A device and system for early risk identification of fetal hypoxia based on macro-pulse phenotype

Abstract

This invention provides a device and system for early fetal hypoxia risk identification based on large pulse pressure phenotype, belonging to the fields of biomedical signal processing and heart sound signal analysis technology. The device includes: acquiring fetal heart sound signals and surface vibration signals from the pregnant woman's abdominal wall via a multi-channel sensor array; employing preprocessing techniques to improve the signal-to-noise ratio of the target signal and suppress interference signals; extracting the large pulse pressure phenotype feature set and heart sound temporal parameters within a sliding time window; determining the risk of fetal hypoxia based on joint criteria; outputting graded early warning information and supporting playback of risk event segments. This invention uses the actual mechanical vibration of the fetal heart sound as the information source, achieves early hypoxia warning based on the large pulse pressure phenotype, supports individualized baselines (including diurnal stratification) and gestational age adaptive correction, has strong anti-interference capabilities, can meet the needs of 24 / 7 continuous monitoring in hospitals / homes, and improves the sensitivity and specificity of hypoxia identification.

Description

Technical Field

[0001] This invention relates to the field of biomedical signal processing and heart sound signal analysis technology, and in particular to a device and system for early risk identification of fetal hypoxia based on large pulse pressure phenotype. Background Technology

[0002] Intrauterine fetal hypoxia is a common and serious perinatal complication, which can be caused by a variety of factors, including umbilical cord entanglement, compression, and placental insufficiency. If not detected and intervened in a timely manner, it can lead to intrauterine growth restriction, neonatal neurological damage, and even death. Therefore, early identification of fetal hypoxia has always been a core objective of obstetric monitoring.

[0003] Currently, the fetal heart rate monitoring technology widely used in clinical practice mainly relies on the ultrasound / Doppler principle. This technology emits high-frequency ultrasound waves (usually 1-3MHz) into the pregnant woman's abdominal wall and receives the echoes reflected back from the fetal heart and blood flow. By extracting the Doppler frequency shift signal, rhythmic indicators such as fetal heart rate (FHR) and heart rate variability (HRV) are calculated.

[0004] However, this mainstream technological approach has the following inherent drawbacks: (1) The information source is distorted and cannot reflect the true structure of the heart sound and the details of hemodynamics: The "fetal heart sound" output by the ultrasound / Doppler equipment is not the actual mechanical vibration sound wave of the fetal heart, but an artificial simulated heart sound after envelope extraction and rhythm synthesis of the Doppler frequency shift signal. Therefore, this technology cannot distinguish the first heart sound at all ( ) and second heart sound ( Furthermore, it cannot provide information on the contraction period ( , → ), diastolic phase ( , →Next ) and time comparison ( The device lacks crucial temporal structural information, such as heart rate, etc. This missing information is precisely the core indicator characterizing fetal cardiac mechanical activity, myocardial function, and hemodynamic status. Except for the heart rate, all other acoustic features of the simulated heart sounds output by the device are artificially synthesized, failing to objectively and realistically reflect the actual physiological state of the fetus, and may even be misleading.

[0005] (2) Insufficient sensitivity, easily missing short-term transient abnormal events: This technology uses the average heart rate after time averaging and macroscopic trends such as variable deceleration as the main basis for interpretation, and its observation window is usually long (e.g., 5-10 minutes). This processing method will "dilute" or "smooth out" short-term, transient physiological abnormal events (e.g., a sudden increase in pulsating energy lasting only a few seconds to a few minutes), resulting in low sensitivity to early hypoxia events. The risk of false negatives (missed reports) and false positives (false alarms) coexist, making it difficult to meet the dual requirements of sensitivity and specificity for early warning.

[0006] (3) Complex operation and difficulty in achieving continuous automated monitoring: Existing equipment often requires professional medical staff to hold the probe and repeatedly search for and maintain the best signal incident angle on the pregnant woman's abdominal wall, which is extremely sensitive to the coupling state. This operational dependence severely limits its application in home scenarios, and even in hospital environments, it is difficult to achieve continuous and automated monitoring without human intervention 24 hours a day, which makes it impossible to capture many transient hypoxic events.

[0007] In recent years, some studies have attempted to collect fetal heart sounds using passive acoustic sensors (such as microphones and accelerometers). The publicly available technologies primarily focus on sensor array design, signal noise reduction, heart sound segmentation, or fusion with other modalities (such as electrocardiograms). However, none of the existing publicly available technologies have yet addressed the integration of "heart sound time-course structure" (WMC). , The short-term coupling changes of “large pulse pressure” and “pulse envelope energy / pulse pressure proxy” are used as a specific physiological phenotype (i.e., “big pulse pressure phenotype”) and are systematically used as the core joint criterion for early risk of fetal hypoxia, forming a complete and engineered monitoring and early warning scheme.

[0008] Therefore, there is an urgent need in this field for a novel solution that can non-invasively, continuously, and automatically capture the mechanical vibration details of the fetal heart sounds and, based on its unique acoustic phenotype, achieve high sensitivity and high specificity in identifying early risks of fetal hypoxia. Summary of the Invention

[0009] To address this, embodiments of the present invention provide a device and system for early fetal hypoxia risk identification based on large pulse pressure phenotyping, which solves the problems in the prior art where the reliance on Doppler ultrasound to simulate heart sounds fails to capture the energy and temporal structure changes of real heart sounds, resulting in low sensitivity, high false alarm rate, and inability to continuously and automatically monitor early fetal hypoxia identification.

[0010] To address the aforementioned problems, embodiments of the present invention provide a device for early fetal hypoxia risk identification based on a large pulse pressure phenotype, the device comprising: Fetal heart sounds and body surface vibration signals were collected from the abdominal wall of pregnant women using a multi-channel sensor array. The fetal heart sound signal and body surface vibration signal are preprocessed to improve the signal-to-noise ratio of the target signal and suppress interference signals; Within a sliding time window, the large pulse pressure phenotypic feature set is extracted from the preprocessed signal, and the heart sound phase parameters are extracted. Based on the large pulse pressure phenotypic feature set and heart sound phase parameters, the risk of fetal hypoxia is determined by combining the preset joint criteria. Based on the risk assessment results, tiered early warning information is output.

[0011] Preferably, the large pulse pressure phenotypic feature set includes the relative increase in envelope. Pulse pressure proxy Low-frequency energy and the slope of energy rise ; wherein, the relative increase The pulse pressure surrogate is the increase in the peak value of the inner sound envelope relative to the individual baseline envelope value within a sliding time window. For the first heartbeat With the second heart sound The amplitude difference or amplitude ratio, the low-frequency energy The energy rise slope represents the spectral peak intensity in the low-frequency band of the signal envelope. for The first derivative with respect to time; The heart sound timing parameters include systolic duration. Duration of diastole timely comparison ,in The first heart sound within the same cardiac cycle To the second heart sound The time interval, The second heart sound of the current cardiac cycle The first heart sound until the next cardiac cycle The time interval, .

