Magnetocardiogram-based pulmonary arterial hypertension feature extraction method and system

Abstract

This invention relates to the field of cardiac electrophysiological diagnostic technology, and discloses a method and system for extracting pulmonary hypertension features based on magnetocardiography. The method includes: acquiring the cardiac magnetic field strength distribution matrix of the subject and calculating the magnetic field gradient direction vector; extracting the zero-position feature trajectory of the magnetic field gradient direction vector to determine the respiratory displacement compensation value and correct the spatial coordinates of the magnetic field distribution; analyzing the distribution symmetry of the magnetic field gradient direction vector to establish a rotation axis reference; calculating the angular displacement offset of the magnetic field gradient direction vector relative to the rotation axis reference and generating gradient rotation feature parameters; determining the angular distribution consistency index of the magnetic field gradient direction vector and establishing the phase-locked characteristic time, and calibrating the gradient rotation feature parameters. This invention, by constructing a geometric reference for the topological evolution of the magnetic field, eliminates the influence of anatomical position deviations of the subject, captures the rotational distortion of the right ventricular electrical conduction path, and improves the accuracy of pathological feature extraction.

Description

Technical Field

[0001] This invention belongs to the field of cardiac electrophysiological diagnostic technology, and particularly relates to a method and system for extracting pulmonary hypertension features based on magnetocardiography. Background Technology

[0002] Currently, the use of sensor arrays to collect the distribution of magnetic flux in the chest and characterize cardiac magnetic field signals plays an important role in the evaluation of cardiac function and the identification of early electrophysiological changes. The electromechanical coupling process of the heart is accompanied by the continuous evolution of the current field in three-dimensional space, and the spatial orientation characteristics of the magnetic field gradient have a physical correspondence with the myocardial conduction path. Pulmonary hypertension causes chronic pressure overload of the right ventricle, which in turn leads to asymmetric distortion of myocardial electromechanical coupling and conduction path torsion. Conventional feature extraction methods focus on global mapping of magnetic signals or assessment of single-point peak intensity. Due to the differences in the subjects' axillary axis, chest wall anatomy, and physical displacement caused by respiration, the absolute magnetic field strength parameter fluctuates drastically, resulting in a decrease in the specificity of feature recognition, and requiring the magnetic field acquisition device to have extremely high installation accuracy.

[0003] The dependence on detection conditions and the anatomical state of the subject also exists in conventional imaging diagnostic procedures. Existing technologies attempt to introduce signal quality algorithm correction mechanisms. For example, Chinese invention patent CN120360592A discloses a method and related device for non-invasive diagnosis of pulmonary hypertension. It constructs a diagnostic model based on the quality grading of the degree of missing tricuspid regurgitation spectrum and matching differentiated calculation formulas. Although such methods alleviate the impact of data quality fluctuations through grading logic, the core technology still does not escape the absolute dependence on the clarity of the physical acoustic window and the blood flow signal of a specific anatomical plane. Subjects may have limited acoustic windows due to lung diseases such as emphysema, or be in the early stage of disease before the formation of regurgitation dynamics. It is difficult to remove the co-frequency interference introduced by respiratory displacement from the source by weighted correction methods based on posterior data, and it is difficult to detect microtopological distortions in the initial stage of myocardial electromechanical coupling in the absence of a stable physical benchmark.

[0004] Therefore, the technical problem to be solved by this invention is how to extract specific topological features reflecting the pathological evolution of pulmonary hypertension without relying on the physical location of the heart and a specific magnetic shielding environment, and how to construct an evaluation mechanism that can adaptively eliminate respiratory and environmental interference. Summary of the Invention

[0005] This invention provides a method for extracting pulmonary hypertension features based on magnetocardiography, comprising the following steps:

[0006] Step S1: Acquire the multi-channel spatiotemporal data of cardiac magnetic induction intensity distribution of the subject, and map the spatiotemporal data of cardiac magnetic induction intensity distribution to the sensor array coordinate system to construct the magnetic induction intensity distribution matrix;

[0007] Step S2: Perform spatial difference operation on the magnetic induction intensity distribution matrix to calculate the magnetic field gradient direction vector corresponding to each measuring point position;

[0008] Step S3: Extract the zero-position feature trajectory of the magnetic field gradient direction vector in the sensor array plane, calculate the displacement offset of the zero-position feature trajectory with the cardiac cycle to determine the respiratory displacement compensation value, and correct the spatial coordinates of the magnetic field gradient direction vector according to the respiratory displacement compensation value.

[0009] Step S4: Analyze the spatial distribution symmetry of the corrected magnetic field gradient direction vector within the cardiac cycle, determine the rotational symmetry axis of the magnetic field intensity distribution, and use the rotational symmetry axis as the polar reference for calculating the angular displacement offset of the magnetic field gradient direction vector.

[0010] Step S5: Calculate the angular displacement offset of each magnetic field gradient direction vector relative to the rotational symmetry axis to obtain the gradient rotation characteristic parameters characterizing the electromechanical coupling characteristics of the heart.

[0011] Step S6: Divide the rotational symmetry axis circumferentially into multiple angular intervals, statistically analyze the distribution probability of the magnetic field gradient direction vector in each angular interval, determine the angular distribution consistency index based on the distribution probability, identify the minimum point of the angular distribution consistency index as the phase-locked characteristic moment of the cardiac cycle, and use the phase-locked characteristic moment to periodically calibrate the gradient rotation feature parameters, and output the pulmonary hypertension feature extraction results.

