WO2023134253A1 - 一种对心音信号的预测方法及智能心音和心电测量仪 - Google Patents

一种对心音信号的预测方法及智能心音和心电测量仪 Download PDF

Info

Publication number
WO2023134253A1
WO2023134253A1 PCT/CN2022/126866 CN2022126866W WO2023134253A1 WO 2023134253 A1 WO2023134253 A1 WO 2023134253A1 CN 2022126866 W CN2022126866 W CN 2022126866W WO 2023134253 A1 WO2023134253 A1 WO 2023134253A1
Authority
WO
WIPO (PCT)
Prior art keywords
heart sound
heart
signal
sound
unit
Prior art date
Application number
PCT/CN2022/126866
Other languages
English (en)
French (fr)
Inventor
吴先洪
罗育坤
罗祥基
王爱国
卓维华
Original Assignee
吴先洪
罗育坤
罗祥基
王爱国
卓维华
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from CN202210024087.0A external-priority patent/CN114209335A/zh
Priority claimed from CN202210496677.3A external-priority patent/CN115089206B/zh
Application filed by 吴先洪, 罗育坤, 罗祥基, 王爱国, 卓维华 filed Critical 吴先洪
Publication of WO2023134253A1 publication Critical patent/WO2023134253A1/zh

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/25—Bioelectric electrodes therefor
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B7/00—Instruments for auscultation
    • A61B7/02—Stethoscopes
    • A61B7/04—Electric stethoscopes

