Gastrointestinal transit trajectory collection system
Patent Information
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- GUIZHOU HAIQIAN MEDICAL BIG DATA CO LTD
- Filing Date
- 2025-12-09
- Publication Date
- 2026-06-18
AI Technical Summary
Existing gastrointestinal motility testing systems suffer from high positioning data errors and inaccurate precision due to motion artifact interference, making it difficult to effectively balance data accuracy and algorithm complexity, thus affecting subsequent analysis and applications.
A system including wearable positioning devices, edible positioning markers, and auxiliary positioning markers is adopted, combined with magnetic sensor arrays and terminal equipment. Multi-reference point artifact elimination is performed through neural networks and equivalent magnetization current magnetic field models. The position and orientation of a single target point are corrected by using the correlation and weighted model of multiple reference points.
It achieves non-invasive, real-time monitoring of gastrointestinal motility, accurately locates the gastrointestinal transit trajectory, improves the accuracy and reliability of data, adaptability and flexibility, and reduces computational latency and data acquisition costs.
Smart Images
Figure CN2025141170_18062026_PF_FP_ABST
Abstract
Description
Gastrointestinal transit trajectory acquisition system Technical Field
[0001] This invention relates to a gastrointestinal motility detection system, and more particularly to a gastrointestinal transit trajectory acquisition system. Background Technology
[0002] Gastrointestinal motility refers to the peristaltic ability of the gastrointestinal organs, which is of great importance to human health. Traditional gastrointestinal motility tests usually require invasive examination equipment, such as endoscopes, which can cause pain and inconvenience to patients. Sometimes, anesthesia is required to reduce patient discomfort, which carries certain risks.
[0003] CN202011344388.9 discloses a gastrointestinal motility testing system, including a gastrointestinal motility capsule, a positioning device, and a terminal device. The gastrointestinal motility capsule includes a shell and a magnet, and the positioning device includes at least two magnetic sensors. The gastrointestinal motility capsule generates a changing magnetic field as it moves within the subject's body with the peristalsis of the gastrointestinal tract. The magnetic sensors collect magnetic field intensity data and transmit the intensity data to the terminal device. The terminal device determines the position, orientation, and movement frequency of the gastrointestinal motility capsule within the gastrointestinal tract based on the intensity data.
[0004] However, the positioning device in this system is fixed to the subject's abdomen, and the subject needs to wear it continuously for 24-72 hours. During this process, changes in the subject's body posture will cause changes in the relative position and posture between the positioning device and the subject's body, i.e. motion artifact interference. As a result, the position, posture and motion frequency information of the gastrointestinal motility capsule estimated by the system includes not only the normal movement of the gastrointestinal motility capsule caused by gastrointestinal peristalsis, but also the relative movement caused by motion artifact interference. Therefore, the positioning data obtained in this way has high error and inaccuracy.
[0005] Traditional motion artifact processing methods are mostly designed for image data, while the means for processing simply acquired data are relatively limited. Some existing methods struggle to fully utilize information from multiple reference points, and when eliminating motion artifacts at a single target point, they cannot effectively balance the requirements of data accuracy and algorithm complexity, resulting in significant errors in the processed single target point data, which affects subsequent analysis and applications. Summary of the Invention
[0006] The technical problem to be solved by the present invention is to provide a gastrointestinal transit trajectory acquisition system that reduces errors caused by motion artifacts and provides accurate positioning.
[0007] Furthermore, this invention also provides a method for eliminating artifact interference from a single target point based on multiple reference points, which can accurately eliminate motion artifact interference and improve the accuracy and reliability of the data.
[0008] This invention relates to a gastrointestinal transit trajectory acquisition system, comprising: a wearable positioning device equipped with a magnetic sensor array including at least two magnetic sensors; a single ingestible positioning marker containing a magnetic material, the ingestible positioning marker being movable within the gastrointestinal tract; m auxiliary positioning markers, m≥1, the auxiliary positioning markers containing magnetic material, the auxiliary positioning markers being positioned at fixed locations on the human body; and a terminal device, wherein the magnetic sensors are capable of acquiring the magnetic field strength of the ingestible positioning marker and the auxiliary positioning markers through magnetic induction, and the terminal device receives the magnetic field strength data from the magnetic sensors to determine the real-time position of the ingestible positioning marker within the human gastrointestinal tract, wherein the terminal device calculates the positional orientation parameters of the ingestible positioning marker and the auxiliary positioning markers based on the received magnetic field strength data, and corrects the positional orientation parameters of the ingestible positioning marker using a motion artifact interference elimination method based on the positional orientation parameters of the auxiliary positioning markers, thereby determining the real-time position of the ingestible positioning marker within the human gastrointestinal tract.
[0009] Preferably, the magnetic sensor can collect the magnetic field strength at multiple time points to determine the real-time location of the edible positioning marker at each time point, thereby obtaining the transport trajectory of the edible positioning marker in the gastrointestinal tract.
[0010] Preferably, the terminal device obtains the initial magnetic field strength of the edible positioning marker and the auxiliary positioning marker collected by the magnetic sensor array at an initial time, and calculates the initial position orientation parameters of the edible positioning marker and the auxiliary positioning marker based on the initial magnetic field strength to determine the initial position of the edible positioning marker; the terminal device obtains the magnetic field strength generated by the edible positioning marker and the auxiliary positioning marker collected by the magnetic sensor array at time i, and calculates the position orientation parameters of the edible positioning marker and the auxiliary positioning marker at time i based on the magnetic field strength, and determines the position and orientation of the edible positioning marker based on the difference between the position orientation parameters of the auxiliary positioning marker and its initial position orientation parameters.