[0012] Preferably, the joint criteria include at least candidate feature exceeding the threshold, heart sound comparison with the individual baseline, determination result persistence, and multi-channel consistency verification; Specifically, the candidate feature exceeding the threshold refers to the relative increase in the envelope of the large pulse pressure phenotypic feature set. Pulse pressure proxy Low-frequency energy Simultaneously exceeding the preset adaptive threshold; The heart sound time-to-supreme baseline specifically refers to the time-to-supreme baseline calculated in real time. Compared with the preset individual steady-state baseline The ratio is greater than a preset threshold; The persistence of the determination result is specifically defined as follows: the duration of the state exceeding the individual baseline when the candidate feature exceeds the threshold and the heart sound is not less than a preset duration. The multi-channel consistency verification specifically means that the proportion of channels in the multi-channel sensor array that meet the judgment conditions is not less than a preset proportion.

[0013] Preferably, the joint criterion also includes the magnitude of the heart rate decrease. Verification; the magnitude of the heart rate decrease The verification is as follows: the decrease in fetal heart rate within the sliding time window is not greater than the preset negative threshold.

[0014] Preferably, the preprocessing includes at least one of the following: bandpass filtering, notch filtering, adaptive beamforming, blind source separation, sparse representation reconstruction, and adaptive filtering; wherein, the adaptive beamforming includes a minimum variance distortionless response algorithm or a generalized sidelobe cancellation algorithm, the blind source separation includes an independent component analysis algorithm, and the adaptive filtering includes a minimum mean square algorithm or a recursive least squares algorithm.

[0015] Preferably, the device further includes establishing an individual steady-state baseline: when the pregnant woman is at rest and without significant body movement or uterine contractions, signals of a preset duration are collected, and the baseline values ​​of the large pulse pressure phenotypic feature set and the heart sound phase parameters are calculated to form an individual steady-state baseline; the individual steady-state baseline also includes a diurnal stratified baseline, corresponding to the signal baseline states during the day and night respectively; and the preset threshold in the joint criterion is adaptively corrected based on gestational age, and the threshold interpolation adjustment is achieved by fitting the correlation curve between gestational age and the baseline values ​​of the heart sound phase parameters.

[0016] Preferably, the device further includes anti-interference control: when body motion interference is detected, the determination of large pulse pressure phenotypic features is frozen, and only the heart sound phase parameters are tracked; when a uterine contraction signal is detected, a uterine contraction-related label is added to the risk warning information; when maternal heart sound interference exists, the same frequency component of maternal heart sounds is removed by spatial difference or blind source separation algorithm.

[0017] This invention also provides a fetal hypoxia early risk identification system based on large pulse pressure phenotype, comprising: The signal acquisition module is used to acquire fetal heart sound signals and body surface vibration signals at the pregnant woman's abdominal wall through a multi-channel sensor array; The signal preprocessing module is used to preprocess the fetal heart sound signal and body surface vibration signal to improve the signal-to-noise ratio of the target signal and suppress interference signals. The feature extraction module is used to extract the large pulse pressure phenotypic feature set from the preprocessed signal within a sliding time window, and to extract the heart sound phase parameters. The criterion calculation module is used to determine the risk of fetal hypoxia based on the large pulse pressure phenotypic feature set and heart sound phase parameters, combined with a preset joint criterion. The early warning output module is used to output graded early warning information based on the risk assessment results.

[0018] Preferably, the signal acquisition module includes a multi-channel sensor array and a wearable carrier; the multi-channel sensor array includes at least two sensors, which are at least one of MEMS microphones, piezoelectric sensors, or accelerometers; the array type of the multi-channel sensor array is a linear array or a circular array, and the spacing between the sensors in the array is a preset spacing; the wearable carrier is a flexible abdominal band, and the inner side of the flexible abdominal band is provided with a medical coupling layer and a shielding damping structure.

[0019] Preferably, the early warning output module includes a user terminal and a medical terminal; the user terminal is used to display fetal heart rate, heart sound signal waveform, heart sound phase parameter trend and graded early warning information in real time, and to display a gating status prompt when interference is detected; the medical terminal is used to display detailed data of risk event timeline, large pulse pressure phenotype characteristics and heart sound phase parameters, support the playback of signal segments of preset duration before and after the risk event, and provide data export function for verification.

[0020] As can be seen from the above technical solutions, this invention application has the following beneficial effects: (1) High fidelity of information source: Breaking through the limitations of traditional Doppler ultrasound "simulated heart sounds", it directly collects the real mechanical vibration signal of the fetal heart, and can clearly distinguish the first heart sound. With the second heart sound and extract The timing parameters fully reflect the changes in fetal hemodynamics and details of myocardial function, avoiding the loss and misleading of physiological information caused by artificially synthesized signals; (2) High sensitivity of early identification: Based on the large pulse pressure phenotype that appears transiently after hypoxia (which occurs before the heart rate mean abnormality and lasts for several seconds to several minutes), a multi-feature joint criterion is constructed, which can accurately capture short-term abnormalities that are easily "diluted" by the traditional time averaging method, significantly reduce the false negative and false positive rates, and gain clinical golden time for early intervention of fetal hypoxia. (3) Supports continuous automated monitoring: It adopts a flexible wearable abdominal belt design, combined with anti-interference strategies such as body movement / uterine contraction gating and multi-channel consistency verification, which can be adapted to 24 / 7 unattended monitoring in hospital / home scenarios; at the same time, it establishes an individual baseline (including day and night stratified baseline) based on the pregnant woman's resting home state, and determines the threshold through gestational age adaptive correction, effectively overcoming monitoring deviations caused by individual differences and fetal development, and improving the robustness of monitoring in different scenarios. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly described below. Referring to the drawings will make the features and advantages of the present invention clearer. The drawings are illustrative and should not be construed as limiting the present invention in any way. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 A flowchart of an early risk identification device for fetal hypoxia based on a large pulse pressure phenotype provided by the present invention; Figure 2 This is a schematic diagram of the Big Pulse in this invention; Figure 3 This is a schematic diagram of the overall architecture of the device of the present invention; Figure 4 This invention provides a block diagram of an early risk identification system for fetal hypoxia based on a large pulse pressure phenotype. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Example 1: To address the problems in existing technologies where reliance on Doppler ultrasound to simulate heart sounds fails to capture the energy and temporal structural changes of real heart sounds, resulting in low sensitivity, high false alarm rate, and inability to provide continuous automatic monitoring in the early detection of fetal hypoxia. For example... Figure 1 As shown, this invention proposes a device for early fetal hypoxia risk identification based on a large pulse pressure phenotype. The device includes: S1: Collect fetal heart sound signals and body surface vibration signals at the pregnant woman's abdominal wall using a multi-channel sensor array; S2: Preprocess the fetal heart sound signal and body surface vibration signal to improve the signal-to-noise ratio of the target signal and suppress interference signals; S3: Within the sliding time window, extract the large pulse pressure phenotypic feature set from the preprocessed signal and extract the heart sound phase parameters; S4: Based on the large pulse pressure phenotypic feature set and heart sound phase parameters, the risk of fetal hypoxia is determined by combining the preset joint criteria. S5: Output graded early warning information based on the risk assessment results.