[0012] Preferably, the process of determining the respiratory displacement compensation value in step S3 includes: constructing a cardiac magnetic dipole model based on the magnetic induction intensity distribution matrix, extracting the spatial center coordinates of the cardiac magnetic dipole model, and determining the respiratory displacement compensation value by monitoring the quasi-periodic drift of the spatial center coordinates within the sampling period, wherein the frequency of the sampling period is not less than 1000Hz.

[0013] Preferably, the process of determining the rotational symmetry axis in step S4 includes: collecting background magnetic field observation data of the edge region of the sensor array, using the background magnetic field observation data to correct the zero-point gain of the sensor array, and determining the spatial orientation of the rotational symmetry axis based on the corrected vector field distribution law.

[0014] Preferably, the process of calculating the angular distribution consistency index in step S6 follows the following formula: ,in, Indicates time The angular distribution consistency index, The direction vector of the magnetic field gradient is represented in the th case. The probability distribution of each angle interval. The total number of angles is determined by a preset angle.

[0015] Preferably, after outputting the pulmonary hypertension feature extraction results, the method further includes: extracting the high-frequency signal component in the magnetic induction intensity distribution, calculating the residual energy projection value of the high-frequency signal component on the normal plane, and extracting the spatial heterogeneity features of the subject's cardiac electrical activity based on the spatial distribution of the residual energy projection value.

[0016] Preferably, the process of extracting spatial heterogeneity features of the subject's cardiac electrical activity includes: calculating the spatial energy gradient of the residual energy projection value in the normal plane, and identifying the local perturbation features of the magnetic field vector during myocardial depolarization based on the spatial energy gradient.

[0017] Preferably, after outputting the pulmonary hypertension feature extraction results, the method further includes: extracting the gradient rotation feature parameters corresponding to the cardiac systolic phase, and performing correlation calculations between the gradient rotation feature parameters corresponding to the cardiac systolic phase and the pre-excitation magnetic field data at the end of diastole to generate a cross-phase electromagnetic coupling coefficient.

[0018] Preferably, the process of generating the transphase electromagnetic coupling coefficient includes: extracting the magnetic principal axis deviation parameter during the repolarization process, and using the magnetic principal axis deviation parameter to perform gain correction on the transphase electromagnetic coupling coefficient in order to determine the spatial stability index of the electrocardiogram conduction vector.

[0019] Preferably, the process of calculating the magnetic field gradient direction vector in step S2 includes: direction and Spatial partial derivatives of the magnetic flux density distribution matrix are calculated along the direction to obtain the horizontal magnetic gradient components. With vertical magnetic gradient components ,Depend on and The synthesized vector determines the magnetic field gradient direction vector, and the pulmonary hypertension feature extraction results are used to construct a quantitative mapping logic for pulmonary artery pressure status to output a feature index indicating changes in pulmonary vascular resistance.

[0020] A pulmonary hypertension feature extraction system based on magnetocardiography includes a data acquisition unit, a gradient calculation unit, a displacement compensation unit, a benchmark calibration unit, a feature extraction unit, a phase-locked output unit, and a prediction modeling unit.

[0021] The data acquisition unit is used to acquire multi-channel spatiotemporal data of cardiac magnetic induction intensity distribution of the subject, and to map the spatiotemporal data of cardiac magnetic induction intensity distribution to the sensor array coordinate system to construct a magnetic induction intensity distribution matrix;

[0022] The gradient calculation unit is used to perform spatial difference operations on the magnetic induction intensity distribution matrix and calculate the magnetic field gradient direction vector corresponding to each measurement point.

[0023] The displacement compensation unit is used to extract the zero-position feature trajectory of the magnetic field gradient direction vector in the sensor array plane, and determine the respiratory displacement compensation value based on the displacement offset of the zero-position feature trajectory with the cardiac cycle, so as to correct the spatial coordinates of the magnetic field gradient direction vector.

[0024] The reference calibration unit is used to analyze the spatial distribution symmetry of the corrected magnetic field gradient direction vector within the cardiac cycle to determine the rotational symmetry axis, and uses the rotational symmetry axis as the polar reference.

[0025] The feature extraction unit is used to calculate the angular displacement offset of the magnetic field gradient direction vector relative to the rotational symmetry axis to generate gradient rotation feature parameters.

[0026] The phase-locked output unit is used to statistically analyze the distribution probability of the magnetic field gradient direction vector in each angular interval and determine the angular distribution consistency index. The minimum point of the angular distribution consistency index is used to establish the phase-locked characteristic time, and the gradient rotation characteristic parameters are calibrated based on the phase-locked characteristic time to output the feature extraction results.

[0027] The predictive modeling unit is used to perform quantization mapping logic based on the feature extraction results to output a feature index characterizing the pulmonary artery pressure state.

[0028] Compared with existing technologies, the pulmonary hypertension feature extraction method based on magnetocardiography of the present invention has the following advantages:

[0029] 1. In the pulmonary hypertension feature, a magnetic flux virtual axis of symmetry is constructed using magnetic gradient tensor flow. The individual differences in the cardiac anatomy of the subjects and the installation position deviation of the chest magnetic field acquisition device are transformed into the geometric rotation of the magnetic flux field. By calculating the angular displacement offset of the gradient vector relative to the virtual axis of symmetry, feature parameters reflecting pulmonary artery pressure can be extracted without determining the physical center of the heart, thus eliminating the interference of individual anatomical differences on the feature extraction process.

