CN215349053U - Congenital heart disease intelligent screening robot - Google Patents

Abstract

The utility model discloses an intelligent congenital heart disease screening robot which comprises a base, a robot main body, a robot head, a human-computer interaction module, a heart sound pickup module, an embedded main board, a blood oxygen saturation monitoring module and a power management module, wherein the robot head is connected with the base through a power supply; the embedded mainboard comprises an A/D conversion module, a signal comparison module and a data storage module, wherein a congenital heart disease heart sound sample library is stored in the data storage module, the heart sound signals are converted into digital signals which can be identified by a computer through the A/D conversion module, and the acquired target user heart sound signals are compared with the heart sound sample library through the signal comparison module; the congenital heart disease automatic identification system is perfected by combining blood oxygen saturation detection in parallel, the automatic identification system and the robot are carried to be suitable for being used in campus and community scenes, a highly intelligent screening system integrating face identification and human-computer interaction is realized, and closed-loop research from scientific research design → software and hardware design → scene application → scientific verification → comprehensive popularization can be realized.

Description

Congenital heart disease intelligent screening robot

The technical field is as follows:

the utility model relates to the technical field of medical artificial intelligent robots, in particular to a robot with an intelligent congenital heart disease screening system.

Background art:

congenital heart disease (called congenital heart disease for short) refers to abnormal anatomical structure caused by dysgenesis of heart and great vessels during embryonic development, and is the most common birth defect at present; in the birth defect disease spectrum of China, the number of children with congenital heart disease is increased from 20 ten thousand to more than 30 ten thousand, the disease is on the rising trend year by year, becomes a Chinese birth defect list-top disease after 2005, is also the first cause of death of infants aged 0-5 in the city of China, seriously influences the quality and the overall health level of the birth population, and becomes a major public health problem influencing the health quality of the whole people; therefore, it is important to fully develop the diagnosis and treatment and screening of congenital heart disease.

In the embodiment of professor Huang national English of the university pediatric hospital of 'Fudan' in 2013, the newborn is screened by using an echocardiogram in the Shanghai area, the incidence rate of the congenital heart disease of the newborn in the Shanghai area is found to be about 26.6 per thousand, the sensitivity of the ultrasonic detection on the severe congenital heart disease reaches 93.06%, and the specificity reaches 97.98%, but the screening of the congenital heart disease by means of the echocardiogram is difficult to be promoted all over the country, particularly in underdeveloped areas, on one hand, because the machine is expensive, and on the other hand, the other children heart color ultrasonic doctor who lacks in profession is not a very economic and efficient method; the four most common congenital heart diseases are ventricular septal defect, atrial septal defect, patent ductus arteriosus and Fallo tetrad, which account for more than 80 percent of congenital heart diseases, and the four congenital heart diseases all have characteristic heart murmurs (heart sound signals are sound signals generated by vibration when a heart valve is opened or closed due to blood flow and are important biological blood flow signals of a human body, and the heart murmurs are abnormal sounds generated by the vibration of a heart chamber wall, a valve or a blood vessel caused by turbulent flow of blood in the heart or the blood vessel during systole or diastole besides heart sound and extra heart sound); meanwhile, the blood oxygen saturation is an index for reflecting the oxygen-containing concentration of human blood, is a specific gravity coefficient of oxyhemoglobin and hemoglobin, and is used for measuring the oxygen conveying capacity of blood, the blood oxygen saturation of normal people is not lower than 94% under general conditions, most congenital heart diseases are accompanied by mixed flow of arterial blood and venous blood or oxygen deficiency with different degrees to cause abnormal blood oxygen saturation, and a plurality of researches prove that the examination rate of congenital heart diseases can be obviously improved by utilizing the blood oxygen saturation screening, and the important theoretical basis for realizing early identification of the congenital heart diseases through heart murmurs and the blood oxygen saturation is also provided.

Although the method for realizing early recognition of congenital heart disease through heart murmurs and blood oxygen saturation is simple, time-saving and high in accuracy, the method depends excessively on the subjective experience of doctors, and basically realizes that each child receives one-time regular screening of heart disease in a relatively poor area, the screening is difficult and does not have a complete diagnosis and treatment system, including a mature heart color ultrasonography technology, generally only by means of manual auscultation of doctors, the average daily auscultation amount of each doctor participating in auscultation screening is huge, repeated auscultation actions are repeated for many times, the injury and infection of an external auditory canal are often caused, the screening speed is too slow, and particularly for doctors with shallow experience, the method still has high misdiagnosis rate and often delays correct treatment of the disease; therefore, how to utilize effective technological means to solve the problem of congenital heart disease screening based on community, rural area and campus population is very important.