[0011] Preferably, there are at least three auxiliary positioning markers.
[0012] Preferably, the fixed position of the human body can be at the anterior superior iliac spine, the posterior superior iliac spine, and / or the coccyx.
[0013] Preferably, the wearable positioning device is a belt or waist support suitable for fixing to the waist or abdomen of a person.
[0014] Preferably, the position and orientation parameters of the drivable positioning marker and the auxiliary positioning marker are calculated based on the received magnetic field strength data, and obtained by a multi-magnetic target positioning method based on neural networks and an equivalent magnetization current magnetic field model. The multi-magnetic target positioning method includes the following steps:
[0015] S1. Generate a dataset based on the equivalent magnetization current magnetic field model, which includes:
[0016] S11. Determine the positioning space of each magnetic target based on the expected motion range of each magnetic target relative to the sensor array.
[0017] S12. Perform uniformly distributed sampling of position and orientation within the positioning space to obtain a series of magnet position and orientation labels;
[0018] S13. Using the equivalent magnetization current magnetic field model, generate corresponding sensor array prediction values for the position and orientation of each magnet.
[0019] S14. Combine the sensor array predictions with the magnet position and orientation labels to form a labeled dataset.
[0020] S2, Training the pose estimation network, which includes:
[0021] S21. Divide the generated dataset into a training set and a test set;
[0022] S22. Iteratively train the pose estimation network based on back gradient propagation and the training dataset;
[0023] S23. In each round of iterative training, the localization accuracy of the pose estimation network is verified using the test set.
[0024] S24. When the localization accuracy of the pose estimation network in the test set is lower than a certain threshold, or when the maximum number of training iterations is reached, training is terminated.
[0025] S3. Real-time localization based on pose estimation network, including:
[0026] S31. Obtain the measurement values of each magnetic sensor from the actual sensor array in real time;
[0027] S32. Perform preprocessing operations such as filtering on the measured values;
[0028] S33. The preprocessed sensor array measurements are passed to the pose estimation network to estimate the position and orientation of each magnetic target in real time.
[0029] Preferably, in step S1, the magnetic sensor in the sensor array is a Hall effect-based magnetic sensor, an anisotropic magnetoresistive (AMR) sensor, a giant magnetoresistive (GMR) sensor, and / or a tunnel magnetoresistive (TMR) sensor.
[0030] Preferably, the data obtained by each sensor from measuring the magnetic field can be 1-dimensional, 2-dimensional, or 3-dimensional.
[0031] Preferably, in step S2, the basic network layer of the pose estimation network adopts a convolutional neural network (CNN) to better capture the local spatial correlation between adjacent sensors. The pose estimation network is an end-to-end convolutional neural network model that maps sensor measurements to the position and orientation of the magnet. A combined loss function integrating position loss and orientation loss is used to train the pose estimation network. The loss function is:
[0032] L p L represents the location regression loss. o The directional regression loss is represented by the scaling factor β. i To balance the regression losses of position and orientation with different dimensions, the two losses are defined as follows:
[0033] Among them, ||·|| δ Representing the δ-norm, R(·) transforms the attitude Euler angle mapping into a rotation matrix, ||·|| F Represents the Frobenius norm, symbol This represents the output of the pose estimation network PoseNet.
[0034] Preferably, uniformly distributed sampling is performed within the workspace of the positioning system to generate a dataset, with a position sampling interval of 1-50mm and an angle interval of 1-50°.
[0035] Preferably, the position sampling interval is 10 mm, and the angle interval is 10°.
[0036] Preferably, during training, noise is randomly added to the sensor measurements in the training dataset to improve the robustness of the network.
[0037] Preferably, the motion artifact interference elimination method includes: using the edible positioning marker as a single target point and the auxiliary positioning marker as multiple reference points, and performing the following steps:
[0038] A1. Data Acquisition and Processing Steps: Acquire motion data sequences including single target points and multiple reference points;
[0039] A2. Analyze the correlation between a single target point and multiple reference points: Calculate the correlation index between each reference point and the single target point, and determine the time delay matching point and the degree of correlation;
[0040] A3. Construction of multi-reference point weighted model: Construct a weighted model based on correlation, introduce a stability index to correct the weighting coefficients, and construct a single target point data estimation model based on multiple reference points;
[0041] A4. Model Optimization and Artifact Removal: The weighted model is optimized using the least squares method to obtain a single target point data sequence after motion artifact interference is eliminated.
[0042] Preferably, the motion artifact interference elimination method further includes:
[0043] A5. Data Validation and Quality Assessment: The root mean square error and correlation coefficient are used to assess the quality of single target point data before and after processing, verify the effectiveness of the algorithm, and adjust parameters or models as needed.
[0044] Preferably, in step A1, a motion data sequence containing a single target point and multiple reference points is acquired, wherein the acquired single target point data sequence is S(t) = {s1, s2, ..., s...}. n}, where t represents the time series, s i The data value of the drivable positioning marker (2) at time i; simultaneously, data sequences of m reference points are collected, and the data sequence of the j-th reference point is denoted as R. j (t)={r j1 ,r j2 ,...,r jn}, j = 1, 2, ..., m; perform preliminary processing on the collected data to remove obviously abnormal data points.
[0045] Preferably, the obvious abnormal data points are judged and removed by setting a data threshold range.