Definitions

  • the invention relates to an algorithm of a medical device, in particular to a method for predicting heart sound signals and a heart auscultation device using the method.
  • cardiovascular disease has become the number one killer of human beings, and it tends to be younger.
  • cardiovascular diseases especially structural heart diseases (various congenital heart diseases and acquired valvular heart diseases) cannot be detected by traditional electrocardiogram machines, and can only be listened to by a professional doctor through a stethoscope.
  • the diagnosis can be made after preliminary screening and inspection by related equipment (such as cardiac color Doppler ultrasound, etc.).
  • the first-generation smart stethoscope is based on the traditional stethoscope, using high-sensitivity sensors to collect the sound from body organs, amplifying and filtering, and transmitting effective sound to wireless earphones or other receiving devices through Bluetooth.
  • the second-generation heart sound and ECG integrated measuring instrument is developed on the basis of the first-generation smart stethoscope. It measures heart sound and ECG synchronously and displays them intuitively to users, allowing doctors to obtain heart beating conditions more conveniently.
  • arrhythmia such as atrial fibrillation, ventricular premature beats, atrial premature beats, paroxysmal supraventricular tachycardia, severe bradycardia, etc.
  • arrhythmias are mostly paroxysmal attacks, and timely electrocardiography in the hospital is also helpful. It may not be possible to catch the abnormal heart rhythm in time, and the physical examination at home and early self-discovery are particularly important.
  • ECG and heart sound measuring instruments The general promotion of household ECG and heart sound measuring instruments has made the use environment of ECG and heart sound measuring instruments more complicated. Multiple noise sources generally exist, and various phenomena such as interference and diffraction may occur between sound waves. The occurrence of interference may result in the failure to collect sound signals in certain positions, or distort the heart sound signals collected by the heart sound sensor. Therefore, it is necessary to intelligently adjust the position of the sound sensor to be suitable for use in complex home environments.
  • the present invention has designed a kind of prediction method to heart sound signal and intelligent heart sound and electrocardiogram measuring instrument, and the technical problem that it solves is (1) the existing algorithm can't fully tap the time and space high-level semantic information contained in the original sound signal of auscultation , nor can a sufficient association be established between auscultatory sound signals and cardiac disease. (2) The use environment of the heart sound measuring instrument in the existing technology center has become more complicated. Multiple noise sources generally exist, and various phenomena such as interference and diffraction may occur between sound waves. Due to interference, it may cause In some positions, sound signals cannot be collected, or the heart sound signals collected by the heart sound sensor will be distorted.
  • the present invention adopts the following scheme:
  • a method for predicting heart sound signals comprising the following steps: step 1, obtaining from an intelligent stethoscope the original heart sound signal of an individual's auscultation for noise reduction, and forming a data segment; step 2, mining the data segment based on multi-resolution deep learning The high-level temporal semantic information and translation invariance contained in the signal, and finally output the prediction result.
  • the denoised data y (y 1 ,y 2 ,...,y n ); considering that the duration of abnormal sound caused by heart disease has multi-scale characteristics, in order to fully mine the information in y, from y Extract M data fragments of different sizes ⁇ s (1) ,s (2) ,...,s (M) ⁇ , where the kth (1 ⁇ k ⁇ M) data fragment is Where
  • the above process is repeated to obtain the output ⁇ o (1) ,w (1 ) ⁇ , ⁇ o ( 2) , w (2 ) ⁇ ,..., ⁇ o (M) ,w (M) ⁇ , and finally use weighted voting to get the final prediction result
  • the j in represents the index of the most likely heart disease voted by M predictors in L
  • c j represents the predicted heart disease
  • L is equal to the number of types of heart disease
  • the heart disease set represents the set of different types/categories of heart disease, corresponding to L, and corresponding to the prediction result c j .
  • the final prediction result is broadcast through the Bluetooth device.
  • the Bluetooth device is a Bluetooth headset or a Bluetooth speaker.
  • a heart auscultation device using the above analysis and prediction method uses the above analysis and prediction method.
  • the silicone sleeve at the bottom of the stethoscope body is a closed-loop structure composed of multiple monomers, and each silicone sleeve monomer is connected to the stethoscope body through a telescopic structure, and each telescopic structure can make the silicone sleeve monomer move so that the entire silicone sleeve It forms a structure with different heights, so as to ensure that the entire silicone sleeve fits closely on the human body to form a closed sound collection environment for the vibrating diaphragm to collect heart sounds.
  • a bluetooth speaker or earphone wherein the final output prediction result obtained by the above analysis and prediction method is output in voice through the bluetooth speaker or the bluetooth earphone.
  • An automatically adjusted intelligent heart sound and ECG measuring instrument comprising a heart sound unit and an ECG unit, the ECG unit including an ECG electrode piece (9), and the ECG electrode piece (9) and a main body (6) form a sound collecting cavity ( 7), the main microphone (5) in the heart sound unit is located in the sound collecting chamber (7), and it is characterized in that: it also includes a driving device and a transmission structure, and the driving device enables the main microphone (5) to be able to use the transmission structure through the transmission structure.
  • the different height positions in the sound cavity (7) are moved to avoid the interaction of multiple external sound sources causing the main microphone (5) to fail to collect the heart sound signal or the heart sound signal is too weak.
  • the driving device is a micro motor (1)
  • the transmission mechanism includes a rotating threaded rod (2), an axially moving bearing sleeve (3) and a connecting piece (4), and one end of the rotating threaded rod (2) Connected with the micro motor (1), the rotating threaded rod (2) is sleeved with the axially moving bearing sleeve (3), and the rotating threaded rod (2) and the axially moving bearing sleeve (3) form a ball screw structure
  • the inner ring of the axially moving bearing sleeve (3) is threadedly connected with the rotating threaded rod (2), the outer ring of the axially moving bearing sleeve (3) is connected with the main microphone (5) through the connecting piece (4), and the axially moving bearing
  • the inner ring of the sleeve (3) rotates, the outer ring of the axially moving bearing sleeve (3) remains relatively stationary, and the main microphone (5) moves axially along the rotating threaded rod (2).
  • the driving device is a micro motor (1)
  • the transmission mechanism includes a rotating threaded rod (2), a nut and a connector (4), and one end of the rotating threaded rod (2) is connected to the micro motor (1)
  • the rotating threaded rod (2) is socketed with a nut
  • the rotating threaded rod (2) and the nut form a ball screw structure
  • the nut is connected with the main microphone (5) through a connecting piece (4)
  • the axially moving bearing sleeve When the inner ring of 3) rotates, the main microphone (5) rotates and moves axially along the rotating threaded rod (2).
  • a control method for an automatically adjusted intelligent heart sound and electrocardiogram measuring instrument comprising the following steps: step 1, sticking the electrocardiogram electrode piece (9) on the skin; step 2, starting the heart sound unit and the electrocardiogram unit to collect heart sound signals respectively and the ECG signal; step 3, judging whether the strength of the heart sound signal collected by the heart sound unit meets the set standard, and when the standard is met, the heart sound signal and the ECG signal are collected normally.
  • step 4 judging whether the strength of the heart sound signal collected by the heart sound unit meets the set standard, and when the heart sound signal is too weak, start the micro motor so that the main microphone moves in the vertical direction of the sound collecting chamber (7), and in different Stop at the height of the heart sound signal to collect whether the strength of the heart sound signal at the position meets the set standard. After comparison, the main microphone is finally moved to the best position for heart sound signal collection.
  • the audio amplification unit amplifies the heart sound signal collected by the main microphone and the environmental noise signal collected by the secondary microphone, and the filtering unit removes the amplified environmental noise signal from the amplified heart sound signal (containing noise signal) to obtain
  • the analog-to-digital conversion unit converts the heart sound analog signal into a digital signal and outputs it.
  • the noise source changes, and re-judging whether the strength of the heart sound signal collected by the heart sound unit meets the set standard to determine whether to repeat step 4.
  • the method proposed by the present invention comprehensively utilizes the convolutional neural network and the long-short-term memory network, which can fully excavate the time and space high-level semantic information contained in the original sound signal of auscultation, and better establish the relationship between the sound signal of auscultation and heart disease.
  • the association between; the confidence level is introduced to quantify the discriminative ability of different members in the heart disease prediction model to predict different heart diseases, so that the designed heart disease prediction model has strong adaptability and robustness.
  • the present invention obtains by arranging the results ⁇ o (1) ,o (2) ,...,o (M) ⁇ of M predictors
  • the method can improve the accuracy and stability of the prediction results and reduce the variance of the prediction results.
  • the present invention cooperates with a plurality of temperature sensors and the mobile terminal, so that when the heart sound is measured in cold weather, the user's discomfort will not be caused by the temperature of the stethoscope being too low, and the accuracy and comfort of the measurement will not be affected.
  • the present invention ensures that when the stethoscope is not pressed by hand, the silicone ring of the stethoscope can be closely attached to human skin to form an airtight sound collection environment for the vibrating diaphragm to collect heart sounds to ensure anti-interference heart sound data collection.
  • the present invention enables the main microphone to be adjusted at the height position, so that in a complex sound source environment, interference between different sound sources is avoided and qualified heart sound signals cannot be collected, thereby ensuring the collection quality of the heart sound signals.
  • the present invention enables the main microphone to stop at the nearest height position to collect heart sound signals through self-detection and comparison, and adjust the best position dynamically.
  • Fig. 1 the present invention is to the input and output signal schematic diagram in the prediction method of heart sound signal
  • Fig. 2 the outward appearance schematic diagram of cardiac auscultation device of the present invention
  • Fig. 3 schematic diagram of the circuit connection block diagram of the cardiac auscultation device of the present invention
  • Figure 4 The first schematic diagram of the telescopic structure of the cardiac auscultation device of the present invention.
  • Figure 5 The second schematic diagram of the telescopic structure of the cardiac auscultation device of the present invention.
  • Figure 6 The third schematic diagram of the telescopic structure of the cardiac auscultation device of the present invention.
  • Fig. 7 Schematic diagram I of telescopic structure of the cardiac auscultation device of the present invention.
  • Fig. 8 Schematic diagram II of the telescopic structure of the cardiac auscultation device of the present invention.
  • Fig. 9 A schematic diagram of adjusting the main microphone at the lowest point of the intelligent heart sound and ECG measuring instrument automatically adjusted in the present invention
  • Figure 10 A schematic diagram of adjusting the main microphone at the highest point of the intelligent heart sound and ECG measuring instrument automatically adjusted in the present invention
  • Fig. 11 A schematic block diagram of sound processing in the present invention.