The utility model content is as follows:

the present invention is directed to solve the above-mentioned problems, and an object of the present invention is to provide an intelligent screening robot for congenital heart disease, which is suitable for school and community applications, and which can effectively extract the features of heart sound signals through artificial intelligence deep learning, develop a heart sound recognition device, combine an automatic congenital heart disease recognition system for blood oxygen saturation detection, and mount the automatic congenital heart disease recognition system on an AI robot.

The utility model adopts the following technical scheme to realize the purpose of the utility model: the utility model provides a congenital heart disease intelligence screening robot which characterized in that: the robot comprises a base, a robot main body, a robot head, a human-computer interaction module, a heart sound pickup module, an embedded mainboard and a power management module; a pickup circuit and an amplifying circuit are arranged in the heart sound pickup module, and the amplifying circuit clearly amplifies weak heart sound signals received by the pickup circuit; the embedded mainboard comprises an A/D conversion module, a signal comparison module and a data storage module, wherein a congenital heart disease heart sound sample library is stored in the data storage module, and the amplified heart sound signal is converted into a digital signal which can be recognized by a computer through the A/D conversion module and is used for comparing the acquired heart sound signal of the target user with the heart sound sample library through the signal comparison module; the human-computer interaction module acquires the interaction record of the target user, and stores the interaction record, the heart sound signal of the target user and the comparison result in the data storage module.

Preferably, a neural network processor is further integrated in the embedded motherboard.

Furthermore, the embedded mainboard further comprises a signal preprocessing module for performing XGboost combination classification on the heart sound signals after wavelet denoising, sliding window segmentation, artificial feature extraction and depth feature extraction.

Further, this congenital heart disease intelligence screening robot still includes oxyhemoglobin saturation monitoring module, and oxyhemoglobin saturation monitoring module locates on the robot main part, oxyhemoglobin saturation monitoring module is for pointing double-layered oxyhemoglobin saturation monitor and/or binding type oxyhemoglobin saturation monitor, through data line or bluetooth and embedded mainboard communication connection.

Preferably, the finger-clipped oxyhemoglobin saturation monitor and/or the bundled oxyhemoglobin saturation monitor are provided with two in total.

Furthermore, the heart sound pickup module is arranged on the robot main body and comprises a pickup head and a sound guide tube, one end of the sound guide tube is connected to the pickup head, and the other end of the sound guide tube is connected to the A/D conversion module; the pickup listening head adopts a zinc alloy nickel plating listening head; the pickup circuit and the amplifying circuit are arranged in the pickup head.

Furthermore, the embedded main board is arranged in the robot main body or the robot head and comprises a USB interface for leading in and leading out the heart sound data stored in the data storage module, and the USB interface is arranged on the robot main body or the robot head shell.

Furthermore, the robot adopts a 12V lithium battery for power supply, and the power management module is arranged on the robot main body or the base and used for completing charging and discharging and controlling system power consumption.

Furthermore, the human-computer interaction module is arranged on the robot main body or the robot head and comprises a high-definition capacitive touch screen, a camera and a voice system; the high-definition capacitive touch screen is used for displaying target user information, inputting instructions, guiding correct heart sound data acquisition and displaying details of heart sound signal waveforms; the camera is used for acquiring a facial image of a target user and carrying out identity recognition through the facial image; the voice system is used for voice guidance and voice broadcasting.

Furthermore, the intelligent screening robot for congenital heart diseases also comprises a weight measuring module, a height measuring module and a body temperature measuring module; the weight measuring module is used for measuring the weight of a target user and storing the weight in the data storage module; the height measuring module is used for measuring the height of a target user and storing the height in the data storage module; the body temperature measuring module is used for measuring the body temperature of a target user and storing the body temperature in the data storage module.

