Non-invasive intrauterine sound field reconstruction method and system facing uterine sound environment construction

Abstract

The present application relates to the field of biomedical acoustics and computational acoustics, and discloses a non-invasive intrauterine sound field reconstruction method and system for uterine sound environment construction. Individual anatomical structure data of a target pregnant woman is acquired, and a parameterized biophysical sound propagation model including multiple tissue regions, an intrauterine target region, and to-be-identified sound propagation parameters is established. Reference signals of at least two types of sound sources and sound vibration observation signals of multiple positions outside the maternal body are synchronously acquired, wherein at least one type is a physiological sound source. The measured propagation characteristics and the model predicted propagation characteristics of each sound source to each external observation position are determined respectively, and the difference between the two is jointly constrained to identify individualized sound propagation parameters. Based on the parameters, the acoustic transfer operator of the sound source to the intrauterine target region is calculated, and the intrauterine target region sound field is reconstructed in combination with the estimated sound source signal. The present application is used for uterine sound environment construction and fetal acoustic stimulation research.

Description

Technical Field

[0001] This invention relates to the fields of biomedical acoustics, computational acoustics, maternal-fetal sound propagation modeling, and non-invasive sound field reconstruction, specifically to a non-invasive intrauterine sound field reconstruction method and system for constructing the uterine sound environment. Background Technology

[0002] The acoustic stimuli received by the fetus in utero are related to various sound sources, including external airborne sounds, maternal vocalizations, maternal heartbeats, respiration, intestinal activity, and the fetus's own physiological vibrations. Because the maternal abdominal wall, muscles, uterine wall, placenta, amniotic fluid, and fetal tissues have different acoustic impedances and frequency-dependent attenuation characteristics, external or internal maternal sound sources undergo complex multipath propagation, scattering, reflection, and local standing wave effects when reaching the fetal head, fetal ears, or other target areas within the uterus.

[0003] In existing technologies, partial frequency transmission curves can be obtained through intrauterine hydrophone experiments on humans or animals. Alternatively, spectral corrections for surface or external sounds can be performed based on gestational age, abdominal wall thickness, fat layer thickness, amniotic fluid volume, or phantom data. However, these methods typically rely on population average curves or preset tissue parameters, making it difficult to reflect the actual propagation characteristics of the target pregnant woman in the audio frequency band.

[0004] Existing technologies can also establish individualized geometric models of pregnant women based on medical imaging data and use numerical acoustic methods such as boundary element and finite element methods to calculate the propagation of external sound within maternal tissues, the uterus, and amniotic fluid. While these methods can utilize individual geometric structures, they typically still rely on literature statistics or pre-defined tissue acoustic parameters. Individual geometric models are not equivalent to individually measured sound propagation models, and they are particularly difficult to address uncertainties related to different tissue interfaces, sensor contact coupling, and individual frequency-dependent attenuation parameters.

[0005] Furthermore, spatial sound field generation technologies such as multi-channel loudspeaker arrays, wavefield synthesis, and phased arrays can form sound fields with predetermined wavefronts or propagation directions; abdominal wall multi-sensor technology can also be used for fetal heart rate, fetal movement, fetal position, maternal movement, or multi-source signal separation. However, existing spatial sound field technology and body surface multi-sensor technology themselves have not solved how to incorporate measured propagation characteristics from different sound source locations, different propagation boundary conditions, or different body surface observation paths into the joint identification process of the same set of individual biophysical sound propagation parameters.

[0006] The urgent technical problem to be solved is to jointly calibrate an individualized biophysical sound propagation model using acoustic vibration observation signals from multiple locations outside the target pregnant woman and at least two different propagation paths, without relying on intrauterine sensors as a clinical application step, and to further calculate the sound field of the target intrauterine region. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the prior art and provide a non-invasive intrauterine sound field reconstruction method and system for constructing the uterine acoustic environment. This method utilizes the measured propagation characteristics of at least two types of sound sources with different sound source locations or boundary conditions generated at multiple observation points outside the mother's body to perform multi-path constraint optimization on the same set of acoustic parameters to be identified in the parameterized biophysical sound propagation model. This results in propagation parameters that better reflect the actual tissue acoustic characteristics of the target pregnant woman and reconstructs the sound field of the target intrauterine region accordingly. This solves the problems of poor individual adaptability and high reconstruction uncertainty caused by relying on average curves or preset parameters.

[0008] To achieve the above objectives, a non-invasive intrauterine sound field reconstruction method for constructing the uterine acoustic environment is designed, comprising the following steps: acquiring individual anatomical data of the target pregnant woman, and establishing a parametric biophysical sound propagation model based on the individual anatomical data; the parametric biophysical sound propagation model includes multiple divided tissue regions, at least one intrauterine target region, and a set of acoustic propagation parameters to be identified; the set of acoustic propagation parameters to be identified includes at least sound velocity and frequency-related attenuation parameters, and the sound propagation in each tissue region is described by wave equations or Helmholtz equations and solved numerically; simultaneously acquiring sound source reference signals from at least two different sound sources and sound vibration observation signals set at multiple different locations outside the mother's body, wherein at least one sound source is naturally generated by the mother or fetus. The physiological sound vibration sources are identified. For each type of sound source, the measured propagation characteristics from the sound source to each external observation location are determined based on the sound source reference signal and the sound vibration observation signals. In the parameterized biophysical sound propagation model, the predicted propagation characteristics from the sound source to each external observation location are calculated according to the sound source location and sound propagation boundary conditions. Using the same set of acoustic propagation parameters to be identified, the differences between the measured propagation characteristics and the predicted propagation characteristics corresponding to at least two types of sound sources are jointly constrained to obtain the individualized sound propagation parameters of the target pregnant woman. Based on the individualized sound propagation parameters, the acoustic transfer operator from at least one type of sound source to the target area in the uterus is calculated. Based on the acoustic transfer operator and the estimated sound source signal of the corresponding sound source, the reconstructed sound field of the target area in the uterus is calculated.

[0009] Preferably, the present invention further includes: the at least two different sound sources include: sound vibration sources generated by maternal vocalization, sound vibration sources generated by maternal heartbeat, sound vibration sources generated by maternal respiration, sound vibration sources generated by fetal heartbeat, and at least two external sound fields generated by a spatially controllable sound field generation device outside the maternal body, wherein, for maternal vocalization, a contact-type sound vibration sensor disposed in the larynx is used to obtain a maternal vocalization reference signal; for maternal heartbeat, a sound vibration sensor in the precordial region and sternal region and an electrocardiogram signal are used to obtain a heartbeat reference signal; for the external sound field, at least one reference acoustic sensor disposed in the space outside the maternal body surface is used to obtain an actual incident sound field reference signal.

[0010] Preferably, the present invention further includes: the plurality of external observation locations comprising multiple locations selected from a set consisting of the larynx, sternum, xiphoid process region, left precordial region, upper abdomen, mid-abdomen, lower abdomen, left and right lateral abdomen, lumbosacral region, iliac crest region, fetal head surface projection region determined by medical imaging, fetal chest surface projection region, placental surface projection region, and the region opposite to the placenta; establishing a surface coordinate system with predetermined anatomical landmarks and recording the spatial positions of each acoustic vibration sensor; the external observation locations are functionally divided into a set of proximal reference points, a set of trunk propagation reference points, a set of abdominal wall spatial observation points, a set of individual anatomical feature projection points, and... The airborne sound field reference point set includes: a near-source reference point set for obtaining near-source references of natural physiological sound vibration sources; a trunk propagation reference point set for characterizing the intermediate response of natural physiological sound vibrations propagating along the maternal trunk; an abdominal wall space observation point set for covering the uterine surface projection area; an individual anatomical feature projection point set for determining the surface projection positions of the fetal head, fetal chest, and placenta based on medical imaging; and an airborne sound field reference point set for measuring the actual sound field boundary of the space outside the maternal body surface. The sound vibration observation signals are acquired by a contact microphone, an accelerometer, a piezoelectric sensor, an inertial acoustic sensor, a fiber optic vibration sensor, or a combination thereof.

[0011] Preferably, the present invention further includes: for the mother's vocalization, simultaneously acquiring the laryngeal contact acoustic vibration reference signal and the air acoustic reference signal corresponding to the same mother's vocalization event, respectively determining the volume conduction propagation characteristics of the mother's vocalization propagating through the trunk tissue to the abdominal wall and the air conduction propagation characteristics coupled to the abdominal wall after propagation through the air, and using the volume conduction propagation characteristics and the air conduction propagation characteristics to jointly constrain the same set of acoustic propagation parameters to be identified; the volume conduction propagation path includes at least the intermediate propagation reference point of the larynx and the sternum and xiphoid process regions.

[0012] Preferably, the present invention further includes: determining the placental surface projection position based on medical imaging data, and setting a placental contralateral observation position in a non-placental region that has a predetermined spatial correspondence with the placental surface projection position. The placental contralateral observation position is a non-placental surface position that is at a similar cephalotail height to the placental surface projection position and where the main propagation path from this surface position to the intrauterine target region does not pass through the main region of the placenta. Based on the difference in measured propagation characteristics between the placental surface projection position and the placental contralateral observation position, the local non-uniform tissue propagation parameters in the parameterized biophysical sound propagation model are constrained. Furthermore, the fetal head and fetal chest surface projection positions are determined based on medical imaging to form observational constraints on the propagation of natural sound sources inside the fetus to the outside of the mother's body.

[0013] Preferably, the present invention further includes: establishing a set of candidate external observation locations based on individual anatomical structure data; calculating the sensitivity of the model-predicted propagation characteristics corresponding to each candidate external observation location to at least some of the acoustic propagation parameters to be identified; determining a parameter observability index based on a sensitivity matrix composed of the sensitivity; and selecting a target observation location for acoustic vibration observation from the candidate external observation locations; the parameter observability index includes at least one of the condition number, determinant, minimum singular value, Fisher information, and parameter estimation covariance of the sensitivity matrix; in the case of using an externally controllable sound field, the propagation direction and frequency of the external calibration sound field are also selected based on the parameter observability index.