[0024] As can be seen from the above technical solution, this invention proposes an early fetal hypoxia risk identification device based on the large pulse pressure phenotype. Step S1 involves a multi-channel sensor array acquiring fetal heart sound signals and body surface vibration signals from the pregnant woman's abdominal wall to obtain real mechanical vibration information of the heart sound, avoiding the information distortion problem of Doppler ultrasound simulation of heart sounds. Step S2 involves preprocessing the acquired signals to improve the signal-to-noise ratio of the target signal and suppress interference from maternal heart sounds, body movements, and environmental noise. Step S3 involves extracting the large pulse pressure phenotype feature set and heart sound temporal parameters within a sliding time window to construct the core feature basis for hypoxia risk determination. Step S4 involves risk determination based on the above features and parameters combined with preset joint criteria to improve the sensitivity of early fetal hypoxia identification and reduce the false alarm rate. Step S5 involves outputting graded early warning information based on the determination results to clarify the risk level and provide a basis for timely intervention. Ultimately, this solves the problems of low sensitivity, high false alarm rate, and inability to continuously and automatically monitor early fetal hypoxia in existing technologies.

[0025] To better understand the technical solution of this invention, the medical principle of the Big Pulse phenotype in this invention will first be explained: The Big Pulse phenotype is a specific transient response of the cardiovascular system during acute intrauterine hypoxia in the fetus, providing a new fetal physiological entry point for the early identification of intrauterine hypoxia. Its medical principle can be explained from three aspects: phenotype discovery, hemodynamics and physiological mechanisms, and correlation with fetal heart sound structure. 1. Phenotypic Discovery and Core Features Long-term systematic research (based on China's first dynamic fetal cardiovascular function detection model) and double-blind verification have confirmed that within a specific time window following an acute intrauterine hypoxic event (such as umbilical cord compression or fetal nuchal cord), the fetal cardiovascular system exhibits a transient phenotype with significantly increased pulsation amplitude and energy, termed "Big Pulse." Figure 2 As shown.

[0026] This phenotype has three significant characteristics: ① It is short-lived, lasting only a few seconds to a few minutes; ② It has a high incidence rate, occurring in more than 80% of large-scale intrauterine hypoxia sample studies; ③ It is accompanied by a sudden drop in heart rate, often occurring synchronously with a brief drop in heart rate to a specific level, suggesting a highly coupled stress regulation mechanism between chemoreceptors, autonomic nervous system, and peripheral resistance.

[0027] Due to its transient nature, traditional monitoring methods based on "time averaging" (such as the average trend of fetal heart rate per unit time) are unable to capture such transient characteristics, which can easily lead to missed detections. The explicit proposal and naming of this phenotype has opened up a new path for the early identification of intrauterine hypoxia.

[0028] 2. Fetal hemodynamics and physiological mechanisms Under acute hypoxic stress, the fetus's physiological response triggers specific hemodynamic changes, ultimately leading to a large pulse pressure phenotype: First, hypoxia activates fetal chemoreceptors, triggering excitation of the sympathetic nervous system, causing constriction of peripheral arterioles, leading to increased peripheral resistance and decreased arterial compliance. At the same time, myocardial contractility increases rapidly, resulting in a significant increase in the amplitude of arterial pressure (PP) and a steeper slope of the ascending limb of the pulsation (i.e., a faster rate of increase in arterial pressure).

[0029] This "high-amplitude pulsation" is transmitted to the pregnant woman's body surface through the vibration of the ventricular wall and valves in the form of mechanical vibration. Finally, it forms an observable envelope with a significant increase in pressure in the heart sound (Phonocardiography, PCG) signal and body surface vibration (such as acceleration and piezoelectric signals)—that is, the "big pulse segment" that reflects intrauterine hypoxia.

[0030] At the same time, the hypoxia reflex pathway activates the vagus nerve, causing a brief drop in heart rate; and the contraction phase of the fetal cardiac cycle ( The first heart sound within the same cardiac cycle. To the second heart sound The time interval between diastole and diastole ( The second heart sound in the current cardiac cycle. The first heart sound until the next cardiac cycle The time interval will increase synchronously, but The growth rate usually exceeds This leads to a comparison with the time It is significantly higher than the individual's normal baseline.

[0031] In summary, under hypoxic stress, the fetus will develop "pulsating energy jump (instantaneous high-amplitude pulsation of the envelope signal) + abnormal temporal linkage ( , Simultaneous rise and The triple physiological coupling characteristics of "extra-individual baseline) + sudden drop in heart rate (fetal heart rate drops to a specific level in a short period of time)" constitute the specific physiological triad of hypoxic stress, which is also the core basis of the detection mechanism of this invention.

[0032] 3. Structure of fetal heart sounds ( , The correlation and discriminative value of ) This invention is the first to define "duration of the systolic phase". "Duration of diastole" The concept of "fetal heart sound analysis" is systematically introduced, and its diagnostic value in hypoxic conditions is clarified. , The temporal division of the heart originates from the acoustic mapping of the fetal cardiac mechanical processes. It has a clear fetal physiological basis and can be observed through engineering means (such as heart sound signal processing).

[0033] To improve the accuracy of the time-phase index, the study used the individual's own normal steady state as the control baseline (rather than relying on the population average), which can better cope with the interference of noise factors such as individual differences and sensor attachment changes.

[0034] Actual measurement data show that during the large pulse pressure phase or the acute hypoxia induction period, and Although all will rise, the magnitude of the rise will not be synchronized. The growth rate significantly exceeded This "disproportionate increase" will make the time ratio... Compared to individual baseline values A significant increase, its relative increment can be expressed as: This can further enhance the accuracy of identifying abnormal states.

[0035] Based on this, compared to (or its relative increment) The joint criteria introduced into the early warning system are related to "the increase in the amplitude of the heart sound envelope". "Pulse pressure proxy volume" The synergistic effect of pulsatile energy characteristics has become a key linkage characteristic for confirming hypoxia risk events from "candidate events of large pulse pressure".

[0036] Furthermore, this device is based on the medical principle of the aforementioned Big Pulse phenotype—after acute hypoxia is triggered, the fetal cardiovascular system exhibits a triple physiological coupling characteristic of "pulsating energy jump, abnormal temporal linkage, and sudden drop in heart rate," which precedes the appearance of abnormal mean heart rate. The device implementation process follows the logic of "signal acquisition → preprocessing → feature extraction → risk assessment → early warning output," and the overall architecture is as follows: Figure 3 (Sensor array → Fetal heart signal preprocessing →) Detection → / / The process (extraction → joint discrimination of large pulse pressure events → terminal display) is shown below, and it also incorporates individual baseline establishment and anti-interference control, as detailed below: In step S1, multi-channel signal acquisition: fetal heart sound signals (Phonocardiography, PCG) and body surface vibration signals at the pregnant woman's abdominal wall are acquired through a multi-channel sensor array. The core is to obtain high-fidelity real fetal heart mechanical vibration information, avoiding the information loss problem of traditional Doppler "simulated heart sounds".

[0037] Sensor array configuration: A multi-channel array is composed of MEMS microphones, piezoelectric films or triaxial accelerometers. The channel configuration is "M channels of microphones (4~8 channels) + K channels of accelerometers (3~6 channels)". The microphones are responsible for collecting fetal heart sound signals, and the accelerometers collect body surface vibration signals to help suppress body movement interference. The frequency response of all sensors covers the main frequency band of fetal heart signals (20~200Hz) and the sensitivity is ≥-40dBFS to ensure the effective capture of weak fetal heart signals.

