historical records, recall voice data, and record diagnosis and treatment information related to treatment and rehabilitation; Terminal algorithm upgrade: The mobile APP realizes automatic online upgrade of terminal algorithms; Data annotation and management: The online annotation platform is used for annotation and review of heart and respiratory sound data. Doctors upload the collected heart and respiratory sound data to the platform. The platform provides an annotation interface and tools, supports multiple annotation methods, and the annotation results are saved uniformly. Intelligent analysis: The artificial intelligence algorithm module performs intelligent analysis on heart sound and respiratory sound data, including signal preprocessing, filtering, noise reduction and normalization of heart sound and respiratory sound signals; extracts cardiac murmur and dry and wet rales features through time domain and frequency domain analysis; and uses machine learning algorithms such as support vector machines to intelligently identify lesion features.

5. The method of processing heart and respiratory sound data based on artificial intelligence algorithm according to claim 4, characterized in that: The identification of dry and wet rales features in the intelligent analysis specifically includes: constructing a sound library containing real dry and wet rales and a sound library containing pseudo dry and wet rales; extracting time domain features, including waveform zero-crossing rate, short-time energy, peak factor, etc., extracting frequency domain features, and using FFT to calculate the spectrum centroid and frequency band energy ratio; using support vector machines for binary classification training, optimizing kernel functions and regularization parameters through grid search, and introducing cross-validation to improve the generalization ability of the model; using resampling and bandpass filtering for signal preprocessing, dynamically calculating thresholds based on Gaussian mixture models, and screening potential dry and wet rales candidate segments.

6. A method for processing heart and respiratory sound data based on artificial intelligence algorithm according to claim 4, characterized in that: The cardiac murmur feature recognition in the intelligent analysis specifically includes: using wavelet packet transform to perform multi-layer decomposition of cardiac sound signals to obtain sub-signals of different frequency bands; extracting the energy proportion, entropy value, kurtosis, and skewness of each frequency band sub-signal, calculating the marginal spectral entropy through Hilbert-Huang transform, and using a sliding window to analyze signal complexity; designing a dual-threshold formula, balancing sensitivity and specificity by adjusting the parameters in the formula, and using the linear kernel of a support vector machine for classification.

Images