By adopting the technical scheme, the intelligent congenital heart disease screening robot better achieves the aim of the utility model, utilizes the existing robot technology, effectively extracting the heart sound signal characteristics by an artificial intelligence deep learning method, collecting the heart sound data of the heart disease, establishing a heart sound sample library, the artificial intelligent recognition to the heart sound sample library is realized through the steps of dimensionality reduction and denoising, data segmentation, artificial feature extraction, depth feature extraction and heart sound classification, the congenital heart disease automatic identification system is perfected by combining blood oxygen saturation detection in parallel, the congenital heart disease automatic identification system and a robot are carried to establish an intelligent congenital heart disease detection system suitable for campus and community scenes, a highly intelligent screening system integrating face identification and human-computer interaction is realized, and closed-loop research from scientific research design → software and hardware design → scene application → scientific verification → comprehensive popularization can be realized.

Description of the drawings:

fig. 1 is a schematic diagram of the overall structure of the front face of the robot according to the present invention.

Fig. 2 is a schematic diagram of the overall structure of the back of the robot according to the present invention.

Fig. 3 is a schematic structural diagram of the heart sound pickup module 5 according to the present invention.

FIG. 4 is a flow chart illustrating the principle of screening congenital heart disease in the present invention.

FIG. 5 is a comparison graph before and after dimension reduction and denoising of a heart sound signal.

Fig. 6 is a schematic diagram of the conversion of an audio signal into a spectrogram by artificial feature extraction in the present invention.

Fig. 7 is a ResNet netlist for deep feature extraction in accordance with the present invention.

FIG. 8 is a CBAM inferred attention diagram in the present invention.

FIG. 9 is a schematic diagram of the learning framework of the XGboost classifier of the present invention.

The reference numerals are illustrated in the following table.

Name of label	Number of mark	Name of label	Number of mark
				1	Base 1	51	Sound pickup head 51
2	Robot main body 2	52	Sound conducting tube 52
				3	Robot head 3	6	Blood oxygen saturation monitoring module 6
4	Human-computer interaction module 4	7	Power management module 7
				5	Heart sound pickup module 5

The specific implementation mode is as follows:

in order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts, and the contents of the utility model will be further described with reference to the drawings:

example 1:

referring to fig. 1 and 2, an intelligent congenital heart disease screening robot comprises a base 1, a robot main body 2, a robot head 3, a human-computer interaction module 4, a heart sound pickup module 5, an embedded mainboard and a power management module 7; a pickup circuit and an amplifying circuit are arranged in the heart sound pickup module 5, the amplifying circuit clearly amplifies weak heart sound signals received by the pickup circuit, and the amplifying circuit designs a signal conditioning circuit by utilizing a triode or an operational amplifier and comprises a signal amplifier, a filter, a voltage follower and the like; storing original heart sound data in a WAV format, wherein the sampling frequency is 5000 Hz; the embedded mainboard comprises an A/D conversion module, a signal comparison module, a signal preprocessing module and a data storage module, wherein the signal comparison module and the signal preprocessing module adopt a Rayleigh core Microcompany RK3399Pro, the RK3399Pro is a high-performance AI processing chip, a large-size core processor framework of ARM dual-core Cortex-A72+ quad-core Cortex-A53 is adopted, the main frequency is up to 1.8GHz, a neural network processor NPU is integrated, and the calculation power is up to 3.0 Tops; a congenital heart disease heart sound sample library is stored in the data storage module, heart sound signals of patients with confirmed congenital heart diseases are stored in the heart sound sample library, the amplified heart sound signals are converted into digital signals which can be recognized by a computer through the A/D conversion module, the heart sound signals are subjected to wavelet de-noising, sliding window segmentation, artificial feature extraction and depth feature extraction and then subjected to BooXSt combination classification through the signal preprocessing module, the heart sound signals of the target user are compared with the heart sound sample library through the signal comparison module, and the human-computer interaction module 4 acquires interaction records of the target user and stores the interaction records, the heart sound signals of the target user and comparison results in the data storage module; the screening basis of congenital heart disease is as follows: the collected heart sound signals of the target user are processed by the embedded mainboard and then compared with the heart sound sample library of the congenital heart disease stored in the data storage module, once the matching is successful, the possibility that the target user may have four congenital heart diseases, namely ventricular septal defect, atrial septal defect, patent ductus arteriosus and Fallo tetrads, is shown, and the verification is carried out through the heart color ultrasound.