[0038] Array type and layout: Linear array: suitable for short-term monitoring with small changes in fetal heart position (such as a single examination in the hospital). The sensors are evenly arranged along a straight line with a spacing of 5~15cm (preferably 8cm) to enhance the fetal heart signal in a fixed direction through directionality. Circular array: Suitable for long-term monitoring of dynamic changes in fetal heart position (such as home 7×24-hour monitoring). The sensors are arranged in a circle (10~15cm in diameter) with a center distance of 40mm, which can capture fetal heart signals in 360° and avoid signal loss caused by fetal turning over.

[0039] Adhesion and channel selection: Adhesion site: Following clinical guidelines, under the guidance of a doctor, the array is arranged in the "area where the fetal heart can be heard" on the pregnant woman's abdominal wall (usually 2-4 cm below the navel, with slight adjustment when in the lateral position), allowing for slight rotation / displacement of ±5°; Optimal channel determination: After the system starts, it first collects the initial signal for 30 seconds, calculates the "envelope signal-to-noise ratio (eSNR)" and "acceleration-heart sound signal coherence" of each channel, selects the top 2 to 4 channels (preferably 3 channels) as the "main determination set", and uses the remaining channels as backups to ensure the reliability of subsequent feature extraction.

[0040] In step S2, signal preprocessing is performed: the goal of preprocessing is to improve the signal-to-noise ratio of the fetal heart signal and suppress "array target-end interference" (maternal heart sounds, amniotic fluid fluctuations, intestinal peristalsis, body movement, fetal movement) and "external environmental interference" (household appliance noise, traffic noise, electronic power frequency interference (50Hz)). Specifically, a multi-dimensional processing strategy is adopted: 1) Basic filtering: Bandpass filtering: FIR filter is used, with a passband of 20~200Hz (covering the main frequency band of fetal heart signal) and a stopband attenuation of ≥40dB, to filter out low-frequency noise of body movement below 20Hz and high-frequency noise of environment above 200Hz; Notch filtering: IIRnotch filter is used with a center frequency of 50Hz (power frequency interference) and a 3dB bandwidth of 2Hz to suppress periodic interference introduced by the power grid.

[0041] 2) Adaptive Beamforming: For spatially distributed interference (such as maternal heart sounds and lateral environmental noise), the Minimum Variance Distortionless Response (MVDR) algorithm or the Generalized Sidelobe Cancellation (GSC) algorithm is employed. MVDR Implementation: Constructing a Multi-Channel Signal Covariance Matrix ( (For the original signal vector), solve for the weight vector. ( (Guiding vector for the direction of fetal heart incidence), output enhanced signal. Improved signal-to-noise ratio; GSC switching: When there is strong sidelobe interference (such as on the side near the maternal heart), the sidelobe signal is canceled by the blocking matrix to further optimize the anti-interference effect.

[0042] 3) Identification by blind source (ICA): Targeting strong periodic interference from maternal heart sounds. The beam output signal is whitened (PCA is reduced to 2D to remove signal correlation). The FastICA algorithm is used to separate the "fetal heart signal source" and the "maternal heart sound source" with the objective function of "maximizing non-Gaussianity". Remove the maternal heart sound source and retain the fetal heart signal source.

[0043] 4) Sparse representation reconstruction (OMP): Addressing the impact noise generated by fetal movement: Constructing a dedicated dictionary of fetal heart sounds (trained based on 1000 normal fetal heart sound samples, including...) Feature primitives); Sparse decomposition of noisy signals ( It is a signal to be processed. For a dictionary, (where is the sparsity coefficient). pass ( yes norm ， Used to control sparsity (The regularization parameter) is used to solve for the sparse coefficients, discard the small coefficients corresponding to noise, and reconstruct the pure fetal heart signal.

[0044] 5) Adaptive Filtering: Least Mean Square (LMS) Algorithm: For stationary noise such as amniotic fluid fluctuations and intestinal peristalsis, the filter coefficients are updated in real time using the acceleration signal as the reference noise. ( (Step size) Recursive Least Squares (RLS) Algorithm: For non-stationary noise such as fetal kicking and pregnant woman turning over, it quickly tracks noise changes by using the exponentially weighted least squares criterion.

[0045] 6) Body Movement / Contraction Gating: Body Movement Gating: Calculate the root mean square (ACC-RMS) of the acceleration signal. If ACC-RMS > ( =Baseline ACC - RMS + 3 , (where the baseline standard deviation is used), then freeze. , The determination of energy-related characteristics only involves tracking Phase parameters; Contraction gating: If a contraction sensor (such as a TOCO tension sensor) is connected, when the rising edge of the TOCO signal exceeds the clinical contraction threshold, the system adds a "contraction-related" label. This does not change the core judgment threshold, but clearly indicates the correlation between the risk and the contraction.

[0046] In step S3, the large pulse pressure phenotypic features and heart sound phase parameters are extracted: within a sliding time window. Within this process, the large pulse pressure phenotypic feature set and heart sound phase parameters are extracted from the preprocessed signal. All parameter calculations are based on the individual's steady-state baseline. 1. Set of characteristics of large pulse pressure phenotype: ① Relative increase in envelope : First, calculate the Hilbert envelope of the preprocessed signal. ( (for Hilbert transform) Extract the envelope peak value of each heart sound within the sliding window. ; calculate ,in The median of the envelope of the individual baseline.

[0047] ②Pulse pressure proxy : First identify the first heart sound ( The systolic phase begins, with large amplitude and long duration, and is associated with the second heart sound (…). (Diastolic phase begins, with small amplitude and short duration): Combining wavelet packet decomposition and peak-valley tracking algorithms, annotation is performed. , Time point , ; calculate envelope amplitude , ,in For the first heartbeat The envelope amplitude or energy, The second heart sound The envelope amplitude or energy; Using the form of "amplitude difference" Or in the form of "amplitude ratio" Simulating the clinical concept of "pulse pressure," the large pulse pressure segment It is more than 3σ higher than the baseline.

[0048] ③ Low-frequency energy : Perform Fourier transform on the envelope and extract the spectral peak amplitude in the low-frequency band (corresponding to the slow fluctuation of fetal heart beat).

[0049] ④ Energy rising slope : The central difference method is used to calculate the first derivative, to characterize the steepness of energy climbing.

[0050] 2. Heart sound phase parameters: ① Systolic duration : In the same cardiac cycle, the time interval from the first heart sound to the second heart sound , .

[0051] ② Diastolic duration : the time interval from the second heart sound of the current cardiac cycle to the first heart sound of the next cardiac cycle , ( is the th time point, is the th time point).

[0052] ③ Phase ratio : .

[0053] 3. Fetal heart rate (FHR): converted from the adjacent time interval, and the large pulse pressure segment is often accompanied by a sudden drop in FHR.

[0054] 4. Quality index (QI) screening: QI is calculated for each heart sound (comprehensively including peak shape, peak spacing consistency, and spectral centroid drift). Low QI segments do not participate in feature extraction to avoid feature distortion caused by noise.