[0030] 2. By extracting the spatial zero invariant from the magnetic flux data sequence and identifying its geometric centroid trajectory, the spatial geometric characteristics of the magnetocardiogram signal itself are used to achieve real-time compensation for the subject's respiratory motion. By applying dynamic spatial weight correction to the gradient vector flow, the benchmark consistency of the topological torque calculation is maintained even when the subject has respiratory distress or chest rise and fall, ensuring the continuity of the detection process.

[0031] 3. By cross-phase coupling of the time-varying characteristics of topological torque during systole with the characteristics of pre-excitation magnetic field at end-diastole, and combining this with the spatial principal axis deviation during repolarization for consistency verification, a closed-loop electromagnetic coupling evaluation covering the entire cycle of cardiac depolarization, repolarization, and diastolic filling is formed. This multi-mechanism synergy not only enhances the ability to capture early compliance changes in pulmonary hypertension, but also distinguishes signal distortions caused by physical position drift and pathological electrophysiological remodeling through consistency residual assessment of spatial manifold, thereby improving the specificity of diagnostic features. Attached Figure Description

[0032] Figure 1 This is the main flowchart of the feature extraction method for introducing the respiratory displacement compensation mechanism in this invention;

[0033] Figure 2 This is a schematic diagram of the functional unit architecture and data interaction logic of the prediction system of the present invention. Detailed Implementation

[0034] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0035] It should be noted that all directional and positional terms used in this invention, such as: up, down, left, right, front, back, vertical, horizontal, inner, outer, top, bottom, transverse, longitudinal, center, etc., are only used to explain the relative positional relationship and connection between components in a specific state (as shown in the accompanying drawings). They are only for the convenience of describing this invention and do not require that this invention be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention. In addition, the descriptions of "first," "second," etc., in this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated.

[0036] In the description of this invention, unless otherwise explicitly specified and limited, the terms installation, connection, and linking should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections; they can refer to direct connections or indirect connections through an intermediate medium; they can refer to the internal communication between two components. For those skilled in the art, the specific meaning of the above terms in this invention can be understood according to the specific circumstances.

[0037] In the description of this specification, references to the terms "an embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example, and the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0038] A method and system for pulmonary hypertension feature extraction based on magnetocardiography (MCG) is proposed. The system acquires the cardiac magnetic field intensity distribution matrix of the subject, calculates the magnetic field gradient direction vector, extracts the zero-position feature trajectory of the magnetic field gradient direction vector to determine the respiratory displacement compensation value, corrects the spatial coordinates of the magnetic field distribution, analyzes the distribution symmetry of the magnetic field gradient direction vector to establish a rotation axis reference, calculates the angular displacement offset of the magnetic field gradient direction vector relative to the rotation axis reference, generates gradient rotation feature parameters, determines the angular distribution consistency index of the magnetic field gradient direction vector and establishes the phase-locked feature time, calibrates the gradient rotation feature parameters, and outputs the pulmonary hypertension feature extraction results. The system includes a data acquisition unit, a gradient calculation unit, a displacement compensation unit, a reference calibration unit, a feature extraction unit, a phase-locked output unit, and a prediction modeling unit. To establish a mechanism for eliminating anatomical position deviations in the subject, the system utilizes the data acquisition unit to acquire multi-channel spatiotemporal data of the subject's cardiac magnetic field intensity distribution, which is then distributed in the anterior chest region of the subject. Sensor array acquisition, sampling frequency set to The data acquisition unit maps the collected magnetic flux data sequence to the sensor array coordinate system to construct a magnetic induction intensity distribution matrix. ,in, and These represent the physical location indices of the sensors within the array plane. As a time index, this distribution matrix serves as the initial input for subsequent spatial topology analysis; based on the influence of spindle deflection on magnetic signal intensity, the system utilizes gradient operation units to perform spatial difference operations on the magnetic induction intensity distribution matrix. direction and Calculate the spatial partial derivatives in each direction to obtain the horizontal magnetic gradient components. With vertical magnetic gradient components ,Depend on and The composite vector determines the direction vector of the magnetic field gradient. The system calculates the difference in magnetic field strength between adjacent measuring points and divides it by the physical distance to obtain the magnetic field gradient in the local space. This direction vector reflects the projection of the current field onto the sensor plane.

[0039] To eliminate physical displacement interference caused by the subject's breathing, the displacement compensation unit extracts the zero-position feature trajectory of the magnetic field gradient direction vector in the sensor array plane, extracts the spatial manifold formed by connecting sampling points with zero magnetic induction intensity, and monitors the quasi-periodic drift of this zero-position feature trajectory within the sampling period to determine the breathing displacement compensation value. When the geometric centroid of the zero-position manifold deviates due to breathing, the system performs real-time correction on the spatial coordinates of the magnetic field gradient direction vector based on the offset, ensuring the consistency of spatiotemporal coordinates in subsequent feature extraction processes. Considering individual differences in the subject's anatomical structure, the benchmark calibration unit analyzes the corrected magnetic field gradient direction vector. By identifying the center of rotation of the magnetic field distribution and establishing a rotation axis reference based on the spatial distribution symmetry during isochoric contraction, the system acquires background magnetic field observation data from the edge region of the sensor array, constructs an environmental multipole distribution model, and calculates the background interference vector. This background interference vector is then used to correct the spatial orientation of the rotation axis reference, making it independent of the physical location of the heart and serving as a reference axis for topological evolution. To capture rotational distortion of the right ventricular conduction path, the feature extraction unit calculates the angular displacement offset of each magnetic field gradient direction vector relative to the rotation axis reference, generating gradient rotation feature parameters. These parameters represent the topological torque value characterizing the electromechanical coupling features of the heart. The system calculates the change in the angle between the gradient vector direction and the reference feature vector direction, and obtains the angular displacement curve that evolves over time. By monitoring the transient slope of this curve at the beginning of ventricular systole, a physical quantity reflecting the change in right ventricular compliance is obtained.