[0046] Preferably, in step A2: a cross-correlation function is used to measure the correlation. For the j-th reference point and the single target point, the cross-correlation function C... s,Rj The formula for calculating (τ) is as follows:
[0047] in The mean of a single target point data sequence. Let τ be the mean of the data sequence at the j-th reference point, and τ be the time delay parameter, with a value ranging from [-τ]. max ,τ max By analyzing the peak position and magnitude of the cross-correlation function, the optimal matching point and correlation degree between each reference point and the single target point in terms of time delay can be determined.
[0048] Preferably, in step A3: let the weighting coefficient be ω j And introduce the stability index St j Stability indices can be calculated by measuring the variance of the reference point data sequence. Measured by the reciprocal, that is
[0049] Furthermore, the weighting coefficients are corrected, and the corrected weighting coefficients ω' j The calculation formula is:
[0050] Construct a single-target point data estimation model based on multiple reference points:
[0051] Where Δt j This is the time delay compensation value between the j-th reference point and the single target point, determined based on the cross-correlation function.
[0052] Preferably, the initial values are initially allocated based on the degree of correlation; the reference points with a higher degree of correlation have a higher weighting coefficient ω. j The larger.
[0053] Preferably, in step A4: the weighted model is optimized using the least squares method to minimize the error between the estimated value S′(t) and the original single-target point data S(t), wherein the error function is set as follows: By analyzing E with respect to the weighting coefficient ω' j Taking the partial derivatives and setting them equal to zero yields a set of equations. Solving this set of equations yields the optimized weighting coefficients ω. j Substituting the optimized weighting coefficients into the data estimation model, we obtain the single-target point data sequence S"(t) after eliminating motion artifact interference.
[0054] Preferably, in step A5: the root mean square error (RMSE) and correlation coefficient (CC) are used to evaluate the quality of single-target point data before and after eliminating motion artifact interference, wherein the RMSE calculation formula is:
[0055] The formula for calculating the relevant CC coefficient is:
[0056] The effectiveness of this algorithm is verified based on the evaluation results. If the results do not meet the expected standards, return to step A2 or step A3 to adjust the parameters or rebuild the model.
[0057] Advantages of this invention:
[0058] 1. No invasive examination equipment is required. It is easy to use and can monitor the patient's gastrointestinal motility in real time, accurately reflecting the state of intestinal movement.
[0059] 2. By wearing a magnetic induction positioning device and swallowing a magnetic positioning marker (capsule), the examinee can understand their gastrointestinal health status at any time and promptly detect and prevent potential gastrointestinal problems.
[0060] 3. By using auxiliary positioning markers fixed on specific parts of the human body, errors caused by changes in human posture can be corrected and motion artifacts can be eliminated.
[0061] 4. In terms of localization method, an end-to-end pose estimation network based on deep learning is adopted to perform real-time localization of edible magnetic markers and auxiliary magnetic markers. Compared with traditional optimization-based localization methods, it has stronger robustness (does not depend on initial values and avoids the risk of getting trapped in local optima) and lower computational latency (the pose of the magnetic marker is directly regressed from the feedforward pose estimation network, rather than iteratively optimized).
[0062] 5. The training dataset is generated and acquired using an equivalent magnetization current magnetic field model, rather than by collecting data from a physical system, which reduces the time and manpower costs of data collection.
[0063] 6. This invention employs a more optimized algorithm for eliminating motion artifact interference, further improving positioning accuracy. This algorithm utilizes multiple reference points: by fully leveraging information from multiple reference points, it can more comprehensively capture motion information related to a single target point compared to traditional methods that rely solely on a single reference point or do not utilize reference points at all, thus more effectively eliminating motion artifact interference. Furthermore, through steps such as correlation analysis, weighted model construction and optimization, this algorithm accurately determines the relationship between reference points and single target points, and rationally allocates weighting coefficients, improving the accuracy and reliability of single target point data processing, providing stronger data support for subsequent analysis and decision-making based on single target point data.
[0064] 7. Adaptability and Flexibility: The selection of reference points, correlation analysis parameters, and weighted model parameters can be adjusted according to different data acquisition scenarios and requirements. It has strong adaptability and flexibility and can be widely used in various single-target point data processing fields involving motion artifact interference. Attached Figure Description
[0065] Figure 1 is a schematic diagram of an embodiment of the gastrointestinal transit trajectory acquisition system of the present invention.
[0066] Figure 2 is a side view of an embodiment of the gastrointestinal transit trajectory acquisition system of the present invention.
[0067] Figure 3 is a rear view of an embodiment of the gastrointestinal transit trajectory acquisition system of the present invention.
[0068] Figure 4 shows the accurate modeling of the magnetic field of a cylindrical permanent magnet, where the X-axis is along the length of the magnet, the origin of the coordinate system is located at the center of the magnet, and the magnet coordinate system and the current direction of the equivalent magnetization current model are shown.
[0069] Figure 5 shows the framework for training and predicting the pose estimation network.
[0070] Figure 6 is a flowchart of the method for eliminating motion artifact interference according to the present invention.
[0071] For ease of understanding, the same reference numerals are used where possible to denote common elements in the figures. It is contemplated that elements disclosed in one embodiment may be advantageously used in other embodiments without specific description. Detailed Implementation
[0072] Referring to Figures 1-3, a gastrointestinal transit trajectory acquisition system of the present invention includes a wearable positioning device 1, which is equipped with multiple magnetic sensors 3. The positioning device 1 can be a type of belt or waist support that is fixed to the waist or abdomen of the human body. The magnetic sensors 3 are arranged in an array, for example, in 1×2, 2×2, 3×3, 3×4, 5×6, etc.