In a specific embodiment, the specific method for establishing the heart sound sample library is as follows: all clinical trials were conducted in compliance with the declaration of helsinki, approved by the medical ethics committee, and patients signed written informed consent for inclusion in the selection criteria: the children 1 day to 17 years old who are seen by a specialized hospital for children at trimethyl are firstly clear of whether heart malformation exists through heart color ultrasonography and are less than 4 years old with poor cooperation, under the calm state of heart color ultrasonography, heart sound collection is immediately completed according to an apical area, a pulmonary valve area, an aortic valve second auscultation area and a tricuspid valve area under the calm environment, and when the cooperation degree is more than 4 years old, the heart sound collection is completed according to the five auscultation areas under the calm state, collected heart sound data are processed by an embedded mainboard, and a heart sound sample library is established by completing matching of the heart sound data and the heart color ultrasonography by a cardiovascular physician and a cardiovascular surgeon.

The XGboost combined classification of the heart sound signals after wavelet denoising, sliding window segmentation, artificial feature extraction and depth feature extraction is carried out by adopting the existing method, and the specific method is as follows:

in a specific embodiment, the method for reducing dimensions and denoising includes: the frequency of the normal heart sound signal is within the range of 5-600 Hz, the frequency of some pathological noise in the heart sound signal can reach 1500Hz, and generally, the frequency of more than 2000Hz does not contain effective information basically; according to the nyquist sampling theorem, the sampled discrete digital signal can retain the target information of the original signal without distortion as long as the condition that the sampling frequency is more than twice of the highest frequency in the signal is met during signal sampling. Noise interference inevitably exists in the heart sound signal in the acquisition process, for example, noise interference such as skin friction sound with a sensor, background noise of the acquisition environment, respiratory disturbance sound of a patient and the like, so that the heart sound signal needs to be denoised to obtain the heart sound signal with less noise; the heart sound signal is a random non-stationary signal, and the wavelet transformation is a method which is widely applied and has obvious effect in the denoising processing process of the non-stationary signal, so the embodiment is to select the wavelet transformation to denoise the heart sound data; the basic idea of wavelet denoising is to decompose layer by layer according to signal frequency, wherein each layer of decomposition is from the initial frequency of an original signal to one half frequency of the signal; because the sampling frequency of the original heart sound data is 5000Hz, the four-layer wavelet can filter out noise information of most of heart sound signals, and the denoised heart sound signals are basically similar to the original heart sound signals in appearance, namely, effective heart sound information (shown in figure 5) is reserved to the maximum extent, meanwhile, the dimension reduction operation of reducing the sampling frequency is completed, and the scale of input data can be effectively reduced.

In a specific embodiment, the method for sliding window segmentation includes: the traditional method for identifying the heart sound is to firstly segment the heart sound signals periodically, then classify the heart sound signals by extracting the characteristics of each periodic signal and statistically analyzing the data, and to ensure the effectiveness of the used segmentation method, a large amount of data sets marked with the periodic positions of the heart sound or the positions of all components are needed for measuring the accuracy of the segmentation method. The marking operation consumes manpower resources, especially when a machine learning model is used for segmentation, training samples with marks are too few, the effect of the trained model is not ideal, incorrect segmentation can result in poor recognition effect, common segmentation methods belong to automatic segmentation, the segmentation result cannot be completely correct, the segmentation result of an unknown newly-measured sample does not have a position label to measure the accuracy of segmentation, and once segmentation is wrong, errors occur in extracted features, so that the subsequent classification process is influenced; in fact, when a heart sound segment is intercepted, there is no requirement that a certain basic heart sound is located, and whether the heart sound needs to be periodically segmented or not, there is no clear description, and from published documents, classification of heart sound signals can be realized by periodic segmentation and non-segmentation.