[0014] Preferably, the present invention further includes: the measured propagation characteristics include at least one of complex frequency response, amplitude frequency response, phase frequency response, cross power spectrum, transfer function estimate, amplitude squared coherence function, time delay, group delay, impulse response, and time-frequency propagation characteristics; for the sound source s and the i-th observation position, the measured propagation characteristics are determined by the cross spectrum of the sound source reference signal and the acoustic vibration observation signal, and the amplitude squared coherence function is calculated; in the joint constraints, the weights are determined based on coherence, signal-to-noise ratio, measurement confidence, and model sensitivity, and the parameter constraint terms are established based on at least one of human tissue acoustic parameters, measured data of sound propagation in the uterus of humans or animals, tissue equivalent phantom data, and individual anatomical measurement data; when the coherence, signal-to-noise ratio, contact coupling quality, or model residual does not meet the predetermined conditions, the weights of the corresponding sound source, observation position, or frequency band are reduced or excluded from the joint constraints.

[0015] Preferably, the present invention further includes: the parameterized biophysical sound propagation model determines geometric parameters based on at least one of ultrasound, three-dimensional ultrasound, and magnetic resonance imaging; the geometric parameters include at least one of abdominal wall contour, subcutaneous fat layer thickness, muscle layer thickness, uterine boundary, amniotic fluid region, placental position, fetal position, fetal orientation, fetal head position, and maternal spine or pelvic structure; sound propagation in each tissue region is described by wave equation, Helmholtz equation, or their numerical equivalent model, and calculated using boundary element method, finite element method, finite difference time-domain method, spectral method, reduced-order model, or a combination thereof; wherein, the complex wave number of each tissue region is determined by sound velocity and frequency-related attenuation parameters, and the sound pressure satisfies the Helmholtz equation or its discrete equivalent form.

[0016] This invention also provides a non-invasive intrauterine sound field reconstruction system for constructing the uterine acoustic environment, comprising: an individual anatomical structure data acquisition module for acquiring individual anatomical structure data of the target pregnant woman; a multi-source reference acquisition module for simultaneously acquiring sound source reference signals of at least two different sound sources; an in vitro observation array including acoustic vibration sensors disposed at multiple different locations outside the mother's body for simultaneously acquiring multiple acoustic vibration observation signals; a synchronization and preprocessing module for synchronously calibrating and preprocessing the sound source reference signals and acoustic vibration observation signals; a measured propagation feature extraction module for determining the measured propagation features from each sound source to each in vitro observation location based on the sound source reference signals of various sound sources and each acoustic vibration observation signal; and a parameterized biophysical sound propagation model module for establishing a model containing multiple tissue regions and at least one intrauterine target region based on the individual anatomical structure data. The system includes a parameterized biophysical sound propagation model for the set of acoustic propagation parameters to be identified, and calculates the predicted propagation characteristics from each sound source to each in vitro observation location according to the sound source location and sound propagation boundary conditions corresponding to each type of sound source. The sound propagation in each tissue region is described by the wave equation or the Helmholtz equation and solved numerically. A multi-source joint parameter identification module is used to use the same set of acoustic propagation parameters to jointly constrain the differences between the measured propagation characteristics and the predicted propagation characteristics corresponding to at least two types of sound sources to obtain the individualized sound propagation parameters of the target pregnant woman. An acoustic transfer operator calculation module is used to calculate the acoustic transfer operator from at least one type of sound source to the target region in the uterus based on the individualized sound propagation parameters. A sound field reconstruction output module is used to calculate the reconstructed sound field of the target region in the uterus based on the acoustic transfer operator and the estimated sound source signal of the corresponding sound source.

[0017] Preferably, the present invention further includes an observation point adaptive selection module and an external space controllable sound field generation device; the observation point adaptive selection module is configured to select a target observation position for joint parameter identification based on individual anatomical structure data, model sensitivity of candidate observation positions, and parameter observability index; the external space controllable sound field generation device controls multiple spatially distributed sound sources to form multiple calibrated sound fields with different propagation directions, spectra, and spatial phase relationships in a predetermined area including the mother's abdomen, and measures the actual incident boundary conditions through a reference acoustic sensor set in the space outside the mother's body surface, using the actual incident boundary conditions as the boundary input of the parameterized biophysical sound propagation model.

[0018] Compared with the prior art, the advantages of this invention are: First, this invention significantly improves the observability and identification accuracy of individualized acoustic propagation parameters by jointly constraining the same set of acoustic propagation parameters to be identified using measured propagation characteristics corresponding to at least two different types of sound sources. Different sound sources have different source locations and sound propagation boundary conditions, and their respective propagation paths have different spatial coverage ranges in the maternal tissue, thus resulting in varying sensitivities to the same set of tissue acoustic parameters. By incorporating these multi-source observation information with complementary sensitivities into the same parameter identification framework, cross-constraints can be effectively formed on key acoustic propagation parameters such as tissue sound velocity and frequency-related attenuation parameters. This overcomes the problem of parameter ambiguity caused by insufficient information from a single sound source or a single observation path, enabling the finally obtained individualized acoustic propagation parameters to more realistically reflect the acoustic characteristics of various tissue regions of the target pregnant woman.

[0019] Secondly, this invention distinguishes between different sound source locations and boundary conditions, avoiding the technical bias of simply equating different propagation paths such as the mother's voice, heartbeat, and external sound field to the same average filter curve. Existing technologies often use a uniform spectral correction curve to filter sound signals from different sources, ignoring the differences in the actual propagation paths of different sound sources within the mother body. This invention calculates and predicts propagation characteristics separately according to the actual sound source locations and boundary conditions corresponding to each type of sound source in the parametric model, and establishes acoustic transfer operators to the target region within the palace for different sound sources during the reconstruction stage. Finally, it synthesizes the propagation based on the estimated sound source signals, so that the reconstructed sound field can more accurately reflect the actual propagation effects of different sound source components in both spectral structure and time-domain waveform.

[0020] Third, this invention enables the reconstruction of the sound field in the target area within the uterus under non-invasive acquisition conditions. All sound source reference signals and sound vibration observation signals are set on the mother's body surface or the space outside the body surface, without the need to implant any sensors into the uterine cavity or amniotic fluid. This avoids the infection risks, fetal safety risks, and ethical restrictions that may be caused by invasive measurements, and has good clinical feasibility and repeatability. It is suitable for intrauterine sound environment assessment in routine prenatal examination scenarios.

[0021] Fourth, this invention incorporates prior information such as historical human or animal intrauterine sound propagation data and tissue equivalent phantom data as parameter constraints into the identification process, rather than directly replacing individual measured output curves. This mechanism fully utilizes existing prior knowledge of sound propagation to reduce the parameter search space and improve identification stability, while retaining the flexibility to adaptively calibrate based on the individual anatomical structure and actual propagation characteristics of the target pregnant woman, avoiding individual bias problems caused by simply applying group average data.

[0022] This invention systematically improves the accuracy and reliability of non-invasive reconstruction of intrauterine acoustic fields through an individualized parameter identification strategy based on multi-source joint constraints, providing more reliable technical support for the construction of uterine acoustic environment and fetal acoustic stimulation research. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the overall architecture of the reconstruction system of the present invention, in which solid lines represent core processing relationships and dashed lines represent optional implementation methods. At least one of the at least two types of sound sources is a physiological sound vibration source naturally generated by the mother or fetus.

[0024] Figure 2 This is a schematic diagram showing the external observation location, sensor arrangement, and body surface coordinates in this invention. Figure 2 (a) is a schematic diagram of the maternal abdomen from a side view. Figure 2 (b) is a schematic diagram of the anterior viewpoints and coordinate references of the abdominal wall. The center O of the umbilicus is used as the origin of the local coordinate system of the abdominal wall. The x-axis represents the transverse direction of the abdominal wall, and the y-axis represents the cephalothorax. Contact sensors are not required to be installed on the umbilicus itself. The projection positions are determined based on medical imaging, and the diagram does not limit fixed individual coordinates. The placental surface projection position 141 and the observation position 142 on the opposite side of the placenta are at similar cephalothorax heights. The main propagation path from the observation position 142 on the opposite side of the placenta to the target area within the uterus does not pass through the main area of ​​the placenta.

[0025] Figure 3 This is a schematic diagram of the reconstruction method of the present invention.

[0026] Figure 4This diagram illustrates the tissue regions, intrauterine tissues, and intrauterine target regions in the parametric biophysical sound propagation model of this invention. The dashed line indicates the intrauterine target region 307, which may include the fetal head region 308 and the fetal chest region 309. The boundaries of each tissue, fetal position, and placental location are determined based on individual anatomical data. The indicated locations are not intended to limit fixed geometric proportions, fetal position, or placental location.

[0027] Figure 5 This is a schematic diagram illustrating the relationship between the joint constraints of multiple propagation paths and the identification of individualized sound propagation parameters in this invention.

[0028] Figure 6 This is a schematic diagram illustrating the relationship between the controllable acoustic field excitation in external space and external observation in this invention.

[0029] In the picture: 201, Individual anatomical structure data acquisition module; 202, Multi-source reference acquisition module; 203, In vitro observation array; 204, Synchronization and preprocessing module; 205, Measured propagation feature extraction module; 206, Parametric biophysical sound propagation model module; 207, Adaptive selection of observation points module; 208, Multi-source joint parameter identification module; 209, Acoustic transfer operator calculation module; 210, Sound field reconstruction output module; 211, External space controllable sound field generation device; 212, Spatial distribution sound source; 213, Sound field controller; 214, Measured incident boundary acquisition unit.

[0030] 131, Maternal acoustic reference sensor; 132, Precordial and sternal region reference sensor (sternal region); 133, Precordial and sternal region reference sensor (left precordial region); 134, Sternal manubrium / xiphoid process trunk propagation reference sensor; 135, Abdominal wall spatial acoustic vibration sensor (upper abdomen); 135-L, Abdominal wall spatial acoustic vibration sensor (upper left abdomen); 135-R, Abdominal wall spatial acoustic vibration sensor (upper right abdomen); 136, Abdominal wall spatial acoustic vibration sensor (paraumbilical mid-abdomen); 136-L, Abdominal wall spatial acoustic vibration sensor (paraumbilical mid-abdomen left); 136-R, Abdominal wall spatial acoustic vibration sensor (paraumbilical mid-abdomen right); 137, Abdominal wall spatial acoustic vibration sensor (lower abdomen) ); 137-L, Abdominal wall spatial acoustic vibration sensor (lower left abdomen); 137-R, Abdominal wall spatial acoustic vibration sensor (lower right abdomen); 138, Abdominal wall spatial acoustic vibration sensor (left and right sides of the abdomen); 138-L, Abdominal wall spatial acoustic vibration sensor (left side of the abdomen); 138-R, Abdominal wall spatial acoustic vibration sensor (right side of the abdomen); 139, Fetal head / fetal chest / placenta projection observation sensor (fetal head surface projection position); 140, Fetal head / fetal chest / placenta projection observation sensor (fetal chest surface projection position); 141, Fetal head / fetal chest / placenta projection observation sensor (placenta surface projection position); 142, Placenta contralateral observation sensor; 143, Abdominal lateral air acoustic field reference sensor.

