MEMS heart sound and electrocardio detection system, method and equipment

Abstract

The invention provides an MEMS heart sound and electrocardio detection system, method and equipment, and relates to the technical field of medical instruments, the MEMS heart sound and electrocardio detection system comprises a bionic cilium heart sound sensor, an electrocardio acquisition circuit and a signal fusion model; the bionic cilia heart sound sensor comprises hollow bullet type bionic cilia and a cantilever beam microstructure; the hollow bullet type bionic cilia and the cantilever beam microstructure are mechanically coupled, and the cantilever beam microstructure comprises a Wheatstone bridge formed by arranging a plurality of piezoresistors and is used for collecting heart sound signals; the electrocardio acquisition circuit and the bionic cilium heart sound sensor are packaged in an acoustic impedance matching mode, and the electrocardio acquisition circuit comprises an electrocardio acquisition electrode plate, a power supply voltage stabilization module, an instrument amplification module, a band-pass filtering module and a notch filter and is used for acquiring and conditioning electrocardio signals; the signal fusion model comprises a Resnet18 feature extractor, an encoder and a mutual cross attention mechanism, and is used for performing feature extraction, linear transformation and feature deep fusion on the collected electrocardiosignals and electrocardiosignals and outputting a signal classification result.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of medical devices, in particular to a MEMS heart sound and electrocardiogram detection system, method and device. BACKGROUND

[0002] Heart sound signal is an important physiological signal of human body, which reflects the movement of heart and cardiovascular system, and contains physiological information and pathological information of interaction between each part of heart such as atrium, ventricle, cardiovascular system, large blood vessels and each valve. Heart function can be evaluated by using heart sound. Electrocardiogram signal is a signal reflecting the coupling state of heart electrical activity and mechanical function, which contains core physiological information and pathological characteristics such as myocardial cell depolarization and repolarization, heart conduction system function, ventricular muscle electromechanical contraction synchrony and electrolyte balance. By analyzing the timing characteristics and waveform variation of electrocardiogram signal, key functional indicators such as heart rhythm, myocardial ischemia, conduction block and structural lesions can be accurately evaluated.

[0003] With the continuous development of electronic auscultation technology, various electronic stethoscopes have emerged in the market in recent years, but most of the stethoscopes can only detect heart sound or electrocardiogram signal. This separation makes it difficult for doctors to directly associate electrocardiogram rhythm abnormalities with heart sound changes, increasing the risk of missed diagnosis. Even though some devices try to integrate heart sound and electrocardiogram modules, the signal time synchronization error is large, which may mask the key pathological characteristics for detecting short PR interval (such as pre-excitation syndrome) or small ST segment deviation (such as early myocardial ischemia), and it is difficult to meet the demand of full-dimensional evaluation. SUMMARY

[0004] Therefore, the present application provides a MEMS heart sound and electrocardiogram detection system, method and device.

[0005] The technical scheme of the present application is realized as follows: the present application provides a MEMS heart sound and electrocardiogram detection system in the first aspect, which comprises:

[0006] bionic cilium heart sound sensor, electrocardiogram acquisition circuit and signal fusion model;

[0007] The bionic cilium heart sound sensor comprises a hollow bullet type bionic cilium and a cantilever beam microstructure; the hollow bullet type bionic cilium is mechanically coupled with the cantilever beam microstructure; the cantilever beam microstructure comprises a Wheatstone bridge composed of a plurality of pressure sensitive resistors, which is used for collecting heart sound signals;

[0008] The electrocardiogram acquisition circuit is packaged with the bionic cilium heart sound sensor through acoustic impedance matching, and comprises an electrocardiogram acquisition electrode sheet, a power supply voltage stabilizing module, an instrument amplification module, a band pass filter module and a notch filter, which are used for collecting and conditioning electrocardiogram signals;

[0009] The signal fusion model comprises a Resnet18 feature extractor, an encoder and a mutual cross-attention mechanism, and is used for feature extraction, linear transformation and feature depth fusion of the collected electrocardio signals and the electrocardio signals, and outputs a signal classification result.

[0010] Based on the above technical scheme, preferably, the heart sound simulation circuit connected with the biomimetic cilium heart sound sensor comprises a power stabilizing module, an amplification module and a band-pass filtering module.

[0011] The power stabilizing module is used for providing stable voltage for the heart sound simulation circuit.

[0012] The amplification module is used for enhancing the input electrocardio signals.

