High-precision fetal heart monitoring device

Abstract

The invention discloses a high-precision fetal heart monitoring device in the technical field of fetal heart monitoring, and the device comprises a multi-mode sensing module which is used for collecting fetal heart signals, maternal movement data and maternal body surface feature information, and calculating the abdominal fat thickness and skin elasticity coefficient of a pregnant woman; the self-adaptive compensation module is used for carrying out interference classification on signals, tracking position changes of a fetus in a maternal body in real time, compensating position signals of the fetus in combination with a deep learning model, correcting signal deviation caused by movement of the fetus and movement of the maternal body through a prediction algorithm, and outputting the corrected signal deviation. The ultrasonic transmitting power and the time gain compensation curve are dynamically adjusted based on the abdominal fat thickness; and the visualization module is used for visually displaying the fetal heart rate curve, the movement trend and the positioning information. According to the invention, high-precision fetal heart electrical activity monitoring can be realized in complex scenes (such as pregnant woman exercise and obese constitution), and reliable data support is provided for early diagnosis of fetal arrhythmia.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fetal heart monitoring, in particular to a high-precision fetal heart monitoring device. BACKGROUND

[0002] Fetal heart monitoring is a crucial examination method during pregnancy and childbirth, and its core purpose is to assess the health status of the fetus and identify potential risks so that timely intervention measures can be taken. Congenital heart disease is one of the most common fetal malformations, and fetal echocardiography (such as fetal echocardiogram) can screen for cardiac structural abnormalities at an early stage, creating conditions for timely surgery or treatment after birth. During childbirth, changes in fetal heart rate can directly reflect whether the fetus is at risk of hypoxia, for example, fetal bradycardia or variable deceleration may indicate fetal distress, at which point emergency cesarean section and other intervention measures are needed to ensure the safety of the fetus.

[0003] Noise generated by the daily activities, body position changes, and maternal physiological signals (such as breathing, muscle movement) of pregnant women seriously interferes with the collection of fetal electrocardiogram signals. In a non-recumbent state, traditional fetal heart monitors can easily confuse maternal and fetal electrocardiograms, and individual differences such as the thickness of the pregnant woman's abdominal fat and the amount of amniotic fluid can cause signal transmission attenuation. External electromagnetic environments, such as mobile phones and medical equipment electromagnetic waves, can also interfere with monitoring equipment and affect data accuracy. At the same time, the fetus moves randomly in the uterus, and traditional monitoring equipment relies on fixed electrode patches or probes, which cannot dynamically track the position changes. Once the fetal position changes, the monitoring signal may weaken or be interrupted. In addition, individual differences and fetal position changes also affect the accuracy of the monitoring results. The fat layer of obese pregnant women can significantly attenuate the ultrasound signal, reducing the signal-to-noise ratio of the electrocardiogram signal; fetal position changes cause the ultrasound window to shift, and traditional single-probe effective signal collection time is short, and there is a lack of dynamic compensation mechanism for skin elasticity and contact impedance fluctuations.

[0004] Therefore, the present application proposes a high-precision fetal heart monitoring device to solve the above problems while being easy to operate. SUMMARY

[0005] To solve the above problems, the present application provides a high-precision fetal heart monitoring device, which combines the thickness of the pregnant woman's abdominal fat and the skin elasticity coefficient with conventional monitoring data through multi-modal data fusion perception, deep learning-driven adaptive compensation, and dynamic parameter optimization technology, to achieve high-precision fetal cardiac electrical activity monitoring in complex scenarios (such as pregnant women exercising, obese constitution), and to provide reliable data support for early diagnosis of fetal arrhythmia.

[0006] In order to achieve the above purpose, the technical scheme of the present application is as follows: a high-precision fetal heart monitoring device, comprising:

[0007] The multi-modal sensing module is composed of an ultrasonic sensor array, an acceleration sensor, and a bioelectric sensor, and is used to collect fetal heart signals, maternal motion data, and maternal body surface characteristic information; the maternal body surface characteristic information includes abdominal surface pressure distribution data and body surface impedance values; and is also used to calculate the pregnant woman's abdominal fat thickness and skin elasticity coefficient based on the body surface impedance values and the abdominal surface pressure distribution data;

[0008] The adaptive compensation module is used to connect the multi-modal sensing module based on a convolutional neural network and long short-term memory network fusion architecture, analyze and process the signals collected by the multi-modal sensing module, and then classify the processed data by frequency spectrum analysis and pattern recognition algorithms, the interference types including maternal respiration, myoelectric interference, and environmental electromagnetic noise; and is also used to provide a hybrid positioning technology of inertial navigation and signal strength fingerprinting, to track the position changes of the fetus in the mother in real time, and then combine a deep learning model to compensate the fetal position signals, correct the signal deviation caused by fetal movement and maternal activity through a prediction algorithm, and dynamically adjust the ultrasonic transmission power and time gain compensation curve based on the abdominal fat thickness.