[0055] In step S4, fetal hypoxia risk determination: based on the large pulse pressure phenotypic feature set and heart sound phase parameters, the present invention adopts a joint determination mechanism of "logic criterion + scoring function + multi-dimensional verification" to ensure the sensitivity and specificity of the determination: The joint criterion at least includes candidate feature over threshold, heart sound phase ratio over individual baseline, persistence of determination result and multi-channel consistency verification; wherein the candidate feature over threshold is specifically: the relative envelope增幅 in the large pulse pressure phenotypic feature set , pulse pressure proxy , low-frequency energy Simultaneously exceeding the preset adaptive threshold; heart sound time-to-time ratio compared to the individual baseline specifically refers to the real-time calculated time-to-time ratio. Compared with the preset individual steady-state baseline The ratio is greater than the preset threshold; the determination result persistence is specifically defined as: the duration of the state exceeding the individual baseline when the candidate feature exceeds the threshold compared with the heart sound is not less than the preset duration; the multi-channel consistency verification is specifically defined as: the proportion of channels in the multi-channel sensor array that meet the determination conditions is not less than the preset proportion.

[0056] Furthermore, the joint criteria also include the optional magnitude of the heart rate drop. verify; The verification is as follows: the decrease in fetal heart rate within the sliding time window is not greater than the preset negative threshold.

[0057] Specifically, the core components of the joint criteria are: (1) Candidate feature exceeds threshold: , , ,in: 80%-120% (80% for 20-30 weeks of gestation, 120% for 30-40 weeks of gestation); : ( Baseline (standard deviation) 5 to 8 times the baseline (5x for resting scenarios, 8x for contraction scenarios).

[0058] (2) Heart sounds at time compared to individual baseline: real time Compared with individual baseline ratio .

[0059] (3) Result persistence: satisfying "candidate feature exceeds threshold + Duration of "beyond baseline" state ≥ (Home resting scenario) =60s, intra-hospital contraction scene =90s, strong interference scenario =45s).

[0060] (4) Multi-channel consistency verification: the percentage of channels in the main decision set that meet the decision conditions. ≥60% = (Number of channels in the main decision set that meet the conditions × 100%).

[0061] (5) Optional: Verification of the magnitude of heart rate decrease (ΔHR≤ (The smaller the value, the earlier the gestational age).

[0062] Logical indicator function: When the above conditions are met simultaneously, the indicator function outputs 1 (determined as a "large pulse pressure candidate event"): ; Comprehensive scoring function (secondary judgment of boundary cases): For cases where a single feature slightly exceeds the threshold, a scoring function is used: ; in, For the scoring function, , , , These are feature weight coefficients, which can be set by clinical priors or data learning.

[0063] In step S5, graded early warning information is output based on the risk assessment results: according to the degree of satisfaction of the joint criteria, three levels of early warning are output: High / Medium / Low. At the same time, event playback and data logging functions are provided, and differentiated early warning and recording strategies are implemented on user terminals, medical staff terminals, and data storage links. I. High Risk 1. Judgment conditions (must be met simultaneously): ① The logical indicator function returns "1" (meaning that the candidate feature exceeds the threshold, the heart sound exceeds the individual baseline, and the heart rate decrease meets the verification criteria). ② The duration of the state that satisfies the core criteria is greater than or equal to the preset duration. ; ③ In a multi-channel sensor array, the percentage of channels that meet the judgment criteria. .

[0064] 2. Early warning and data processing: User terminal: A red warning pop-up window appears, and a voice prompt plays: "High risk, please contact a doctor immediately"; Medical terminal: Synchronously push event notifications, along with signal segments for a period of time before and after the event (including original waveforms, large pulse pressure phenotypic characteristics, and heart sound phase parameter change curves). Data caching: Automatically saves data for the event period and adds a timestamp for easy backtracking and analysis.

[0065] II. Medium Risk 1. Judgment conditions (must be met simultaneously): ① The logic indicator function returns "1" (but does not include verification of the heart rate drop, i.e., the core features and heart sound phase abnormalities have been triggered, but the heart rate change has not met the additional verification requirements). ② The duration of the state that satisfies the core criteria is greater than or equal to the preset duration. ; ③ In a multi-channel sensor array, the percentage of channels that meet the judgment criteria. ; 2. Early warning and data processing: User terminal: A yellow warning pop-up window appears, along with a voice prompt saying "Medium risk, it is recommended to change position and continue monitoring"; Healthcare terminal: Mark the event as a "suspicious event" for further review by the doctor; Additional prompt: If the risk condition does not improve within 30 minutes, the user is prompted to contact medical personnel.

[0066] III. Low Risk 1. Judgment conditions (meeting any one of them is sufficient): ① The logic indicator function returns "0" (the core feature or heart sound phase does not meet the abnormal criteria); ② The duration of the state that satisfies the core criteria is less than the preset duration. ; ③ In a multi-channel sensor array, the percentage of channels that meet the judgment criteria. ; ④ The system is in the gating period of body movement interference or uterine contraction signals (due to body movement and uterine contraction interference, the large pulse pressure phenotype characteristic determination is temporarily not performed).

[0067] 2. Early warning and data processing: User terminal: No warning pop-up window appears; only the fetal heart rate (FHR) and heart sound phase parameters are displayed normally. , , The trend curve of )); if it is in the gating period, mark "Interference, monitoring paused"; Data recording: Only routine fetal heart rate trend data is retained; no special event segments are saved.

[0068] Furthermore, to overcome the influence of individual differences (gestational age, BMI, fetal position) on the determination, a steady-state baseline of "individualized, diurnal stratification, and gestational age adaptive" is established: (1) Baseline acquisition: Acquisition time: When the system is used for the first time, the pregnant woman should lie flat for 30 minutes (resting state, without body movement or uterine contractions). Data collection time: 5~15 minutes (10 minutes preferred), covering ≥150 beats of heartbeats; Baseline calculation: calculation (Median of envelope) ( (median value) ( (median value) ( The system stores the median and FHR0 (FHR median), and stores the interquartile range (IQR) for robust statistics. The robust z-score method is used to tune the threshold: z = (x - median) / IQR (median is the median, IQR is the interquartile range).

[0069] (2) Day and night stratification baseline: Day and night are divided into daytime (6:00~22:00) and nighttime (22:00~6:00), based on the pregnant woman's daily routine; Layered acquisition: Collect 10 minutes of resting signal during the day and 10 minutes at night to establish a "diurnal stratified baseline" to avoid misjudgment caused by diurnal rhythm (such as lower FHR at night).

[0070] (3) Gestational age adaptive correction: Fitting correlation curve: Based on baseline data of 1000 pregnant women at different gestational ages, the baseline parameters were fitted to the correlation curve with gestational age. Correlation curve: Threshold interpolation: The system uses the current gestational week as a reference. The corrected baseline parameters are obtained through the above curve interpolation, and the judgment threshold is then optimized based on the ROC-Youden criterion.

