Lung sound analysis system

By designing a lung sound analysis system, using multi-channel signal acquisition and advanced artificial intelligence algorithms, the subjectivity and environmental interference problems of traditional auscultation methods are solved, and more accurate and reliable lung sound diagnosis is achieved, reducing equipment cost and complexity.

CN119091931APending Publication Date: 2024-12-06CHENGDU MEDICAL COLLEGE
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411285967.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-13
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

Traditional auscultation methods have problems such as subjectivity, difficulty in quantification and environmental interference in lung sound diagnosis. The complexity of the feature extraction and artificial intelligence analysis algorithms in the prior art is insufficient, resulting in the problems of diagnostic accuracy and high cost.

Method used

A lung sound analysis system was designed, including signal acquisition module, audio preprocessing module, signal processing module, feature extraction module and artificial intelligence analysis module. It uses a microphone array to collect multi-channel audio signals, and remove noise through a bandpass filter, which is short. Time Fourier transform and recursive neural network perform time-frequency mask separation, combining machine learning and deep learning algorithms to extract time domain, frequency domain and intensity features to identify normal and abnormal breathing patterns.

Benefits of technology

It improves the accuracy and reliability of lung sound diagnosis, can accurately capture and analyze lung sounds in noisy environments, provides more comprehensive lung sound analysis, reduces the cost and complexity of the equipment, and improves diagnostic efficiency and automation level.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119091931A_ABST
    Figure CN119091931A_ABST
Patent Text Reader

Abstract

The invention provides a lung sound analysis system. The lung sound analysis system comprises a signal acquisition module, an audio preprocessing module, an analog-digital converter, a signal processing module, a feature extraction module and an artificial intelligence analysis module, the signal acquisition module is used for acquiring multi-channel audio signals at different lung auscultation positions by using a microphone array; the audio preprocessing module is used for amplifying and filtering the collected lung sound signals so as to enhance the signal intensity and remove environmental noise; the analog-digital converter is used for converting the preprocessed analog signal into a digital signal; the signal processing module is used for processing the digital signal, including denoising and signal enhancement; the feature extraction module is used for extracting time domain, frequency domain and intensity feature parameters of the lung sound signals; and the artificial intelligence analysis module is used for analyzing the lung sound signals based on the extracted characteristic parameters so as to identify normal and abnormal breathing modes.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of lung sounds, and in particular to a lung sound analysis system. Background Art

[0002] In the medical field, accurate diagnosis is the key to successful treatment. Traditionally, doctors use stethoscopes to listen to the patient's lung sounds to judge the patient's health status. However, this method is highly dependent on the doctor's experience and subjective judgment, and cannot provide quantitative data support. With the rapid development of information technology, especially the advancement of digital signal processing and artificial intelligence technology, the field of medical diagnosis is undergoing a transformation.

[0003] First of all, global medical care is developing towards informatization and intelligence. This trend is reflected in the following aspects: 1. Digitization of medical data: Traditional paper medical records are being replaced by electronic medical record systems, and medical data are gradually being stored and transmitted digitally.

[0004] 2. The rise of telemedicine: Through the Internet and mobile communication technologies, telemedicine enables medical services to transcend geographical restrictions and improve the accessibility of medical resources.

[0005] 3. Application of artificial intelligence in medicine: Artificial intelligence technology, especially machine learning and deep learning, is being used to assist diagnosis, disease prediction and personalized treatment.

[0006] In specific medical diagnosis practices, the diagnosis of respiratory diseases relies on the accurate identification and analysis of lung sounds. However, the traditional auscultation method has the following limitations: Subjectivity: Different doctors may interpret the same lung sounds differently.

[0007] Difficult to quantify: Traditional auscultation cannot provide quantitative data, which limits the accuracy of diagnosis.

[0008] Environmental interference: A noisy environment may interfere with the auscultation results.