[0009] The visualization module is used to construct a three-dimensional dynamic model of the fetal heart according to biological experimental data, and is connected to the adaptive compensation module to receive data information from the adaptive compensation module, analyze the fetal heart rate curve, motion trend, and positioning information, and visualize them in the three-dimensional dynamic model of the fetal heart.

[0010] Further, in the multi-modal sensing module, the ultrasonic sensor array uses a phased array technology to capture the beating echo signals of the fetal heart through several ultrasonic probes; the acceleration sensor is used to monitor the motion and body position changes of the mother in real time; and the bioelectric sensor is used to detect the myoelectric signal, electrocardiogram signal, and pressure distribution signal of the maternal abdomen.

[0011] Further, in the visualization module, the three-dimensional positioning of the fetal heart is realized based on a time difference positioning algorithm, and the U-Net++ deep learning model is used to segment the abdominal surface contour in real time to construct the three-dimensional coordinate mapping relationship between the fetal position and the body surface.

[0012] Further, in the adaptive compensation module, the time-frequency domain features, statistical characteristics, and correlation parameters of the real-time collected signals are acquired, an interference classification model is constructed through an adaptive learning algorithm, the maternal motion artifacts, environmental electromagnetic noise, myoelectric interference, and baseline drift are identified, and an interference feature vector is outputted, and the filtering strategy is dynamically switched according to the interference feature vector result.

[0013] Further, the filtering strategy includes frequency domain notch filtering for narrowband periodic interference, time domain prediction filtering for motion-related artifacts, and time-frequency domain decomposition and reconstruction algorithm for baseline drift, to realize layered noise reduction processing of the fetal heart signals.

[0014] Further, the filtering strategy specific process includes: when narrowband interference is detected, notch filtering is automatically enabled to suppress the interference of specific frequency components in the signal frequency domain; when maternal motion artifacts are identified, motion trajectory compensation is performed on the ultrasonic blood flow signal or electrocardiogram signal through a prediction filtering algorithm combined with motion data; when baseline drift is detected, a multi-resolution time-frequency analysis method is used to perform hierarchical decomposition on the signal, and feature enhancement and baseline correction are performed on the effective frequency band of the fetal heart signal.

[0015] Further, the hierarchical decomposition determines the minimum decomposition layer number according to the characteristic frequency band of the fetal electrocardiogram signal and the frequency characteristics of the baseline drift through the Nyquist sampling theorem.

[0016] Further, the calculation process of the pregnant woman's abdominal fat thickness and skin elasticity coefficient includes:

[0017] Step one, apply different frequency currents through bioelectric sensors, measure impedance values between different positions, and construct a two-dimensional impedance map;

[0018] Step two, establish a pressure-impedance coupling model combined with the abdominal surface pressure distribution data collected by the bioelectric sensor, and solve the fat layer thickness through the finite element inversion algorithm;

[0019] Step three, apply a step pressure to the same measurement point, record the pressure-deformation response curve, and obtain the skin elasticity coefficient through viscoelastic model parameter identification;

[0020] Step four, input the fat layer thickness and skin elasticity coefficient into the adaptive compensation module to dynamically adjust the ultrasonic transmission power and electrode contact pressure.

[0021] Further, the adaptive compensation module is used to dynamically adjust the time gain compensation curve of the ultrasonic signal based on the fat thickness, increase the gain exponentially in the depth direction, adjust the contact pressure of the bioelectric sensor according to the skin elasticity coefficient, and ensure that the contact impedance is stable within the preset threshold range through closed-loop feedback; construct a multi-dimensional compensation model containing fat thickness, skin elasticity coefficient and fetal position, optimize the model parameters through deep learning algorithm, so that the fetal heart rate monitoring error rate of pregnant women with different body types is controlled within a unified threshold, when the fat thickness exceeds the preset critical value, automatically activate the enhanced mode, increase the ultrasonic transmission power and adjust the phased array focusing depth.