[0031] 301, Subcutaneous fat layer; 302, Abdominal wall muscle layer; 303, Uterine wall; 304, Amniotic fluid region; 305, Placenta; 306, Fetal-related structures; 307, Intrauterine target region; 308, Fetal head region; 309, Fetal chest region. Detailed Implementation

[0032] To make the purpose, principle and structure of the present invention clearer, the following description is provided in conjunction with the accompanying drawings and specific embodiments.

[0033] See Figures 1 to 6 The non-invasive intrauterine sound field reconstruction system, each functional module, the in vitro observation array 203, the external space controllable sound field generation device 211, and the tissue equivalent phantom involved in this invention can all be implemented by hardware, software, firmware, or a combination thereof, and the corresponding steps are executed by at least one processor.

[0034] like Figure 1As shown, the non-invasive intrauterine acoustic field reconstruction system includes an individual anatomical structure data acquisition module 201, a multi-source reference acquisition module 202, an in vitro observation array 203, a synchronization and preprocessing module 204, a measured propagation feature extraction module 205, a parameterized biophysical sound propagation model module 206, an observation point adaptive selection module 207, a multi-source joint parameter identification module 208, an acoustic transfer operator calculation module 209, and an acoustic field reconstruction output module 210. The individual anatomical structure data acquisition module 201 is used to acquire the individual anatomical structure data of the target pregnant woman. The multi-source reference acquisition module 202 is used to simultaneously acquire sound source reference signals from at least two types of sound sources. The in vitro observation array 203 includes acoustic vibration sensors placed at multiple locations on the maternal body surface for simultaneously acquiring multiple acoustic vibration observation signals. The synchronization and preprocessing module 204 is used to perform synchronous calibration and preprocessing of the sound source reference signals and acoustic vibration observation signals. The measured propagation feature extraction module 205 is used to determine the measured propagation features from various sound sources to different in vitro observation locations based on the sound source reference signals and the observed acoustic vibration signals. The parameterized biophysical acoustic propagation model module 206 is used to establish a parameterized biophysical acoustic propagation model based on individual anatomical data, including multiple tissue regions, at least one intrauterine target region, and a set of acoustic propagation parameters to be identified. It then calculates the predicted propagation features from each sound source to different in vitro observation locations according to the sound source location and acoustic propagation boundary conditions corresponding to each type of sound source. The adaptive observation point selection module 207 is configured to determine the target observation location among candidate in vitro observation locations based on individual anatomical structure, model sensitivity, and parameter observability. The multi-source joint parameter identification module 208 uses the same set of acoustic propagation parameters to jointly constrain the differences between the measured and predicted propagation features corresponding to at least two types of sound sources, obtaining individualized acoustic propagation parameters for the target pregnant woman. The acoustic transfer operator calculation module 209 calculates the acoustic transfer operator from at least one type of sound source to the intrauterine target region based on the individualized acoustic propagation parameters. The sound field reconstruction output module 210 is used to calculate the reconstructed sound field of the target area inside the palace based on the acoustic transfer operator and the estimated sound source signal of the corresponding sound source.

[0035] See Figure 6 In a preferred embodiment, the system further includes an external spatial controllable sound field generation device 211, which includes spatially distributed sound sources 212, a sound field controller 213, and a measured incident boundary acquisition unit 214. The sound field controller 213 controls multiple spatially distributed sound sources 212 to form multiple calibrated sound fields with different propagation directions, spectra, and spatial phase relationships within a predetermined area including the mother's abdomen. The measured incident boundary acquisition unit 214 is used to measure the actual incident boundary conditions using a reference acoustic sensor positioned in the space outside the mother's body surface, and uses these actual incident boundary conditions as the boundary input of a parameterized biophysical sound propagation model. As an optional solution, Figure 6This diagram illustrates an external spatial controllable sound field generation device 211 and its relationship with in vitro observation, parameterized biophysical sound propagation models, and multi-source joint parameter identification. The external spatial controllable sound field generation device 211 includes a sound field controller 213, spatially distributed sound sources 212, and a measured incident boundary acquisition unit 214. The sound field controller 213 controls multiple spatially distributed sound sources 212 to form multiple calibration sound fields within a predetermined region encompassing the mother's abdomen. These calibration sound fields have different propagation directions, spectra, and spatial phase relationships, thereby providing external sound excitation with controllable direction, spectrum, and spatial phase relationships. The device can employ wavefield synthesis, phased arrays, multi-speaker arrays, or other spatial sound field generation methods. Multiple calibration sound fields form an actual incident sound field within the predetermined region encompassing the mother's abdomen and reach the space outside the mother's body surface, where it is received by an airborne sound field reference sensor 143 located on the outer side of the abdomen. This sensor, positioned in the space outside the mother's body surface, belongs to the airborne sound field reference point set and is used to measure the actual sound field boundary outside the mother's body surface, serving as a sound source reference signal for the external sound field. The measured incident boundary acquisition unit 214 acquires the actual incident boundary conditions through the reference acoustic sensor and provides these conditions as boundary inputs to the parameterized biophysical sound propagation model module 206. The parameterized biophysical sound propagation model module 206 uses the actual incident boundary conditions as boundary inputs, calculates the predicted propagation characteristics from the sound source to each external observation location according to the sound source location and sound propagation boundary conditions corresponding to the external sound field, and outputs the results to the multi-source joint parameter identification module 208. Simultaneously, the external observation array 203 is positioned on the mother's body surface to acquire multiple acoustic vibration observation signals, which are then sent to the synchronization and preprocessing module 204 and the measured propagation feature extraction module 205. The actual incident sound field reference signal provided by the abdominal lateral air acoustic field reference sensor 143 is also used as a sound source reference signal and sent to the synchronization and preprocessing module 204 and the measured propagation feature extraction module 205. The synchronization and preprocessing module 204 performs synchronization calibration and preprocessing on the sound source reference signal and the acoustic vibration observation signal. The measured propagation feature extraction module 205 determines the measured propagation characteristics from the sound source to each external observation location based on the external sound field sound source reference signal and the acoustic vibration observation signals at each external observation location. The multi-source joint parameter identification module 208 uses the same set of acoustic propagation parameters to be identified to jointly constrain the differences between the measured propagation characteristics and the predicted propagation characteristics corresponding to at least two types of sound sources, thereby obtaining the individualized acoustic propagation parameters of the target pregnant woman. Among the at least two types of sound sources, at least one is a physiological acoustic vibration source naturally generated by the mother or fetus, and the controllable sound field in the external space can be used as the other type of sound source.The instruction manual also points out that for the controllable sound field in the external space, the system does not directly use the theoretical target sound field as the boundary condition of the human body model. Instead, it measures the actual incident boundary condition in the space outside the mother's body surface through the abdominal outer air sound field reference sensor 143 or the measured incident boundary acquisition unit 214. This reduces the deviation caused by the limited aperture of the loudspeaker, spatial discretization, room reflection, and spatial aliasing on the model calibration. When using the controllable sound field in the external space, the propagation direction and frequency of the external calibration sound field can also be selected according to the observability index of the parameters. The scheme corresponding to the figure makes the external calibration sound field form a boundary constraint from the outside to the inside, which participates in the joint identification of the same set of individualized biophysical sound propagation parameters together with natural physiological sound vibration sources such as maternal vocalization, maternal heartbeat, maternal respiration, and fetal heartbeat. On this basis, at least one type of acoustic transfer operator from the sound source to the intrauterine target area can be calculated based on the individualized sound propagation parameters, and the reconstructed sound field of the intrauterine target area can be calculated based on the acoustic transfer operator and the estimated sound source signal of the corresponding sound source.

[0036] like Figure 2As shown, the external observation array 203 includes acoustic and vibration sensors disposed at multiple locations on the surface of the mother body. These acoustic and vibration sensors can be contact microphones, accelerometers, piezoelectric sensors, inertial acoustic sensors, fiber optic vibration sensors, or combinations thereof. The sensors in the external observation array 203 are arranged spatially according to their functional layers and are attached to or close to the surface of the mother body. The external observation array 203 may include: a maternal acoustic reference sensor 131, a precordial and sternal region reference sensor (sternal region) 132, a precordial and sternal region reference sensor (left precordial region) 133, an abdominal wall spatial acoustic vibration sensor (upper abdomen) 135, an abdominal wall spatial acoustic vibration sensor (paraumbilical and mid-abdomen) 136, an abdominal wall spatial acoustic vibration sensor (lower abdomen) 137, an abdominal wall spatial acoustic vibration sensor (left and right lateral abdomen) 138, an abdominal lateral air acoustic field reference sensor 143, a sternal manubrium / xiphoid process trunk propagation reference sensor 134, a fetal head / fetal chest / placenta projection observation sensor (fetal head surface projection position) 139, a fetal head / fetal chest / placenta projection observation sensor (fetal chest surface projection position) 140, a fetal head / fetal chest / placenta projection observation sensor (placental surface projection position) 141, and a placental contralateral observation sensor 142. A maternal vocalization reference sensor 131 is positioned in the larynx, preferably anterior to the thyroid cartilage or adjacent to the anterior cervical region, to acquire proximal body conduction reference signals of maternal vocalization. Precordial and sternal region reference sensors are positioned in the left precordial region and sternal region to acquire cardiac or respiratory reference signals. Abdominal wall spatial acoustic vibration sensors are positioned at multiple locations on the maternal abdominal wall to acquire acoustic vibration observation signals. Lateral abdominal air acoustic field reference sensor 143 is positioned in the lateral abdominal space of the mother to measure the actual acoustic field boundary on the lateral side of the maternal body surface. A sternal manubrium / xiphoid process trunk propagation reference sensor 134 is positioned at the sternal manubrium, xiphoid process, and thoracoabdominal junction region to characterize the intermediate response of natural physiological acoustic vibrations propagating along the maternal trunk. Fetal head / fetal chest / placental projection observation sensors are positioned according to the surface projection positions of the fetal head, fetal chest, and placenta determined by medical imaging. The placental contralateral observation sensor 142 is located in a non-placental area that has a predetermined spatial correspondence with the projection position of the placental surface. Preferably, it is located at a non-placental surface position that is at a similar cephalo-to-tail height to the projection position of the placental surface and where the main propagation path from the surface position to the target area in the uterus does not pass through the main area of ​​the placenta.