[0009] To overcome these limitations, researchers have begun to explore the application of electronic technology and artificial intelligence to the collection and analysis of lung sounds. Existing technologies have problems such as feature extraction and artificial intelligence analysis algorithms may not be sophisticated enough to accurately identify all types of lung sounds, and the cost of intelligent auscultation systems may be high, limiting their widespread application in clinical practice. Summary of the invention

[0010] In view of the deficiencies in the prior art, the present invention provides a lung sound analysis system, which allows clinicians to use electronic stethoscopes more accurately, more reliably and more easily, thereby improving the efficiency and accuracy of diagnosing respiratory diseases.

[0011] In order to solve the above technical problems, the present invention is solved by the following technical solutions In order to achieve the above object, the present invention adopts the following technical solutions: A lung sound analysis system, comprising: A signal acquisition module that uses a microphone array to collect multi-channel audio signals at different lung auscultation locations; An audio preprocessing module is used to amplify and filter the collected lung sound signals to enhance the signal strength and remove environmental noise; an analog-to-digital converter, used for converting the pre-processed analog signal into a digital signal; A signal processing module, used for processing digital signals, including noise removal and signal enhancement; A feature extraction module is used to extract the time domain, frequency domain and intensity feature parameters of the lung sound signal; An artificial intelligence analysis module is used to analyze lung sound signals based on the extracted feature parameters to identify normal and abnormal breathing patterns.

[0012] Preferably, the audio preprocessing module includes a bandpass filter for removing low-frequency noise and high-frequency interference in the lung sound signal.

[0013] Preferably, the signal processing module uses short-time Fourier transform and recursive neural network model to achieve time-frequency mask separation of lung sounds and noise signals.

[0014] Preferably, the feature extraction module further comprises an algorithm for separating the expiratory phase and the inspiratory phase.

[0015] Preferably, the algorithm for splitting the expiratory phase and the inspiratory phase is based on the analysis of short-time energy, and the expiratory phase and the inspiratory phase are preliminarily divided by energy threshold, and the final inspiratory phase and expiratory phase judgment is made in combination with time-frequency features and machine learning classifier.

[0016] Preferably, the feature extraction step includes performing short-time Fourier transform and wavelet transform on the processed lung sound signal, and extracting its time domain, frequency domain and intensity feature parameters.

[0017] Preferably, the artificial intelligence analysis module includes a convolutional neural network or a recurrent neural network model for learning and identifying normal and abnormal patterns of lung sounds and providing preliminary diagnostic results.

[0018] Preferably, the device further comprises a user interface for displaying a waveform diagram, a spectrum diagram and characteristic parameters of the lung sound signal and allowing a user to perform interactive operations.

[0019] Preferably, the user interface allows the physician to adjust parameters, view historical data and analyze results.

[0020] Compared with the prior art, the present invention has the following beneficial effects: The present invention aims to extract more comprehensive and detailed lung sound characteristic parameters, including but not limited to time domain, frequency domain and intensity parameters, through advanced signal processing technology, so as to provide a more comprehensive lung sound analysis.

[0021] The present invention aims to accurately capture and analyze lung sounds even in noisy environments by improving noise filtering and signal enhancement technology, thereby improving the stability and reliability of the device.

[0022] The present invention will adopt optimized algorithms, including machine learning and deep learning technologies, to improve the accuracy and automation level of lung sound feature extraction and provide doctors with more intelligent diagnostic support.