[0022] Further, the visualization module is also used to realize local storage and remote wireless transmission of monitoring data, and supports docking with the hospital cloud platform, which is convenient for medical staff to view and analyze monitoring data in real time.

[0023] The above scheme has the following beneficial effects: 1. The multi-modal sensing fusion and intelligent compensation technology solves the performance bottleneck of the traditional fetal monitoring device in a complex scene. In the obese pregnant woman group (BMI>30), the bioelectric impedance and pressure sensing fusion technology realizes real-time quantitative detection of the abdominal fat thickness (accuracy ±0.5 cm), and combines the dynamic ultrasonic power adjustment mechanism to improve the signal penetration depth to 15 cm, effectively solving the strong attenuation problem of the ultrasonic signal by the fat layer. In the maternal motion scene, the inertial navigation and signal strength fingerprint hybrid positioning technology controls the fetal heart three-dimensional tracking error within 3 mm, and cooperates with the motion artifact prediction compensation algorithm to improve the signal effective acquisition rate of the pregnant woman in the daily activity state from 60% of the traditional technology to 95%.

[0024] 2. The scheme stabilizes the contact impedance between the electrode and the skin in the 5-10kΩ safe interval through the closed-loop feedback system, significantly reduces the distortion of the electrocardiosignal caused by the difference in skin characteristics, and automatically optimizes the monitoring parameters based on the dynamic correlation of the fat thickness, skin elasticity and fetal position through the deep learning algorithm, so that the monitoring error rate of pregnant women of different body types is uniformly controlled within 5%. In multiple pregnancy monitoring, the heart positions of two fetuses can be tracked at the same time, and the positioning accuracy rate reaches 98%, providing a reliable tool for high-risk pregnancy monitoring.

[0025] Additional aspects and advantages of the application will be in part apparent and in part pointed out hereinafter. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 The flowchart of the high-precision fetal heart monitoring device embodiment of the application is shown. DETAILED DESCRIPTION

[0027] The technical solutions of the application will be described below in detail with reference to the drawings. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0028] In the description of the present application, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application. In addition, the terms "first", "second", "third" are only for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0029] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium; it can be the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0030] The specific embodiments are described in detail below:

[0031] Embodiments:

[0032] As shown in the accompanying drawings, a high-precision fetal heart monitoring device comprises: Figure 1

[0033] A multi-modal sensing module, which is composed of an ultrasonic sensor array, an acceleration sensor and a bioelectricity sensor.

[0034] Among them, the ultrasonic sensor array adopts phased array technology, and 16 ultrasonic probes can be preferably configured, the working frequency of the probes is all set to 3-5MHz, the direction and focusing depth of the ultrasonic beam can be adjusted through electronic control, and in the actual monitoring process, when the pregnant woman is in a supine position, the ultrasonic probe array can accurately capture the beating echo signal of the fetal heart in different cardiac cycles, providing raw data for fetal heart structure and hemodynamics analysis.

[0035] The acceleration sensor is used to monitor the motion and body position changes of the mother in real time. A three-axis acceleration sensor is preferably selected, the range is set to ±8g, and the sampling frequency is 100Hz, which can monitor the motion acceleration and body position change information of the mother during daily activities (such as turning over and walking) in real time. For example, when the mother turns over, the acceleration sensor can quickly detect the change of the motion direction and acceleration, and transmit the data to the adaptive compensation module.

[0036] ​The bioelectric sensor is used to detect the myoelectric signal, electrocardiogram and pressure distribution signal (e.g. abdominal surface pressure distribution data and body surface impedance value) of the maternal abdomen; meanwhile, the maternal abdominal fat thickness and skin elasticity coefficient are calculated based on the body surface impedance value and abdominal surface pressure distribution data. Preferably, the bioelectric sensor is composed of a flexible electrode array of 8-12 Ag / AgCl electrodes, and the electrode surface is coated with a conductive polymer coating to reduce the contact impedance with the skin. The sensor can detect the myoelectric signal, electrocardiogram of the maternal abdomen, and simultaneously collect the abdominal surface pressure distribution data and body surface impedance value. In actual operation, the bioelectric sensor is closely attached to the maternal abdominal skin, and by applying a weak current of different frequencies (10 kHz-100 kHz), the impedance values between different positions are measured to construct a two-dimensional impedance map.