[0013] The band-pass filtering module is used for denoising the electrocardio signals after enhancement and extracting electrocardio signals in a preset frequency band.

[0014] Based on the above technical scheme, preferably, the biomimetic cilium heart sound sensor is connected with the heart sound simulation circuit through wire bonding, and is fixed by using an internal card slot of a shell wrapping the biomimetic cilium heart sound sensor, the electrocardio collection circuit and the signal fusion model.

[0015] Based on the above technical scheme, preferably, the synchronous collector comprises an analog-digital conversion module and a communication module.

[0016] The analog-digital conversion module is used for converting the electrocardio signals collected by the biomimetic cilium heart sound sensor and the electrocardio signals collected by the electrocardio collection circuit from analog signals into digital signals.

[0017] The communication module is used for realizing communication connection with the upper computer through a USB interface, and transmitting the collected synchronous heart sound electrocardio signals to the signal fusion model in the upper computer.

[0018] Based on the above technical scheme, preferably, the Resnet18 feature extractor comprises a 7*7 convolution layer, a 3*3 maximum pooling layer and four continuous basic blocks, each of which comprises two residual blocks; the encoder comprises position encoding, a multi-head attention mechanism, a feedforward network and a sum normalization structure; and the mutual cross-attention mechanism comprises two one-way cross-attention mechanisms.

[0019] On the basis of the above technical scheme, preferably, the mutual cross attention mechanism is used to generate corresponding query matrix, key matrix and value matrix through linear transformation of an input pair of heart sound signal features and electrocardiogram signal features, and the mutual cross attention mechanism is used to realize bidirectional dot product attention operation and output signal classification results.

[0020] On the basis of the above technical scheme, preferably, the method further comprises an analysis module, wherein the analysis module is used to:

[0021] The signal classification results are analyzed in time-frequency domain, and the synchronous heart sound electrocardiogram signals in the signal classification results are visually processed by wavelet transformation to obtain a time-frequency graph and / or an envelope graph.

[0022] Further preferably, the second aspect of the present application provides a MEMS heart sound electrocardiogram detection method, which is applied to the MEMS heart sound electrocardiogram detection system of the first aspect and comprises the following steps:

[0023] The heart sound signal of the target to be detected is collected by using the bionic cilium heart sound sensor, and the electrocardiogram signal of the target to be detected is collected by using the electrocardiogram collection circuit.

[0024] The heart sound signal and the electrocardiogram signal are subjected to feature extraction and feature enhancement based on a Resnet18 feature extractor and an encoder to generate heart sound signal features and electrocardiogram signal features.

[0025] The heart sound signal features and the electrocardiogram signal features are subjected to linear transformation and feature depth fusion by using a mutual cross attention mechanism to output signal classification results.

[0026] On the basis of the above technical scheme, preferably, after the heart sound signal of the target to be detected is collected by using the bionic cilium heart sound sensor and the electrocardiogram signal of the target to be detected is collected by using the electrocardiogram collection circuit, the method further comprises the following steps:

[0027] The heart sound signal and the electrocardiogram signal are subjected to down-sampling, and the heart sound signal is subjected to filtering processing by using a Butterworth band-pass filter, and the electrocardiogram signal is subjected to filtering processing by using a Butterworth band-pass filter and a notch filter.

[0028] The filtered heart sound signal and the filtered electrocardiogram signal are subjected to segmentation processing to obtain a plurality of signal segments with the same length.

[0029] The signal segments are subjected to normalization processing, and the signal segments are converted into time-frequency graphs by using continuous wavelet transformation.

[0030] Further preferably, the third aspect of the present application provides a MEMS heart sound electrocardiogram detection device, which comprises the MEMS heart sound electrocardiogram detection system of the first aspect.

[0031] The MEMS heart sound and electrocardio detection system, method and device of the present application have the following beneficial effects relative to the prior art:

[0032] 1. By adopting the mechanical coupling design of the hollow bullet type bionic cilium, the detection sensitivity of the system is improved. The electrocardio collection circuit and the bionic cilium heart sound sensor are packaged through acoustic impedance matching, which reduces the problem of signal attenuation and interference caused by the impedance mismatch of the transmission medium of the heart sound sound wave and the electrocardio electrical signal, and reduces the detection error. By concentrating the heart sound signal and the electrocardio signal to the signal fusion model for synchronous fusion, zero delay synchronization and pathological level fusion of the heart sound signal and the electrocardio signal are realized, greatly improving the accuracy of detection.