[0037] See Figure 2In one embodiment, a body surface coordinate system is established using predetermined anatomical landmarks, and the spatial positions of each acoustic vibration sensor are recorded. The anatomical landmarks may include reproducible landmarks such as the umbilicus center, xiphoid process, sternal manubrium, iliac crest, and pubic symphysis. Preferably, the umbilicus center serves as the reference origin of the local abdominal wall coordinate system, without requiring contact-type acoustic vibration sensors to be installed on the umbilicus itself, thus reducing differences in local curvature and contact coupling. The external observation positions are functionally divided into a near-source reference point set, a trunk propagation reference point set, an abdominal wall spatial observation point set, an individual anatomical feature projection point set, and an airborne acoustic field reference point set. The near-source reference point set is used to obtain near-source references for natural physiological acoustic vibration sources; the trunk propagation reference point set is used to characterize the intermediate response of natural physiological acoustic vibrations propagating along the maternal trunk; the abdominal wall spatial observation point set is used to cover the uterine surface projection area; the individual anatomical feature projection point set determines the surface projection positions of the fetal head, fetal chest, and placenta based on medical imaging; and the airborne acoustic field reference point set is used to measure the actual acoustic field boundary of the space outside the maternal body surface.

[0038] The above five-layer point system can be expressed as follows: R=R source ∪R trunk ∪R abdomen ∪R anatomy ∪R air ; Where R source Let R represent the set of near-source reference points. trunk R represents the set of reference points for trunk propagation. abdomen R represents the set of observation points in the ventral wall space. anatomy R represents the set of projection points of an individual's anatomical features. air This represents the set of reference points for the airborne sound field. The actual target observation locations participating in a joint identification can be a subset of this set.

[0039] like Figure 3As shown, in one embodiment, the method provided by the present invention can be executed by at least one processor, including the following main operations: acquiring individual anatomical structure data of the target pregnant woman; establishing a parameterized biophysical sound propagation model based on the individual anatomical structure data, wherein the parameterized biophysical sound propagation model includes multiple tissue regions, at least one intrauterine target region, and at least one set of acoustic propagation parameters to be identified; simultaneously acquiring sound source reference signals corresponding to at least two types of sound sources and multiple acoustic vibration observation signals set at different locations outside the mother's body, wherein the at least two types of sound sources correspond to different sound source locations and different sound propagation boundary conditions, and at least one type of sound source is a natural physiological acoustic vibration source generated by the mother or fetus; for each type of sound source, according to the corresponding... Based on the sound source reference signal and multiple acoustic vibration observation signals, the measured propagation characteristics of this type of sound source to multiple in vitro observation locations are determined. In the parameterized biophysical sound propagation model, according to the sound source location and sound propagation boundary conditions corresponding to each type of sound source, the predicted propagation characteristics of this type of sound source to multiple in vitro observation locations are calculated. Using the same set of acoustic propagation parameters to be identified, the differences between the measured propagation characteristics and predicted propagation characteristics corresponding to at least two types of sound sources are jointly constrained to obtain individualized sound propagation parameters for the target pregnant woman. Based on the individualized sound propagation parameters, an acoustic transfer operator from at least one type of sound source to the intrauterine target region is calculated. Based on the acoustic transfer operator and the corresponding sound source signal, the reconstructed sound field of the intrauterine target region is calculated. For ease of explanation, the above main operations are described below as steps S101 to S107. The numbering of steps S101 to S107 is only for the elaboration of this specific embodiment and does not constitute a limitation on the numbering, order, or division of the main steps of the method.

[0040] like Figure 1 , 3 As shown, in step S101, the individual anatomical structure data acquisition module 201 acquires the individual anatomical structure data of the target pregnant woman. The term "anatomy" in this invention refers to a decompositional analysis of the target pregnant woman's individual body structure, rather than a live dissection or limb dissection. This data is used to characterize the geometry and distribution of tissue regions in the target pregnant woman. This data can come from conventional obstetric ultrasound, 3D ultrasound, magnetic resonance imaging, surface scanning, or other non-invasive measurement methods. The individual anatomical structure data may include the outer contour of the abdominal wall, the thickness of the local subcutaneous fat layer, the thickness of the abdominal wall muscle layer, the uterine boundary, the amniotic fluid region, the placental region, the position and orientation of the fetus, the target area of ​​the fetal head or fetal ear, and the spatial position of the external observation array 203. The individual anatomical structure data is used to establish a parametric biophysical sound propagation model corresponding to the actual anatomical structure of the target pregnant woman.

[0041] In step S102, the processor establishes a parameterized biophysical sound propagation model based on the individual anatomical data. This model divides maternal tissues, the uterine wall, amniotic fluid, placenta, and fetal-related structures into multiple tissue regions and assigns acoustic propagation parameters to be identified for each region. Limited to this embodiment, initial search ranges or prior ranges for different tissue regions are provided. These values ​​are only used for parameter identification initialization and constraint ranges and are not considered as fixed physiological constants for all pregnant women. Specifically, the initial range for the sound velocity in the subcutaneous fat / fat equivalent region is 1450 to 1480 m / s, and the initial range for density is 850 to 1000 kg / m³. 3 The initial range of sound velocity in the abdominal wall muscle layer is 1540 to 1620 m / s, and the initial range of density is 1000 to 1180 kg / m³. 3 The initial range of sound velocity in the uterine myometrium / general soft tissue area is 1480 to 1600 m / s, and the initial range of density is 1000 to 1100 kg / m³. 3 The initial range of sound velocity in the amniotic fluid region is 1480 to 1540 m / s, and the initial range of density is 990 to 1030 kg / m³. 3 The initial range of sound velocity in the equivalent region of the placenta and fetal soft tissue is 1480 to 1600 m / s, and the initial range of density is 1000 to 1100 kg / m³. 3 .

[0042] like Figure 4 As shown, the subcutaneous fat layer 301, abdominal wall muscle layer 302, uterine wall 303, amniotic fluid region 304, placenta 305, and fetal-related structures 306 correspond to the aforementioned individual anatomical data and multiple tissue regions; the intrauterine target region 307 may include the aforementioned target regions such as the fetal head region 308 and fetal chest region 309. The figure only illustrates the relationship between tissues and target regions; the geometric boundaries, fetal position, and placental location of each region are determined based on the individual anatomical data of the target pregnant woman.

[0043] The frequency-dependent attenuation parameter uses α j (f)=a j ×(f / f0) bj The non-negative power-law form, where f0 is the reference frequency, with the same unit as f, i.e., Hz; a j Let be the frequency-dependent attenuation parameter of the j-th tissue region at the reference frequency f0, and its unit is α. j (f) Same; b j is the dimensionless frequency power-law exponent. Where b j The initial constraint range is from 0.5 to 2.0; a jThe upper bound is determined according to the target analysis frequency band and the actual thickness of the tissue, ensuring that the cumulative attenuation of a single layer per pass does not exceed 60dB. The initial parameter θ0 can be taken as the median of the above range; if there is already human, animal or phantom data in the same frequency band, the mean of the data is used as the prior center, and the parameter is restricted to the above range.

[0044] In one implementation, the complex wave number of the j-th tissue region is expressed according to the following formula: k j (f)=(2πf / c j )-iα j (f); Where c j Let α be the speed of sound in the j-th tissue region. j (f) represents the frequency-dependent attenuation parameter, where f is the frequency and i is the imaginary unit. This representation can be used for Helmholtz-type frequency domain solutions, and can also be replaced by equivalent parameter representations in time-domain wave equations, finite-difference time-domain models, or reduced-order models.

[0045] In one implementation, the sound pressure in each tissue region satisfies the following formula or its discrete equivalent form: ; Where p j Let q be the sound pressure level in the j-th organizational region. j For equivalent sound source terms; The Laplace operator, acting on spatial coordinates, describes the second-order spatial variation of sound pressure; j represents the organization region number, i.e., the j-th organization region among multiple organization regions; k j The complex wave number of the j-th tissue region is given by the previous equation k. j (f) Determine; This formula indicates that the sound pressure in the j-th tissue region satisfies the Helmholtz equation or its discrete equivalent form, and is used for sound propagation calculation of each tissue region in the parameterized biophysical sound propagation model.

[0046] At the interface between adjacent tissues, continuous sound pressure, continuous normal velocity, impedance boundary, contact coupling boundary, or experimentally calibrated equivalent boundary conditions can be used. The model can be solved using the boundary element method, finite element method, finite difference time-domain method, spectral method, or a combination thereof.

[0047] For ease of explanation, the above digital twin model of sound propagation can be expressed as the following formula: D={G,θ,Qs,Bs,ri,rt}; Where G represents the individual anatomical geometric model of the target pregnant woman, θ represents the set of acoustic propagation parameters to be identified, Qs represents the s-th type of sound source model, Bs represents the corresponding sound propagation boundary conditions, ri represents the i-th external observation position, and rt represents the position of the target area within the uterus.

[0048] In step S103, the multi-source reference acquisition module 202 simultaneously acquires sound source reference signals from at least two types of sound sources, and the external observation array 203 simultaneously acquires multiple acoustic vibration observation signals from different locations outside the mother's body. The at least two types of sound sources correspond to different sound source locations and different sound propagation boundary conditions, and at least one type of sound source is a natural physiological acoustic vibration source generated by the mother or fetus. For illustration, the acquisition target of this invention is the pregnant woman and the fetus in utero. The acquisition process is conducted on or outside the mother's body surface. Therefore, the mother's own continuous physiological activities such as heartbeat and respiration, as well as the fetus's own physiological activities such as heartbeat, will always exist and will propagate through maternal tissues, the uterine wall, and the amniotic fluid area to the mother's body surface, thus inevitably entering the external observation signal and constituting a natural physiological acoustic vibration source that is necessarily present in the acquisition environment. These natural physiological acoustic vibration sources are generated by the mother or fetus itself and can appear without external artificial stimulation. Their sound source locations are respectively located in the mother's precordial region, sternal region, thoracic and abdominal cavity, or fetal body. The sound propagation boundary conditions at different locations within the body involve different boundaries such as the maternal trunk, thoracic and abdominal cavities, uterine wall, amniotic fluid area, and fetal tissue. Therefore, these conditions differ from those of external environmental sound or controllable sound fields in terms of both source location and propagation boundary conditions. Thus, among at least two types of sound sources, at least one type of natural physiological sound vibration source from the mother or fetus can be identified as one of the comprehensive sound sources. The reference signals generated by multiple sound sources, including this natural physiological sound vibration source, are confirmed and collected using sensors, providing conditions for subsequently determining the measured propagation characteristics of each sound source to the external observation location. For fetal heartbeat synchronization, the fetal head / chest / placental projection observation sensor at the fetal chest surface projection location and other sensors in the external observation array 203 are clock-synchronized and heartbeat cycle-segmented by the synchronization and preprocessing module 204.