[0071] Furthermore, this invention also designs failure modes and recovery strategies: Long-term strong interference: Trigger "Observation Mode": Pause risk assessment, only display FHR and QI trends, and prompt the user "Please remain silent"; Recovery logic: When the duration of QI≥0.8 is ≥3 minutes, automatically exit observation mode, re-establish the baseline and resume assessment.

[0072] Sensor patch displacement / detachment: Detection logic: When the QI of all channels is less than 0.4 for 10 seconds, or the ACC-RMS does not fluctuate significantly (determined as detachment), trigger the "pattern abnormality" alarm; Recovery logic: prompt the user to reattach the sensor, and after reattachment, automatically execute "optimal channel selection" (calculate the eSNR of each channel, select the first 2-4 channels as the main judgment channels), and quickly collect 2 minutes of data to correct the baseline.

[0073] Communication interruption (e.g., Wi-Fi / BLE disconnection): Local caching: During the communication interruption, data is stored in a local circular cache with a maximum cache capacity of 12 hours; Automatic reconnection: Attempts to reconnect every 30 seconds, and upon successful connection, prioritizes uploading risk event data from the interruption period to ensure no data loss.

[0074] Example 2: Figure 4As shown, this invention provides a fetal hypoxia early risk identification system based on large pulse pressure phenotype. This system is used to implement the steps in the fetal hypoxia early risk identification device based on large pulse pressure phenotype in Embodiment 1 above, specifically including: The signal acquisition module 100 is used to acquire fetal heart sound signals and body surface vibration signals at the abdominal wall of the pregnant woman through a multi-channel sensor array. The signal preprocessing module 200 is used to preprocess the fetal heart sound signal and body surface vibration signal to improve the signal-to-noise ratio of the target signal and suppress interference signals. The feature extraction module 300 is used to extract the large pulse pressure phenotypic feature set from the preprocessed signal within a sliding time window, and to extract the heart sound phase parameters. The criterion calculation module 400 is used to determine the risk of fetal hypoxia based on the large pulse pressure phenotypic feature set and heart sound phase parameters, combined with preset joint criteria. The early warning output module 500 is used to output graded early warning information based on the risk assessment results.

[0075] Furthermore, the signal acquisition module 100 serves as the system's signal input, responsible for acquiring fetal heart sounds and surface vibration signals. Its core consists of a multi-channel sensor array and a wearable flexible abdominal binder. The multi-channel sensor array includes: sensor types: MEMS microphones (4~8 channels, sampling rate ≥48kHz, frequency response 20~200Hz), triaxial accelerometers (3~6 channels, range ±2g, sampling rate ≥1kHz), or optional piezoelectric film sensors (adapted to different attachment scenarios). Array structure: linear array (short-term monitoring, spacing 5~15cm), circular array (long-term monitoring, diameter 10~15cm). The sensor adopts a stethoscope-like structure with a built-in damping layer to enhance mechanical vibration coupling and reduce airborne noise. Channel synchronization: Each acquisition channel is driven by a unified clock (accuracy ≤1μs) to ensure the delay accuracy requirements of array beamforming and avoid signal distortion caused by phase difference between channels.

[0076] The wearable flexible abdominal binder includes: Material and structure: Medical-grade elastic fabric (elongation ≥150%), suitable for pregnant women with a waist circumference of 60~120cm; an inner 5mm thick medical hydrogel coupling layer (acoustic impedance matching human tissue, enhancing signal transmission); and an outer copper mesh shielding layer (suppressing 50Hz power frequency and high-frequency electronic interference). Fixation and Operation: The sensor can be detachably fixed with Velcro and supports ±5° micro-adjustment; the surface of the abdominal binder is printed with "attachment point markings" (such as a positioning line 2cm below the navel) to facilitate the pregnant woman's independent operation; Safety design: Made of medical-grade environmentally friendly materials, non-irritating to the skin; operating voltage 3.7V, current ≤50mA, no risk of electromagnetic radiation.

[0077] Furthermore, the signal preprocessing module 200, as the signal purification center of the system, is responsible for executing the preprocessing algorithm. Its core consists of a hardware processing unit and algorithm firmware. The hardware processing unit includes: core chip: using a system-on-a-chip (SoC) or digital signal processor (DSP), such as TITMS320C6748 (supporting floating-point operations to meet real-time beamforming requirements), or optional FPGA (such as Xilinx Artix-7, adapted for multi-channel parallel processing). Front-end circuitry: Low-noise amplifier (LNA, gain ≥40dB, noise figure ≤1.5dB), anti-aliasing filter (cutoff frequency 200Hz) to ensure effective amplification and filtering of weak signals; Power management: Employs low-power power chips (such as TITPS61230) to support battery power (≥8 hours of battery life, suitable for home mobile monitoring) or external power supply (long-term monitoring in hospitals).

[0078] The algorithm firmware includes: Firmware architecture: with a task queue as the core, it processes signals frame by frame (frame length 100-200ms, 50% overlap), with end-to-end latency ≤300ms, meeting the requirements of real-time monitoring; Algorithm integration: Integrates firmware for algorithms such as bandpass / notch filtering, MVDR / GSC beamforming, ICA blind source separation, OMP sparse representation, LMS / RLS adaptive filtering, and body movement / uterine contraction gating, and supports online parameter adjustment (such as dynamic threshold updates).

[0079] Furthermore, the feature extraction module 300 is used to extract large pulse pressure phenotypic features and heart sound phase parameters, and consists of a feature calculation unit and a quality assessment unit: The feature calculation unit includes the following functions: Real-time calculation: Based on the preprocessed signal, perform real-time calculation. , , , , , , Parameters such as FHR are updated once per frame; Data storage: A circular cache (capacity ≥ 10 minutes) is used to store feature sequences, supporting backtracking analysis (such as comparing features before and after an event).

[0080] The quality assessment unit includes the following functions: QI calculation: calculates the QI value of each heart sound in real time and marks low-quality segments (QI<0.6). Feature filtering: Remove feature values ​​of low quality segments and retain only high quality features for risk assessment to ensure feature reliability.

[0081] Furthermore, the risk assessment module 400 is used for joint criteria and consists of a criteria calculation unit and a baseline management unit: The criterion calculation unit includes the following functions: Real-time criterion: Based on the high-quality features output by the feature extraction module, calculate in real time whether candidate features exceed the threshold. Criteria such as exceeding baseline, persistence, and multi-channel consistency are used to output the judgment result (candidate event / valid event / invalid event). Scoring function: Calculated for boundary cases Value, used for secondary determination of event validity; Risk Classification: Outputs High / Medium / Low risk levels based on the assessment results, triggering an early warning signal.

[0082] Baseline Management Unit: Includes the following functions: Baseline Storage: Stores individual basic baselines and day-night stratified baselines using the pregnant woman's ID and gestational week as the index, and supports baseline query and update; Gestational age correction: Based on the current gestational age, the gestational age correlation curve is automatically invoked to achieve adaptive correction between the baseline and the judgment threshold; Baseline updates: Supports periodic (e.g., every 7 days) baseline re-collection to adapt to baseline changes caused by fetal growth and development.