[0037] Specifically, the calculation process of the maternal abdominal fat thickness and skin elasticity coefficient includes:

[0038] Step one, a weak sinusoidal current I of different frequencies (set the frequency range as f∈10 kHz, 100 kHz) is applied to the maternal abdominal skin surface by the bioelectric sensor, and according to Ohm's law V=I×Z (where V is the voltage between the measurement electrodes, and Z is the impedance). Taking a flexible electrode array composed of 8 Ag / AgCl electrodes as an example, in the flexible electrode array composed of 8 Ag / AgCl electrodes, n measurement points (n is determined by the electrode distribution density, and in this embodiment, n=36) are selected, and the impedance values Z between different positions are measured. ij (i,j represents different measurement point numbers). The impedance values of all measurement points are distributed according to the spatial position to construct a two-dimensional impedance map M, and each element in the map corresponds to the impedance value of a measurement point, i.e. M(i,j)=Z ij .

[0039] Step two, the pressure-impedance coupling model is established by combining the abdominal surface pressure distribution data P(x,y) collected by the bioelectric sensor (x,y is the pressure sensor coordinate) and considering the correlation between the electrical and mechanical properties of biological tissues. It is assumed that the biological tissue can be approximated as a layered medium, and the relationship between the impedance Z and the pressure P can be expressed as:

[0040] Z=Z0+k p ×P+∈

[0041] Where Z p is the basic impedance without pressure, k p is the pressure-impedance coupling coefficient, and ∈ is the error term. Through regression analysis on a large amount of sample data, the k pthe experience value), and then inputting the two-dimensional impedance map M and the pressure distribution data P(x, y) as inputs, dividing the abdominal tissue into a plurality of tiny units by using a finite element method, and based on a pressure-impedance coupling model, establishing a target function:

[0042]

[0043] wherein is an actually measured impedance value, is an impedance value calculated by using a finite element model. The thickness distribution d(x, y) of the abdominal fat layer is solved by minimizing the target function F by using an iterative optimization algorithm (such as a conjugate gradient method).

[0044] Step three, for the same measurement point on the bioelectric sensor, a step pressure with an amplitude of P0 is applied, and a pressure-deformation response curve δ(t) is recorded by using the pressure sensor at a sampling frequency of 1000 Hz, wherein t is time. A standard linear solid model (a common viscoelastic model) is used to describe the mechanical behavior of the skin, and the constitutive equation is:

[0045]

[0046] wherein σ is stress, ∈ is strain, E1 and E2 are elastic moduli, and η1 is a viscosity coefficient. The recorded pressure-deformation response curve δ(t) is converted into stress-strain data (σ, ∈), and the constitutive equation is fitted by using a nonlinear least square method to identify the elastic moduli (E1, E2) and the viscosity coefficient η1, and then the skin elastic coefficient E is calculated: A is a pressure acting area, and l0 is an initial length), and the constitutive equation is fitted by using a nonlinear least square method to identify the elastic moduli (E1, E2) and the viscosity coefficient η1, and then the skin elastic coefficient E is calculated:

[0047]

[0048] Step four, the calculated fat layer thickness d(x, y) and the skin elastic coefficient E skin are input into the adaptive compensation module. According to the fat thickness, the ultrasonic transmission power is dynamically adjusted according to the formula:

[0049] P transmit = P base + α × d avg

[0050] (wherein P transmit is the adjusted ultrasonic transmission power, P base is the basic transmission power, α is a power adjustment coefficient, and d avg is the average fat thickness); and simultaneously, according to the skin elastic coefficient, the electrode contact pressure is adjusted by using a closed-loop feedback system to ensure that the contact impedance is stably within a preset threshold range (the preset threshold in this embodiment is [5 kΩ, 10 kΩ]), so as to optimize the collection quality of the fetal heart signal.