[0040] Considering the impact of arrhythmia on phase recognition, the phase-locked output unit divides the circumferential direction of the rotating axis into... Given a preset angle interval, the probability distribution of the magnetic field gradient direction vector within each angle interval is statistically analyzed. And determine the angular distribution consistency index based on the distribution probability. The calculation formula is as follows: ,in, Indicates time The angular distribution consistency index, The direction vector of the magnetic field gradient is represented in the th case. The probability distribution of each angle interval. To divide the total number of angles, the phase-locked output unit will... The minimum point is identified as the phase-locked characteristic moment of the cardiac cycle, and this moment is used to periodically calibrate the gradient rotation characteristic parameters, outputting the pulmonary hypertension feature extraction results. To provide an assessment dimension reflecting the heterogeneous changes in myocardial tissue, the system extracts the high-frequency signal component from the magnetic induction intensity distribution, calculates the residual energy projection value of this component on the normal plane, analyzes the spatial energy gradient of the projection value, and simultaneously, the prediction modeling unit obtains the pre-excitation magnetic field data of the subject within the end-diastolic time window and extracts the average direction of its gradient vector as the reference polar axis to calculate the gradient rotation characteristic parameters established at the systolic phase-locked characteristic moment. The spatial projection residual between the reference polar axis and the reference polar axis is used, and a recursive mean correction based on a time decay factor is applied to the spatial projection residual to generate the cross-phase electromagnetic coupling coefficient. The pre-excitation sampling window length at the end of diastole is selected as... ms to The system uses a second-order Butterworth low-pass filter to smooth the generated coefficient sequence, and the final output quantitative characteristic index can stably reflect the electromechanical compliance state of the right ventricle. The right ventricular compliance characterization index is calibrated, and the final output is a quantitative characteristic index representing the pulmonary artery pressure state. When processing subject data, the system performs anonymization processing in the local processor, converting the collected raw magnetic signals into neutral physical quantity characteristics. All data processing tasks are executed locally on the sensor acquisition terminal to avoid transmitting raw data involving the subject's personal information to the cloud, ensuring privacy protection while outputting technical indicators.

[0041] Example 1: In monitoring subjects with differences in thoracic anatomy and accompanied by shallow and rapid breathing, the spatial orientation of the principal vector of the cardiac magnetic field shifts quasi-periodically with the rise and fall of the thoracic cavity. Conventional fixed spatial coordinate mapping methods cannot distinguish between physical displacement and pathological electrophysiological distortions. Especially when the long axis of the subject's heart anatomy is three-dimensionally deflected relative to the sensor plane, fluctuations in absolute intensity parameters can mask the characteristic evolution of the right ventricle under pressure load. The system acquires the magnetic induction intensity distribution matrix. And calculate the magnetic field gradient direction vector. The displacement compensation unit extracts the zero-position feature trajectory in the vector field, determines the respiratory displacement compensation value by monitoring the drift trajectory of the geometric centroid of the zero-position manifold with the cardiac cycle, and uses the compensation value to correct the spatial coordinates of the magnetic field distribution, so that the magnetic field observation coordinate system is anchored to the zero-position physical boundary of the cardiac magnetic field itself. Under the premise of maintaining the consistency of the spatial topology, the measurement noise caused by respiratory motion is eliminated, providing a stable physical reference for capturing the right ventricular torsion characteristics.

[0042] The benchmark calibration unit analyzes the spatial distribution symmetry of the corrected magnetic field gradient direction vector during isochoric contraction. It determines the rotation axis reference by searching for the axis with the highest directional consistency in the gradient vector field. The feature extraction unit generates gradient rotation feature parameters reflecting the torsional intensity of the right ventricular conduction path based on the angular displacement offset of the magnetic field gradient direction vector relative to the rotation axis reference. The phase-locked output unit statistical magnetic field gradient direction vector is in The probability distribution of a preset angle interval Calculate time Angular distribution consistency index The formula for calculating this indicator is: ,in, The direction vector of the magnetic field gradient is at the 1st... The probability distribution of each angle interval. The phase-locked output unit identifies the total number of angle interval divisions. The minimum point is used to establish the phase-locked loop characteristic time and to calibrate it. At the established phase-locked characteristic moment, the system outputs the calibrated gradient rotation characteristic parameters as the pulmonary hypertension feature extraction result. The interference of individual anatomical pose differences is eliminated by dynamic analysis of the inherent topological evolution of the magnetic field, and the original magnetic signal is anonymized in the feature extraction process.