[0023] By integrating advanced artificial intelligence algorithms, the present invention aims to achieve an automated diagnostic function that can automatically identify and classify lung sounds and provide doctors with preliminary diagnostic suggestions. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0025] Figure 1 It is a schematic diagram of the overall system frame of the present invention; Figure 2 It is a schematic diagram of the signal acquisition module system frame of the present invention; Figure 3 It is a schematic diagram of the signal processing module system frame of the present invention; Figure 4 This is a schematic diagram of the lung sound signal before noise reduction of the present invention; Figure 5 It is a schematic diagram of the signal after the lung sound noise reduction of the present invention; Figure 6 It is a schematic diagram of the expiratory phase and inspiratory phase cycle segmentation signal of the present invention; Figure 7 A signal diagram of a sound extracted from the features of the present invention; Figure 8 The figure is a schematic diagram showing the structure of the interactive interface of the present invention. DETAILED DESCRIPTION

[0026] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art can think of other obvious variations. The basic principles of the present invention defined in the following description can be used for other embodiments, variations, improvements, equivalents, and other technical solutions that do not deviate from the spirit and scope of the present invention.

[0027] Those skilled in the art should understand that, in the disclosure of the present invention, the directions or positions indicated by the terms "longitudinal", "lateral", "up", "down", "left", "right", "front", "back", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc. are based on the directions or positional relationships shown in the accompanying drawings, which are only simplified descriptions for the convenience of describing the present invention, and do not indicate or imply that the device or component referred to must have a specific direction, be constructed and operated in a specific direction. Therefore, the above terms should not be understood as limitations on the present invention.

[0028] It is to be understood that the term "one" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of an element may be one, while in another embodiment, the number of the element may be multiple, and the term "one" should not be understood as a limitation on the quantity.

[0029] The present invention is further described in detail below in conjunction with the accompanying drawings: Figure 1 The system frame of the present invention provides a lung sound analysis system, which includes a signal acquisition module, an audio preprocessing module, an analog-to-digital converter, a signal processing module, a feature extraction module, and an artificial intelligence analysis module; The signal acquisition module uses a microphone array to collect multi-channel audio signals at different lung auscultation positions; the audio preprocessing module is used to amplify and filter the collected lung sound signals to enhance the signal strength and remove environmental noise; the analog-to-digital converter is used to convert the preprocessed analog signal into a digital signal; the signal processing module is used to process the digital signal, including denoising and signal enhancement; the feature extraction module is used to extract the time domain, frequency domain and intensity feature parameters of the lung sound signal; the artificial intelligence analysis module is used to analyze the lung sound signal based on the extracted feature parameters to identify normal and abnormal breathing patterns.

[0030] Reference Figure 2 The signal acquisition module of the present invention is integrated into the electronic stethoscope to collect the patient's lung sounds. Before use, the stethoscope is sterilized and the probe is placed on the palm for 1-2 minutes. The following steps are taken during use: After making sure the surrounding environment is quiet; The tester exposes his chest, sits, stands or lies flat, inserts the plug into the earphone jack of the mobile phone, places the stethoscope head at the position where auscultation is required, applies appropriate pressure, turns on the mobile phone recorder, and starts recording. The recording time should be at least 2 respiratory cycles for each lung sound auscultation area (because there will be a short and loud noise at the beginning, the recording time can be appropriately extended by about 10 seconds); Six lung auscultation sites were located, namely, the bilateral apex, hilum, and base: (1) right apex: second intercostal space on the right midclavicular line; (2) left apex: second intercostal space on the left midclavicular line; (3) right hilum: fourth intercostal space on the right parasternal line; (4) left hilum: fourth intercostal space on the left parasternal line; (5) right base: sixth intercostal space on the right midaxillary line; (6) left base: sixth intercostal space on the left midaxillary line.

[0031] The audio preprocessing module of the present invention amplifies the collected lung sound signal to enhance the signal strength, and then removes low-frequency noise and high-frequency interference in the signal through a bandpass filter to improve the signal-to-noise ratio.

[0032] The analog-to-digital converter (ADC) of the present invention converts the pre-processed analog signal into a digital signal to facilitate subsequent digital signal processing.