[0033] 2. The signals of the bionic cilium heart sound sensor and the electrocardio collection circuit are simultaneously collected by the synchronous collector, and are synchronously converted into digital signals and transmitted to the signal fusion model in the host computer, which can accurately capture the time relationship between the heart sound and electrocardio signals, and provide high-quality input data for the signal fusion model. The signal fusion model can integrate the multi-dimensional information of the heart sound signal and the electrocardio signal, mine the internal relationship between the signals through advanced algorithms, improve the sensitivity and specificity of disease diagnosis, and provide more valuable reference for clinical diagnosis and treatment.

[0034] 3. The mutual cross-attention mechanism realizes the bidirectional information interaction between the heart sound signal features and the electrocardio signal features. By generating a query matrix, a key matrix and a value matrix, and using dot product attention operation, important associated information between the heart sound and electrocardio features can be automatically learned, more discriminative fusion features can be generated, the limitations of single modal information are reduced, and the accuracy and reliability of classification are improved. BRIEF DESCRIPTION OF DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0036] Figure 1 A structural schematic diagram of a MEMS heart sound and electrocardio detection system provided by an embodiment of the present application;

[0037] Figure 2 A structural schematic diagram of a MEMS heart sound and electrocardio synchronous stethoscope provided by an embodiment of the present application;

[0038] Figure 3This is a schematic diagram of the structure of the biomimetic ciliary heart sound sensor provided in an embodiment of the present invention;

[0039] Figure 4 A schematic diagram of the algorithm framework structure provided in the embodiments of the present invention;

[0040] Figure 5 This is a flowchart illustrating a MEMS heart sound and electrocardiogram detection method provided in an embodiment of the present invention. Detailed Implementation

[0041] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0042] In some embodiments, such as Figure 1 As shown, Figure 1 This is a schematic diagram of a MEMS heart sound and electrocardiogram detection system provided in an embodiment of the present invention. The MEMS heart sound and electrocardiogram detection system 100 provided by the present invention includes: a biomimetic ciliary heart sound sensor 110, an electrocardiogram acquisition circuit 120, and a signal fusion model 130.

[0043] The biomimetic ciliary heart sound sensor 110 includes a hollow bullet-shaped biomimetic ciliary 111 and a cantilever beam microstructure 112; the hollow bullet-shaped biomimetic ciliary 111 and the cantilever beam microstructure 112 are mechanically coupled, and the cantilever beam microstructure 112 includes a Wheatstone bridge composed of multiple varistors for the acquisition of heart sound signals.

[0044] The electrocardiogram (ECG) acquisition circuit 120 is packaged with the bionic ciliary heart sound sensor 110 through acoustic impedance matching. It includes an ECG acquisition electrode 121, a power supply regulator module 122, an instrumentation amplifier module 123, a bandpass filter module 124, and a notch filter 125, and is used for ECG signal acquisition and conditioning.

[0045] The signal fusion model 130 includes a ResNet18 feature extractor 131, an encoder 132, and a cross-attention mechanism 133, which is used to extract features, perform linear transformations and deep feature fusion on the acquired ECG signals and output signal classification results.

[0046] In this embodiment, the hollow bullet-shaped biomimetic cilium 111 has low hydrodynamic resistance in the acoustic waves generated by the flow of fluid environment such as blood flow, and can more efficiently capture the slight changes of heart sound signals. The hollow design helps to optimize the acoustic characteristics, enhance the response to specific frequency heart sound components, and improve the sensitivity of heart sound detection. When the cantilever beam microstructure 112 is subjected to stress generated by heart sound vibration, the resistance value of the piezoresistor will change, and this resistance value change is converted into an electrical signal output through the Wheatstone bridge, realizing accurate detection of heart sound signals. The ECG acquisition circuit 120 is packaged with the biomimetic cilium heart sound sensor 110 through acoustic impedance matching. This packaging method can effectively reduce the reflection and loss of heart sound signals between the sensor and the circuit, and improve the transmission efficiency of heart sound signals. The ECG acquisition electrode sheet 121 is directly in contact with the human body to obtain the ECG signal. The power stabilizing module 122 provides stable power supply for the entire ECG acquisition circuit. The instrument amplification module 123 has high input impedance and high common-mode rejection ratio, which can effectively amplify the ECG signal and suppress the common-mode interference signal, thereby improving the signal-to-noise ratio of the ECG signal. The band-pass filter module 124 allows ECG signals within a specific frequency range to pass through while suppressing interference signals of other frequencies. The notch filter 125 can further filter out interference of a specific frequency, such as power frequency interference. Through the combined use of these two filter modules, the ECG signal can be further purified, and the signal quality can be improved. The Resnet18 feature extractor 131 has strong feature extraction capability and can automatically learn complex features in heart sound signals and ECG signals. The residual connection solves the gradient vanishing problem in deep network training. The encoder 132 encodes the extracted features to compensate for the lack of attention to long-distance semantic information by the Resnet18 feature extractor. Through multi-head attention and other operations, the attention to long-distance feature information is enhanced. The mutual cross-attention mechanism 133 realizes bidirectional information interaction and deep fusion between heart sound signal features and ECG signal features.