  • 1 miniature motor
  • 2 rotating threaded rod
  • 3 axially moving bearing sleeve
  • 4 connecting
  • 5 main microphone
  • 6 main body
  • 7 acoustic cavity
  • 8 sound insulation layer
  • a kind of prediction method to heart sound signal comprises the following steps: step 1, obtains from intelligent stethoscope about the heart original sound signal of auscultation of individuality and carries out noise reduction, and forms data segment; Step 2, based on multi-resolution High rate deep learning mines the high-level temporal semantic information and translation invariance contained in the data segment signal, and finally outputs the prediction result.
  • the final data y (y 1 ,y 2 ,...,y n ); considering that the duration of abnormal sounds caused by heart disease has multi-scale characteristics, in order to fully mine the information in y, M pieces of data are extracted from y Data fragments of different sizes ⁇ s (1) ,s (2) ,...,s (M) ⁇ , where the kth (1 ⁇ k ⁇ M) data fragment is Where
  • the j in represents the index of the most likely heart disease voted by M predictors in L
  • c j represents the predicted heart disease
  • L is equal to the number of types of heart disease
  • Step 4 the final prediction result is broadcasted through the bluetooth device.
  • the Bluetooth device is a Bluetooth headset or a Bluetooth speaker.
  • the heart auscultation device of the present invention includes a stethoscope body 1, a mobile terminal and a remote computer.
  • the stethoscope body 1 sends the heart sound to the mobile terminal through the Bluetooth communication module after the heart sound is collected, and the mobile terminal sends the heart sound signal to the remote computer through the Internet.
  • Terminal computer for doctors to listen to and use.
  • the first temperature sensor 12 is arranged on the stethoscope body 1, and it monitors the ambient temperature T1;
  • the bottom of the stethoscope body 1 is provided with a silicone sleeve 11, and a heating wire 14 is arranged in the silicone sleeve 11, and the heating wire 14
  • the mobile terminal 2 monitors that the ambient temperature T1 collected by the first temperature sensor 12 is lower than the normal body temperature T2 of the human body, and starts the heating wire 14 to heat, so that the silicone sleeve 11 approaches or reaches the normal body temperature of the human body, ensuring that the silicone When the cover 11 contacts the human body, the temperature of the silicone cover 11 reaches the comfortable temperature sensed by the human body.
  • the second temperature sensor 13 monitors the temperature T3 of the silicone sleeve 11. When T3 is higher than the preset value T4, the mobile terminal turns off the heating function of the heating wire 14 to prevent the temperature of the silicone sleeve 11 from being too high. Discomfort.
  • the stethoscope body 1 is provided with an installation chamber, and the bottom of the installation chamber is provided with a vibrating diaphragm.
  • the installation chamber there is a pickup based on Bluetooth communication.
  • the bluetooth module communicates wirelessly with the mobile terminal 2, and the high-fidelity audio sensor, lithium battery and USB interface are all connected to the bluetooth mainboard.
  • the silicone sleeve 11 at the bottom of the stethoscope body 1 is a closed-loop structure composed of multiple monomers, and each silicone sleeve monomer is connected to the stethoscope body 1 through a telescopic structure 15, and each telescopic structure 15 can The silicone sleeve is moved so that the entire silicone sleeve 11 forms a structure with different heights, thereby ensuring that the entire silicone sleeve 11 is closely attached to the human body to form a closed sound collection environment for the vibrating diaphragm to collect heart sounds.
  • All the telescopic structures 15 in FIG. 5 are in the maximum stretched state, the leftmost silicone sleeve monomer in FIG. 5 is in the maximum stretched state, and the other silicone sleeve monomers are in the semi-stretched state.
  • Fig. 6 the leftmost silicone sleeve monomer and the middle silicone sleeve monomer are in the maximum stretched state, and the other silicone sleeve monomers are in a semi-stretch state.
  • the above telescopic position can be freely adjusted according to the measurement position, so that the entire silicone ring can fit on the skin.
  • the telescopic structure 15 includes an upper horizontal bracket 151 and a lower horizontal bracket 153, the upper horizontal bracket 151 is connected with the stethoscope body 1, the lower horizontal bracket 153 is connected with the silicone sleeve monomer, and the upper horizontal bracket 151 and the lower horizontal bracket 153 A plurality of moving connecting pieces 152 are arranged between them, the moving connecting pieces 152 are arranged obliquely and one end is fixedly connected with the upper horizontal support 151, and the other end of the moving connecting pieces 152 is movably connected with the lower horizontal support 153, or the moving connecting pieces 152 are arranged obliquely and One end is movably connected with the upper horizontal bracket 151, and the other end of the mobile connecting piece 152 is fixedly connected with the lower horizontal bracket 153; when the distance between the upper horizontal bracket 151 and the lower horizontal bracket 153 changes, the silicone sleeve monomer moves up and down, and the mobile connection The angle of inclination of the sheet 152 also changes accordingly.
  • the structure of movable connection comprises that the upper horizontal support 151 or the lower horizontal support 153 has a chute, the opening of the chute is narrow, and the inside of the chute is wide. thick, so that the moving connecting piece 152 cannot break away from the chute but can move along the extending direction of the chute.
  • the telescopic structure 15 is multi-layered, the uppermost telescopic structure 15 is connected with the stethoscope body 1, and the lowermost telescopic structure 15 is connected with the silicone sleeve monomer;
  • the cover 11 is provided with a second sound-insulating baffle 17, the first sound-insulating baffle 16 and the second sound-insulating baffle 17 are misplaced and both can move relative to each other with the movement of the silicone sleeve 11.
  • the first sound-insulating The baffle 16 and the second sound-insulating baffle 17 always form a shielding structure to prevent the vibrating diaphragm at the bottom of the installation cavity from being affected by external noise.
  • the control principle of cardiac auscultation device of the present invention is as follows:
  • Step 1 Turn on the power of the stethoscope body
  • Step 2 The mobile terminal connects to the stethoscope body via Bluetooth, and starts the first temperature sensor.
  • the temperature T1 data collected by the first temperature sensor is sent to the mobile terminal.
  • the mobile terminal judges that T1 is higher or lower than the normal body temperature T2 of the human body; if lower, Then start the heating wire 14 in the silicone sleeve 11 to generate heat, if it is higher than that, then directly enter step 3;
  • Step 3 Place the stethoscope body on the skin of the aortic valve area or mitral valve area, and ensure that the entire silicone sleeve 11 fits closely on the human body through the telescopic structure to form a closed sound-collecting environment for the vibrating diaphragm to collect heart sounds;
  • Step 4 The stethoscope body 1 collects the heart sound and sends it to the mobile terminal through the Bluetooth communication module, and the mobile terminal sends the heart sound signal to the remote computer through the Internet for the doctor to listen to and use.
  • an automatically adjusted intelligent heart sound and ECG measuring instrument includes a heart sound unit and an ECG unit, and the ECG unit includes an ECG electrode piece 9, and the ECG electrode piece 9 forms a sound collection with the main body 6 Cavity 7, the main microphone 5 in the heart sound unit is located in the sound collecting cavity 7, and is characterized in that: it also includes a driving device and a transmission structure, and the driving device enables the main microphone 5 to be positioned at different heights in the sound collecting cavity 7 through the transmission structure Move to avoid the mutual influence of multiple external sound sources causing the main microphone 5 to fail to collect the heart sound signal or the heart sound signal is too weak.
  • the first structure can be: the driving device is a micro motor 1, the transmission mechanism includes a rotating threaded rod 2, an axially moving bearing sleeve 3 and a connecting piece 4, one end of the rotating threaded rod 2 is connected with the micro motor 1, and the rotating threaded rod 2
  • the axially moving bearing sleeve 3 is sleeved, the rotating threaded rod 2 and the axially moving bearing sleeve 3 form a ball screw structure, the inner ring of the axially moving bearing sleeve 3 is threadedly connected with the rotating threaded rod 2, and the axially moving bearing sleeve 3
  • the outer ring is connected with the main microphone 5 through the connecting piece 4, when the inner ring of the axially moving bearing sleeve 3 rotates, the outer ring of the axially moving bearing sleeve 3 remains relatively stationary, and the main microphone 5 moves axially along the rotating threaded rod 2 .
  • the second structure can be: the driving device is a micro motor 1, the transmission mechanism includes a rotating threaded rod 2, a nut and a connector 4, one end of the rotating threaded rod 2 is connected to the micro motor 1, and the rotating threaded rod 2 is sleeved with a nut, and the rotating The threaded rod 2 and the nut form a ball screw structure, and the nut is connected to the main microphone 5 through the connecting piece 4.
  • the driving device is a micro motor 1
  • the transmission mechanism includes a rotating threaded rod 2, a nut and a connector 4
  • one end of the rotating threaded rod 2 is connected to the micro motor 1
  • the rotating threaded rod 2 is sleeved with a nut
  • the rotating The threaded rod 2 and the nut form a ball screw structure
  • the nut is connected to the main microphone 5 through the connecting piece 4.
  • the second structure is simpler than the first structure, but the first structure can prevent the main microphone 5 from rotating and moving, thus ensuring that it is always on the axis of the threaded rod, so that the position of sound collection can be known.
  • the main microphone 5 is located at the lowermost end, almost touching the electrocardiographic electrode piece 9 . As shown in Fig. 2, the main microphone 5 is located at the uppermost end. Multiple stop points can be set between the uppermost end and the lowermost end as required to collect the best heart sound signal.
  • the top of the main body 6 is provided with a through hole for the rotating threaded rod 2 to pass through, and a sound insulation layer 8 is provided between the micro motor 1 and the main body 6 to prevent the noise generated by the micro motor from interfering with the acquisition of the main microphone.
  • the audio amplifier unit amplifies the heart sound signal collected by the main microphone and the environmental noise signal collected by the auxiliary microphone
  • the filter unit amplifies the amplified heart sound signal.
  • the amplified environmental noise signal is removed to obtain a more realistic heart sound signal
  • the analog-to-digital conversion unit converts the heart sound analog signal into a digital signal and outputs it.
  • the control method of the intelligent heart sound of the present invention and the electrocardiogram measuring instrument of automatic regulation comprises the following steps:
  • Step 1 Stick the ECG electrode sheet 9 on the skin.
  • Step 2 start the heart sound unit and the ECG unit to collect the heart sound signal and ECG signal respectively; Step 3, judge whether the strength of the heart sound signal collected by the heart sound unit meets the set standard, and when the standard is met, collect the heart sound signal and ECG normally Signal.
  • Step 4 Determine whether the strength of the heart sound signal collected by the heart sound unit meets the set standard.
  • the heart sound signal is too weak, start the micro motor so that the main microphone moves vertically in the sound collection chamber 7 and stops at different heights. Whether the strength of the heart sound signal collected at this position meets the set standard, through comparison, the main microphone is finally moved to the best position for heart sound signal collection.
  • Step 5 The audio amplification unit amplifies the heart sound signal collected by the main microphone and the environmental noise signal collected by the auxiliary microphone, and the filter unit removes the amplified environmental noise signal from the amplified heart sound signal containing noise signal to obtain a more realistic heart sound signal, the analog-to-digital conversion unit converts the heart sound analog signal into a digital signal and outputs it.