[0043] Example 2: This experiment utilizes a magnetic field multipole simulation model to construct a simulated physical platform, verifying the stability of the prediction system in extracting right ventricular electromechanical coupling characteristics under respiratory motion disturbances. The physical platform is pre-set with a spatiotemporal magnetic induction intensity distribution matrix. And superimposed signal-to-noise ratio is dB of Gaussian white noise, while introducing frequency of Hz and peak displacement is The quasi-periodic respiratory drift component in mm, the data acquisition unit selects the sampling frequency. for Hz, sampling frequency The settings are designed to cover the bandwidth range of high-frequency components of the magnetocardiogram and suppress signal aliasing at the end of depolarization. This experiment sets up the sample group of the present invention and a control group with the displacement compensation unit disabled, and monitors the angular distribution consistency index. The evolution trend of the minimum point under different respiratory displacement amplitudes was recorded, along with the locking deviation at characteristic moments and gradient rotation characteristic parameters. The root mean square error.

[0044] Example 3: This example combines Figures 1 to 2 This section describes the method and system for feature extraction of pulmonary hypertension based on magnetocardiography, such as... Figure 1As shown, step S1 acquires multi-channel spatiotemporal data of cardiac magnetic field intensity distribution of the subject and maps the spatiotemporal data of cardiac magnetic field intensity distribution to the sensor array coordinate system to construct a magnetic field intensity distribution matrix. Step S2 performs spatial difference operation on the magnetic field intensity distribution matrix to calculate the magnetic field gradient direction vector corresponding to each measurement point. Then, step S3 extracts the zero-position feature trajectory of the magnetic field gradient direction vector in the sensor array plane, calculates the displacement offset of the trajectory with the cardiac cycle to determine the respiratory displacement compensation value, and corrects the spatial coordinates of the magnetic field gradient direction vector accordingly. Based on this, step S... 4. Analyze the spatial distribution symmetry of the corrected magnetic field gradient direction vector within the cardiac cycle, determine the rotational symmetry axis of the magnetic field intensity distribution, and use it as the polar reference for calculating the angular displacement offset. Then, execute step S5 to calculate the angular displacement offset of each magnetic field gradient direction vector relative to the rotational symmetry axis, obtain the gradient rotation feature parameters characterizing the electromechanical coupling characteristics of the heart, and finally execute step S6 to statistically analyze the distribution probability of the magnetic field gradient direction vector in each angular interval to determine the angular distribution consistency index, establish the phase-locked characteristic time, periodically calibrate the gradient rotation feature parameters, and output the pulmonary hypertension feature extraction results.

[0045] like Figure 2 As shown, the medical operator, as the main body of system interaction, intervenes in the operation process. The system acquires the spatiotemporal data of magnetic induction intensity. This data stream points to the processing module that constructs the magnetic induction intensity distribution matrix. The system calculates the magnetic field gradient direction vector and triggers the function module that performs respiratory displacement compensation. By extracting the zero-position feature trajectory, the system corrects respiratory interference. Then, the system establishes the rotation axis reference. This process relies on the calculation logic of analyzing the spatial distribution symmetry. Based on this, the system generates gradient rotation feature parameters and enters the phase-locked calibration and result output stage, which includes two sub-processes: determining the angular distribution consistency index and establishing the phase-locked feature time. Finally, the system outputs the pulmonary artery pressure feature index.

[0046] Example 4: Under dynamic monitoring conditions where the external environmental magnetic field gradient and the subject's intrinsic cardiac magnetic field gradient are linearly superimposed, the magnetic field gradient direction vector... The spatial orientation of the magnetic field is deflected due to external environmental interference. The system constructs a second-order multipole magnetic field distribution model using the magnetic induction intensity values ​​at the outer edge of the sensor array, and fits the spatial tensor components of the background magnetic field using the least squares method. It then calculates the projection amplitude of the background interference vector onto the sensing plane and uses the background interference vector to correlate with the magnetic induction intensity distribution matrix. Spatial domain subtraction is performed to obtain the corrected net-center magnetic signal; to determine the physical coordinates of the rotation axis reference, the system utilizes the corrected magnetic field gradient direction vector during the isochoric contraction period. The gradient vector consistency search procedure is executed, starting from the geometric center of the sensor array, with a radial search step size set as [value missing]. In 3D space, the orthogonal deviation between the vector direction of each measurement point and the line connecting the current search center is calculated. The physical rotation center of the gradient vector flow is found by minimizing the objective function. When the center coordinate deviation between two adjacent iterations is less than [a certain value], [the value is determined]. The search stops when mm is reached, where, and This represents the coordinate components of the rotation center within the sensor plane.

[0047] Angular distribution consistency index caused by background noise To address pseudo-minimum interference, the phase-locked loop (PLL) output unit utilizes an adaptive sliding window method to smooth the instantaneous directional entropy sequence, setting the window length to [value missing]. ms, the system calculates after smoothing sequence pair time The second derivative ,in, As an index of angular distribution consistency, For time indexing, identify the second derivative. Greater than zero and first derivative The coordinates of zero are used as candidate phase-locked feature moments, and an entropy jump constraint is introduced, i.e. The magnitude of the decrease must exceed the average fluctuation amplitude within the sampling period. The system uses this characteristic moment to calibrate the gradient rotation characteristic parameters. The system outputs pulmonary hypertension feature extraction results. During background field subtraction and iterative search tasks, it uses a local computing unit to perform matrix operations. The output pulmonary hypertension feature extraction results are converted into a quantitative risk index without biometric features after local encryption. This index reflects the degree of electromechanical coupling deviation of the right ventricle after excluding environmental noise. This improves the extraction accuracy in complex electromagnetic environments while protecting privacy.