[0047] In one example, the size of the hollow bullet-shaped biomimetic cilium is shown in Table 1 below:

[0048] Dimensional parameters Values Bullet head height (pm) 750 Bullet head base radius (mm) 1.5 Hollow cylinder height (mm) 3.5 Hollow cylinder outer diameter (mm) 0.35 Hollow cylinder inner diameter (mm) 0.1

[0049] Table 1

[0050] The height of the bullet head, the radius of the bottom of the bullet head, the height of the hollow cylinder, the outer diameter of the hollow cylinder, and the inner diameter of the hollow cylinder corresponding to the hollow bullet-shaped bionic cilium can fluctuate within a small range, and Table 1 above is exemplary. The micron and millimeter level dimensions can significantly reduce the contact area of the cilium with the fluid and reduce the viscous drag. The hollow structure can accelerate the deformation recovery process by reducing the inertial mass, increase the dynamic response frequency, and thus improve the detection accuracy of the heart sound signal. Compared with a simple cylindrical cilium, the hollow cylinder and bullet head structure reduce the weight of the cilium, improve the sensitivity of the cilium, and ensure the low-frequency characteristics of the sensor, making it more suitable for heart sound sensing. The specific size of the hollow bullet-shaped bionic cilium can be obtained by using COMSOL software to perform parameterized scanning simulation and characteristic frequency analysis on the cilium size. After the characteristic frequency simulation, the first-order natural frequency of the hollow bullet-shaped bionic cilium with the size in Table 1 above in air is 1332.7 Hz, and the first-order natural frequency in fluid is 898.64 Hz, meeting the design requirements of the 20-600 Hz working bandwidth of the heart sound signal. The hollow bullet-shaped bionic cilium can be made by a high-precision 3D printing device with a precision of 20 μm, and can be integrated with the cantilever beam microstructure by adhesion.

[0051] In another example, the amplification factor of the instrumentation amplification module is 1100 times. The passband frequency of the bandpass filter module is set to 0.5-100 Hz to filter out the electromyographic noise in the electrocardiogram signal, and the center frequency of the stopband of the notch filter is 50 Hz to filter out the power frequency interference.

[0052] In some embodiments, the MEMS heart sound and electrocardiogram detection system further comprises a heart sound analog circuit connected to the bionic cilium heart sound sensor; the heart sound analog circuit comprises: a power stabilizing module, an amplification module, and a bandpass filter module;

[0053] The power stabilizing module is used to provide stable voltage for the heart sound analog circuit;

[0054] The amplification module is used to enhance the input electrocardiogram signal;

[0055] The bandpass filter module is used to denoise the enhanced electrocardiogram signal and extract the electrocardiogram signal in the preset frequency band.

[0056] The core function of the power stabilizing module is to eliminate the fluctuation and noise of the input power voltage, and to provide stable and clean DC voltage for the entire heart sound analog circuit. The electrocardiogram signal is usually very weak, with an amplitude generally in the millivolt level, and needs to be enhanced by the amplification module to meet the subsequent circuit acquisition and analysis requirements. The amplification module can be realized by using an operational amplifier, which can amplify the input electrocardiogram signal by using the high gain characteristic of the operational amplifier. The function of the band-pass filter module is to allow the electrocardiogram signal in a specific frequency range to pass through, while suppressing other frequency interference signals. It can be composed of a low-pass filter and a high-pass filter in series, and the passband range of the band-pass filter can be determined by reasonably setting the cutoff frequency of the low-pass filter and the cutoff frequency of the high-pass filter.

[0057] For example, the amplification factor in the heart sound analog circuit is set to 495 times, and the filter passband frequency is 20-600Hz, covering the main frequency band of the heart sound signal, to filter out environmental noise, respiratory artifacts and other noise mixed in the signal acquisition process, and to ensure the quality of the collected signal.