[0049] The at least two types of sound sources include at least two of the following: sound vibration sources generated by maternal vocalization, sound vibration sources generated by maternal heartbeat, sound vibration sources generated by maternal respiration, sound vibration sources generated by fetal heartbeat, and external sound fields generated by a spatially controllable sound field generation device outside the mother's body. For maternal vocalization, a contact-type sound vibration sensor placed in the larynx is used to obtain a maternal vocalization reference signal; for maternal heartbeat, sound vibration sensors in the precordial and sternal regions and electrocardiogram signals are used to obtain a heartbeat reference signal; for the external sound field, at least one reference acoustic sensor placed in the space outside the mother's body surface is used to obtain an actual incident sound field reference signal. The sound source reference signal may also include ambient sound or actual incident sound field references obtained from external reference acoustic sensors, as well as other reference signals that can characterize the activity of the sound source.

[0050] The maternal respiratory reference signal was acquired by the precordial region, sternal region reference sensor and sternal manubrium / xiphoid process trunk propagation reference sensor 134, and obtained by respiratory cycle segmentation.

[0051] The external observation array 203 may include multiple acoustic and vibration sensors disposed on the surface of the mother's body. The system can establish a surface coordinate system based on repeatable anatomical landmarks such as the umbilicus center, xiphoid process, manubrium of the sternum, iliac crest, and pubic symphysis, and record the position ri of each sensor. The external observation locations can be divided into five layers according to function: the first layer is the proximal reference point layer, with the maternal sound source proximal reference point preferably set in front of the thyroid cartilage or in the adjacent anterior cervical region, and the cardiac reference point preferably set in the left anterior ventricular region and the sternal region; the second layer is the trunk propagation reference point layer, preferably including the manubrium of the sternum, the xiphoid process, and the thoracoabdominal junction region, and optionally including the iliac crest and lumbosacral region; the third layer is the abdominal wall space observation point layer, preferably setting multiple observation positions near the surface projection of the uterine fundus, the mid-abdomen beside the umbilicus, the lower abdomen above the pubic symphysis, and the left and right lateral abdomen; the fourth layer is the individual anatomical feature projection point layer, including the surface projection positions of the fetal head, fetal chest, and placenta determined by medical imaging, as well as the position of the placenta on the opposite side; the fifth layer is the airborne sound field reference point layer, including one or more reference acoustic sensors set in the lateral space of the maternal abdomen.

[0052] In a preferred embodiment, the abdominal wall spatial observation point layer may be equipped with 8 to 12 contact acoustic vibration sensors, preferably arranged in a spatial distribution on the upper abdomen (left and right), around the umbilicus (left and right), lower abdomen (left and right), and on the left and right sides of the abdomen, avoiding the umbilicus itself; at least some observation positions are located in the left and right lateral regions of the mother's abdomen to increase the lateral propagation baseline. The above quantity and distribution are only examples and do not constitute a limitation on the number of channels of the external observation array 203.

[0053] Combination Figure 2 and Figure 5For maternal vocalization, the maternal vocalization reference sensor 131 is preferably located in front of the thyroid cartilage or adjacent to the anterior cervical region to provide proximal body guidance reference; the precordial and sternal region reference sensors can be located in the sternal region and the left precordial region, respectively corresponding to the precordial and sternal region reference sensor (sternal region) 132 and the precordial and sternal region reference sensor (left precordial region) 133; the sternal manubrium / xiphoid process trunk propagation reference sensor 134 is preferably located in the sternal manubrium, xiphoid process and the thoracoabdominal junction region; the abdominal wall spatial acoustic vibration sensor includes the abdominal wall spatial acoustic vibration sensor (upper abdomen) 135, the abdominal wall spatial acoustic vibration sensor (paraumbilical and mid-abdomen) 136, the abdominal wall spatial acoustic vibration sensor (lower abdomen) 137 and the abdominal wall spatial acoustic vibration sensor (left and right lateral abdomen) 138, which are used to provide propagation end observation. Simultaneously, the abdominal lateral air acoustic field reference sensor 143 acquires the air acoustic reference of the same maternal acoustic event, thereby obtaining the body conduction propagation characteristics and air conduction propagation characteristics of the maternal sound, and using the two types of boundary conditions to jointly constrain the same set of acoustic propagation parameters to be identified. For the individual anatomical feature projection point layer, the projection positions of the fetal head, fetal chest, and placenta on the maternal abdominal wall can be determined by ultrasound, three-dimensional ultrasound, or magnetic resonance imaging, and fetal head / fetal chest / placental projection observation sensors are set at the corresponding positions, including fetal head surface projection position 139, fetal chest surface projection position 140, and placental surface projection position 141, respectively; wherein, the placental contralateral observation sensor 142 is preferably set at a control abdominal wall position at a similar cephalo-tail height to the placental projection point and whose propagation path does not pass through the main placental area, so as to form a paired spatial observation of the placental path and the non-placental path. Preferably, a body surface coordinate system is established with the umbilicus center, and the cephalo-tail coordinate difference between the observation position on the opposite side of the placenta and the placental projection point is no greater than 30 mm. Candidate points on the opposite side of the abdomen that meet the above conditions are preferentially selected. In another embodiment, the fetal heartbeat can serve as a natural physiological sound source propagating from inside the fetus to outside the mother's body, and combined with the fetal chest surface projection position and multiple abdominal wall observation positions, it forms a propagation constraint from the inside out; external environmental sound or a spatially controllable calibrated sound field forms a boundary constraint from the outside in. The observations in different propagation directions are all interpreted using the same set of individualized biophysical sound propagation parameters.

[0054] As a preferred embodiment of the abdominal wall spatial acoustic vibration sensor (upper abdomen) 135, the abdominal wall spatial acoustic vibration sensor (near the umbilicus) 136, the abdominal wall spatial acoustic vibration sensor (lower abdomen) 137, and the abdominal wall spatial acoustic vibration sensor (left and right lateral abdomen) 138, see [link to relevant documentation]. Figure 2(b) Taking the center of the umbilicus O as the reference origin of the local coordinate system of the abdominal wall, the horizontal axis x represents the transverse direction of the abdominal wall, and the head-to-tail axis y represents the head-to-tail direction. The abdominal wall spatial acoustic vibration sensors are further arranged in pairs on the left and right sides to form: abdominal wall spatial acoustic vibration sensor (upper left) 135-L and abdominal wall spatial acoustic vibration sensor (upper right) 135-R; abdominal wall spatial acoustic vibration sensor (next to the umbilicus, middle left) 136-L and abdominal wall spatial acoustic vibration sensor (next to the umbilicus, middle right) 136-R; abdominal wall spatial acoustic vibration sensor (lower left) 137-L and abdominal wall spatial acoustic vibration sensor (lower right) 137-R; abdominal wall spatial acoustic vibration sensor (lateral left) 138-L and abdominal wall spatial acoustic vibration sensor (lateral right) 138-R.

[0055] Preferably, the abdominal wall spatial acoustic vibration sensors (upper left abdomen) 135-L and (upper right abdomen) 135-R, the abdominal wall spatial acoustic vibration sensors (next to the umbilicus, middle left abdomen) 136-L and (next to the umbilicus, middle right abdomen) 136-R, the abdominal wall spatial acoustic vibration sensors (lower left abdomen) 137-L and (lower right abdomen) 137-R, and the abdominal wall spatial acoustic vibration sensors (lateral left abdomen) 138-L and (lateral right abdomen) 138-R are located at the same or similar head-to-tail height and are distributed left and right relative to the umbilicus midline. This forms not only head-to-tail propagation sampling along the head-to-tail axis y, but also lateral propagation sampling along the horizontal axis x in the abdominal wall spatial observation point layer, thus constituting a two-dimensional spatial observation baseline. The aforementioned paired abdominal wall spatial acoustic vibration sensors can be used in conjunction with the fetal head / fetal chest / placenta projection observation sensors (fetal head surface projection position) 139, 140, 141, and the placental contralateral observation sensor 142 to form an abdominal wall forward-looking observation array. The umbilical center O serves only as the reference origin for abdominal wall coordinates, and it is not required to install contact acoustic vibration sensors on the umbilical fossa itself.

[0056] Furthermore, in the joint parameter identification, the measured propagation characteristics corresponding to the abdominal wall spatial acoustic vibration sensor (upper left abdomen) 135-L, abdominal wall spatial acoustic vibration sensor (upper right abdomen) 135-R, abdominal wall spatial acoustic vibration sensor (middle left abdomen near the umbilicus) 136-L, abdominal wall spatial acoustic vibration sensor (middle right abdomen near the umbilicus) 136-R, abdominal wall spatial acoustic vibration sensor (lower left abdomen) 137-L, abdominal wall spatial acoustic vibration sensor (lower right abdomen) 137-R, abdominal wall spatial acoustic vibration sensor (left side abdomen) 138-L, and abdominal wall spatial acoustic vibration sensor (right side abdomen) 138-R can be compared with the predicted propagation characteristics calculated by the same parameterized biophysical acoustic propagation model. The ability to distinguish between left and right asymmetry of abdominal wall tissue, placental deviation, fetal position deviation, and sensor contact coupling differences can be enhanced by cross-spectrum analysis, transfer function ratio, coherence function analysis, differential characteristics, or common-mode / differential-mode decomposition between the left and right paired sensors. Meanwhile, the aforementioned paired observations can also provide redundant spatial constraints, improve the observability index of parameters determined by the sensitivity matrix, including reducing the condition number of the sensitivity matrix, increasing the minimum singular value, or increasing Fisher information, thereby reducing the impact of unilateral observation noise, respiratory and cardiac common-mode interference, and local coupling fluctuations on parameter identification, and helping to improve the stability of individualized sound propagation parameter identification and the accuracy of sound field reconstruction in the target area within the uterus.

[0057] In step S104, the measured propagation feature extraction module 205 determines the measured propagation features from various sound sources to multiple external observation locations based on the sound source reference signal and multiple acoustic vibration observation signals. The measured propagation features can be complex frequency response, amplitude frequency response, phase frequency response, cross power spectrum, transfer function estimate, amplitude squared coherence function, time delay, group delay, impulse response, or time-frequency propagation features.