[0083] Furthermore, the early warning output module 500 is used for tiered early warning. Its core consists of a user terminal, a medical staff terminal, and a data transmission unit, enabling interactive functions such as real-time viewing by users and remote review by medical staff. (1) The data transmission unit includes: short-range transmission: using BLE5.0 (Bluetooth Low Energy) to transmit lightweight data such as FHR, risk level, and characteristic values ​​(rate ≥2Mbps), suitable for home scenarios; Long-distance transmission: Uses Wi-Fi 802.11n or 4G / 5G (optional) to transmit event fragments, raw data, and other batch data at a rate of ≥10Mbps; Secure transmission: Transmission uses TLS 1.3 end-to-end encryption, and data storage uses AES-256 encryption; message verification is supported to avoid data transmission errors.

[0084] (2) User terminal (pregnant woman terminal) includes: hardware form: mobile APP (compatible with iOS / Android system) or dedicated portable display (touch screen); Real-time display: Main area: Fetal heart rate waveform (labeled) / ), FHR values ​​(green: 110-160 bpm, red: <110 bpm or >160 bpm); auxiliary area: / / Trend bar (comparison of real-time value and baseline value, highlighted in red if exceeding the baseline), QI value (green ≥ 0.6, yellow 0.4-0.6, red < 0.4); Warning zones: High / Medium risk will display a red / yellow pop-up window with a voice prompt (such as "High risk, please contact a doctor"); Low risk will not display a pop-up window. Historical data: Displays FHR trends and risk event records by day (timestamp + level + triggering conditions), and supports exporting to PDF (for doctors to view during prenatal checkups).

[0085] (3) Medical care terminals (hospital end) include: hardware form: hospital cloud platform (web end) or doctor-specific APP; Event View: Timeline: Displays risk events for pregnant women in chronological order (High / Medium events are marked with red / yellow dots), and supports filtering by gestational week and risk level; Event Details: Click on the event to view signal segments (original waveform + ...) within a certain period before and after the event. / / FHR characteristic curve), judgment criteria, duration, and gestational age correction parameters; Verification function: Supports doctors to annotate confirmed hypoxia, suspected hypoxia, and false alarms, and the annotation results are synchronized to the user terminal; the anonymized annotated data is used for iterative optimization of the system threshold (included in the algorithm training set). Data export: Supports exporting monitoring reports (including feature tables, event records, and review results) for individual pregnant women or statistical data for batches of pregnant women (for clinical research).

[0086] Furthermore, the workflow of the system of the present invention includes the following steps: (1) Initialization: The user wears a wearable abdominal belt. After the system starts, it performs “sensor self-test → optimal channel selection → 3-minute initial signal acquisition” to establish a temporary baseline. (2) Baseline establishment: After the user enters a resting state, a 15-minute steady-state signal acquisition is performed to establish the individual basic baseline and the day-night stratified baseline; (3) Real-time monitoring: The signal acquisition module 100 continuously acquires signals and transmits them to the signal preprocessing module 200; The signal preprocessing module purifies the signal and outputs a high-quality fetal heart rate signal. The feature extraction module 300 extracts and filters features, outputting high-quality features; The risk assessment module 400 determines the risk level based on joint criteria and outputs an early warning signal; The early warning output module 500 displays early warning information and data through user terminals / medical staff terminals; (4) Event handling: When a High / Medium risk event is triggered, the system automatically saves the event fragment and pushes it to the medical terminal for doctors to review; (5) Data management: Monitoring data is stored synchronously on local and cloud, and data export, deletion and anonymization are supported.

[0087] This embodiment provides a fetal hypoxia early risk identification system based on large pulse pressure phenotyping, used to implement the steps in the aforementioned fetal hypoxia early risk identification device based on large pulse pressure phenotyping. Therefore, the specific implementation of the fetal hypoxia early risk identification system based on large pulse pressure phenotyping can be found in the previous embodiment section of the fetal hypoxia early risk identification device based on large pulse pressure phenotyping. For example, the signal acquisition module 100, signal preprocessing module 200, feature extraction module 300, risk determination module 400, and early warning output module 500 are respectively used to implement steps S1, S2, S3, S4, and S5 in the aforementioned fetal hypoxia early risk identification device based on large pulse pressure phenotyping. Therefore, its specific implementation can be referred to the description of the corresponding embodiments. To avoid redundancy, it will not be repeated here.

[0088] The following are three representative cases covering different monitoring scenarios to further verify the system's versatility and stability: Case A: Quiet Scenario (Home Environment) Acquisition configuration: A ring array (4 MEMS microphones + 4 accelerometers) with a center distance of 40mm is used, and the abdominal binder is attached to the left side below the navel (fetal heart auscultation area). Preprocessing flow: bandpass filtering + MVDR beamforming + sparse representation (OMP algorithm), without body movement / uterine contraction interference; Segmentation: Envelope + Wavelet Packet + Peak-Valley Tracking; Parameter settings: Sliding time window s, threshold , (Differential type is based on baseline +3) ), Θ3= ×5 times, =1.4, ≥60%; Monitoring results: Appearance ↑+ ↑+ For brief events like the ↑ signal, the system issues a High-level alert; it also caches and reports the timestamped waveform and feature sequence for doctors to review.

[0089] Case B: Scenario of uterine contractions (hospital environment) Data acquisition configuration: linear array (4-channel MEMS microphone + 4-channel accelerometer), connected to TOCO uterine contraction sensor, abdominal binder attached to the right side below the navel; Preprocessing flow: bandpass filtering + GSC beamforming + LMS adaptive filtering, motion gating enabled; Segmentation: Envelope + Wavelet Packet + Peak-Valley Tracking; Parameter settings: s, =120%, Θ3= ×8 times, =1.5, η≥60%; Monitoring results: TOCO rising edge + ↑ and ↑ Appears simultaneously; the system outputs "High-level warning (contraction-related)"; if TOCO is stable and the same pattern appears, it is marked as "non-contraction-related" to prompt further review.

[0090] Case C: High-interference scenario (walking around / talking) Data acquisition configuration: circular array (4-channel MEMS microphone + 4-channel accelerometer), home living room environment (with TV sound and conversation). Preprocessing flow: bandpass filtering + MVDR + ICA + volumetric gating; Segmentation: Envelope + Wavelet Packet + Peak-Valley Tracking; Parameter settings: =45s, =100%, Θ3= ×6 times, =1.3, ≥60%; Monitoring results: No triggering during strong body movement; if "large pulse pressure candidate + ρ exceeds the threshold and meets the requirement of persistence" appears after body movement is relieved, triggering will be performed again, and false alarms will be controlled.