[0058] In some embodiments, the bionic cilium heart sound sensor is connected to the heart sound analog circuit through wire bonding, and is fixed by using the internal card slot of the shell wrapping the bionic cilium heart sound sensor, the electrocardio acquisition circuit and the signal fusion model.

[0059] In this embodiment, the bionic cilium heart sound sensor after secondary integration is connected to the heart sound analog circuit through wire bonding, and is fixed by means of the internal card slot of the shell. The weak signal collected by the hollow bullet-shaped bionic cilium and the cantilever beam microstructure is sent to the heart sound analog circuit for voltage stabilization, amplification and filter noise reduction. Here, the bionic cilium heart sound sensor, the heart sound analog circuit and the electrocardio acquisition circuit can be packaged through acoustic impedance matching to form a MEMS heart sound electrocardio synchronous stethoscope. The MEMS heart sound electrocardio synchronous stethoscope can also include a stethoscope shell, a sound-transmitting membrane, an internal support framework and a medical coupling agent. The sound-transmitting membrane is made of e-PTFE porous sound-transmitting material, and the coupling agent is a medical coupling agent with acoustic impedance close to human soft tissue, so as to reduce the loss of heart sound during transmission.

[0060] In one example, please refer to Figure 2 , Figure 2The structural schematic diagram of the MEMS heart sound and electrocardiogram synchronous stethoscope provided by the embodiment of the present application mainly comprises: a stethoscope shell 2-1 made of polished aluminum alloy material, which has high mechanical strength and good electromagnetic compatibility and can shield external environmental noise; a bionic cilium heart sound sensor 2-2 is connected to a heart sound analog circuit 2-4 through wire bonding to realize electrical connection between the microstructure and the circuit board; an electrocardio collection circuit 2-3, a stethoscope inner support skeleton 2-5, an e-PTFE sound transmission film 2-6, a stethoscope cavity 2-7, an embedded heart sound conditioning circuit board, and filling with medical coupling agent; and an electrocardio electrode sheet 2-8 extends outward through a connecting line.

[0061] In another example, refer to Figure 3 , Figure 3 The structural schematic diagram of the bionic cilium heart sound sensor provided by the embodiment of the present application; the cantilever beam microstructure 3-1 is a cross-shaped cantilever beam, the hollow bullet-shaped bionic cilium 3-2 supports the column of the bullet head, and the Wheatstone bridge 3-3 is arranged by piezoresistors. The bionic cilium is integrated onto the cross beam microstructure by adhesive bonding.

[0062] In some embodiments, the MEMS heart sound and electrocardio detection system further comprises: a synchronous collector; the synchronous collector comprises an analog-to-digital conversion module and a communication module;

[0063] The analog-to-digital conversion module is used to convert the electrocardio signals collected by the bionic cilium heart sound sensor and the electrocardio signals collected by the electrocardio collection circuit from analog signals to digital signals;

[0064] The communication module is used to realize communication connection with the upper computer through a USB interface, and transmit the collected synchronous heart sound and electrocardio signals to a signal fusion model in the upper computer.

[0065] In the embodiment, the signals of the bionic cilium heart sound sensor and the electrocardio collection circuit are collected simultaneously by the synchronous collector, and are converted into digital signals and transmitted to the signal fusion model in the upper computer. In addition to the above-mentioned modules, the synchronous collector further comprises a main control module and an isolation module. The main control module is connected with other modules and is the core of the whole collector. The isolation module respectively isolates the power supply and the USB to ensure the stability of their respective work and the safety of the user. Here, the sampling rate of the synchronous collector heart sound channel and the electrocardio channel can be set to 4000Hz.

[0066] In some embodiments, the Resnet18 feature extractor comprises one 7x7 convolutional layer, one 3x3 max pooling layer, and four consecutive basic blocks, each basic block comprising two residual blocks; the encoder comprises position encoding, multi-head attention mechanism, feedforward network, and sum normalization structure; the mutual cross-attention mechanism comprises two one-way cross-attention mechanisms.