[0051] The adaptive compensation module is used for connecting the multi-modal sensing module based on a convolutional neural network and a long short-term memory network fusion architecture, analyzing and processing signals collected by the multi-modal sensing module, classifying the processed data through a spectrum analysis and a pattern recognition algorithm, and providing an inertial navigation and signal strength fingerprint hybrid positioning technology, real-time tracking of the position change of the fetus in the mother, combination of a deep learning model for compensation of the position signal of the fetus, correction of signal deviation caused by fetal movement and maternal activity through a prediction algorithm, and dynamic adjustment of ultrasonic transmission power and time gain compensation curve based on abdominal fat thickness.

[0052] The specific process of classifying the signals in the adaptive compensation module is as follows: real-time collection of time-frequency domain features (such as spectrum feature after short-time Fourier transform), statistical characteristics (mean, variance, kurtosis, etc.) and correlation parameters of the signals, construction of an interference classification model through an adaptive learning algorithm, identification of maternal motion artifacts, environmental electromagnetic noise, electromyographic interference and baseline drift, output of an interference feature vector, and dynamic switching of a filtering strategy according to the interference feature vector result.

[0053] Specifically, the filtering strategy includes frequency domain notch filtering (50Hz) for narrowband periodic interference, time domain prediction filtering for motion-related artifacts and time-frequency domain decomposition reconstruction algorithm for baseline drift, to realize layered noise reduction processing of the fetal heart signal.

[0054] In addition, the specific process of the filtering strategy includes: when narrowband interference is detected, automatically enabling notch filtering (second-order IIR notch filter) to suppress the interference of specific frequency components in the signal frequency domain; when maternal motion artifacts are identified, combining motion data to compensate for the motion trajectory of the ultrasonic blood flow signal or the electrocardiogram signal through a prediction filtering algorithm; when baseline drift is detected, using a multi-resolution time-frequency analysis method to perform layered decomposition on the signal, and enhancing the characteristics and correcting the baseline of the effective frequency band of the fetal heart signal. Among them, the minimum decomposition layer L is determined according to the characteristic frequency band of the fetal electrocardiogram signal and the frequency characteristics of the baseline drift through the Nyquist sampling theorem:

[0055]

[0056] Where f max is the highest frequency of baseline drift (0.5Hz in this embodiment). For the decomposed coefficients of each layer, a soft threshold function W j = sgn(W j )(|W j |- λ) (λ is the threshold value) is used to remove low-frequency drift components, and the reconstructed signal realizes baseline correction.

[0057] a visualization module for constructing a three-dimensional dynamic model of a fetal heart according to biological experimental data; and connecting an adaptive compensation module to receive data information from the adaptive compensation module, and analyzing to obtain a fetal heart rate curve, a motion trend and positioning information and visualizing in the three-dimensional dynamic model of the fetal heart.

[0058] Wherein, the biological experimental data for constructing the three-dimensional dynamic model of the fetal heart is obtained as follows:

[0059] Step one, using two-dimensional (2D) or three-dimensional (3D) differentiation technology, by regulating WNT, Nodal / Activin, BMP and other signaling pathways, hPSC / hESC is induced into cardiac progenitor cells (including FHF, aSHF, pSHF). Activin A and CHIR99021 are used to inhibit WNT and Nodal signals; pSHF (posterior second heart field): add retinoic acid (RA) to activate HOXB1, TBX5 and other transcription factors; and then the progenitor cells are aggregated into 3D structures in ultra-low adsorption culture plates, and through the time sequence addition of signal factors (such as BMP4, FGF2, insulin), cardiac organoids with chamber structure are induced to simulate the morphogenesis and functional maturation of heart development;

[0060] Step two, use an optical microscope to record the size and chamber expansion dynamics (such as left ventricular / right ventricular cavity area change) of the cardiac organoid, use TBX1 (aSHF marker), NR2F2 (atrial marker), IRX1 (right ventricular marker) and other antibodies to locate the spatial distribution and differentiation state of the progenitor cells; then perform calcium transient imaging, record the signal propagation path and speed of the myocardial cells through a wide-field microscope; and then perform patch clamp recording, i.e. measure the action potential (AP) of a single myocardial cell to obtain parameters such as AP amplitude and repolarization time (APD); 2+

[0061] Step three, analyze the gene expression profile of FHF / aSHF / pSHF progenitor cells and cardiac organoids, screen chamber-specific markers (such as IRX4 for LV and IRX1 for RV), construct a human embryonic heart single-cell transcriptome map, and analyze the molecular characteristics of myocardial cells in different chambers; and use CRISPR / Cas9 technology to construct hPSC / hESC lines with ISL1, TBX5, FOXF1 and other transcription factor knockouts, and observe the development defects of the cardiac organoids (such as TBX5 knockout leading to ventricular differentiation failure).