[0058] In the multi-source observation model, the observation signal at the i-th external observation location can be represented as: Y i (f)=∑ s∈S G s,i (f;θ)S s (f)+N i (f); Where S represents the set of sound source categories, S s (f) represents the s-th type of sound source signal, G s,i (f;θ) represents the propagation characteristics from the s-th type of sound source to the i-th external observation location, N i (f) indicates observation noise or unmodeled components.

[0059] For a sound source s and the i-th observation location, the measured propagation characteristics can be determined by the cross spectrum of the sound source reference signal and the observed acoustic vibration signal: G s,i obs(f)=(S YiXs (f)) / (S XsXs (f)); Where Xs(f) represents the reference signal of the s-th type of sound source, Yi(f) represents the ith sound vibration observation signal, and S represents the cross power spectrum or self power spectrum.

[0060] To improve identification stability, the system can further calculate the amplitude squared coherence function: γ s,i 2 (f)=(|S XsYi (f)| 2 ) / (S XsXs (f)×S YiYi (f)).

[0061] When the coherence, signal-to-noise ratio, contact coupling quality, or model sensitivity of a sound source, observation location, or frequency combination falls below predetermined conditions, the system can reduce its weight or remove it from joint identification, thereby avoiding the adverse effects of unsteady noise, coughing, body position changes, or sensor loosening on parameter identification. In one specific embodiment, only the amplitude squared coherence γ is included. 2 Sound sources, observation points, and frequency combinations with a signal-to-noise ratio (SNR) ≥ 0.5 and a signal-to-noise ratio (SNR) ≥ 6 dB are eligible; other combinations are not eligible for this round of joint identification.

[0062] Step S105: The parameterized biophysical sound propagation model calculates the predicted propagation characteristics of each type of sound source to multiple in vitro observation locations based on the sound source location and sound propagation boundary conditions.

[0063] In one implementation, the predicted propagation characteristics can be represented as follows: G s,i model (f;θ,B s )=M s,i (f;θ,B s ); Where Bs represents the actual or estimated boundary conditions corresponding to the s-th type of sound source. For the mother body's sound, Bs may include the boundaries related to the larynx, chest cavity, airway, and bone conduction; for heartbeat or respiration, Bs may include the mechanical vibration boundaries of the chest and abdominal cavities; for a controllable sound field in external space, Bs is preferably obtained by actual measurement from a reference acoustic sensor in the space outside the mother body surface.

[0064] like Figure 5 As shown, in step S106, the multi-source joint parameter identification module 208 uses the same set of acoustic propagation parameters θ to jointly constrain the differences between the measured propagation characteristics and the predicted propagation characteristics corresponding to at least two types of sound sources, and obtains the individualized sound propagation parameters θ* of the target pregnant woman.

[0065] In one implementation, the individualized sound propagation parameters can be determined as follows: θ*=argθ min[∑ s ∑ i ∑ f w s,i,f |G s,i obs (f)-G s,i model (f;θ)| 2 +λR(θ)]; Wherein, θ* represents the individualized acoustic propagation parameters of the target pregnant woman, i.e., the parameter identification result obtained after joint constraints; θ represents the set of acoustic propagation parameters to be identified, including tissue sound velocity, frequency-related attenuation parameters, etc.; s represents the sound source category, taking one of at least two sound source categories; i represents the in vitro observation location, taking one of multiple different locations outside the mother's body; f represents the frequency; w s,i,f The weights corresponding to the s-th type of sound source, the i-th external observation location, and the frequency f can be determined based on coherence, signal-to-noise ratio, measurement confidence, and / or model sensitivity; G s,i obs (f) represents the measured propagation characteristics of the sound source of type s to the external observation position of the i-th individual at frequency f; G s,i model (f;θ) represents the predicted propagation characteristics of the s-th type of sound source to the i-th individual's outdoor observation location at frequency f, given the acoustic propagation parameter θ to be identified; w represents the weight, R(θ) represents the parameter constraint term, and λ represents the regularization coefficient. This formula represents the search for the θ that minimizes the sum of the weighted squared differences between the measured and predicted propagation characteristics and the sum of the parameter constraint term in the value space of the acoustic propagation parameter θ to be identified, which is taken as the individualized acoustic propagation parameter θ* for the target pregnant woman.

[0066] The parameter constraints can be established based on human tissue acoustic parameters, measured data of intrauterine sound propagation in humans or animals, tissue equivalent phantom data, individual anatomical measurement data, or equipment calibration data. In this embodiment, historical intrauterine sound propagation data is preferably used as a priori constraints or regularization information, rather than directly as the intrauterine output curve of the target pregnant woman.

[0067] The weight of a data item can be w=(γ) 2 ) / (σ 2 +ε), where σ 2 γ represents the repeated measures variance of the corresponding observed features. 2 Let γ represent the squared amplitude coherence function, specifically the squared amplitude coherence function of the s-th type of sound source and the i-th external observation position at frequency f, which can be denoted as γ. s,i 2(f), and can be calculated according to the previous formula; ε represents the small quantity to prevent the denominator from being zero; all weights are normalized and then participate in the objective function; this formula indicates that the weights of the data items are determined according to the amplitude squared coherence function and the variance of repeated measurements of the corresponding observation features, and all weights are normalized and then participate in the objective function.

[0068] The regularization term can be taken as R(θ) = Σ[(θ) k -μ k ) / (U k -L k )] 2 , where μ k As a priori center, U k L k θ represents the upper and lower bounds of the parameters; R(θ) represents the parameter constraint term, used to regularize or impose prior constraints on individualized sound propagation parameters in the joint constraints; θ k μ represents the k-th acoustic propagation parameter to be identified in the parameter vector θ. k Let U represent the prior center of the k-th acoustic propagation parameter to be identified; Σ represents the summation over each acoustic propagation parameter to be identified that participates in the constraint; this formula represents the difference U between the upper and lower bounds of the k-th parameter. k -L k The parameter deviates from the prior center μ k The degree of normalization is then normalized, the square is calculated, and the summation is applied over each parameter to obtain the parameter constraint term R(θ). A specific example is taken where λ=0.01 in the normalized objective function.

[0069] The same maternal vocalization event can simultaneously form a body conduction path propagating through the larynx, sternum / xiphoid process, and abdominal wall, as well as an air conduction path formed by coupling the air reference point and the abdominal wall; the fetal heartbeat can form a path propagating from the fetal interior to the abdominal wall; the placental projection point and the contralateral point of the placenta can form paired spatial observations for locally non-uniform tissues. The multi-source joint parameter identification module 208 requires the same set of parameters θ to simultaneously interpret the measured propagation characteristics of at least two of the above propagation paths, and the observation point adaptive selection module 207 can also determine the target observation points participating in the joint identification based on the observability of the parameters, thus distinguishing it from schemes that only perform spectral compensation or inverse filtering based on a single abdominal sensor signal.

[0070] See Figure 3 and Figure 5 In step S107, the acoustic transfer operator calculation module 209 calculates the acoustic transfer operator from the sound source to the target region within the uterus based on the individualized sound propagation parameter θ*. In one embodiment, the transfer operator from the sound source s to the target region t is expressed as follows: H s→t (f;θ * )=T(r s ,r t ,f;θ* ); Among them, H s→t (f;θ * ) represents the individualized sound propagation parameter θ * Below, the acoustic transfer operator from the s-th type of sound source to the target region t within the palace at frequency f; s represents the sound source type, t represents the target region within the palace, and f represents the frequency; θ * This represents the individualized sound propagation parameters of the target pregnant woman. s Indicates the location of the sound source or the equivalent sound source region, r t This represents the fetal head, fetal ear, or other intrauterine target areas. This formula indicates that spatially dependent acoustic transfer operators are established for different sound source locations, different external observation locations, and different intrauterine target locations, rather than coarsely equating all paths to a single frequency response curve. Even if the acoustic medium locally satisfies reciprocity, the system still establishes spatially dependent transfer operators for different source locations, different external observation locations, and different intrauterine target locations, rather than coarsely equating all paths to a single frequency response curve. When using frequency domain finite element method as the main implementation method, a three-dimensional or local three-dimensional geometric model of subcutaneous tissue, abdominal wall muscle layer, uterine wall, amniotic fluid, placenta, and related fetal areas is established based on medical imaging data, and initial values ​​or identification results of the acoustic propagation parameters to be identified are assigned.

[0071] At the highest analysis frequency f of the target max Below, using the lowest sound speed c in each region min Calculate the shortest wavelength λ min =c min / f max The maximum element size of the volumetric mesh is no greater than λ. min / 6, encrypt the area near the organization interface to a value no greater than λ. min / 10.

[0072] For the external calibration sound field, the measured complex sound pressure amplitude and phase obtained from the air reference array are used as the actual incident boundary conditions outside the mother's body surface; for the natural sound source of the mother or fetus, equivalent source terms are set in the corresponding sound source region and normalized with the corresponding reference signal. The outer boundary of the model adopts an absorbing boundary or PML (perfectly matched layer, an artificial absorbing layer set at the outer boundary of the computational domain to absorb outwardly propagating sound waves and suppress boundary reflections).

[0073] Solve for the complex sound pressure field p(r,f) for each discrete frequency, and take the volume average or the sound pressure p at a specified target point for the target region Ω within the palace. t (f). Define the transfer operator H from the sound source s to the target region t. s,t (f)=p t (f) / X s (f), where X s(f) is the calibrated reference signal for this sound source; H s,t (f) represents the acoustic transfer operator from the s-th type sound source to the target region t within the palace at frequency f; p t (f) represents the complex sound pressure at the target region t inside the uterus, obtained by solving the parameterized biophysical sound propagation model at frequency f. It can be the volume-average sound pressure of the target region inside the uterus or the sound pressure at a specified target point; X s (f) represents the calibrated reference signal of the s-th sound source, used to normalize the complex sound pressure at the target region within the palace; s represents the sound source category, t represents the target region within the palace, and f represents the frequency. This formula represents the normalization of the complex sound pressure at the target region t within the palace using the calibrated reference signal of the s-th sound source, thereby obtaining an acoustic transfer operator from the sound source to the target region that is independent of the sound source intensity.