[0067] In the present embodiment, the two groups of high-dimensional features extracted by Resnet18 are sent to the encoder in the form of 512-dimensional vectors. The encoder structure can effectively expand the model's attention to long-distance semantic information, expand the model's perception field, and to some extent make up for the problem of insufficient attention to overall rhythm abnormalities of heart sound and electrocardiogram by the model based on convolution structure. The output of the encoder is a pair of 512-dimensional feature vectors, and then the pair of feature vectors are sent to the mutual cross-attention mechanism for deep fusion between the heart sound and electrocardiogram modalities. The mutual cross-attention mechanism consists of two one-way cross-attention (CA) mechanisms. Let the input pair of heart sound and electrocardiogram feature vectors be F pgc and F egc , respectively, after linear transformation, the corresponding query (Q), key (K) and value (V) matrices are generated, and the two one-way cross-attention operations are as follows:

[0068]

[0069]

[0070] where d k is the dimension of K, Q pcg , K pcg and V pcg represent the query, key and value matrices corresponding to the heart sound feature vector, respectively, and Q ecg , K ecg and V ecg represent the query, key or value matrices corresponding to the electrocardiogram feature vector, respectively. On the basis of the above scheme, the calculation process of mutual cross-attention can be represented as:

[0071] MCA=CA1(Q pcg , K ecg , V ecg )+CA2(Q ecg , K pcg , V pcg ).

[0072] The fused feature vectors are classified by a fully connected layer.

[0073] In an optional embodiment, please refer to Figure 4 , Figure 4The algorithm framework structure schematic diagram provided by the embodiment of the present application; the input of the framework is the time-frequency diagram of the electrocardiosignal and the heart sound signal. The ResNet18 feature extractor extracts a 512-dimensional feature vector from the input heart sound electrocardio time-frequency diagram. The encoder structure expands the perception field of the model and captures long-distance semantic information in the heart sound electrocardio time-frequency diagram. The mutual cross-attention mechanism performs bidirectional dot product attention operation and addition on a pair of heart sound electrocardio feature vectors output by the encoder, and realizes deep fusion of the feature vectors.

[0074] In some embodiments, the mutual cross-attention mechanism is used to generate corresponding query matrices, key matrices and value matrices by linear transformation of the input pair of heart sound signal features and electrocardio signal features respectively, and to realize bidirectional dot product attention operation by using the mutual cross-attention mechanism, and output a signal classification result.

[0075] In some embodiments, the MEMS heart sound electrocardio detection system further comprises an analysis module; the analysis module is used for:

[0076] The time-frequency domain analysis is performed on the signal classification result, and the time-frequency diagram of the synchronous heart sound electrocardio signal in the signal classification result is visualized by using wavelet transform to obtain a time-frequency envelope diagram.

[0077] After obtaining the classified heart sound electrocardio signal, the change characteristics of the signal over time, such as amplitude and periodicity, are observed, the frequency components of the signal are analyzed by Fourier transform FFT, and the time and frequency information of the signal can be captured simultaneously by using wavelet transform. On this basis, a suitable wavelet basis function is selected, the time-frequency distribution of the signal is calculated, the time-frequency envelope is extracted, and then the time-frequency envelope diagram can be drawn to compare the time-frequency characteristics of different classified signals.

[0078] In an optional embodiment, the application scenario of the MEMS heart sound and electrocardiogram detection system is taken as an example for illustration. After a user logs in the software through an account and a password, enters the host computer homepage, the homepage mainly includes real-time display of synchronous heart sound and electrocardiogram waveforms and data storage functions. Before the display, the software will first check the device connection state of the bionic cilium heart sound sensor, the electrocardiogram acquisition circuit and the signal fusion model, and give a prompt. After successful connection, the software will automatically display the collected signal waveforms. The second page is a historical playback function. The user needs to select and import the data that the user wants to view. After clicking start, the selected synchronous heart sound and electrocardiogram signals can be played. The waveform amplitude size can be adjusted. At the same time, the user can add note information to the playback waveform. The third page is a data analysis page. The selected data can be analyzed in the time and frequency domains. The wavelet transform is further used to visualize the time-frequency diagram of the synchronous heart sound and electrocardiogram signals. At the same time, the time-frequency envelope diagram is provided to show the signal characteristics. The selected signal is sent to the built deep learning model for calculation and returns the classification probability as the diagnosis result. The software gives the corresponding health advice according to the diagnosis result. The data analysis and intelligent diagnosis result can be written into an analysis report and exported in the format of PDF.

[0079] In some embodiments, referring to Figure 5 , Figure 5 A flowchart of a MEMS heart sound and electrocardiogram detection method provided by an embodiment of the application; the MEMS heart sound and electrocardiogram detection method is applied to the MEMS heart sound and electrocardiogram detection system described above, and includes the following steps.