[0073] The system of the present invention also includes an ingestible positioning marker 2, which comprises a magnetic material. The positioning marker 2, for example, is made into an oral capsule that moves within the gastrointestinal tract C with peristalsis, and contains a permanent magnet. As the positioning marker 2 moves within the gastrointestinal tract, its position and orientation change, resulting in a change in the strength of the magnetic field. A magnetic sensor 3 is capable of detecting the change in the magnetic field strength of the positioning marker 2 through magnetic induction.
[0074] The system of the present invention also includes at least one auxiliary positioning marker, preferably three auxiliary positioning markers P1, P2, and P3. The auxiliary positioning markers also contain magnetic materials, such as magnets. The auxiliary positioning markers P1, P2, and P3 are positioned at fixed locations on the human body, such as the anterior superior iliac spine, posterior superior iliac spine, and / or coccyx. The magnetic sensor 3 is capable of detecting the magnetic field strength of the auxiliary positioning markers P1, P2, and P3 through magnetic induction.
[0075] The system also includes terminal devices (not shown in the figure), such as mobile devices like computers, mobile phones, or iPads, or dedicated control instruments. The terminal devices are connected to the magnetic sensor 3, providing power for its operation and receiving magnetic field strength data from the positioning marker 2 and auxiliary positioning markers P1, P2, and P3 collected by the magnetic sensor 3. Based on the magnetic field strength of the edible positioning marker 2 and the auxiliary positioning markers P1, P2, and P3, the terminal devices execute instructions and use algorithms to determine the relative positional relationship between the auxiliary positioning markers P1, P2, and P3 and the edible positioning marker 2, thereby obtaining the real-time position of the edible positioning marker 2.
[0076] In one embodiment, the system includes a drivable positioning marker 2, namely magnetic target 1#; and three auxiliary positioning markers P1, P2, and P3, namely magnetic target 2#, magnetic target 3#, and magnetic target 4#, respectively.
[0077] By measuring the magnetic field strength (in three dimensions) at the first time (initial time) by each magnetic sensor 3 in the magnetic sensor array, the algorithm obtains the four real-time position and direction parameters of the magnetic targets 1#-4# at the first time, and thus obtains the initial positions of the traversable positioning marker 2 and the three auxiliary positioning markers P1, P2, and P3.
[0078] Then, the magnetic field strength at the second time (time j) is measured, and the algorithm yields the four real-time position orientation parameters of magnetic targets 1#-4# at the second time (time j). Because the three auxiliary positioning markers P1, P2, and P3 are set at fixed positions on the human body, such as the anterior superior iliac spine, posterior superior iliac spine, and / or coccyx, their actual position orientation remains essentially unchanged regardless of changes in the human body's posture. The difference between the position orientation parameters of these three auxiliary positioning markers for magnetic targets 2#-4# at time j and their position orientation parameters at the initial time can be used to infer the position orientation changes of each magnetic sensor 3 due to the human body's posture, thereby correcting the obtained position orientation parameters of magnetic target 1# (i.e., ingestible positioning marker 2), and obtaining the second (j) real-time position of positioning marker 2. Through the above correction, the obtained second real-time position of positioning marker 2 is more accurate with smaller errors.
[0079] By repeating the above steps, a series of real-time locations of the oral positioning marker 2 can be obtained, thus obtaining a series of time-related trajectories, which can be displayed or recorded on the terminal device for further diagnostic reference.
[0080] The position and orientation parameters of the drivable positioning marker 2 and the auxiliary positioning markers P1, P2, and P3 are calculated based on the received magnetic field strength data using a multi-magnetic target positioning method based on neural networks and an equivalent magnetization current magnetic field model.
[0081] In one embodiment, the above-described multi-magnetic target localization method includes the following steps:
[0082] S1. Generate a dataset based on the equivalent magnetization current magnetic field model, which includes:
[0083] S11. Determine the positioning space of each magnetic target based on the expected motion range of each magnetic target relative to the sensor array.
[0084] S12. Perform uniformly distributed sampling of position and orientation within the positioning space to obtain a series of magnet position and orientation labels;
[0085] S13. Using the equivalent magnetization current magnetic field model, generate corresponding sensor array prediction values for the position and orientation of each magnet.
[0086] S14. Combine the sensor array predictions with the magnet position and orientation labels to form a labeled dataset.
[0087] S2, Training the pose estimation network, which includes:
[0088] S21. Divide the generated dataset into a training set and a test set;
[0089] S22. Iteratively train the pose estimation network based on back gradient propagation and the training dataset;
[0090] S23. In each round of iterative training, the localization accuracy of the pose estimation network is verified using the test set.
[0091] S24. When the localization accuracy of the pose estimation network in the test set is lower than a certain threshold, or when the maximum number of training iterations is reached, training is terminated.
[0092] S3. Real-time localization based on pose estimation network, including:
[0093] S31. Obtain the measurement values of each magnetic sensor from the actual sensor array in real time;
[0094] S32. Perform preprocessing operations such as filtering on the measured values;
[0095] S33. The preprocessed sensor array measurement values are passed to the pose estimation network to perform real-time pose estimation for each magnetic target.
[0096] In step S1 above, the magnetic sensor in the sensor array is a Hall effect-based magnetic sensor, or it can be an anisotropic magnetoresistive (AMR) sensor, a giant magnetoresistive (GMR) sensor, or a tunnel magnetoresistive (TMR) sensor.
[0097] The data obtained from the magnetic field measurements by each sensor can be 1-dimensional, 2-dimensional, or 3-dimensional.