[0091] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0092] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0093] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0094] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A device for identifying early risk of fetal hypoxia based on large pulse pressure phenotyping, characterized in that, The device includes: Fetal heart sounds and body surface vibration signals were collected from the abdominal wall of pregnant women using a multi-channel sensor array. The fetal heart sound signal and body surface vibration signal are preprocessed to improve the signal-to-noise ratio of the target signal and suppress interference signals; Within a sliding time window, the large pulse pressure phenotypic feature set is extracted from the preprocessed signal, and the heart sound phase parameters are extracted. Based on the large pulse pressure phenotypic feature set and heart sound phase parameters, the risk of fetal hypoxia is determined by combining the preset joint criteria. Output graded early warning information based on the risk assessment results; The set of large pulse pressure phenotypic features includes the relative increase in envelope. Pulse pressure proxy Low-frequency energy and the slope of energy rise ; wherein, the relative increase The pulse pressure surrogate is the increase in the peak value of the inner sound envelope relative to the individual baseline envelope value within a sliding time window. For the first heartbeat With the second heart sound The amplitude difference or amplitude ratio, the low-frequency energy The energy rise slope represents the spectral peak intensity in the low-frequency band of the signal envelope. for The first derivative with respect to time; The heart sound timing parameters include systolic duration. Duration of diastole timely comparison ,in The first heart sound within the same cardiac cycle To the second heart sound The time interval, The second heart sound of the current cardiac cycle The first heart sound until the next cardiac cycle The time interval, ; The joint criteria include at least candidate feature exceeding the threshold, heart sound time compared to the individual baseline, persistence of the judgment result, multi-channel consistency verification, and the magnitude of heart rate decrease. verify; Specifically, the candidate feature exceeding the threshold refers to the relative increase in the envelope of the large pulse pressure phenotypic feature set. Pulse pressure proxy Low-frequency energy Simultaneously exceeding the preset adaptive threshold; The heart sound time-to-supreme baseline specifically refers to the time-to-supreme baseline calculated in real time. Compared with the preset individual steady-state baseline The ratio is greater than a preset threshold; The persistence of the determination result is specifically defined as follows: the duration of the state exceeding the individual baseline when the candidate feature exceeds the threshold and the heart sound is not less than a preset duration. The multi-channel consistency verification specifically means that the proportion of channels in the multi-channel sensor array that meet the judgment conditions is not less than a preset proportion. The decrease in heart rate The verification specifically states that the decrease in fetal heart rate within the sliding time window should not exceed a preset negative threshold.

2. The fetal hypoxia early risk identification device based on large pulse pressure phenotype according to claim 1, characterized in that, The preprocessing includes at least one of the following: bandpass filtering, notch filtering, adaptive beamforming, blind source separation, sparse representation reconstruction, and adaptive filtering; wherein, the adaptive beamforming includes a minimum variance distortionless response algorithm or a generalized sidelobe cancellation algorithm, the blind source separation includes an independent component analysis algorithm, and the adaptive filtering includes a minimum mean square algorithm or a recursive least squares algorithm.

3. The fetal hypoxia early risk identification device based on large pulse pressure phenotype according to claim 1, characterized in that, The device further includes establishing an individual steady-state baseline: when the pregnant woman is at rest and there is no significant body movement or uterine contraction, signals of a preset duration are collected, and the baseline values ​​of the large pulse pressure phenotypic feature set and the heart sound phase parameters are calculated to form an individual steady-state baseline; the individual steady-state baseline also includes a diurnal stratified baseline, corresponding to the signal baseline states during the day and night respectively; and the preset threshold in the joint criterion is adaptively corrected based on gestational age, and the threshold interpolation adjustment is achieved by fitting the correlation curve between gestational age and the baseline values ​​of heart sound phase parameters.

4. The fetal hypoxia early risk identification device based on large pulse pressure phenotype according to claim 1, characterized in that, The device also includes anti-interference control: when body movement interference is detected, the determination of large pulse pressure phenotypic characteristics is frozen, and only the heart sound phase parameters are tracked; when uterine contraction signals are detected, uterine contraction-related tags are added to the risk warning information; when maternal heart sound interference exists, the same frequency components of maternal heart sounds are removed by spatial difference or blind source separation algorithms.

5. A fetal hypoxia early risk identification system based on large pulse pressure phenotype, characterized in that, include: The signal acquisition module is used to acquire fetal heart sound signals and body surface vibration signals at the pregnant woman's abdominal wall through a multi-channel sensor array; The signal preprocessing module is used to preprocess the fetal heart sound signal and body surface vibration signal to improve the signal-to-noise ratio of the target signal and suppress interference signals. The feature extraction module is used to extract the large pulse pressure phenotypic feature set from the preprocessed signal within a sliding time window, and to extract the heart sound phase parameters. The criterion calculation module is used to determine the risk of fetal hypoxia based on the large pulse pressure phenotypic feature set and heart sound phase parameters, combined with a preset joint criterion. The early warning output module is used to output graded early warning information based on the risk assessment results; The set of large pulse pressure phenotypic features includes the relative increase in envelope. Pulse pressure proxy Low-frequency energy and the slope of energy rise ; wherein, the relative increase The pulse pressure surrogate is the increase in the peak value of the inner sound envelope relative to the individual baseline envelope value within a sliding time window. For the first heartbeat With the second heart sound The amplitude difference or amplitude ratio, the low-frequency energy The energy rise slope represents the spectral peak intensity in the low-frequency band of the signal envelope. for The first derivative with respect to time; The heart sound timing parameters include systolic duration. Duration of diastole timely comparison ,in The first heart sound within the same cardiac cycle To the second heart sound The time interval, The second heart sound of the current cardiac cycle The first heart sound until the next cardiac cycle The time interval, ; The joint criteria include at least candidate feature exceeding the threshold, heart sound time compared to the individual baseline, persistence of the judgment result, multi-channel consistency verification, and the magnitude of heart rate decrease. verify; Specifically, the candidate feature exceeding the threshold refers to the relative increase in the envelope of the large pulse pressure phenotypic feature set. Pulse pressure proxy Low-frequency energy Simultaneously exceeding the preset adaptive threshold; The heart sound time-to-supreme baseline specifically refers to the time-to-supreme baseline calculated in real time. Compared with the preset individual steady-state baseline The ratio is greater than a preset threshold; The persistence of the determination result is specifically defined as follows: the duration of the state exceeding the individual baseline when the candidate feature exceeds the threshold and the heart sound is not less than a preset duration. The multi-channel consistency verification specifically means that the proportion of channels in the multi-channel sensor array that meet the judgment conditions is not less than a preset proportion. The decrease in heart rate The verification specifically states that the decrease in fetal heart rate within the sliding time window should not exceed a preset negative threshold.

6. The fetal hypoxia early risk identification system based on large pulse pressure phenotype according to claim 5, characterized in that, The signal acquisition module includes a multi-channel sensor array and a wearable carrier; the multi-channel sensor array includes at least two sensors, which are at least one of MEMS microphones, piezoelectric sensors, or accelerometers; the array type of the multi-channel sensor array is a linear array or a circular array, and the spacing between the sensors in the array is a preset spacing; the wearable carrier is a flexible abdominal band, and the inner side of the flexible abdominal band is provided with a medical coupling layer and a shielding damping structure.

7. The fetal hypoxia early risk identification system based on large pulse pressure phenotype according to claim 5, characterized in that, The early warning output module includes a user terminal and a medical terminal; the user terminal is used to display fetal heart rate, heart sound signal waveform, heart sound phase parameter trend and graded early warning information in real time, and to display a gating status prompt when interference is detected. The medical terminal is used to display detailed data on the timeline of risk events, large pulse pressure phenotypic characteristics, and heart sound phase parameters. It supports playback of signal segments for preset durations before and after risk events and provides a data export function for verification.

Images