In a specific embodiment, the method for extracting artificial features includes: the waveform data of the audio signal is usually not suitable for the depth feature extraction directly, because there is no waveform information at many positions in the waveform data, which easily causes the input data to be too sparse, and is not beneficial to the network training, in order to facilitate the audio data to be used for the depth feature extraction, the conversion from the audio signal to the effective vector data needs to be completed, the most common conversion method is to convert the audio signal into a spectrogram (refer to fig. 6), which is a three-dimensional spectrum image and is a graph representing the change of a voice spectrum along with time, the vertical axis of the graph is frequency, the horizontal axis of the graph is time, the strength of a given frequency component at a given moment is represented by the gray scale or shade of the tone of a corresponding point, the spectrogram displays a large amount of information related to the audio characteristic, integrates the characteristics of the spectrogram and a time domain waveform, and obviously displays the change condition of the voice spectrum along with time, or a dynamic spectrum, and in addition, in order to verify the heart sound classification effect of different artificial features, the embodiment also proposes to adopt Mel Frequency Cepstrum Coefficients (MFCCs) widely applied in the field of speech recognition to extract the feature parameters of the heart sound signals, wherein the mel frequencies are proposed according to the characteristics of human auditory sense, which reflects that the human auditory sense and the sound frequency form a nonlinear corresponding relationship, and the response sensitivities of the signals with different frequencies are different. Through research on human ears, it can be found that if the frequency difference between two tones is smaller than a certain bandwidth, the human ears cannot be distinguished, which is called a masking effect, the bandwidth is called a critical bandwidth, the mel scale can represent the critical bandwidth, and the MFCC is a cepstrum parameter extracted in the mel scale frequency domain and represented as a one-dimensional vector.

In a specific embodiment, the depth feature extraction method specifically includes: to extract the depth features of the heart sound signal, by comparing a plurality of depth convolutional neural network structures and comprehensively considering the classification effect and the inference speed, the embodiment is intended to select a ResNet network (refer to FIG. 7) for depth feature extraction, wherein a ResNet structure diagram gives 5 depth ResNet structures, respectively 18, 34, 50, 101 and 152, firstly passes through a convolution layer of 7x7, then a maximum pooling is carried out, and then a stacking residual block is carried out, the residual blocks used by the residual networks of the 50, 101 and 152 layers are bottleneck structures, the number of the residual blocks in each network is 8, 16, 33 and 50 from left to right, and finally a global average pooling is usually connected at the end of the network, so that the advantages of no parameter optimization, effective prevention of overfitting, the space transformation of input and output is more robust, and the consistency of feature mapping and categories is enhanced; since the heart sound signal contains time sequence information, in order to verify whether the extraction effect of the heart sound features can be enhanced, the embodiment considers adding a rolling block attention module (CBAM) on the basis of a ResNet network later, which is a simple and effective attention module for a feedforward convolutional neural network, and given an intermediate feature map, the CBAM sequentially deduces an attention diagram along two independent dimensions (channel and space, refer to fig. 8), and then multiplies the attention diagram with an input feature map to perform adaptive feature optimization, since CBAM is a lightweight, general-purpose module, the overhead of this module can be ignored and seamlessly integrated into any CNN architecture, and can be trained end-to-end with the underlying CNN, experiments have shown that, using this module on various models, and continued improvements in classification and detection performance demonstrate the broad applicability of CBAM. The method does not need to modify the network structure of ResNet, and is convenient for addition and verification; in the embodiment, only the feature extraction function of the ResNet network is used to convert each spectrogram into a 2048-dimensional feature vector for the input of the classifier.

In a specific embodiment, the XGBoost combination classification method specifically includes: the XGboost classifier is a simple and efficient integrated learning framework (refer to figure 9), is a combination of a series of classification regression trees, and has the advantages that overfitting is not easy, the training speed is high, the interpretability is strong, a weak classifier model is trained in each step of the XGboost, a new strong classifier is obtained through weighting, finally the strong classifier is used for classifying the features, the algorithm is the strongest in the current common machine learning and the most complex, the XGboost achieves the purpose of improving the performance of the overall model by adopting an addition model and a method of continuously reducing the difference between a predicted value and a real value of a sample in the training process, a classification and regression tree is adopted by a base classifier, for a classification task, the XGboost generates a series of CART trees, each tree is generated according to the gradient value of training data on the previous tree, a newly generated tree structure is continuously adjusted according to the difference between a predicted value and a target value of the sample in the training process, finally, outputting results through the last tree, after obtaining a plurality of depth features, calculating a classification result by each feature by the XGboost classifier, combining the results into a new feature vector, and calculating by using the XGboost classifier again, thereby obtaining a final heart sound classification result.