[0098] In step S2 above, the basic network layer of the pose estimation network adopts a convolutional neural network (CNN) to better capture the local spatial correlation between adjacent sensors. This pose estimation network is an end-to-end convolutional neural network model that maps sensor measurements to the position and orientation of the magnet. During training, noise is randomly added to the sensor measurements in the training dataset to improve the robustness of the network. A combined loss function that integrates position loss and orientation loss is used to train the pose estimation network. The loss function will be explained in detail below.
[0099] The specific explanation of the permanent magnet magnetic field modeling is as follows.
[0100] As shown in Figure 4, a cylindrical permanent magnet is used as the target to be tracked, with a length of 2l and a radius of r, respectively. Several equivalent models used to calculate the spatial magnetic field distribution of the permanent magnet include the magnetic dipole model, the equivalent magnetizing current model, and the equivalent magnetic charge model. The magnetic dipole model is widely used in magnetic tracking systems due to its simple expression, but it can introduce approximation errors for non-spherical magnets, especially in the near-field region. The equivalent magnetizing current model assumes that the magnetic field is generated by surface current, and the direction of the current is indicated by blue arrows in the figure. According to Biot-Savart's law, the magnetic field distribution of the cylindrical permanent magnet can be calculated. Position is determined by surface current I = J s The magnetic field dB generated by dx c J s This represents the current density (A / m). (x,y,z) T It is a position on that current. Position The magnetic field at a location can be estimated using the following formula.
[0101] Where dB c It can be calculated using the Biot-Savart theorem, for
[0102] Where C is the path of current flow, and dl = (0, -sinθ, cosθ)rdθ. Based on the superposition principle, the magnetic field strength generated by the equivalent current of the permanent magnet can be calculated as follows:
[0103] Where K c =rJs μ0 / 4π, y=r cosθ, z=r sinθ. Therefore, the expression for the magnetic field generated by the cylindrical permanent magnet can be obtained through numerical integration.
[0104] In this system, a magnetic sensor array is used to sense the magnetic field generated by the permanent magnet in space. As shown in Figure 4, a system coordinate frame {s} is established at the center of the magnetic sensor array. In Equation (1.3), the magnetic induction intensity model of the magnet is established on the magnet coordinate system {m}. The pose of the permanent magnet describes the translation and rotation of the coordinate system {m} relative to the coordinate system {s}.
[0105] Suppose v = [x, y, z, α, β, γ] is the six-degree-of-freedom pose (position + attitude) of the permanent magnet in the system coordinate system {s}, where t sm = [x, y, z] T Let (α, β, γ) be the translation vector, and (α, β, γ) be the Euler angles of the permanent magnet's rotation around the X, Y, and Z axes in sequence. This rotation sequence can be represented as R. sm =R x (α)R y (β)R z (γ) Equation (1.4)
[0106] Point P i The sensor position at point can be represented in coordinate systems {s} and {m} as follows: and and The relationship between them is as follows:
[0107] When the permanent magnet is in pose v, the magnetic flux density at the i-th magnetic sensor in coordinate system {s} (assuming the sensor array has N magnetic sensors) is:
[0108] The above formula can be used to predict the magnetic field strength at any location in the space surrounding the permanent magnet.
[0109] It should be particularly noted that the radially magnetized cylindrical permanent magnet is wound around X. m When the axis (i.e., the SN axis of the magnet) rotates, the magnetic field strength it generates in space remains unchanged due to the symmetry of the magnetic field. Therefore, α can be fixed at 0.
[0110] More generally, when there are M permanent magnet targets to be tracked, M≥2, according to the superposition theorem, the magnetic field strength generated by the M permanent magnets at the i-th sensor is B=B i (1)+B i (2) + ... + B i(M) Equation (1.7)
[0111] Given an arbitrary position and orientation of a permanent magnet, the measurement values of all magnetic sensors on the magnetic sensor array can be predicted using the magnetic field model in (1.7).
[0112] The following is a detailed explanation of the multi-target pose estimation network.
[0113] The process of magnet pose estimation is essentially to derive the position and orientation of the magnet from the measurements of the sensor array, that is, to find the inverse solution of equation (1.7). Since the calculation of the magnetic field model involves integral operations, it is difficult to derive an analytical solution. Although swarm intelligence optimization algorithms (such as Particle Swarm Optimization (PSO) and Whale Optimization (WOA)) can estimate the magnet pose through swarm search strategies, they cannot guarantee the convergence of the search, and the large amount of computation makes it difficult to meet real-time requirements. Deep neural networks have powerful nonlinear modeling capabilities. This invention uses an end-to-end multi-target pose estimation network, PoseNet, as shown in Figure 5.
[0114] All parameters in the model are learned end-to-end using a combined loss function that combines position and orientation losses, as follows:
[0115] L p L represents the location regression loss. o This represents the direction regression loss. Since position and direction are represented by different units, a scaling factor β is used. i To balance these two regression tasks, the two types of loss are defined as follows:
[0116] Among them, ||·|| δ Representing the δ-norm, R(·) transforms the attitude Euler angle mapping into a rotation matrix, ||·|| F
[0117] Represents the Frobenius norm, symbol This represents the output of the pose estimation network PoseNet.
[0118] Dataset Generation: Uniformly distributed sampling is performed within the workspace of the positioning system. The sampling interval for position and angle is a trade-off. A smaller sampling interval results in a larger data volume and higher accuracy for the trained pose estimation network model. Generally, the position sampling interval can be 1-50 mm, preferably 10 mm, and the angle interval is 1-50°, preferably 10°. Predicted values for the magnetic sensor array are generated using the magnetic field model in Equation (1.7), thus obtaining a dataset for training the pose estimation network.