Abstract

一种对心音信号的预测方法及使用其的心脏听诊装置，该预测方法包括以下步骤：步骤1、从智能听诊器获得关于个体的听诊的原始声音信号进行降噪，并且形成数据片段；步骤2、基于多分辨率深度学习挖掘数据片段信号中蕴含的高层时序语义信息和平移不变性，并最终输出预测结果。该心脏听诊装置包括听诊器本体(1),听诊器本体(1)底部的硅胶套(11)为多个单体组成的闭环结构，每个硅胶套单体通过一个伸缩结构(15)与听诊器本体(1)连接，每个伸缩结构(15)能够使得硅胶套单体移动从而使得整个硅胶套(11)形成高低不同的结构，从而确保整个硅胶套(11)紧密贴合在人体上形成密闭的采音环境供振动膜片采集心音。一种自动调节的智能心音和心电测量仪及其控制方法，包括心音单元和心电单元，心电单元包括心电电极片（9），心电电极片（9）与主体（6）形成采音腔（7），心音单元中的主麦克风（5）位于采音腔（7）中，还包括一驱动装置和传动结构，该驱动装置通过传动结构使得主麦克风（5）能够在采音腔（7）中的不同高度位置进行移动，避免外部多声源相互影响造成主麦克风（5）采集不到心音信号或心音信号过弱。