[0062] After obtaining the biological experimental data of the heart, the heart model is constructed, and the three-dimensional dynamic model of the fetal heart is divided into two parts of anatomical structure modeling and electrophysiological activity modeling, which are as follows:

[0063] ​1. Anatomical structure model construction (with immunostaining images and RNA-seq screened chamber-specific gene expression data of heart organoids as input data): using U-Net++ deep learning model, based on ultrasound images or fluorescent staining data, segment the left ventricle (LV), right ventricle (RV), atrium (Atrial), outflow tract (OFT) and atrioventricular canal (AVC) of the heart organoid, according to the development timing of the heart organoid (such as SHF heart organoid chamber formation delay), set the growth rate parameters of each chamber (such as RV volume daily growth 15%-20%), and introduce "progenitor cell sorting algorithm", simulate the spatial self-organization process of FHF / aSHF / pSHF progenitor cells (such as aSHF progenitor cells migrate to RV area), finally, input the morphological data of gene knockout or teratogen treatment (such as ISL1 knockout leads to atrial shrinkage), construct a pathological model library, which is used to compare the ultrasound images of fetal heart;

[0064] 2. Electrophysiological activity model construction (with signal propagation velocity measured by transient experiment and ion channel expression data recorded by patch clamp as input data): based on finite element analysis algorithm, divide the heart into multiple electrophysiological units (LV / RV / Atrial, etc.), define the electrical conductance parameters between units (such as GJA1 mediated intercellular connection conductance), then perform pacing signal simulation (early pregnancy model: left ventricular pacing is mainly used; mid-late pregnancy model: switch to atrial pacing), finally, introduce the electrical signal abnormal data in teratogenic experiment (such as salidomide treatment leads to AVC area conduction block), adjust the ion channel parameters (such as reduce GJA1 expression), simulate arrhythmia (such as premature beat, conduction delay).

[0065] After acquiring the biological experimental data and completing the construction of the three-dimensional dynamic model of the fetal heart, the device realizes the spatial positioning and visual presentation of the monitoring data through the following technologies, improves the clinical application value, and specifically, realizes the three-dimensional positioning of the fetal heart based on the time difference positioning algorithm, realizes the real-time segmentation of the surface contour of the pregnant woman's abdomen through the U-Net++ deep learning model, and constructs the three-dimensional coordinate mapping relationship between the fetal position and the body surface. The fetal heart rate curve, motion trend and positioning information are displayed on the screen in real time, and are presented to medical staff and pregnant women in the form of intuitive charts and graphs. For example, the fetal heart rate curve is dynamically drawn with time as the horizontal axis and heart rate value as the vertical axis, and when the fetal heart rate fluctuates abnormally, the curve color automatically changes to red to warn. In addition, preferably, the visual module is built-in with a large-capacity storage device, which can realize local storage of monitoring data for up to 60 days. At the same time, it supports Bluetooth, Wi-Fi and 4G full network wireless transmission methods, and through the data interface compatible with HL7 V3.0 and DICOM 3.0 standards, the monitoring data is synchronized to the hospital cloud platform, which is convenient for medical staff to view and analyze the monitoring data in real time on any terminal device in the hospital, and provides convenience for remote consultation and real-time evaluation of fetal health status.

[0066] Based on the above embodiment, a comparative test is carried out, specifically as follows:

[0067] I. Experimental purpose: to verify the performance of the scheme in different scenarios, including signal acquisition accuracy, interference suppression ability, positioning accuracy and adaptive compensation effect.

[0068] II. Experimental object: 30 pregnant women in 28-36 weeks of pregnancy are recruited and divided into three groups (10 people in each group): normal body type group (BMI 18.5-23.9); obesity group (BMI≥28); exercise group (simulating daily activities such as standing, walking and turning over). The comparison device is the mainstream fetal monitor in the market.

[0069] III. Experimental steps

[0070] (I) Static monitoring experiment

[0071] The pregnant woman lies on the examination bed and wears the device and the comparison device respectively.

[0072] Record the fetal heart rate and electrocardio signal for 10 minutes, calculate the average error rate and signal-to-noise ratio (SNR).