[0119] Referring to Figure 6, the following algorithm is used to correct motion artifact interference using auxiliary positioning markers.
[0120] Step A1. Data Acquisition and Processing: Collect motion data sequences containing single target points and multiple reference points, perform preliminary data processing, and remove outlier data points;
[0121] Step A2. Correlation analysis between reference points and target points: Calculate the correlation index between each reference point and a single target point to determine the time delay matching point and the degree of correlation;
[0122] Step A3. Construction of multi-reference point weighted model: Construct a weighted model based on correlation, introduce a stability index to correct the weighting coefficients, and construct a single target point data estimation model based on multiple reference points;
[0123] Step A4. Model optimization and artifact removal: The weighted model is optimized using the least squares method to obtain a single target point data sequence after motion artifact interference is eliminated;
[0124] Step A5. Data Validation and Quality Assessment: Use indicators such as root mean square error and correlation coefficient to assess the quality of single target point data before and after processing, verify the effectiveness of the algorithm, and adjust parameters or models as needed.
[0125] In the above steps, the single target point can be a monitoring point for the displacement of contents in the colon, which in this invention can be the positioning marker 2. The reference point can be a monitoring point in other relatively stable parts of the body, which in this invention can be the auxiliary positioning markers P1, P2, and P3.
[0126] In step A1: collect motion data sequences containing a single target point and multiple reference points, wherein the collected single target point data sequence is S(t)={s1,s2,...,s n}, where t represents the time series, s i The data value of the drivable positioning marker 2 at time i; simultaneously, data sequences from m reference points are acquired, and the data sequence of the j-th reference point is denoted as R. j (t)={r j 1,r j 2, ..., r j Let n}, j = 1, 2, ..., m; Perform preliminary processing on the collected data to remove obviously abnormal data points (such as sudden noise interference causing abnormal data points during data acquisition due to equipment vibration, electromagnetic interference, etc.). The obvious abnormal data points can be judged and removed by setting a data threshold range (which can be determined based on prior knowledge or previous experimental data).
[0127] In step A2: Calculate the correlation index between each reference point and the single target point. A cross-correlation function is used to measure this correlation. For the j-th reference point and the single target point, the cross-correlation function C... s,Rj The formula for calculating (τ) is as follows:
[0128] in The mean of a single target point data sequence. Let τ be the mean of the data sequence at the j-th reference point, and τ be the time delay parameter, with a value ranging from [-τ]. max ,τ max By analyzing the peak position and magnitude of the cross-correlation function, the optimal matching point and correlation degree between each reference point and the single target point in terms of time delay can be determined. For example, when calculating the cross-correlation function, τ can be set... max A certain percentage (e.g., 10%) of the data sequence length is used to comprehensively search for the best time delay matching point.
[0129] Step A3, in constructing the weighted model based on the correlation between the reference point and the single target point: let the weighting coefficient be ω. j Its initial value is initially allocated based on the degree of correlation; the reference point with a higher degree of correlation has a higher weighting coefficient ω. j The larger the value, the better. Meanwhile, considering the stability of data at different reference points, a stability index St is introduced. j Stability indices can be calculated by measuring the variance of the reference point data sequence. Measured by the reciprocal, that is Furthermore, the weighting coefficients are corrected, and the corrected weighting coefficients ω' j The calculation formula is:
[0130] Construct a single-target point data estimation model based on multiple reference points:
[0131] Where Δt j This is the time delay compensation value between the j-th reference point and the single target point, determined based on the cross-correlation function.
[0132] Step A4, in model optimization and artifact removal: the least squares method is used to optimize the weighted model to minimize the error between the estimated value S′(t) and the original single-target point data S(t), where the error function is set. By analyzing E with respect to the weighting coefficient ω' j Taking the partial derivatives and setting them equal to zero yields a set of equations. Solving this set of equations yields the optimized weighting coefficients ω. j Substituting the optimized weighting coefficients into the data estimation model, we obtain the single-target point data sequence S"(t) after eliminating motion artifact interference.
[0133] Step A5, Data Validation and Quality Assessment: Root Mean Square Error (RMSE) and Correlation Coefficient (CC) are used to assess the quality of single-target point data before and after eliminating motion artifact interference. The formula for calculating RMSE is:
[0134] The formula for calculating the relevant CC coefficient is:
[0135] The effectiveness of this algorithm is verified based on the evaluation results. If the results do not meet the expected standards, return to step A2 or step A3 to adjust the parameters or rebuild the model.
[0136] In the embodiments of the above method, during the data acquisition and processing stage, the accuracy and stability of the data acquisition equipment are ensured, and data from a single target point and multiple reference points are collected according to a predetermined sampling frequency. For example, in biomedical monitoring, a single target point might be a monitoring point for the displacement of dissolved substances in the colon, and reference points might be monitoring points at other relatively stable parts of the body. For the acquired data, abnormal data points caused by equipment vibration, electromagnetic interference, etc., are removed by setting a reasonable data threshold (which can be determined based on prior knowledge or previous experimental data).
[0137] In the correlation analysis step between reference points and target points, for each reference point, its cross-correlation function with a single target point under different time delays is calculated. For example, when calculating the cross-correlation function, τ can be set. max A certain percentage (e.g., 5-15%, preferably 10%) of the data sequence length is used to comprehensively search for the optimal time delay matching point. Δt is determined by analyzing the peak value of the cross-correlation function. j The degree of correlation provides a basis for constructing the weighted model.