Description

一种对心音信号的预测方法及智能心音和心电测量仪 技术领域
本发明涉及一种医疗器械的算法，尤其是涉及一种对心音信号的预测方法及使用其的心脏听诊装置。
背景技术
最新数据显示，心血管疾病已经成为人类的第一杀手，并且趋于年轻化。在基层，不少心血管疾病特别是结构性心脏病(各种的先天性心脏病和后天的瓣膜性心脏病)大多无法应用传统的心电图机检测出来，只能通过听诊器先由专业的医生听到异常杂音后初筛再经过相关设备(如心脏彩超等)检查后得以确诊。
第一代智能听诊器是在传统听诊器的基础上，采用高灵敏度感应器来收集身体器官发出的声音，通过放大、过滤，将有效的声音通过蓝牙传输给无线耳机或其他接收设备。
第二代心音心电一体测量仪则是在第一代智能听诊器的基础上研发的，将心音和心电同步测量，直观展示给使用者，让医生更为方便的获得心脏的跳动情况。
心律失常的患者，如房颤、室性早搏、房性早搏，阵发性室上性心动过速，严重的心动过缓等其心律不齐多为阵发性发作，及时在医院做心电图也不一定能及时捕捉到异常心律，在家的体检和自我及早发现显得特别的重要。
家用的心电心音测量仪的普遍推广，使得心电心音测量仪使用环境变得也更加复杂，多个噪音声源普通存在，声波之间可能就会发生干涉,衍射等各种现象,由于会发生干涉就有可能造成在某些位置无法到采集声音信号，或会使得心音传感器采集的心音信号失真，因此需要对声音传感器的位置进行智能调整以适用于复杂家用环境使用。
发明内容
本发明设计了一种对心音信号的预测方法及智能心音和心电测量仪，其解决的技术问题是(1)现有算法无法充分挖掘听诊的原始声音信号所蕴含的时间和空间高层语义信息，也不能建立听诊声音信号与心脏疾病之间的充分关联。(2)现有技术中心电心音测量仪使用环境变得也更加复杂，多个噪音声源普通存在，声波之间可能就会发生干涉,衍射等各种现象,由于会发生干涉就有可能造成在某些位置无法到采集声音信号，或会使得心音传感器采集的心音信号失真。
为了解决上述存在的技术问题，本发明采用了以下方案：
一种对心音信号的预测方法，包括以下步骤：步骤1、从智能听诊器获得关于个体的听诊的心脏原始声音信号进行降噪，并且形成数据片段；步骤2、基于多分辨率深度学习挖掘数据片段信号中蕴含的高层时序语义信息和平移不变性，并最终输出预测结果。
优选地，步骤1中从智能听诊器获得关于个体的听诊的心脏原始声音信号x＝(x 1,x 2,…,x n)后，首先利用低通滤波器过滤点x中的高频信号，得到去噪后的数据y＝(y 1,y 2,…,y n)；考虑到心脏疾病所引起的异常声音持续的时间具有多尺度的特点，为充分挖掘y中的信息，从y中提取M个大小不同的数据片段{s (1),s (2),…,s (M)}，其中第k(1≤k≤M)个数据片段是
Figure PCTCN2022126866-appb-000001
其中|s (k)|表示s (k)中的声音数据的样本点个数。
优选地，步骤2中对于第k个数据片段s (k)，首先利用具有一维卷积核的卷积神经网络CNN对s (k)进行处理，得到z (k)＝CNN(s (k))；然后采用具有若干个隐含层的长短时记忆网络LSTM k对进行处理，得到[o (k),w (k)]＝LSTM k(z (k))，并且
Figure PCTCN2022126866-appb-000002
表示LSTM k预测z (k)属于第i类心脏病的概率，
Figure PCTCN2022126866-appb-000003
Figure PCTCN2022126866-appb-000004
表示LSTM k对于做出预测
Figure PCTCN2022126866-appb-000005
的置信度，将[o (k),w (k)]作为s (k)的预测结果进行输出；[o (k),w (k)]是LSTM k(z (k))的输出，其中o (k)是一个L维的向量
Figure PCTCN2022126866-appb-000006
它的第j个元素
Figure PCTCN2022126866-appb-000007
表示LSTM k(z (k))预测z (k)的标签是c j的概率；w (k)也是一个L维的向量
Figure PCTCN2022126866-appb-000008
它的第j个元素
Figure PCTCN2022126866-appb-000009
量化了 LSTM k(z (k))将z (k)的标签预测为c j的置信度。
优选地，重复上述过程，得到关于{s (1),s (2),…,s (M)}的输出{o (1),w (1)},{o (2),w (2)},…,{o (M),w (M)}，最后采用加权投票的方式，得到最终的预测结果
Figure PCTCN2022126866-appb-000010
Figure PCTCN2022126866-appb-000011
中的j表示M个预测器投票得到的最有可能的心脏病在L中的索引，c j表示预测的心脏病，L等于心脏病的种类数，
Figure PCTCN2022126866-appb-000012
的每个分量对应一种心脏病。
优选地，待识别的心脏病集合是C＝{c 1,c 2,…,c L}，L表示心脏病中的种类数。其中，心脏病集合表示不同类型/种类的心脏病的集合，与L对应，并与预测结果c j相对应。
优选地，步骤4、最终的预测结果通过蓝牙设备进行播报。
优选地，所述蓝牙设备为蓝牙耳机或蓝牙音箱。
一种心脏听诊装置，使用上述分析预测方法。
优选地，听诊器本体底部的硅胶套为多个单体组成的闭环结构，每个硅胶套单体通过一个伸缩结构与听诊器本体连接，每个伸缩结构能够使得硅胶套单体移动从而使得整个硅胶套形成高低不同的结构，从而确保整个硅胶套紧密贴合在人体上形成密闭的采音环境供振动膜片采集心音。
一种蓝牙音箱或耳机，上述分析预测方法获得的最终输出预测结果通过所述蓝牙音箱或所述蓝牙耳机以语音的方式进行输出。
一种自动调节的智能心音和心电测量仪，包括心音单元和心电单元，心电单元包括心电电极片(9)，心电电极片(9)与主体(6)形成采音腔(7)，心音单元中的主麦克风(5)位于采音腔(7)中，其特征在于：还包括一驱动装置和传动结构，所述驱动装置通过传动结构使得主麦克风(5)能够在采音腔(7)中的不同高度位置进行移动，避免外部多声源相互影响造成主麦克风(5)采集不到心音信号或心音信号过弱。
优选地，所述驱动装置为微型电机(1)，所述传动机构包括旋转螺纹杆(2)、轴向移动轴承套(3)以及连接件(4)，旋转螺纹杆(2) 一个端部与微型电机(1)连接，旋转螺纹杆(2)套接轴向移动轴承套(3)，所述旋转螺纹杆(2)与所述轴向移动轴承套(3)形成滚珠丝杆结构，轴向移动轴承套(3)的内圈与旋转螺纹杆(2)螺纹连接，轴向移动轴承套(3)的外圈通过连接件(4)与主麦克风(5)连接，轴向移动轴承套(3)的内圈转动时，轴向移动轴承套(3)的外圈保持相对静止，并且主麦克风(5)沿着旋转螺纹杆(2)轴向移动。
优选地，所述驱动装置为微型电机(1)，所述传动机构包括旋转螺纹杆(2)、螺母以及连接件(4)，旋转螺纹杆(2)一个端部与微型电机(1)连接，旋转螺纹杆(2)套接螺母，所述旋转螺纹杆(2)与所述螺母形成滚珠丝杆结构，螺母通过连接件(4)与主麦克风(5)连接，轴向移动轴承套(3)的内圈转动时，主麦克风(5)旋转并沿着旋转螺纹杆(2)轴向移动。
一种自动调节的智能心音和心电测量仪的控制方法，包括以下步骤：步骤1、将心电电极片(9)贴在皮肤上；步骤2、启动心音单元和心电单元分别采集心音信号和心电信号；步骤3、判断心音单元采集的心音信号的强弱是否符合设定标准，当符合标准时，正常采集采集心音信号和心电信号。
优选地，步骤4、判断心音单元采集的心音信号的强弱是否符合设定标准，当心音信号过弱时，启动微型电机，使得主麦克风在采音腔(7)垂直方向移动，并且在不同的高度停止，分别采集该位置的心音信号的强弱是否符合设定标准，通过比对，最终将主麦克风移动至最佳的位置进行心音信号的采集。
优选地，步骤5、音频放大单元将主麦克风采集的心音信号和副麦克风采集的环境噪音信号进行发大，滤波单元将放大后的心音信号(含噪音信号)除去放大后的环境噪音信号，获得的更加真实的心音信号，模数转换单元将心音模拟信号转换为数字信号并且输出。
优选地，当副扬声器采集的声音信号变化较大时，噪音声源有改变，重新判断心音单元采集的心音信号的强弱是否符合设定标准从而 决定是否重复步骤4。
该对心音信号的预测方法及智能心音和心电测量仪具有以下有益效果：
(1)本发明所提出的方法综合利用了卷积神经网络和长短时记忆网络，能够充分挖掘听诊的原始声音信号所蕴含的时间和空间高层语义信息，更好地建立听诊声音信号与心脏疾病之间的关联；引入置信度，量化了心脏病预测模型中的不同成员预测不同心脏疾病的判别能力，使得所设计的心脏病预测模型具有较强的适应性和鲁棒性。
(2)本发明相对于使用单个预测器得到最终的预测结果，通过整理M个预测器的结果{o (1),o (2),…,o (M)}得到
Figure PCTCN2022126866-appb-000013
的方式，能够提高预测结果的准确率和稳定性，降低预测结果的方差。
(3)本发明通过多个温度传感器与移动终端的配合，使得寒冷天气测量心音时，不会因为听诊器温度过低而导致使用者不适，影响测量的准确性和舒适性。
(4)本发明通过伸缩机构确保在没有人手按压听诊器时，也可以使得听诊器的硅胶圈与人体皮肤紧密贴合形成密闭的采音环境供振动膜片采集心音，确保心音数据采集抗干扰。
(5)本发明通过使得主麦克风可以在高度位置进行调节，使得在复杂声源环境下避免不同声源之间出现干涉造成无法采集到合格的心音信号，确保了心音信号的采集质量。
(6)本发明通过自行检测和比对的方式，使得主麦克风可以停在最近高度位置进行心音信号的采集，并且在动态中调整最佳的位置。
附图说明
图1：本发明对心音信号的预测方法中输入输出信号示意图；
图2：本发明心脏听诊装置的外观示意图；
图3：本发明心脏听诊装置电路连接方框示意图；
图4：本发明心脏听诊装置伸缩结构第一示意图；
图5：本发明心脏听诊装置伸缩结构第二示意图；
图6：本发明心脏听诊装置伸缩结构第三示意图；
图7：本发明心脏听诊装置伸缩结构示意图I；
图8：本发明心脏听诊装置伸缩结构示意图II。
附图标记说明：
1—听诊器本体；11—硅胶套；12—第一温度传感器；13—第二温度传感器；14—发热丝；15—伸缩结构；151—上水平支架；152—移动连接片；1521—固定端；1522—移动端；153—下水平支架；16—第一隔音挡片；17—第二隔音挡片。
图9：本发明中自动调节的智能心音和心电测量仪将主麦克风调整在最低点示意图；
图10：本发明中自动调节的智能心音和心电测量仪将主麦克风调整在最高点示意图；
图11：本发明中声音处理示意方框图。
附图标记说明：