Referring to the attached drawing 2, the intelligent congenital heart disease screening robot further comprises an oxyhemoglobin saturation monitoring module 6, the oxyhemoglobin saturation monitoring module 6 is arranged on the robot main body 2, and the oxyhemoglobin saturation monitoring module 6 is a finger-clipped oxyhemoglobin saturation monitor and/or a bundled oxyhemoglobin saturation monitor and is in communication connection with the embedded main board through a data line or Bluetooth; specifically, the number of the finger-clipped oxyhemoglobin saturation monitors and/or the number of the bundled oxyhemoglobin saturation monitors are two; the judgment basis of the abnormity of the percutaneous blood oxygen saturation is as follows: (ii) those with an oxygen saturation of less than 90% in the blood of the right hand and any foot meridian skin; ② the oxygen saturation degree of the blood of the right hand or any foot meridian skin is 90-94%, or the difference of the oxygen saturation degree of the blood of the right hand and any foot meridian skin is more than 3%; the measurement is repeated once after 2 to 4 hours for the abnormal person, the positive person is clearly diagnosed by the heart color Doppler ultrasound, and the advanced heart disease, aortic stenosis, aortic arch interruption, or late ductus arteriosus, ventricular septal defect and Faluo tetrad disease with hypoxemia can be screened by screening the blood oxygen saturation of the upper and lower limbs.

Referring to fig. 1, 2 and 3, the heart sound pickup module 5 is disposed on the robot main body 2, and includes a pickup head 51 and a sound guide tube 52, one end of the sound guide tube 52 is connected to the pickup head 51, and the other end is connected to the a/D conversion module; the pickup listening head 51 adopts a zinc alloy nickel plating listening head; a sound pickup circuit and an amplifier circuit are built in the sound pickup head 51.

In a preferred embodiment, the embedded motherboard is built in the robot main body 2 or the robot head 3, and includes a USB interface for importing and exporting the heart sound data stored in the data storage module, and the USB interface is mounted on the housing of the robot main body 2 or the robot head 3.

In a preferred embodiment, the robot is powered by a 12V lithium battery, and the power management module 7 is disposed on the robot main body 2 or the base 1, and is used for completing charging and discharging and controlling system power consumption, fully ensuring reliability of system operation, and improving cruising ability of the instrument.

In one embodiment, the human-computer interaction module 4 is arranged on the robot main body 2 or the robot head 3 and comprises a high-definition capacitive touch screen, a camera and a voice system; the high-definition capacitive touch screen is used for displaying target user information, inputting instructions, guiding correct heart sound data acquisition and displaying details of heart sound signal waveforms, has high display resolution and color saturation on one hand, and can well display the details of the signal waveforms on the other hand, and can replace physical buttons on the other hand, so that the instrument design is more concise and attractive; the camera is used for acquiring a facial image of a target user and carrying out identity recognition through the facial image; the voice system is used for voice guidance and voice broadcasting.

In one embodiment, the intelligent screening robot for congenital heart disease further comprises a weight measuring module, a height measuring module and a body temperature measuring module; the weight measuring module is used for measuring the weight of a target user and storing the weight in the data storage module; the height measuring module is used for measuring the height of a target user and storing the height in the data storage module; the body temperature measuring module is used for measuring the body temperature of a target user and storing the body temperature in the data storage module.

The use method of the intelligent congenital heart disease screening robot comprises the following steps: firstly, a heart sound sample library is led into a data storage module through a USB interface, a target user stands facing a robot, a camera carries out identity recognition through facial images, the congenital heart disease intelligent screening mode is entered after the identity recognition is successful, a high-definition capacitive touch screen plays videos to guide the target user to respectively place a heart sound pickup module 5 and a blood oxygen saturation monitoring module 6 at proper positions of a body after being disinfected by medical disinfection wet tissues for heart sound collection and blood oxygen saturation monitoring, a voice system broadcasts relevant conditions, the target user is guided to put the heart sound pickup module 5 and the blood oxygen saturation monitoring module 6 back to the original position after the heart sound collection and the blood oxygen saturation monitoring are finished, the robot acquires heart sound signals of the target user, processes the heart sound signals through an embedded mainboard, then compares the heart sound signals with the heart sound sample library, broadcasts comparison results through the voice system, and displays the comparison results and the relevant conditions of the blood oxygen saturation through the high-definition capacitive touch screen, if there is no abnormality, the operation is finished, and if there is abnormality, the operation is prompted to retry later or to suggest medical verification.