[0033] Reference Figure 3-Figure 5 The signal processing module of the present invention: firstly, preprocess the mixed signal of heart and lung sounds with noise, including unifying the sampling rate and normalizing it, then convert it into a spectrum representation by using short-time Fourier transform (STFT), process the STFT amplitude-mode spectrum by a recurrent neural network (RNN) model, and obtain the time-frequency mask of the heart sound signal and the noise signal; then, multiply the STFT amplitude-mode spectrum by the mask output by the RNN, reconstruct the signal by using the inverse STFT, and obtain the estimation of the noisy heart sound signal; then, filter the heart sound signal, extract features by using the Hilbert transform, calculate the heart rate and duration, and obtain different sub-state sequences of the heart sound signal by using the hidden Markov model (HMM); after denoising and splicing the sub-state sequence, repeat the iteration until the set number of iterations is reached, and finally realize the effective separation and reconstruction of the noisy heart and lung sound signal.

[0034] Reference Figure 6-Figure 7 The feature extraction module of the present invention includes an algorithm for splitting the exhalation phase (S1) and the inspiration phase (S2), which splits S1 and S2 by calculating the start and end positions; after locating and segmenting the breathing cycle, the audio data analysis tool identifies the intensity features, time domain and frequency domain features of S1 and S2 as follows: (I) The duration of S1 is longer than that of S2 and the amplitude of S1 is also much higher than that of S2; (II) S1 and S2 have a certain frequency range physiologically; (III) S2 occurs between adjacent S1 signals.

[0035] By combining the above features, the start and end positions of S1 and S2 are determined.

[0036] The feature extraction module of the present invention uses S1 and S2 as segments to extract digital signals, and displays the signal information through a two-dimensional plane, so that the analysis of the signal can obtain the specific time and specific frequency of the signal. It is an important method that is indispensable for the analysis and processing of non-stationary signals. The heart sound signal is windowed and the signal in each window is Fourier transformed. Usually, the window function uses a real even function whose energy is concentrated at low frequencies, so as to ensure that the short-time Fourier transform has localization functions in both the time domain and the frequency domain. It effectively extracts information from the signal and is suitable for analyzing non-stationary heart sound signals. The basis of the Hilbert space can be generated by the translation operation of the mother wavelet function, and different Hilbert subspaces can be formed by the combined compression and translation operation of the wavelet mother function. Signal analysis in different Hilbert subspaces reveals the approximation characteristics of the signal from coarse to fine.

[0037] The feature extraction module of the present invention extracts the signal features of the relevant sound by extracting some important features of the cough sound signal, including energy, frequency, time domain and frequency domain features; the purpose of feature extraction is to convert the cough sound signal into a mathematical form for subsequent analysis and processing.

[0038] Among them, feature extraction methods include short-time Fourier transform, wavelet packet transform, wavelet transform, cepstral coefficients and other methods.

[0039] The artificial intelligence analysis module of the present invention includes a convolutional neural network or a recurrent neural network model, which uses deep learning models such as convolutional neural network (CNN) or recurrent neural network (RNN) to train the extracted features, learn the normal and abnormal patterns of lung sounds, predict the patient's lung sound status through the model, and provide preliminary diagnostic results.