1—微型电机；2—旋转螺纹杆；3—轴向移动轴承套；4—连接件；5—主麦克风；6—主体；7—采音腔；8—隔音层；9—心电电极片。
具体实施方式
实施例1：
下面结合图1至图8，对本发明做进一步说明：
如图1所示，一种对心音信号的预测方法，包括以下步骤：步骤1、从智能听诊器获得关于个体的听诊的心脏原始声音信号进行降噪，并且形成数据片段；步骤2、基于多分辨率深度学习挖掘数据片段信号中蕴含的高层时序语义信息和平移不变性，并最终输出预测结果。
步骤1中从智能听诊器获得关于个体的听诊的心脏原始声音信号x＝(x 1,x 2,…,x n)后，首先利用低通滤波器过滤点x中的高频信号，得到去噪后的数据y＝(y 1,y 2,…,y n)；考虑到心脏疾病所引起的异常声音持续的时间具有多尺度的特点，为充分挖掘y中的信息，从y中提取M个大小不同的数据片段{s (1),s (2),…,s (M)}，其中第k(1≤k≤M)个数据片段是
Figure PCTCN2022126866-appb-000014
其中|s (k)|表示s (k)中的声音数据的样本点个数。
步骤2中对于第k个数据片段s (k)，首先利用具有一维卷积核的卷积神经网络CNN对s (k)进行处理，得到z (k)＝CNN(s (k))；然后采用具有若干个隐含层的长短时记忆网络LSTM k对进行处理，得到[o (k),w (k)]＝LSTM k(z (k))，并且
Figure PCTCN2022126866-appb-000015
表示LSTM k预测z (k)属于第i类心脏病的概率，
Figure PCTCN2022126866-appb-000016
表示LSTM k对于做出预测
Figure PCTCN2022126866-appb-000017
的置信度，将[o (k),w (k)]作为s (k)的预测结果进行输出；[o (k),w (k)]是LSTM k(z (k))的输出，其中o (k)是一个L维的向量
Figure PCTCN2022126866-appb-000018
它的第j个元素
Figure PCTCN2022126866-appb-000019
表示LSTM k(z (k))预测z (k)的标签是c j的概率；w (k)也是一个L维的向量
Figure PCTCN2022126866-appb-000020
它的第j个元素
Figure PCTCN2022126866-appb-000021
量化了LSTM k(z (k))将z (k)的标签预测为c j的置信度。
重复上述过程，得到关于{s (1),s (2),…,s (M)}的输出{o (1),w (1)},{o (2),w (2)},…,{o (M),w (M)}，最后采用加权投票的方式，得到最终的预测结果
Figure PCTCN2022126866-appb-000022
Figure PCTCN2022126866-appb-000023
中的j表示M个预测器投票得到的最有可能的心脏病在L中的索引，c j表示预测的心脏病，L等于心脏病的种类数，
Figure PCTCN2022126866-appb-000024
的每个分量对应一种心脏病。
待识别的心脏病集合是C＝{c 1,c 2,…,c L}，L表示心脏病中的种类数,心脏病集合表示不同类型/种类的心脏病的集合，与L对应，并与预测结果c j相对应。
步骤4、最终的预测结果通过蓝牙设备进行播报。蓝牙设备为蓝牙耳机或蓝牙音箱。
如图2所示，本发明心脏听诊装置，包括听诊器本体1、移动终端以及远端电脑，听诊器本体1将心音采集后通过蓝牙通讯模块发送至移动终端，移动终端将心音信号通过互联网发送至远端电脑供医生听取和使用。
如图3和图4所示，听诊器本体1上设有第一温度传感器12，其监测环境温度T1；听诊器本体1下方设有硅胶套11，在硅胶套11中设有发热丝14，发热丝14利用听诊器本体1中蓄电池发热，移动终端 2监测到第一温度传感器12采集的环境温度T1小于人体正常体温T2时，启动发热丝14加热，使得硅胶套11接近或达到人体正常体温，确保硅胶套11接触人体时，硅胶套11温度达到人体感应的舒适温度。
还包括第二温度传感器13，第二温度传感器13监测硅胶套11的温度T3，当T3高于预设值T4，移动终端关闭发热丝14的加热功能，避免硅胶套11温度过高给人体产生不适感。
听诊器本体1设有安装腔，在安装腔底部设有振动膜片，安装腔中有基于蓝牙通讯的拾音器，拾音器包括蓝牙主板，蓝牙主板上设有高保真音频传感器、锂电池、USB接口和蓝牙模块，蓝牙模块与移动终端2无线通信，高保真音频传感器、锂电池和USB接口均与蓝牙主板连接。
如图5-图7所示，听诊器本体1底部的硅胶套11为多个单体组成的闭环结构，每个硅胶套单体通过一个伸缩结构15与听诊器本体1连接，每个伸缩结构15能够使得硅胶套单体移动从而使得整个硅胶套11形成高低不同的结构，从而确保整个硅胶套11紧密贴合在人体上形成密闭的采音环境供振动膜片采集心音。
图5中全部的伸缩结构15都处于最大伸展状态，图5中的最左侧的硅胶套单体处于最大伸展状态，其他硅胶套单体处于半伸缩状态。图6中的最左侧的硅胶套单体和中间的硅胶套单体处于最大伸展状态，其他硅胶套单体处于半伸缩状态。上述伸缩位置可以按照测量位置进行自由调节，使得整个硅胶圈贴合在皮肤上即可。
如图8所示，伸缩结构15包括上水平支架151和下水平支架153，上水平支架151与听诊器本体1连接，下水平支架153与硅胶套单体连接，上水平支架151和下水平支架153之间设有多个移动连接片152，移动连接片152倾斜设置并且一端与上水平支架151固定连接，移动连接片152另一端与下水平支架153活动连接，或者，移动连接片152倾斜设置并且一端与上水平支架151活动连接，移动连接片152另一端与下水平支架153固定连接；上水平支架151和下水平支架153之 间的距离改变时从而使得硅胶套单体上下移动，并且移动连接片152倾斜角度也发生对应的变化。
活动连接的结构包括在上水平支架151或下水平支架153开有滑槽，滑槽的开口处窄，滑槽的内部宽，移动连接片152一端位于滑槽开口处的薄，位于滑槽内部的厚，从而使得移动连接片152不能脱离滑槽但能够沿着滑槽延伸方向移动。
如图9所示，伸缩结构15为多层，最上层伸缩结构15与听诊器本体1连接，最下层伸缩结构15与硅胶套单体连接；听诊器本体1上设有第一隔音挡片16，硅胶套11设有第二隔音挡片17，第一隔音挡片16和第二隔音挡片17错位设置并且两者能够随着硅胶套11运动相对运动，伸缩结构15伸展或收缩时，第一隔音挡片16和第二隔音挡片17始终形成屏蔽结构避免安装腔底部的振动膜片受外部噪音影响。
本发明心脏听诊装置的控制原理如下：
步骤1、开启听诊器本体电源；
步骤2、移动终端通过蓝牙连接听诊器本体，并且启动第一温度传感器，第一温度传感器采集的温度T1数据发送至移动终端，移动终端判断T1高于或低于人体正常体温T2；如果低于，则启动硅胶套11中的发热丝14发热，如果高于，则直接进入步骤3；
步骤3、将听诊器本体放置在主动脉瓣区或二尖瓣区的皮肤上，通过伸缩结构确保确保整个硅胶套11紧密贴合在人体上形成密闭的采音环境供振动膜片采集心音；
步骤4、听诊器本体1将心音采集后通过蓝牙通讯模块发送至移动终端，移动终端将心音信号通过互联网发送至远端电脑供医生听取和使用。
实施例2：
下面结合图1至图3，对本发明做进一步说明：
如图1和2所示，一种自动调节的智能心音和心电测量仪，包括心音单元和心电单元，心电单元包括心电电极片9，心电电极片9与主 体6形成采音腔7，心音单元中的主麦克风5位于采音腔7中，其特征在于：还包括一驱动装置和传动结构，驱动装置通过传动结构使得主麦克风5能够在采音腔7中的不同高度位置进行移动，避免外部多声源相互影响造成主麦克风5采集不到心音信号或心音信号过弱。
第一种结构可以为：驱动装置为微型电机1，传动机构包括旋转螺纹杆2、轴向移动轴承套3以及连接件4，旋转螺纹杆2一个端部与微型电机1连接，旋转螺纹杆2套接轴向移动轴承套3，旋转螺纹杆2与轴向移动轴承套3形成滚珠丝杆结构，轴向移动轴承套3的内圈与旋转螺纹杆2螺纹连接，轴向移动轴承套3的外圈通过连接件4与主麦克风5连接，轴向移动轴承套3的内圈转动时，轴向移动轴承套3的外圈保持相对静止，并且主麦克风5沿着旋转螺纹杆2轴向移动。
第二种结构可以为：驱动装置为微型电机1，传动机构包括旋转螺纹杆2、螺母以及连接件4，旋转螺纹杆2一个端部与微型电机1连接，旋转螺纹杆2套接螺母，旋转螺纹杆2与螺母形成滚珠丝杆结构，螺母通过连接件4与主麦克风5连接，轴向移动轴承套3的内圈转动时，主麦克风5旋转并沿着旋转螺纹杆2轴向移动。
第二种结构相对于第一种结构更加简单，但是第一种结构可以使得主麦克风5不会旋转移动，因而确保其始终位螺纹杆的轴线上，因而可以知晓采音的位置。
如图1所示，主麦克风5位于最下端，几乎接触心电电极片9。如图2所示，主麦克风5位于最上端。可以按照需要在最上端与最下端之间设置多个停止点，采集最佳的心音信号。
主体6顶部设有供旋转螺纹杆2穿过的通孔，微型电机1与主体6之间设有隔音层8，避免微型电机工作时产生的噪音对主麦克风的采集进行干扰。
还包括副麦克风，副麦克风位于主体6外收集环境噪音。
如图3所示，还包括音频放大单元、滤波单元以及模数转换单元，音频放大单元将主麦克风采集的心音信号和副麦克风采集的环境噪音 信号进行发大，滤波单元将放大后的心音信号除去放大后的环境噪音信号，获得的更加真实的心音信号，模数转换单元将心音模拟信号转换为数字信号并且输出。
本发明自动调节的智能心音和心电测量仪的控制方法，包括以下步骤：
步骤1、将心电电极片9贴在皮肤上。
步骤2、启动心音单元和心电单元分别采集心音信号和心电信号；步骤3、判断心音单元采集的心音信号的强弱是否符合设定标准，当符合标准时，正常采集采集心音信号和心电信号。
步骤4、判断心音单元采集的心音信号的强弱是否符合设定标准，当心音信号过弱时，启动微型电机，使得主麦克风在采音腔7垂直方向移动，并且在不同的高度停止，分别采集该位置的心音信号的强弱是否符合设定标准，通过比对，最终将主麦克风移动至最佳的位置进行心音信号的采集。
步骤5、音频放大单元将主麦克风采集的心音信号和副麦克风采集的环境噪音信号进行发大，滤波单元将放大后的心音信号含噪音信号除去放大后的环境噪音信号，获得的更加真实的心音信号，模数转换单元将心音模拟信号转换为数字信号并且输出。
当副扬声器采集的声音信号变化较大时，噪音声源有改变，重新判断心音单元采集的心音信号的强弱是否符合设定标准从而决定是否重复步骤4。
上面结合附图对本发明进行了示例性的描述，显然本发明的实现并不受上述方式的限制，只要采用了本发明的方法构思和技术方案进行的各种改进，或未经改进将本发明的构思和技术方案直接应用于其它场合的，均在本发明的保护范围内。