[0048] Example 5: In a system deployment scenario requiring pre-emptive elimination of environmental background interference, the data acquisition unit collects data for a duration before the subject enters the detection area. The background magnetic field signal flow of s is used to calculate the zero-position offset of each sensor channel using the local processor, and then... The initial readings of the sensor array are reset to zero. Background magnetic field intensity data from measurement points in the array edge region are extracted. A second-order multipole magnetic field distribution model is established based on the spatial distribution characteristics of the magnetic field in the sensor plane. The coefficient matrix of the background interference vector in the sensing coordinate system is calculated. This coefficient matrix is ​​used to determine the background field cancellation weight under the current physical environment, which is then used as the correction magnetic induction intensity distribution matrix. The input data; when the subject is in a resting state and the prediction system initiates the initial calibration procedure, the displacement compensation unit acquires data including... Magnetic field gradient direction vector for each consecutive cardiac cycle The data stream extracts the geometric centroid of the zero-position feature trajectory within each cycle and calculates the statistically average physical coordinates of this centroid sequence on the sensing plane coordinate axes. This is used as the reference origin for subsequent calculations of respiratory displacement compensation values. The benchmark calibration unit monitors the search residual amplitude during the dynamic optimization process of the rotation axis benchmark within the isovolumetric contraction period. When the physical spatial coordinate deviation of the rotation axis benchmark within two adjacent cardiac cycles is less than... When the deviation limit of mm is reached, the system is confirmed to have entered a stable spatiotemporal topology alignment state and begins to output the pulmonary hypertension feature extraction results.

[0049] In the standardized engineering procedure for performing background noise benchmark testing and initial system range setting in the detection environment, the data acquisition unit performs the test on the vacant detection area for a duration of [duration missing]. The system acquires background interference signals, calculates the standard deviation of magnetic induction intensity for each sensor channel to determine the background noise power spectral density, and sets the global signal-to-noise ratio threshold as follows: dB and adjust the sensor sensitivity gain when the sampling signal-to-noise ratio is below this threshold, according to The spatial sampling density of the sensor array determines the total number of angle intervals. The value of is determined based on the angular resolution calculated from the spacing between array measuring points. for This establishes a logical correlation between the accuracy of angular rotation consistency assessment and the physical sampling frequency of the hardware array. When the subject enters the testing station and the system performs pre-deployment alignment calibration, the displacement compensation unit acquires the spatiotemporal sequence data of the zero-position feature trajectory within the complete respiratory cycle, setting the upper limit of the number of iterations in the gradient vector consistency search procedure to [missing information]. This time, the convergence threshold for the rotation axis coordinate offset of adjacent iteration steps is determined as follows: mm, when processing high-frequency signal components reflecting local heterogeneity changes in myocardial tissue, an adaptive passband filter is used to filter out... Hz to Baseline drift and power frequency interference outside the Hz frequency range, selected after the onset of cardiac depolarization. ms to The residual energy projection value is calculated within a ms time window. The system performs a one-way hash mapping-based identifier processing on the pulmonary hypertension feature extraction results to generate a quantitative risk index that reflects the degree of deviation of right ventricular electromechanical coupling.

[0050] Example 6: To further illustrate the core algorithms and data processing procedures involved in the above examples, this example provides a specific engineering implementation path, which covers three key stages: zero-position feature trajectory extraction, rotation axis reference search, and quantization index generation; a zero-position feature trajectory sub-pixel level extraction and breathing compensation procedure is adopted. Addressing the insufficient accuracy of zero-position feature recognition due to the limited physical resolution of the sensor array, and the nonlinear attenuation of magnetic field strength caused by changes in the vertical distance between the heart and the sensor array due to respiratory motion, a procedure based on bicubic interpolation sub-pixel reconstruction and magnetic field strength normalization is used. Based on the principle of quasi-static magnetic field spatial continuity, the procedure defines the initial input state: acquiring the sampling time. Original magnetic flux density distribution matrix Dimension The spatial step size is determined by the physical spacing between the sensors. It was decided to construct a high-resolution magnetic field distribution matrix and use a bicubic interpolation algorithm to... matrix Mapped to High-resolution matrix The calculation process is for Middle adjacent Neighborhood grid, constructing cubic polynomial functions Fit the magnetic field surface to ensure that the first and second derivatives of the interpolated magnetic field distribution remain continuous in space. Extract the zero-position contour lines and geometric centroid. Perform contour tracing operations within the plane, traversing... For all grid cells, identify the sign of the magnetic induction intensity value, flip the grid edges, determine the zero-point coordinates through linear interpolation, and connect all zero points to form a set of closed or open zero-position contour lines. Calculate the geometric centroid coordinates of this set of zero contour lines. The calculation formula is as follows: ,in To determine the total number of discrete points forming the zero contour line, take a positive integer. , The first on the zero contour line By establishing the horizontal and vertical coordinates, the respiratory displacement compensation vector and intensity normalization factor are determined, and the average centroid coordinates during the respiratory quiescent period are set as the reference point. Calculate the translation compensation vector at the current moment. To eliminate the influence of Z-axis distance changes caused by respiration, the root mean square value of the total magnetic field intensity was calculated. Based on the Biot-Savart law, which describes the nonlinear decay relationship between magnetic field strength and distance, a strength normalization factor is constructed. : ,in, The dimensionless intensity normalization factor. The system utilizes the root mean square value of the full-field magnetic flux density under the reference breathing phase. Correction of spatial coordinate translation of magnetic field gradient direction vector, using The modulus amplitude is corrected, and the corrected magnetic field gradient data is output.