[0080] S510, collecting heart sound signals of a target to be detected by using a bionic cilium heart sound sensor and collecting electrocardiogram signals of the target to be detected by using an electrocardiogram acquisition circuit;

[0081] S520, performing feature extraction and feature enhancement on the heart sound signals and the electrocardiogram signals based on a Resnet18 feature extractor and an encoder to generate heart sound signal features and electrocardiogram signal features;

[0082] S530, performing linear transformation and feature depth fusion on the heart sound signal features and the electrocardiogram signal features by using a mutual cross-attention mechanism to output a signal classification result.

[0083] In some embodiments, after the bionic cilium heart sound sensor is used to collect the heart sound signals of the target to be detected and the electrocardiogram acquisition circuit is used to collect the electrocardiogram signals of the target to be detected, the MEMS heart sound and electrocardiogram detection method further includes the following steps.

[0084] Down-sampling the heart sound signals and the electrocardiogram signals, filtering the heart sound signals by using a Butterworth band-pass filter, and filtering the electrocardiogram signals by using a Butterworth band-pass filter and a notch filter;

[0085] The filtered heart sound signal and the electrocardiogram signal are segmented to obtain multiple signal segments with the same length.

[0086] The signal segments are normalized and converted into time-frequency diagrams through continuous wavelet transform.

[0087] Exemplarily, the original data is first subjected to down-sampling, the sampling rate of the heart sound signal is reduced to 2000 Hz, and the sampling rate of the electrocardiogram signal is reduced to 500 Hz. Then, the heart sound data is further processed using a 3rd order Butterworth band-pass filter of 20-600 Hz, and the electrocardiogram data is further processed using a 3rd order Butterworth band-pass filter of 0.5-100 Hz and a notch filter of 50 Hz. Subsequently, the heart sound and electrocardiogram signals are segmented, all samples are segmented into 2s segments for subsequent processing and model learning, and normalized. Finally, the synchronous heart sound and electrocardiogram signals are converted into time-frequency diagrams through continuous wavelet transform and input into the corresponding model for training.

[0088] It should be noted that the MEMS heart sound and electrocardiogram detection method provided by the embodiments of the present application is based on the same application concept as the MEMS heart sound and electrocardiogram detection system provided by the embodiments of the present application. Therefore, the specific implementation of this embodiment can be referred to the implementation of the aforementioned MEMS heart sound and electrocardiogram detection system, and the repeated parts will not be described again.

[0089] In some embodiments, the MEMS heart sound and electrocardiogram detection device provided by the embodiments of the present application comprises the MEMS heart sound and electrocardiogram detection system described above.

[0090] The present application also provides a computer readable medium having a computer program stored thereon, which, when executed by a processor, implements the MEMS heart sound and electrocardiogram detection method described above. The computer readable medium can be included in the device / apparatus / system described in the above embodiments; or it can exist separately and not be assembled into the device / apparatus / system. The computer readable medium carries one or more programs, which, when executed, implement the method according to the embodiments of the present application.

[0091] According to embodiments of the present application, the computer readable medium can be a computer readable signal medium or a computer readable storage medium or any combination thereof. The computer readable storage medium can be, for example but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the computer readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present application, the computer readable storage medium can be any tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. In the present application, the computer readable signal medium can include a computer readable program code that can be propagated as a data signal in a baseband or as a carrier wave in a propagated signal, where the computer readable program code can be propagated in a variety of forms, including but not limited to, an electromagnetic signal, an optical signal, or any suitable combination of the foregoing. The computer readable signal medium can also be any computer readable medium that can be transitory, propagating, or transmitting a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer readable medium can be transmitted as a carrier wave or other propagated signal, where the computer readable program code can be propagated in a variety of forms, including but not limited to, an electromagnetic signal, an optical signal, or any suitable combination of the foregoing.

[0092] Those skilled in the art will appreciate that features recited in the various embodiments and / or claims of the present application can be combined and / or interchanged, even if this is not explicitly stated in the present application. In particular, the features of the various embodiments and / or claims of the present application can be combined and / or interchanged, without departing from the spirit and teachings of the present application. All combinations and / or interchanges of the features of the various embodiments and / or claims of the present application are expressly encompassed by the present application. Accordingly, the scope of the present application should not be limited to the above-described embodiments, but should be determined by the appended claims and their equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the present application.