[0074] The sound field reconstruction output module 210 can synthesize the estimated signals of one or more sound sources after propagating them to the target area through corresponding transfer operators, according to the following formula: P rec (r t ,f,τ)=∑ s∈S H s→t (f;θ*)S s,est (f,τ); Where P rec Represents the reconstructed sound field of the target area within the palace; r t The location of the target area within the uterus is indicated, such as at least one of the following: fetal head region, fetal left ear region, fetal right ear region, fetal chest region, placental adjacency region, or a predetermined spatial region in the amniotic fluid; f represents frequency; τ represents time frame or time-frequency analysis window; S represents the set of sound source categories; s represents the sound source category, taking one type of sound source from set S; H s→t (f;θ*) represents the acoustic transfer operator at frequency f from the s-th type of sound source to the target region t within the palace, under the individualized sound propagation parameter θ*; S s,est (f,τ) represents the estimated sound source signal of the s-th type of sound source at frequency f and time frame τ. The fetal heartbeat sound source reference or estimated sound source signal uses the sensor signal at the fetal chest surface projection as the sound source reference signal, or separates the fetal heartbeat component from the acoustic vibration sensor observation signals at that location and the abdominal wall space as the estimated sound source signal, corresponding to S in the formula. s,est (f,τ).

[0075] See Figure 1 and Figure 6The output may include time-domain sound pressure signal, spectrum, sound pressure level, dual-channel sound field for left and right ears, spatial sound pressure distribution in the target area, and model reliability. In an optional embodiment, the external spatial controllable sound field generation device 211 includes multiple spatially distributed sound sources 212 and a sound field controller 213. This device can use wave field synthesis, phased array, multi-speaker array, or other spatial sound field generation methods to form multiple calibration sound fields with different propagation directions, spectra, or spatial phase relationships within a predetermined area including the abdomen of the mother body. For the external spatial controllable sound field, the system does not directly use the theoretical target sound field as the boundary condition of the human body model, but instead measures the actual incident boundary conditions in the space outside the mother body surface through the abdominal outer air sound field reference sensor 143 or the measured incident boundary acquisition unit 214. This reduces the deviations caused by the limited aperture of the loudspeakers, spatial discretization, room reflection, and spatial aliasing in the model calibration.

[0076] The system can establish source-specific transfer operators for maternal speech, heartbeat / respiration, external environmental sound, and external calibration sound sources, and synthesize their reconstruction results. For the left and right ear regions of the fetus, the system can output dual-channel sound fields respectively; for the spatial region near the fetal head, the system can output three-dimensional or two-dimensional sound pressure distribution. In one embodiment, the synchronization and preprocessing module 204 can perform clock synchronization, sensor gain calibration, bandpass filtering, motion artifact detection, contact coupling state judgment, sound source activity segment identification, respiratory or heartbeat cycle segmentation, and reference signal quality assessment. Motion artifacts can be determined by inertial sensors, sensor saturation state, low-frequency abnormal energy, or multi-channel consistency anomalies.

[0077] In one implementation, the system can generate confidence levels for the sound source, observation location, and frequency band according to the following formula: C s,i,f =γ s,i 2 (f)×SNR s,i (f)×Q model (s,i,f); Where SNR represents the signal-to-noise ratio, and Q... model This indicates the model quality factor or model sensitivity evaluation. (C) s,i,f γ represents the confidence level corresponding to the s-th type of sound source, the i-th individual's outdoor observation location, and the frequency f; s,i 2 (f) represents the squared coherence function of the amplitude of the s-th sound source and the i-th external observation position at frequency f, which can be calculated according to the previous formula; SNR s,i (f) represents the signal-to-noise ratio of the s-th type of sound source and the i-th external observation location at frequency f; Q model(s,i,f) represents the model quality factor or model sensitivity evaluation corresponding to the s-th type of sound source, the i-th external observation location, and the frequency f; this formula represents the confidence level of generating the sound source, observation location, and frequency band based on the amplitude squared coherence function, signal-to-noise ratio, and model quality factor or model sensitivity evaluation; this confidence level can be used to eliminate low-quality observations, adjust weights, suggest rearranging sensors, or output the uncertainty range of the reconstruction results.

[0078] Model confidence can be used to eliminate low-quality observations, adjust weights, suggest sensor rearrangements, or output the uncertainty range of reconstruction results.

[0079] In one implementation, the observation point adaptive selection module 207 calculates the model sensitivity matrix for candidate observation locations and selects the target observation location from the candidate set based on parameter observability. An exemplary selection relationship can be expressed as: ; J R J represents the model sensitivity matrix corresponding to the set of candidate observation locations. R H J represents R The conjugate transpose of , W represents the weight matrix, Φ represents the observability evaluation function constructed based on condition number, determinant, minimum singular value, Fisher information or parameter estimation covariance; R* represents the set of target observation locations selected from the set of candidate in vitro observation locations; arg R max represents the set of observation locations R when the subsequent observability evaluation function reaches its maximum value; G R model This represents the propagation characteristics of the model prediction under the candidate observation location set R; The model predicts the propagation feature G. R model The partial derivative of the acoustic propagation parameter vector θ to be identified; this formula represents: calculating the model sensitivity matrix J for the candidate external observation location set R. R And construct the observability evaluation function Φ(J) R H WJ R From this set of candidate observation locations, select the set of observation locations that maximizes the observability evaluation function as the target observation location set R*.

[0080] In addition to the observation location on the body surface, when configuring the external space controllable sound field generation device 211, the propagation direction, frequency or spatial phase relationship of the external calibration sound field can also be selected as candidate experimental conditions.

[0081] In a specific embodiment employing finite difference sensitivity and greedy point selection, let the set of candidate observation locations be R. cand={r1,…,r M The parameter vector to be identified is θ=[θ1,…,θ] K At the current prior parameter θ0, a perturbation Δθ is applied to the k-th parameter. k =0.01×(U k -L k ), calculate θ0+Δθ respectively k With θ0-Δθ k The propagation characteristics predicted by the model for each sound source, each candidate point, and each frequency.

[0082] Sensitivity is calculated based on the central difference: J k ≈(G(θ0+Δθ k )-G(θ0-Δθ k )) / (2Δθ k ); It is then normalized according to the parameter range and the variance of the observation noise. Wherein, J k The sensitivity of the model's predicted propagation features to the k-th acoustic propagation parameter to be identified can be represented by a sensitivity matrix J. R The column or component corresponding to the k-th parameter; k represents the k-th parameter in the acoustic propagation parameter vector θ to be identified; θ k Δθ represents the k-th parameter component in the acoustic propagation parameter vector θ to be identified, i.e., the acoustic propagation parameter vector to be identified at the parameter identification initialization or the current iteration; k This represents the perturbation step size of the k-th parameter, preferably Δθ. k =0.01×(U k -L k ), where U k and L k These are the upper and lower bounds of the k-th acoustic propagation parameter to be identified, respectively; G(θ0+Δθ) k This indicates that a positive perturbation Δθ is applied to the k-th parameter based on θ0. k At that time, the model predicts the propagation characteristics of each sound source, each candidate observation location, and each frequency; G(θ0-Δθ) k This indicates that a negative perturbation Δθ is applied to the k-th parameter based on θ0. k At that time, the model predicts the propagation characteristics of each sound source, each candidate observation location, and each frequency; this formula represents the sensitivity of the model predicting the propagation characteristics to the k-th parameter using the central difference approximation, that is, the difference between the model predicting the propagation characteristics under positive and negative perturbations divided by twice the perturbation step size; after calculation, it can be normalized according to the parameter range and the variance of the observed noise to form the sensitivity matrix J. R And participate in the observability evaluation function Φ(J) R H WJ R ).

[0083] Calculate F(R) = J(R) for the current set of observation points R. H WJ(R)+εI; The maximum or minimum singular value of logdetF(R) is used as the observability index. Starting from the near-source reference point that must be retained, a candidate point that maximizes the increment of the above index is added each time; in a specific implementation, up to 8 off-site observation points are selected, or the process stops when the relative gain of the index brought by the newly added points is less than 5%.

[0084] In the tissue equivalent phantom verification implementation, a tissue equivalent phantom can be constructed including a skin / fat equivalent layer, a muscle equivalent layer, a uterine wall equivalent layer, and an amniotic fluid equivalent cavity, with an internal verification sensor installed inside the phantom. The skin / fat equivalent layer is located on the outermost layer, the muscle equivalent layer is located inside the skin / fat equivalent layer, the uterine wall equivalent layer is located inside the muscle equivalent layer, and the amniotic fluid equivalent cavity is located within the cavity enclosed by the uterine wall equivalent layer. The internal verification sensor is positioned at a target location within the amniotic fluid equivalent cavity. An external observation array 203 and a reference sound source are arranged externally. The system reconstructs the sound field at the target location inside the phantom based on the external signals and compares it with the measurement results of the internal verification sensor.

[0085] In phantom validation or research validation, the frequency response error can be calculated according to the following formula: E(f) = 20log 10 [A rec (f) / A ref (f)]; Where A rec (f) represents the amplitude of the reconstructed sound field at frequency f, A ref (f) represents the amplitude of the reference sound field at frequency f as measured by the internal validation sensor or standard hydrophone. In addition to the above formula, waveform correlation coefficient, time delay error, sound pressure level error, mean square error of the frequency band, or spatial distribution error can also be used to evaluate the accuracy of the model. The internal validation sensor is only used as a phantom or research validation tool and does not constitute a necessary step for intrauterine insertion in clinical use.

[0086] In another embodiment, the system can utilize publicly available or self-built intrauterine sound propagation data from equivalent human, animal, or tissue phantoms to form a priori database of intrauterine sound propagation. This database can record metadata such as species, gestational age, amniotic membrane status, excitation direction, measurement location, tissue parameters, sensor model, and calibration conditions, and is used to define parameter ranges, construct prior parameter distributions, or train reduced-order models. In yet another embodiment, when the target pregnant woman only has conventional ultrasound data and has not undergone magnetic resonance imaging, the system can employ an axisymmetric or locally simplified three-dimensional geometric model, using abdominal wall thickness, placental location, amniotic fluid depth, fetal head location, and sensor surface location as the main geometric inputs; when three-dimensional ultrasound or magnetic resonance imaging data is available, the system can establish a more refined three-dimensional boundary element or finite element model. The output of this invention is preferably limited to acoustic engineering quantities such as digital reconstruction results of the intrauterine sound field, acoustic transfer operators, sound propagation parameters, sound pressure level, spectrum, or model confidence. This invention does not use disease diagnosis or treatment methods as necessary technical steps.