[0051] Furthermore, based on the curl symmetry energy minimization rotation axis benchmark search procedure, in order to accurately capture the rotational distortion characteristics of the right ventricular conduction path, it is necessary to determine the intrinsic rotational symmetry center of the magnetic field gradient field without knowing the anatomical center of the heart, and construct a curl symmetry energy minimization search procedure. The input data is defined as: time after respiratory compensation. Magnetic field gradient direction vector field Covering the planar area of ​​the sensor array Define the candidate center rotationally symmetric energy function for the plane. Any candidate rotation center point Define the rotationally symmetric energy function as follows: ,in Candidate Center The rotational symmetry energy index, the smaller the value, the higher the confidence level of its use as a center of rotation; To be included in the calculation of the total number of valid measurement points; To be from candidate centers Pointing to the measuring point Displacement vector, perform gradient descent iterative search, initialize search center. Calculate the geometric center of the sensor array. about and Partial derivatives, along Update center coordinates in the negative gradient direction: ,in For the first The center coordinates of the next iteration; The spatial search step size coefficient is set to 0.5mm to 2.0mm. Convergence criteria and baseline output are determined when the Euclidean distance between the center coordinates of two adjacent iterations is... The search stops when the value is less than a preset threshold, such as 0.05mm, and the final converged coordinates are obtained. The axis of rotation for this cardiac cycle is established as the reference, and a polar coordinate system is constructed using this as the origin to calculate the angular displacement offset.

[0052] A quantification mapping and index generation procedure for pulmonary artery pressure status is established to transform extracted gradient rotation feature parameters into quantitative technical indicators for assessing pulmonary vascular resistance status. This procedure, based on a log-linear regression model, processes discrete physical signal data and outputs dimensionless risk assessment parameters, excluding medical diagnostic conclusions. It incorporates time integration and normalization of feature parameters, selecting 50ms time windows before and after the phase-locked feature time, and applies gradient rotation feature parameters... Integrating and dividing by the average magnetic field strength within the window yields the normalized unit topological torque density. : ,in Unit topological torque density; These are the start and end times of the integration time window; To calculate the average magnetic flux density modulus of the entire array within this time window, a nonlinear mapping calculation is performed. Using pre-set benchmark coefficients obtained from multi-center historical sample calibration, the... Mapped to pulmonary artery pressure characteristic index The mapping follows the Weber-Fechner law to describe the logarithmic relationship between stimulus and response: ,in The pulmonary artery pressure characteristic index has a normalized range of 0 to 100. The higher the value, the heavier the electromechanical load on the right ventricle caused by pulmonary vascular resistance. This is the sensitivity gain coefficient, typically ranging from 15.0 to 25.0. This is the bias compensation constant, typically with a value of 1.0; Using the baseline calibration constant and confidence-weighted output, calculate the angular distribution consistency index within this time window. average value ,like If the noise level exceeds a preset noise threshold (e.g., 0.6), the current data is considered severely affected by environmental interference. The weight of the measurement result is automatically reduced, or a retest prompt is triggered. The final output is... These are confidence-weighted technical parameters, serving as objective physical quantities indicating changes in pulmonary vascular resistance.

[0053] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for extracting features of pulmonary hypertension based on magnetocardiography, characterized in that, Includes the following steps: Step S1: Acquire the multi-channel spatiotemporal data of cardiac magnetic induction intensity distribution of the subject, and map the spatiotemporal data of cardiac magnetic induction intensity distribution to the sensor array coordinate system to construct the magnetic induction intensity distribution matrix; Step S2: Perform spatial difference operation on the magnetic induction intensity distribution matrix to calculate the magnetic field gradient direction vector corresponding to each measuring point position; Step S3: Extract the zero-position feature trajectory of the magnetic field gradient direction vector in the sensor array plane, calculate the displacement offset of the zero-position feature trajectory with the cardiac cycle to determine the respiratory displacement compensation value, and correct the spatial coordinates of the magnetic field gradient direction vector according to the respiratory displacement compensation value. Step S4: Analyze the spatial distribution symmetry of the corrected magnetic field gradient direction vector within the cardiac cycle, determine the rotational symmetry axis of the magnetic field intensity distribution, and use the rotational symmetry axis as the polar reference for calculating the angular displacement offset of the magnetic field gradient direction vector. Step S5: Calculate the angular displacement offset of each magnetic field gradient direction vector relative to the rotational symmetry axis to obtain the gradient rotation characteristic parameters characterizing the electromechanical coupling characteristics of the heart. Step S6: Divide the rotational symmetry axis circumferentially into multiple angular intervals, statistically analyze the distribution probability of the magnetic field gradient direction vector in each angular interval, determine the angular distribution consistency index based on the distribution probability, identify the minimum point of the angular distribution consistency index as the phase-locked characteristic moment of the cardiac cycle, and use the phase-locked characteristic moment to periodically calibrate the gradient rotation feature parameters, and output the pulmonary hypertension feature extraction results.