It is clear that modifications and/or additions of parts may be made to the intelligent congenital heart disease screening robot as described heretofore, and to the corresponding method, without departing from the field and scope of the present invention.

It is also clear that, although the present invention has been described in detail with reference to this intelligent screening robot for congenital heart diseases, a person of skill in the art shall certainly be able to achieve many other equivalent forms of intelligent screening robot for congenital heart diseases and corresponding method, having the characteristics as set forth in the claims and hence all coming within the field of protection defined thereby.

Claims (9)

1. The utility model provides a congenital heart disease intelligence screening robot which characterized in that: the robot comprises a base, a robot main body, a robot head, a human-computer interaction module, a heart sound pickup module, an embedded mainboard and a power management module; a pickup circuit and an amplifying circuit are arranged in the heart sound pickup module, and the amplifying circuit clearly amplifies weak heart sound signals received by the pickup circuit; the embedded mainboard comprises an A/D conversion module, a signal comparison module and a data storage module, wherein a congenital heart disease heart sound sample library is stored in the data storage module, and the amplified heart sound signal is converted into a digital signal which can be recognized by a computer through the A/D conversion module and is used for comparing the acquired heart sound signal of the target user with the heart sound sample library through the signal comparison module; the human-computer interaction module acquires an interaction record of a target user, and stores the interaction record, the heart sound signal of the target user and a comparison result in the data storage module; the heart sound pickup module is arranged on the robot main body and comprises a pickup head and a sound guide tube, one end of the sound guide tube is connected to the pickup head, and the other end of the sound guide tube is connected to the A/D conversion module; the pickup listening head adopts a zinc alloy nickel plating listening head; the pickup circuit and the amplifying circuit are arranged in the pickup head.

2. The intelligent congenital heart disease screening robot of claim 1, wherein: and a neural network processor is also integrated in the embedded mainboard.

3. The intelligent congenital heart disease screening robot of claim 1, wherein: the embedded mainboard further comprises a signal preprocessing module which is used for performing XGboost combination classification on the heart sound signals after wavelet denoising, sliding window segmentation, artificial feature extraction and depth feature extraction.

4. The intelligent congenital heart disease screening robot of claim 1, wherein: still include oxyhemoglobin saturation monitoring module, oxyhemoglobin saturation monitoring module locates on the robot main part, oxyhemoglobin saturation monitoring module is for pointing double-layered oxyhemoglobin saturation monitor and/or bundling oxyhemoglobin saturation monitor, through data line or bluetooth and embedded mainboard communication connection.

5. The intelligent congenital heart disease screening robot as claimed in claim 4, wherein: the finger-clipped oxyhemoglobin saturation monitors and/or the bundled oxyhemoglobin saturation monitors are totally provided with two.

6. The intelligent congenital heart disease screening robot of claim 1, wherein: the embedded mainboard is arranged in the robot main body or the robot head and comprises a USB interface for leading in and leading out the heart sound data stored in the data storage module, and the USB interface is arranged on the robot main body or the robot head shell.

7. The intelligent congenital heart disease screening robot of claim 1, wherein: the robot adopts 12V lithium cell power supply, and power management module locates on robot main part or the base for accomplish charge-discharge and system power consumption control.

8. The intelligent congenital heart disease screening robot of claim 1, wherein: the human-computer interaction module is arranged on the robot main body or the robot head and comprises a high-definition capacitive touch screen, a camera and a voice system; the high-definition capacitive touch screen is used for displaying target user information, inputting instructions, guiding correct heart sound data acquisition and displaying details of heart sound signal waveforms; the camera is used for acquiring a facial image of a target user and carrying out identity recognition through the facial image; the voice system is used for voice guidance and voice broadcasting.

9. The intelligent congenital heart disease screening robot of claim 1, wherein: the intelligent screening robot for the congenital heart disease further comprises a weight measuring module, a height measuring module and a body temperature measuring module; the weight measuring module is used for measuring the weight of a target user and storing the weight in the data storage module; the height measuring module is used for measuring the height of a target user and storing the height in the data storage module; the body temperature measuring module is used for measuring the body temperature of a target user and storing the body temperature in the data storage module.

Images