Claims (20)

  1. 一种对心音信号的预测方法，包括以下步骤：
    步骤1、从智能听诊器获得关于个体的听诊的心脏原始声音信号进行降噪，并且形成数据片段；
    步骤2、基于多分辨率深度学习挖掘数据片段信号中蕴含的高层时序语义信息和平移不变性，并最终输出预测结果。
  2. 根据权利要求1所述的对心音信号的预测方法，其特征在于：步骤1中从智能听诊器获得关于个体的听诊的心脏原始声音信号x＝(x 1,x 2,…,x n)后，首先利用低通滤波器过滤点x中的高频信号，得到去噪后的数据y＝(y 1,y 2,…,y n)；考虑到心脏疾病所引起的异常声音持续的时间具有多尺度的特点，为充分挖掘y中的信息，从y中提取M个大小不同的数据片段{s (1),s (2),…,s (M)}，其中第k(1≤k≤M)个数据片段是
    Figure PCTCN2022126866-appb-100001
    其中|s (k)|表示s (k)中的声音数据的样本点个数。
  3. 根据权利要求2所述的对心音信号的预测方法，其特征在于：步骤2中对于第k个数据片段s (k)，首先利用具有一维卷积核的卷积神经网络CNN对s (k)进行处理，得到z (k)＝CNN(s (k))；然后采用具有若干个隐含层的长短时记忆网络LSTM k对进行处理，得到[o (k),w (k)]＝LSTM k(z (k))，并且
    Figure PCTCN2022126866-appb-100002
    Figure PCTCN2022126866-appb-100003
    Figure PCTCN2022126866-appb-100004
    表示LSTM k预测z (k)属于第i类心脏病的概率，
    Figure PCTCN2022126866-appb-100005
    Figure PCTCN2022126866-appb-100006
    Figure PCTCN2022126866-appb-100007
    表示LSTM k对于做出预测
    Figure PCTCN2022126866-appb-100008
    的置信度，将[o (k),w (k)]作为s (k)的预测结果进行输出；[o (k),w (k)]是LSTM k(z (k))的输出，其中o (k)是一个L维的向量
    Figure PCTCN2022126866-appb-100009
    它的第j个元素
    Figure PCTCN2022126866-appb-100010
    表示LSTM k(z (k))预测z (k)的标签是c j的概率；w (k)也是一个L维的向量
    Figure PCTCN2022126866-appb-100011
    它的第j个元素
    Figure PCTCN2022126866-appb-100012
    量化了LSTM k(z (k))将z (k)的标签预测为c j的置信度。
  4. 根据权利要求3所述的对心音信号的分析预测方法，其特征在于：重复上述过程，得到关于{s (1),s (2),…,s (M)}的输出{o (1),w (1)},{o (2),w (2)},…,{o (M),w (M)}，最后采用加权投票的方式，得到最终的预测结果
    Figure PCTCN2022126866-appb-100013
    Figure PCTCN2022126866-appb-100014
    Figure PCTCN2022126866-appb-100015
    中的j表示M个预测器投票得到的最有可能的心脏病在L中的索引，c j表示预测的心脏病，L等于心脏病的种类数，
    Figure PCTCN2022126866-appb-100016
    的每个分量对应一种心脏病。
  5. 根据权利要求4所述的对心音信号的预测方法，其特征在于：待识别的心脏病集合是C＝{c 1,c 2,…,c L}，L表示心脏病中的种类数；其中，心脏病集合表示不同类型/种类的心脏病的集合，与L对应，并与预测结果c j相对应。
  6. 根据权利要求1-5中任何一项所述的对心音信号的预测方法，其特征在于：步骤4、最终的预测结果通过蓝牙设备进行播报。
  7. 根据权利要求6所述的对心音信号的预测方法，其特征在于：所述蓝牙设备为蓝牙耳机或蓝牙音箱。
  8. 一种心脏听诊装置，其特征在于：使用权利要求1-7中的任何一种预测方法。
  9. 根据权利要求7所述的心脏听诊装置，其特征在于：听诊器本体(1)底部的硅胶套(11)为多个单体组成的闭环结构，每个硅胶套单体通过一个伸缩结构(15)与听诊器本体(1)连接，每个伸缩结构(15)能够使得硅胶套单体移动从而使得整个硅胶套(11)形成高低不同的结构，从而确保整个硅胶套(11)紧密贴合在人体上形成密闭的采音环境供振动膜片采集心音。
  10. 一种蓝牙音箱或耳机，其特征在于：权利要求1-7中的任何一种预测方法获得的最终输出预测结果通过所述蓝牙音箱或所述蓝牙耳机以语音的方式进行输出。
  11. 一种自动调节的智能心音和心电测量仪，包括心音单元和心电单元，心电单元包括心电电极片(9)，心电电极片(9)与主体(6)形成采音腔(7)，心音单元中的主麦克风(5)位于采音腔(7)中，其特征在于：还包括一驱动装置和传动结构，所述驱动装置通过传动结构使得主麦克风(5)能够在采音腔(7)中的不同高度位置进行移动，避免外部多声源相互影响造成主麦克风(5)采集不到心音信号或心音信号过弱。
  12. 根据权利要求11所述的自动调节的智能心音和心电测量仪，其特征在于：所述驱动装置为微型电机(1)，所述传动机构包括旋转螺纹杆(2)、轴向移动轴承套(3)以及连接件(4)，旋转螺纹杆(2)一个端部与微型电机(1)连接，旋转螺纹杆(2)套接轴向移动轴承套(3)，所述旋转螺纹杆(2)与所述轴向移动轴承套(3)形成滚珠丝杆结构，轴向移动轴承套(3)的内圈与旋转螺纹杆(2)螺纹连接，轴向移动轴承套(3)的外圈通过连接件(4)与主麦克风(5)连接，轴向移动轴承套(3)的内圈转动时，轴向移动轴承套(3)的外圈保持相对静止，并且主麦克风(5)沿着旋转螺纹杆(2)轴向移动。
  13. 根据权利要求11所述的自动调节的智能心音和心电测量仪，其特征在于：所述驱动装置为微型电机(1)，所述传动机构包括旋转螺纹杆(2)、螺母以及连接件(4)，旋转螺纹杆(2)一个端部与微型电机(1)连接，旋转螺纹杆(2)套接螺母，所述旋转螺纹杆(2)与所述螺母形成滚珠丝杆结构，螺母通过连接件(4)与主麦克风(5)连接，轴向移动轴承套(3)的内圈转动时，主麦克风(5)旋转并沿 着旋转螺纹杆(2)轴向移动。
  14. 根据权利要求12或13所述的自动调节的智能心音和心电测量仪，其特征在于：主体(6)顶部设有供旋转螺纹杆(2)穿过的通孔，微型电机(1)与主体(6)之间设有隔音层(8)，避免微型电机工作时产生的噪音对主麦克风的采集进行干扰。
  15. 根据权利要求12或13所述的自动调节的智能心音和心电测量仪，其特征在于：还包括副麦克风，副麦克风位于主体(6)外收集环境噪音。
  16. 根据权利要求15所述的自动调节的智能心音和心电测量仪，其特征在于：还包括音频放大单元、滤波单元以及模数转换单元，音频放大单元将主麦克风采集的心音信号和副麦克风采集的环境噪音信号进行发大，滤波单元将放大后的心音信号除去放大后的环境噪音信号，获得的更加真实的心音信号，模数转换单元将心音模拟信号转换为数字信号并且输出。
  17. 一种自动调节的智能心音和心电测量仪的控制方法，包括以下步骤：
    步骤1、将心电电极片(9)贴在皮肤上；
    步骤2、启动心音单元和心电单元分别采集心音信号和心电信号；
    步骤3、判断心音单元采集的心音信号的强弱是否符合设定标准，当符合标准时，正常采集采集心音信号和心电信号。
  18. 根据权利要求17所述的自动调节的智能心音和心电测量仪的控制方法，其特征在于：
    步骤4、判断心音单元采集的心音信号的强弱是否符合设定标准， 当心音信号过弱时，启动微型电机，使得主麦克风在采音腔(7)垂直方向移动，并且在不同的高度停止，分别采集该位置的心音信号的强弱是否符合设定标准，通过比对，最终将主麦克风移动至最佳的位置进行心音信号的采集。
  19. 根据权利要求18所述的自动调节的智能心音和心电测量仪的控制方法，其特征在于：步骤5、音频放大单元将主麦克风采集的心音信号和副麦克风采集的环境噪音信号进行发大，滤波单元将放大后的心音信号(含噪音信号)除去放大后的环境噪音信号，获得的更加真实的心音信号，模数转换单元将心音模拟信号转换为数字信号并且输出。
  20. 根据权利要求19所述的自动调节的智能心音和心电测量仪的控制方法，其特征在于：当副扬声器采集的声音信号变化较大时，噪音声源有改变，重新判断心音单元采集的心音信号的强弱是否符合设定标准从而决定是否重复步骤4。
PCT/CN2022/126866 2022-01-11 2022-10-23 一种对心音信号的预测方法及智能心音和心电测量仪 WO2023134253A1 (zh)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN202210024087.0 2022-01-11
CN202210024087.0A CN114209335A (zh) 2022-01-11 2022-01-11 一种自动调节的智能心音和心电测量仪及其控制方法
CN202210496677.3A CN115089206B (zh) 2022-05-09 2022-05-09 一种对心音信号的预测方法及使用其的心脏听诊装置
CN202210496677.3 2022-05-09