2. The method for extracting pulmonary hypertension features based on magnetocardiography according to claim 1, characterized in that, The process of determining the respiratory displacement compensation value in step S3 includes: constructing a cardiac magnetic dipole model based on the magnetic induction intensity distribution matrix, extracting the spatial center coordinates of the cardiac magnetic dipole model, and determining the respiratory displacement compensation value by monitoring the quasi-periodic drift of the spatial center coordinates within the sampling period, wherein the frequency of the sampling period is not less than 1000Hz.

3. The method for extracting pulmonary hypertension features based on magnetocardiography according to claim 1, characterized in that, The process of determining the rotational symmetry axis in step S4 includes: collecting background magnetic field observation data of the edge region of the sensor array, using the background magnetic field observation data to correct the zero-point gain of the sensor array, and determining the spatial orientation of the rotational symmetry axis based on the corrected vector field distribution law.

4. The method for extracting pulmonary hypertension features based on magnetocardiography according to claim 1, characterized in that, The process of calculating the angular distribution consistency index in step S6 follows the formula below: ,in, Indicates time The angular distribution consistency index, The direction vector of the magnetic field gradient is represented in the th case. The probability distribution of each angle interval. The total number of angles is determined by a preset angle.

5. The method for extracting pulmonary hypertension features based on magnetocardiography according to claim 1, characterized in that, After outputting the pulmonary hypertension feature extraction results, the process also includes: extracting the high-frequency signal component in the magnetic induction intensity distribution, calculating the residual energy projection value of the high-frequency signal component on the normal plane, and extracting the spatial heterogeneity features of the subject's cardiac electrical activity based on the spatial distribution of the residual energy projection value.

6. The method for extracting pulmonary hypertension features based on magnetocardiography according to claim 5, characterized in that, The process of extracting spatial heterogeneity features of cardiac electrical activity in subjects includes: calculating the spatial energy gradient of the residual energy projection value in the normal plane, and identifying the local perturbation features of the magnetic field vector during myocardial depolarization based on the spatial energy gradient.

7. The method for extracting pulmonary hypertension features based on magnetocardiography according to claim 1, characterized in that, After outputting the pulmonary hypertension feature extraction results, the process also includes: extracting the gradient rotation feature parameters corresponding to the cardiac systolic phase, and performing correlation calculations between the gradient rotation feature parameters corresponding to the cardiac systolic phase and the pre-excitation magnetic field data at the end of diastole to generate the cross-phase electromagnetic coupling coefficient.

8. The method for extracting pulmonary hypertension features based on magnetocardiography according to claim 7, characterized in that, The process of generating the transphase electromagnetic coupling coefficient includes: extracting the magnetic principal axis deviation parameter during the repolarization process, and using the magnetic principal axis deviation parameter to perform gain correction on the transphase electromagnetic coupling coefficient in order to determine the spatial stability index of the electrocardiogram conduction vector.

9. The method for extracting pulmonary hypertension features based on magnetocardiography according to claim 1, characterized in that, The process of calculating the magnetic field gradient direction vector in step S2 includes: direction and Spatial partial derivatives of the magnetic flux density distribution matrix are calculated along the direction to obtain the horizontal magnetic gradient components. With vertical magnetic gradient components ,Depend on and The synthesized vector determines the magnetic field gradient direction vector, and the pulmonary hypertension feature extraction results are used to construct a quantitative mapping logic for pulmonary artery pressure status to output a feature index indicating changes in pulmonary vascular resistance.

10. A system for extracting pulmonary hypertension features based on magnetocardiography, used to implement the method for extracting pulmonary hypertension features based on magnetocardiography as described in claim 1, characterized in that, It includes a data acquisition unit, a gradient calculation unit, a displacement compensation unit, a benchmark calibration unit, a feature extraction unit, a phase-locked output unit, and a prediction modeling unit. The data acquisition unit is used to acquire multi-channel spatiotemporal data of cardiac magnetic induction intensity distribution of the subject, and to map the spatiotemporal data of cardiac magnetic induction intensity distribution to the sensor array coordinate system to construct a magnetic induction intensity distribution matrix; The gradient calculation unit is used to perform spatial difference operations on the magnetic induction intensity distribution matrix and calculate the magnetic field gradient direction vector corresponding to each measurement point. The displacement compensation unit is used to extract the zero-position feature trajectory of the magnetic field gradient direction vector in the sensor array plane, and determine the respiratory displacement compensation value based on the displacement offset of the zero-position feature trajectory with the cardiac cycle, so as to correct the spatial coordinates of the magnetic field gradient direction vector. The reference calibration unit is used to analyze the spatial distribution symmetry of the corrected magnetic field gradient direction vector within the cardiac cycle to determine the rotational symmetry axis, and uses the rotational symmetry axis as the polar reference. The feature extraction unit is used to calculate the angular displacement offset of the magnetic field gradient direction vector relative to the rotational symmetry axis to generate gradient rotation feature parameters. The phase-locked output unit is used to statistically analyze the distribution probability of the magnetic field gradient direction vector in each angular interval and determine the angular distribution consistency index. The minimum point of the angular distribution consistency index is used to establish the phase-locked characteristic time, and the gradient rotation characteristic parameters are calibrated based on the phase-locked characteristic time to output the feature extraction results. The predictive modeling unit is used to perform quantization mapping logic based on the feature extraction results to output a feature index characterizing the pulmonary artery pressure state.

Images