[0138] When constructing a multi-reference-point weighted model, first roughly allocate weighting coefficients ω according to the degree of correlation. j For example, the initial weighting coefficient for the reference point with the highest correlation is set to 0.5, and other reference points are assigned between 0 and 0.5 based on their relative correlation. Then, the stability index St of the reference points is calculated. j And adjust the weighting coefficient ω' according to the formula. j Substituting the corrected weighting coefficients and time delay compensation values into the data estimation model, preliminary estimates of single-target point data are obtained.
[0139] In the model optimization and artifact elimination stages, the optimized weighting coefficients ω" are solved using the least squares method. j By constructing an error function and calculating its partial derivatives, a system of equations is obtained, which can be solved using matrix operations and other methods. Substituting the optimized weighting coefficients into the model yields the final single-target point data sequence after eliminating motion artifact interference.
[0140] Finally, in the data validation and quality assessment phase, indices such as RMSE and CC are calculated. If the RMSE is too high or the CC is too low, it indicates that the processing effect is not ideal. At this point, the rationality of the reference point selection should be checked, and the parameters in the correlation analysis (such as τ) should be adjusted. max Then, reconstruct the weighted model or optimize the model parameters until a satisfactory result is obtained.
[0141] The above examples are for illustrative purposes only. The motion artifact removal methods described are not limited to these examples and can be applied to any single-target motion data based on multiple reference points. The following are other application scenarios for single-target motion data acquisition based on multiple reference points.
[0142] - Cardiovascular disease monitoring
[0143] In Holter monitoring, electrode patches are fixed at a relatively stable position on the patient's chest as reference points, with the heart as a single target point. During daily activities, artifacts caused by body movement can interfere with the ECG signal, affecting the accurate diagnosis of heart disease. This algorithm utilizes signals collected from multiple electrode patches as references to eliminate motion artifacts at the single target point (cardiac electrical activity signal) caused by body movements (such as walking, arm swinging, etc.), enabling more accurate analysis of cardiovascular conditions such as arrhythmias.
[0144] For patients with implanted pacemakers, the pacemaker itself serves as a single target point, while surrounding tissues or miniature sensors fixed to locations such as the ribs act as reference points. Changes in patient posture or movement can cause relative displacement between the pacemaker and surrounding tissues, affecting the accuracy of data monitoring pacemaker function. This algorithm, by processing the acquired data, effectively removes motion artifacts, ensuring physicians can accurately assess pacemaker function and the patient's cardiac condition, promptly identifying potential problems and adjusting treatment plans.
[0145] -Sports training and rehabilitation assessment
[0146] During athlete training, sensors are typically attached to key body parts (such as joints and muscle attachment points) to analyze joint movement trajectories and muscle exertion. A specific joint (e.g., the knee) serves as a single target point, while sensors attached to relatively stable areas (such as the pelvis and torso) act as reference points. When athletes perform high-intensity training movements, the overall body movement and vibrations can cause motion artifacts in the target joint's motion data acquisition. This algorithm accurately reconstructs the true movement of the target joint, allowing coaches to optimize training plans, improve training effectiveness, and help prevent sports injuries.
[0147] In the field of rehabilitation therapy, such as limb rehabilitation training for stroke patients, single-target point sensors are placed on the patient's rehabilitation limb (e.g., the arm), while reference point sensors are placed on other parts of the body (e.g., the shoulder, back, etc.). During rehabilitation exercises, limb swaying caused by muscle incoordination can produce motion artifacts, affecting the accurate assessment of rehabilitation progress. This algorithm can obtain more precise limb motion data, allowing rehabilitation therapists to adjust the intensity and method of rehabilitation training based on this data, providing patients with more personalized and effective rehabilitation treatment plans.
[0148] Virtual Reality (VR) and Augmented Reality (AR) Interaction
[0149] In VR / AR systems, the user's handheld interactive device (such as a controller or handheld device) serves as a single target point, while the head-mounted display (HMD) or sensors fixed to the surrounding environment serve as a reference point. When users engage in immersive experiences or interactive operations using VR / AR devices, natural body movements (such as head rotation and body translation) can cause motion data artifacts on the interactive device. Applying this algorithm to process the data allows the system to more accurately track user interactions, improve the matching degree between the virtual environment and real-world actions, and enhance the realism and smoothness of the user experience. For example, it enables more precise aiming and operational feedback in VR games, or provides more accurate operational guidance in AR-assisted repair.
[0150] In VR / AR-based motion capture applications, such as dance instruction and sports motion analysis, the dancer's or athlete's limb joints serve as single target points, while cameras or sensors positioned around the area act as reference points. Due to the dancer's rapid movements and complex posture changes, the collected motion data is susceptible to motion artifacts. This algorithm can reconstruct more realistic and accurate motion postures, which can be used for teaching evaluation, movement correction, or performance optimization, providing valuable data support for professionals.
[0151] Those skilled in the art will understand that the foregoing examples are exemplary and not restrictive. It will be apparent to those skilled in the art, upon reading the specification and studying the accompanying drawings, that all permutations, enhancements, equivalents, and modifications are included within the true spirit and scope of the invention. Therefore, the appended claims are intended to cover all such modifications, permutations, and equivalents that fall within the true spirit and scope of these teachings.
[0152] Explanation of reference numerals in the attached figures: 1. Wearable positioning device; 2. Positioning marker; 3. Magnetic sensor; C. Gastrointestinal tract; P1. Auxiliary positioning marker; P2. Auxiliary positioning marker; P3. Auxiliary positioning marker.