[0040] Further, 1. Convolutional Neural Network (CNN) Model ‌Network Structure‌:‌ - ‌Input layer‌: ‌Accepts a two-dimensional time-frequency spectrum graph after short-time Fourier transform (‌STFT‌), ‌with a shape of (M,N, 1), ‌where M is the time step and ‌N is the frequency component. ‌ - Convolutional Layers: First layer: ‌32 3×3 convolution kernels, ‌ReLU activation function. ‌ Second layer: 64 3×3 convolution kernels, ReLU activation function. Third layer: 128 3×3 convolution kernels, ReLU activation function. -Pooling layers: Each convolution layer is followed by a 2×2 max pooling. - ‌Fully connected layer‌: ‌512 neurons, ‌ReLU activation function, ‌Dropout 50% to prevent overfitting. ‌ - Output layer: 2 neurons, corresponding to the classification of S1 and S2, using Softmax activation function. ‌Model Description‌:‌ The CNN model extracts high-order features of lung sound signals through multi-layer convolution and pooling operations. It is suitable for processing two-dimensional inputs such as time-frequency spectrograms and can capture the time-frequency characteristics of lung sound signals. 2. Recurrent Neural Network (RNN) Model ‌Network Structure‌:‌ - ‌Input layer‌: ‌Accepts one-dimensional time series data, ‌ with a shape of (T, F), ‌ where T is the time step and ‌ F is the number of features at each step. ‌ - ‌Bidirectional LSTM layer‌: ‌128 units, ‌tanh activation function, ‌Dropout 50%. ‌ - ‌LSTM layer‌: ‌64 units, ‌tanh activation function, ‌Dropout 50%. ‌ -‌Fully connected layer‌:‌ 128 neurons,‌ ReLU activation function,‌ Dropout 50%.‌ - Output layer: 2 neurons, corresponding to the classification of S1 and S2, using Softmax activation function. ‌Model Description‌:‌ The RNN model processes the time series characteristics of lung sound signals through LSTM units, which can capture the temporal dependencies of the signals. The bidirectional LSTM layer enhances the model's temporal perception ability and helps to more accurately divide S1 and S2.

[0041] Reference Figure 8 The present invention also includes a user interface for displaying the waveform, spectrum and characteristic parameters of the lung sound signal and allowing the user to perform interactive operations; the user interface allows the doctor to adjust parameters, view historical data and analyze results.

[0042] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are only examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been demonstrated and explained in the embodiments, and the embodiments of the present invention may be deformed or modified in any way without departing from the principles.

Claims

1. A lung sound analysis system, characterized in that: include: A signal acquisition module that uses a microphone array to collect multi-channel audio signals at different lung auscultation locations; An audio preprocessing module is used to amplify and filter the collected lung sound signals to enhance the signal strength and remove environmental noise; an analog-to-digital converter, used for converting the pre-processed analog signal into a digital signal; A signal processing module, used for processing digital signals, including noise removal and signal enhancement; A feature extraction module is used to extract the time domain, frequency domain and intensity feature parameters of the lung sound signal; An artificial intelligence analysis module is used to analyze lung sound signals based on the extracted feature parameters to identify normal and abnormal breathing patterns.

2. A lung sound analysis system according to claim 1, characterized in that: The audio preprocessing module includes a bandpass filter for removing low-frequency noise and high-frequency interference in the lung sound signal.

3. A lung sound analysis system according to claim 1, characterized in that: The signal processing module adopts short-time Fourier transform and recursive neural network model to achieve time-frequency mask separation of lung sounds and noise signals.

4. A lung sound analysis system according to claim 1, characterized in that: The feature extraction module further includes an algorithm for separating the expiratory phase and the inspiratory phase.

5. A lung sound analysis system according to claim 4, characterized in that: The algorithm for splitting the expiratory phase and the inspiratory phase is based on the analysis of short-time energy, and preliminarily divides the expiratory phase and the inspiratory phase through the energy threshold, and makes the final judgment of the inspiratory phase and the expiratory phase by combining the time-frequency characteristics and the machine learning classifier.

6. A lung sound analysis system according to claim 5, characterized in that: The feature extraction step includes performing short-time Fourier transform and wavelet transform on the processed lung sound signal, and extracting its time domain, frequency domain and intensity feature parameters.

7. A lung sound analysis system according to any one of claims 1 to 6, characterized in that: The artificial intelligence analysis module includes a convolutional neural network or a recurrent neural network model, which is used to learn and identify normal and abnormal patterns of lung sounds and provide preliminary diagnostic results.

8. A lung sound analysis system according to claim 7, characterized in that: The device also includes a user interface for displaying a waveform diagram, a spectrum diagram and characteristic parameters of the lung sound signal and allowing a user to perform interactive operations.

9. A lung sound analysis system according to claim 8, characterized in that: The user interface allows the physician to adjust parameters, view historical data, and analyze results.