Claims

1. A MEMS heart sound electrocardio detection system, characterized in that, The application relates to a bionic cilia heart sound sensor, a heart electric signal acquisition circuit and a signal fusion model. The bionic cilia heart sound sensor comprises a hollow bullet type bionic cilia and a cantilever beam microstructure; the hollow bullet type bionic cilia is mechanically coupled with the cantilever beam microstructure; the cantilever beam microstructure comprises a Wheatstone bridge formed by a plurality of pressure-sensitive resistors and is used for collecting heart sound signals. The heart electric signal acquisition circuit is packaged with the bionic cilia heart sound sensor through acoustic impedance matching and comprises a heart electric signal acquisition electrode sheet, a power stabilizing module, an instrument amplification module, a band-pass filter module and a trap wave filter and is used for collecting and processing heart electric signals. The signal fusion model comprises a Resnet18 feature extractor, an encoder and a mutual cross attention mechanism and is used for extracting features, linearly transforming and deeply fusing features of collected heart electric signals and heart sound signals and outputting signal classification results. The application further relates to a heart sound simulation circuit connected with the bionic cilia heart sound sensor.

2. The MEMS PCG-ECG detection system of claim 1, wherein, The power stabilizing module is used for providing stable voltage for the heart sound simulation circuit. The amplification module is used for enhancing input heart electric signals. The band-pass filter module is used for denoising the enhanced heart electric signals and extracting heart electric signals in a preset frequency band. The bionic cilia heart sound sensor is connected with the heart sound simulation circuit through wire bonding and is fixed by using an internal card slot of a shell wrapping the bionic cilia heart sound sensor, the heart electric signal acquisition circuit and the signal fusion model.

3. The MEMS PCG-ECG detection system of claim 2, wherein, The application further relates to a synchronous collector.

4. The MEMS PCG-ECG detection system of claim 1, wherein, The analog-digital conversion module is used for converting analog signals into digital signals. The communication module is used for realizing communication connection with an upper computer through a USB interface and transmitting collected synchronous heart sound and heart electric signals to the signal fusion model in the upper computer. The Resnet18 feature extractor comprises a 7*7 convolution layer, a 3*3 maximum pooling layer and four continuous basic blocks, each of which comprises two residual blocks; the encoder comprises position encoding, a multi-head attention mechanism, a feedforward network and a summation normalization structure; and the mutual cross attention mechanism comprises two one-way cross attention mechanisms.

5. The MEMS heart sound and electrocardiogram detection system of claim 1, wherein, The mutual cross attention mechanism is used for linearly transforming an input pair of heart sound signal features and heart electric signal features to generate corresponding query matrices, key matrices and value matrices, performing bidirectional dot product attention operation through the mutual cross attention mechanism and outputting signal classification results.

6. The MEMS heart sound and electrocardiogram detection system of claim 5, wherein, The application further relates to an analysis module.

7. The MEMS heart sound and electrocardio detection system of claim 6, wherein, The analysis module is used for performing time-frequency domain analysis on the signal classification results and performing visual processing on synchronous heart sound and heart electric signals in the signal classification results through wavelet transformation to obtain a time-frequency graph and / or an envelope graph. The application relates to a bionic cilia heart sound sensor, a heart electric signal acquisition circuit and a signal fusion model. ​ ​ 8. A MEMS heart sound and electrocardio detection method applied to the MEMS heart sound and electrocardio detection system of claims 1 to 7, characterized in that, ​ ​ The heart sound signal and the electrocardio signal are subjected to feature extraction and feature enhancement based on a Resnet18 feature extractor and an encoder to generate heart sound signal features and electrocardio signal features; The heart sound signal features and the electrocardio signal features are subjected to linear transformation and feature deep fusion by using a mutual cross-attention mechanism to output a signal classification result.

9. The MEMS heart sound and electrocardio detection method of claim 8, wherein, After collecting the heart sound signal of a target to be detected by using a bionic cilia heart sound sensor and collecting the electrocardio signal of the target to be detected by using an electrocardio collection circuit, the method further comprises: The heart sound signal and the electrocardio signal are subjected to down-sampling, and the heart sound signal is subjected to filtering processing by using a Butterworth band-pass filter, and the electrocardio signal is subjected to filtering processing by using a Butterworth band-pass filter and a notch filter; The filtered heart sound signal and the filtered electrocardio signal are subjected to segmentation processing to obtain multiple signal segments with the same length; The signal segments are subjected to normalization processing, and the signal segments are converted into time-frequency graphs by using continuous wavelet transform. 10.A MEMS heart sound and electrocardio detection device comprising the MEMS heart sound and electrocardio detection system according to any one of claims 1 to 7.

Images