[0073] Turn on the electromagnetic interference generator, repeat step 2, and observe the interference suppression effect.

[0074] (II) Dynamic monitoring experiment

[0075] The pregnant women in the exercise group perform standing, walking and turning over, and are continuously monitored for 20 minutes.

[0076] The actual position of the fetus is recorded by a three-dimensional motion capture system, and the positioning results of the device are compared.

[0077] The motion artifact suppression effect and the data integrity after signal compensation are analyzed.

[0078] (Three) Special experiment for obese pregnant women

[0079] The obese group of pregnant women wears the device, and adjusts the ultrasonic emission power and electrode pressure.

[0080] The fetal heart rate and electrocardiogram signal are recorded, and the monitoring accuracy under different parameter adjustment strategies is compared.

[0081] Four, experimental data recording and analysis

[0082] (One) Comparison of fetal heart rate monitoring accuracy

[0083] Group Mean error of device (bpm) Mean error of comparative device (bpm) Normal group 1.2 3.5 Obese group 2.1 6.8 Exercise group 1.8 5.2

[0084] (Two) Comparison of signal interference suppression effect

[0085] Interference type SNR improvement of device (dB) SNR improvement of comparative device (dB) 50Hz power frequency noise 15.3 3.2 Electromyographic interference 12.7 2.8

[0086] (Three) Comparison of fetal positioning accuracy

[0087] Motion type Positioning error of device (mm) Three-dimensional motion capture error (mm) Roll over 2.3 0.1 Walk 2.8 0.1

[0088] Five, experimental conclusion

[0089] The experimental data show that in different body types of pregnant women and in different exercise scenarios, the average error of fetal heart rate monitoring of the high-precision fetal heart monitoring device is 1.2bpm for normal body type group, 2.1bpm for obese group, and 1.8bpm for exercise group, which is significantly lower than 3.5bpm, 6.8bpm and 5.2bpm of the comparison equipment, and the error reduction rate in obese and exercise scenarios is more than 60%. When facing 50Hz power frequency noise and electromyographic interference, the signal-to-noise ratio of the device improves by 15.3dB and 12.7dB respectively, which is much higher than 3.2dB and 2.8dB of the comparison equipment, and the interference suppression effect is outstanding. In terms of fetal positioning, the positioning error of the device is 2.3mm and 2.8mm respectively under turning and walking actions, both of which are controlled within 3mm, which can meet the clinical precise positioning requirements. Therefore, the device shows better performance than traditional equipment in terms of signal acquisition accuracy, interference suppression ability and positioning accuracy under complex physiological conditions, and has high clinical application value.

[0090] Obviously, the above embodiments are merely example for clearly illustrating but not limitation to the embodiments. Based on the above description, other different forms of changes or variations can be made by those skilled in the art. Here, all the embodiments need not and can not be enumerated. The obvious changes or variations derived from this still fall within the protection scope of the present application.

Claims

1. A high-precision fetal heart monitoring device, characterized in that, include: The multimodal sensing module consists of an ultrasonic sensor array, an accelerometer, and a bioelectric sensor. It is used to collect fetal heart signals, maternal motion data, and maternal body surface feature information. The maternal body surface feature information includes abdominal surface pressure distribution data and body surface impedance values. It is also used to calculate the thickness of the pregnant woman's abdominal fat and the skin elasticity coefficient based on the body surface impedance values ​​and abdominal surface pressure distribution data. The adaptive compensation module, based on a fusion architecture of convolutional neural networks and long short-term memory networks, connects to the multimodal sensing module to analyze and process the signals collected by the multimodal sensing module. It then uses spectrum analysis and pattern recognition algorithms to classify the processed data for interference types, including maternal respiration, electromyographic interference, and environmental electromagnetic noise. It also provides a hybrid positioning technology combining inertial navigation and signal strength fingerprinting to track fetal position changes in the womb in real time. Combined with a deep learning model, it compensates for the fetal position signal, corrects signal deviations caused by fetal movement and maternal activity through predictive algorithms, and dynamically adjusts the ultrasound transmission power and time gain compensation curve based on abdominal fat thickness. The visualization module is used to construct a three-dimensional animated model of the fetal heart based on biological experimental data; and it is connected to the adaptive compensation module to receive data information from the adaptive compensation module, analyze and obtain the fetal heart rate curve, motion trend and positioning information, and visualize them in the three-dimensional animated model of the fetal heart.