Publications (1)

Publication Number Publication Date
WO2023134253A1 true WO2023134253A1 (zh) 2023-07-20

Family

ID=87280071

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/126866 WO2023134253A1 (zh) 2022-01-11 2022-10-23 一种对心音信号的预测方法及智能心音和心电测量仪

Country Status (1)

Country Link
WO (1) WO2023134253A1 (zh)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3200110U (ja) * 2015-07-16 2015-10-01 經和 ▲高▼階 ダイヤフラム型聴診器
CN111031905A (zh) * 2017-04-07 2020-04-17 南洋理工学院 用于检测心脏异常的ecg和pcg监视系统
CN212816302U (zh) * 2020-02-25 2021-03-30 宁德市闽东医院 一种基于蓝牙无线传输式听诊装置
CN113855063A (zh) * 2021-10-21 2021-12-31 华中科技大学 一种基于深度学习的心音自动诊断系统
CN114209335A (zh) * 2022-01-11 2022-03-22 罗育坤 一种自动调节的智能心音和心电测量仪及其控制方法
CN115089206A (zh) * 2022-05-09 2022-09-23 吴先洪 一种对心音信号的预测方法及使用其的心脏听诊装置

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP3200110U (ja) * 2015-07-16 2015-10-01 經和 ▲高▼階 ダイヤフラム型聴診器
CN111031905A (zh) * 2017-04-07 2020-04-17 南洋理工学院 用于检测心脏异常的ecg和pcg监视系统
CN212816302U (zh) * 2020-02-25 2021-03-30 宁德市闽东医院 一种基于蓝牙无线传输式听诊装置
CN113855063A (zh) * 2021-10-21 2021-12-31 华中科技大学 一种基于深度学习的心音自动诊断系统
CN114209335A (zh) * 2022-01-11 2022-03-22 罗育坤 一种自动调节的智能心音和心电测量仪及其控制方法
CN115089206A (zh) * 2022-05-09 2022-09-23 吴先洪 一种对心音信号的预测方法及使用其的心脏听诊装置

Similar Documents

Publication Publication Date Title
US20180108440A1 (en) Systems and methods for medical diagnosis and biomarker identification using physiological sensors and machine learning
US10117635B2 (en) Electronic acoustic stethoscope with ECG
US9826955B2 (en) Air conduction sensor and a system and a method for monitoring a health condition
CN111657991B (zh) 一种智能阵列传感器电子听诊系统
CN109276272A (zh) 一种多功能智能听诊器
CN201755228U (zh) 无线心电心音一体机
US20130150754A1 (en) Electronic stethoscopes with user selectable digital filters
Yu et al. Automatic detection of atrial fibrillation from ballistocardiogram (BCG) using wavelet features and machine learning
TW201922165A (zh) 訊號同步處理裝置及具有訊號同步處理功能的聽診器、聽診資訊輸出系統與病徵診斷系統
CN115089206B (zh) 一种对心音信号的预测方法及使用其的心脏听诊装置
WO2023134253A1 (zh) 一种对心音信号的预测方法及智能心音和心电测量仪
CN111870273A (zh) 一种远程肺音监测装置、系统及其方法
TWM499203U (zh) 生理訊號感測裝置
CN210019403U (zh) 一种体音监测仪
CN111150422A (zh) 智能听诊系统及其听诊方法
Monika et al. Embedded Stethoscope for Real Time Diagnosis of Cardiovascular Diseases
CN206453785U (zh) 一种智能电子听诊器
CN111883180A (zh) 一种无线心音听诊设备及其数据对比方法
CN205924058U (zh) 一种智能听诊器
CN206414283U (zh) 一种具有视听功能的电子听诊器
US20140012095A1 (en) Storage control apparatus, storage control system, and storage medium
CN211300047U (zh) 一种基于神经网络的智能心肺听诊系统
CN116458916A (zh) 一种用于异步远程医疗的智能听诊器及其控制方法
WO2009039863A1 (en) Computerized medical stethoscope
CN111839582B (zh) 隔离式套耳电子听诊器

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22919887

Country of ref document: EP

Kind code of ref document: A1