[0087] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or changes made by those skilled in the art within the technical scope disclosed in the present invention, based on the technical solution and concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A non-invasive intrauterine sound field reconstruction method for constructing the uterine acoustic environment, characterized in that, Includes the following steps: Obtain individual anatomical data of the target pregnant woman, and establish a parameterized biophysical acoustic propagation model based on the individual anatomical data; the parameterized biophysical acoustic propagation model includes multiple divided tissue regions, at least one intrauterine target region, and a set of acoustic propagation parameters to be identified; The set of acoustic propagation parameters to be identified includes at least sound speed and frequency-related attenuation parameters. Sound propagation in each tissue region is described by wave equation or Helmholtz equation and solved by numerical method. Simultaneously acquire sound source reference signals from at least two different sound sources and sound vibration observation signals set at multiple different locations outside the mother's body, wherein at least one sound source is a physiological sound vibration source naturally generated by the mother or fetus. For each type of sound source, the measured propagation characteristics from the sound source to each external observation location are determined based on the sound source reference signal and each sound vibration observation signal. In the parameterized biophysical sound propagation model, the predicted propagation characteristics from the sound source to each in vitro observation location are calculated according to the sound source location and sound propagation boundary conditions. Using the same set of acoustic propagation parameters to be identified, the differences between the measured propagation characteristics and the predicted propagation characteristics corresponding to at least two types of sound sources are jointly constrained to obtain the individualized sound propagation parameters of the target pregnant woman. Based on the individualized sound propagation parameters, an acoustic transfer operator from at least one type of sound source to the intrauterine target region is calculated. Based on the acoustic transfer operator and the estimated sound source signal of the corresponding sound source, the reconstructed sound field of the intrauterine target region is calculated.

2. The non-invasive intrauterine sound field reconstruction method for constructing the uterine acoustic environment as described in claim 1, characterized in that, The at least two different sound sources include at least two of the following: sound vibration sources generated by maternal vocalization, sound vibration sources generated by maternal heartbeat, sound vibration sources generated by maternal respiration, sound vibration sources generated by fetal heartbeat, and external sound fields generated by a spatially controllable sound field generation device outside the maternal body. Specifically, for maternal vocalization, a contact-type sound vibration sensor located in the larynx is used to obtain a maternal vocalization reference signal; for maternal heartbeat, a sound vibration sensor in the precordial region and sternal region and an electrocardiogram signal are used to obtain a heartbeat reference signal; and for the external sound field, at least one reference acoustic sensor located in the space outside the maternal body surface is used to obtain an actual incident sound field reference signal.

3. The non-invasive intrauterine sound field reconstruction method for constructing the uterine acoustic environment as described in claim 1, characterized in that, Multiple external observation locations include multiple locations selected from a set consisting of the larynx, sternum, xiphoid process region, left precordial region, upper abdomen, mid-abdomen, lower abdomen, left and right lateral abdomen, lumbosacral region, iliac crest region, fetal head surface projection area determined by medical imaging, fetal chest surface projection area, placental surface projection area, and the region opposite the placenta; a surface coordinate system is established using predetermined anatomical landmarks, and the spatial positions of each acoustic vibration sensor are recorded; the external observation locations are functionally divided into a near-source reference point set, a trunk propagation reference point set, an abdominal wall spatial observation point set, an individual anatomical feature projection point set, and an airborne acoustic field parameter set. The test point set includes: a near-source reference point set for obtaining near-source references of natural physiological sound vibration sources; a trunk propagation reference point set for characterizing the intermediate response of natural physiological sound vibrations propagating along the maternal trunk; an abdominal wall space observation point set for covering the uterine surface projection area; an individual anatomical feature projection point set for determining the surface projection positions of the fetal head, fetal chest, and placenta based on medical imaging; and an airborne sound field reference point set for measuring the actual sound field boundary of the space outside the maternal surface. The sound vibration observation signals are acquired by contact microphones, accelerometers, piezoelectric sensors, inertial acoustic sensors, fiber optic vibration sensors, or combinations thereof.

4. The non-invasive intrauterine sound field reconstruction method for constructing the uterine acoustic environment as described in claim 1, characterized in that, For the mother's vocalization, the laryngeal contact acoustic vibration reference signal and air acoustic reference signal corresponding to the same mother's vocalization event are acquired simultaneously. The volume conduction propagation characteristics of the mother's vocalization propagating through the trunk tissue to the abdominal wall and the air conduction propagation characteristics coupled to the abdominal wall after propagation through the air are determined respectively. The volume conduction propagation characteristics and the air conduction propagation characteristics are used to constrain the same set of acoustic propagation parameters to be identified. The volume conduction propagation path includes at least the intermediate propagation reference point in the larynx and the sternum and xiphoid process regions.

5. The non-invasive intrauterine sound field reconstruction method for constructing the uterine acoustic environment as described in claim 1, characterized in that, The placental surface projection position is determined based on medical imaging data, and a contralateral observation position is set in a non-placental region that has a predetermined spatial correspondence with the placental surface projection position. The contralateral observation position is a non-placental surface position that is at a similar cephalotail height to the placental surface projection position and where the main propagation path from this surface position to the target area in the uterus does not pass through the main area of ​​the placenta. Based on the difference in measured propagation characteristics between the placental surface projection position and the contralateral observation position, the local non-uniform tissue propagation parameters in the parameterized biophysical sound propagation model are constrained. Furthermore, the fetal head and fetal chest surface projection positions are determined based on medical imaging to form observational constraints on the propagation of natural sound sources inside the fetus to the outside of the mother.

6. The non-invasive intrauterine sound field reconstruction method for constructing the uterine acoustic environment as described in claim 1, characterized in that, A set of candidate external observation locations is established based on individual anatomical structure data. The sensitivity of the model-predicted propagation characteristics corresponding to each candidate external observation location to at least some of the acoustic propagation parameters to be identified is calculated. The observability index of the parameters is determined according to the sensitivity matrix composed of the sensitivity. The target observation location for acoustic vibration observation is selected from the candidate external observation locations. The observability index of the parameters includes at least one of the condition number, determinant, minimum singular value, Fisher information, and parameter estimation covariance of the sensitivity matrix. In the case of using an externally controllable sound field, the propagation direction and frequency of the external calibration sound field are also selected according to the observability index of the parameters.

7. The non-invasive intrauterine sound field reconstruction method for constructing the uterine acoustic environment as described in claim 1, characterized in that, The measured propagation characteristics include at least one of the following: complex frequency response, amplitude-frequency response, phase-frequency response, cross-power spectrum, transfer function estimate, amplitude squared coherence function, time delay, group delay, impulse response, and time-frequency propagation characteristics. For a sound source s and the i-th observation position, the measured propagation characteristics are determined by the cross-spectrum of the sound source reference signal and the acoustic vibration observation signal, and the amplitude squared coherence function is calculated. In the joint constraints, the weights are determined based on coherence, signal-to-noise ratio, measurement confidence, and model sensitivity. The parameter constraint terms are established based on at least one of the following: acoustic parameters of human tissue, measured data of acoustic propagation in the uterus of humans or animals, tissue equivalent phantom data, and individual anatomical measurement data. When the coherence, signal-to-noise ratio, contact coupling quality, or model residual does not meet the predetermined conditions, the weight of the corresponding sound source, observation position, or frequency band is reduced or it is excluded from the joint constraints.

8. The non-invasive intrauterine sound field reconstruction method for constructing the uterine acoustic environment as described in claim 1, characterized in that, The parameterized biophysical sound propagation model determines geometric parameters based on at least one of ultrasound, three-dimensional ultrasound, and magnetic resonance imaging. These geometric parameters include at least one of the following: abdominal wall contour, subcutaneous fat layer thickness, muscle layer thickness, uterine boundary, amniotic fluid region, placental location, fetal location, fetal orientation, fetal head location, and maternal spine or pelvic structure. Sound propagation in each tissue region is described by the wave equation, the Helmholtz equation, or their numerical equivalent models, and is calculated using the boundary element method, finite element method, finite difference time-domain method, spectral method, reduced-order model, or a combination thereof. The complex wave number of each tissue region is determined by the sound velocity and frequency-dependent attenuation parameters, and the sound pressure satisfies the Helmholtz equation or its discrete equivalent form.

9. A non-invasive intrauterine sound field reconstruction system for constructing the acoustic environment of the uterus, characterized in that, include: The individual anatomical structure data acquisition module is used to acquire the individual anatomical structure data of the target pregnant woman. A multi-source reference acquisition module is used to simultaneously acquire sound source reference signals from at least two different sound sources; An external observation array includes multiple acoustic and vibration sensors located at different positions outside the mother body to simultaneously acquire multiple acoustic and vibration observation signals; The synchronization and preprocessing module is used to synchronize and preprocess the sound source reference signal and the sound vibration observation signal; the measured propagation feature extraction module is used to determine the measured propagation characteristics from the sound source to each external observation position based on the sound source reference signal and the sound vibration observation signal of each type of sound source. The parameterized biophysical acoustic propagation model module is used to establish a parameterized biophysical acoustic propagation model based on individual anatomical data, which includes multiple tissue regions, at least one intrauterine target region, and a set of acoustic propagation parameters to be identified. It calculates the predicted propagation characteristics from each sound source to various in vitro observation locations according to the sound source location and acoustic propagation boundary conditions corresponding to each type of sound source. The acoustic propagation in each tissue region is described by wave equations or Helmholtz equations and solved numerically. The multi-source joint parameter identification module is used to jointly constrain the differences between the measured and predicted propagation characteristics corresponding to at least two types of sound sources using the same set of acoustic propagation parameters to obtain individualized acoustic propagation parameters for the target pregnant woman. The acoustic transfer operator calculation module is used to calculate the acoustic transfer operator from at least one type of sound source to the intrauterine target region based on the individualized acoustic propagation parameters. The sound field reconstruction output module is used to calculate the reconstructed sound field of the target area within the palace based on the acoustic transfer operator and the estimated sound source signal of the corresponding sound source.

10. A non-invasive intrauterine sound field reconstruction system for constructing the acoustic environment of the uterus as described in claim 9, characterized in that, It also includes an adaptive observation point selection module and an external space controllable sound field generation device; the adaptive observation point selection module is configured to select a target observation location for joint parameter identification based on individual anatomical structure data, model sensitivity of candidate observation locations, and parameter observability index; the external space controllable sound field generation device controls multiple spatially distributed sound sources to form multiple calibrated sound fields with different propagation directions, spectra, and spatial phase relationships in a predetermined area including the mother's abdomen, and measures the actual incident boundary conditions through a reference acoustic sensor set in the space outside the mother's body surface, using the actual incident boundary conditions as the boundary input of the parameterized biophysical sound propagation model.

Images