2. The high-precision fetal heart monitoring device according to claim 1, characterized in that: In the multimodal sensing module, the ultrasonic sensor array uses phased array technology to capture the echo signal of the fetal heartbeat through several ultrasonic probes; the accelerometer is used to monitor the mother's movement and position changes in real time; and the bioelectric sensor is used to detect electromyography, electrocardiography and pressure distribution signals in the mother's abdomen.

3. The high-precision fetal heart monitoring device according to claim 2, characterized in that: In the visualization module, the fetal heart is located in three dimensions based on the time difference positioning algorithm. The abdominal surface contour is segmented in real time using the U-Net++ deep learning model to construct the three-dimensional coordinate mapping relationship between the fetal position and the body surface.

4. The high-precision fetal heart monitoring device according to claim 3, characterized in that: In the adaptive compensation module, the time-frequency domain features, statistical characteristics and correlation parameters of the signal are acquired in real time. An interference classification model is constructed through an adaptive learning algorithm to identify maternal motion artifacts, environmental electromagnetic noise, electromyographic interference and baseline drift, and output interference feature vectors. Based on the interference feature vector results, the filtering strategy is dynamically switched.

5. The high-precision fetal heart monitoring device according to claim 4, characterized in that: The filtering strategies include frequency-domain notch filtering for narrowband periodic interference, time-domain predictive filtering for motion-related artifacts, and time-frequency domain decomposition and reconstruction algorithms for baseline drift, to achieve layered noise reduction processing of fetal heart signals.

6. The high-precision fetal heart monitoring device according to claim 5, characterized in that: The specific process of the filtering strategy includes: when narrowband interference is detected, notch filtering is automatically activated to suppress interference of specific frequency components in the signal frequency domain; when maternal motion artifacts are identified, motion trajectory compensation is performed on the ultrasound blood flow signal or electrocardiogram signal by combining motion data and using a predictive filtering algorithm; when baseline drift is detected, the signal is decomposed into layers using a multi-resolution time-frequency analysis method, and feature enhancement and baseline correction are performed on the effective frequency band of the fetal heart signal.

7. The high-precision fetal heart monitoring device according to claim 6, characterized in that: The minimum number of decomposition layers is determined by using the Nyquist sampling theorem based on the characteristic frequency bands of the fetal electrocardiogram signal and the frequency characteristics of baseline drift.

8. The high-precision fetal heart monitoring device according to claim 7, characterized in that: The calculation process for abdominal fat thickness and skin elasticity coefficient in pregnant women includes: Step 1: Apply currents of different frequencies using a bioelectric sensor, measure the impedance values ​​between different locations, and construct a two-dimensional impedance spectrum; Step 2: Combine the abdominal surface pressure distribution data collected by the bioelectric sensor to establish a pressure-impedance coupling model, and solve the fat layer thickness using the finite element inversion algorithm. Step 3: Apply a step pressure to the same measurement point, record the pressure-deformation response curve, and obtain the skin elastic coefficient by identifying the viscoelastic model parameters; Step four: Input the fat layer thickness and skin elasticity coefficient into the adaptive compensation module to dynamically adjust the ultrasound emission power and electrode contact pressure.

9. The high-precision fetal heart monitoring device according to claim 8, characterized in that: The adaptive compensation module is used to dynamically adjust the time gain compensation curve of the ultrasound signal based on fat thickness. The gain increases exponentially in the depth direction. The contact pressure of the bioelectric sensor is adjusted according to the skin elasticity coefficient. Closed-loop feedback ensures that the contact impedance is stable within a preset threshold range. A multi-dimensional compensation model including fat thickness, skin elasticity coefficient and fetal position is constructed. The model parameters are optimized through deep learning algorithms to control the error rate of fetal heart rate monitoring for pregnant women of different body types within a uniform threshold. When the fat thickness exceeds the preset critical value, the enhancement mode is automatically activated, the ultrasound transmission power is increased and the phased array focusing depth is adjusted.

10. The high-precision fetal heart monitoring device according to claim 9, characterized in that: The visualization module is also used to realize local storage and remote wireless transmission of monitoring data, and supports connection with the hospital cloud platform, so that medical staff can view and analyze the monitoring data in real time.

Images