Claims
1. A gastrointestinal transit trajectory acquisition system, characterized in that, include: A wearable positioning device (1) is provided with a magnetic sensor array including at least two magnetic sensors (3). A single edible positioning marker (2), the positioning marker containing a magnetic material, the edible positioning marker (2) being movable within the gastrointestinal tract; m auxiliary positioning markers (P1, P2, P3), where m ≥ 1, wherein the auxiliary positioning markers contain magnetic material, and the auxiliary positioning markers (P1, P2, P3) are used to be placed at fixed positions on the human body; as well as The terminal device includes a magnetic sensor (3) capable of acquiring the magnetic field strength of the edible positioning marker (2) and the auxiliary positioning markers (P1, P2, P3) through magnetic induction, and the terminal device receiving the magnetic field strength data from the magnetic sensor (3) to determine the real-time location of the edible positioning marker (2) within the human gastrointestinal tract. The terminal device calculates the position and orientation parameters of the edible positioning marker (2) and the auxiliary positioning markers (P1, P2, P3) based on the received magnetic field strength data, and corrects the position and orientation parameters of the edible positioning marker (2) using a motion artifact interference elimination method based on the position and orientation parameters of the auxiliary positioning markers (P1, P2, P3), thereby determining the real-time position of the edible positioning marker (2) in the human gastrointestinal tract.
2. The digestive tract transport trajectory acquisition system according to claim 1, characterized in that, The magnetic sensor (3) can collect the magnetic field strength at multiple time points, thereby determining the real-time position of the edible positioning marker (2) at each time point, and thus obtaining the transport trajectory of the edible positioning marker (2) in the gastrointestinal tract.
3. The digestive tract transport trajectory acquisition system according to claim 1, characterized in that, The terminal device obtains the initial magnetic field strength of the edible positioning marker (2) and the auxiliary positioning markers (P1, P2, P3) collected by the magnetic sensor array at the initial time, and calculates the initial position orientation parameters of the edible positioning marker (2) and the auxiliary positioning markers (P1, P2, P3) based on the initial magnetic field strength to determine the initial position of the edible positioning marker (2). The terminal device obtains the magnetic field strength generated by the edible positioning marker (2) and the auxiliary positioning markers (P1, P2, P3) collected by the magnetic sensor array at time i, and calculates the position and orientation parameters of the edible positioning marker (2) and the auxiliary positioning markers (P1, P2, P3) at time i based on the magnetic field strength, and determines the position and orientation of the edible positioning marker (2) based on the difference between the position and orientation parameters of the auxiliary positioning markers (P1, P2, P3) and their initial position and orientation parameters.
4. The digestive tract transport trajectory acquisition system according to claim 1, characterized in that, The auxiliary positioning markers shall be at least three.
5. The digestive tract transport trajectory acquisition system according to claim 4, characterized in that, The fixed position of the human body can be the anterior superior iliac spine, the posterior superior iliac spine, and / or the coccyx.
6. The digestive tract transport trajectory acquisition system according to claim 1, characterized in that, The wearable positioning device is a belt or waist support suitable for fixing to the waist or abdomen of a person.
7. The digestive tract transport trajectory acquisition system according to claim 1, characterized in that, The position and orientation parameters of the drivable positioning marker (3) and the auxiliary positioning markers (P1, P2, P3) are calculated based on the received magnetic field strength data and obtained through a multi-magnetic target positioning method based on neural networks and an equivalent magnetization current magnetic field model. The multi-magnetic target positioning method includes the following steps: S1. Generate a dataset based on the equivalent magnetization current magnetic field model, which includes: S11. Determine the positioning space of each magnetic target based on the expected motion range of each magnetic target relative to the sensor array. S12. Perform uniformly distributed sampling of position and orientation within the positioning space to obtain a series of magnet position and orientation labels; S13. Using the equivalent magnetization current magnetic field model, generate corresponding sensor array prediction values for the position and orientation of each magnet. S14. Combine the sensor array predictions with the magnet position and orientation labels to form a labeled dataset. S2, Training the pose estimation network, which includes: S21. Divide the generated dataset into a training set and a test set; S22. Iteratively train the pose estimation network based on back gradient propagation and the training dataset; S23. In each round of iterative training, the localization accuracy of the pose estimation network is verified using the test set. S24. When the localization accuracy of the pose estimation network in the test set is lower than a certain threshold, or when the maximum number of training iterations is reached, training is terminated. S3. Real-time localization based on pose estimation network, including: S31. Obtain the measurement values of each magnetic sensor from the actual sensor array in real time; S32. Perform preprocessing operations such as filtering on the measured values; S33. The preprocessed sensor array measurements are passed to the pose estimation network to estimate the position and orientation of each magnetic target in real time.
8. The digestive tract transport trajectory acquisition system according to claim 7, characterized in that, In step S1, the magnetic sensors in the sensor array are Hall effect-based magnetic sensors, anisotropic magnetoresistive (AMR) sensors, giant magnetoresistive (GMR) sensors, and / or tunnel magnetoresistive (TMR) sensors.
9. The digestive tract transport trajectory acquisition system according to claim 7, characterized in that, The data obtained from the magnetic field measurements by each sensor can be 1-dimensional, 2-dimensional, or 3-dimensional.
10. The digestive tract transport trajectory acquisition system according to claim 7, characterized in that, In step S2, the basic network layer of the pose estimation network adopts a convolutional neural network (CNN) to better capture the local spatial correlation between adjacent sensors. The pose estimation network is an end-to-end convolutional neural network model that maps sensor measurements to the position and orientation of the magnet. A combined loss function integrating position loss and orientation loss is used to train the pose estimation network. The loss function is: L p L represents the position regression loss. o The directional regression loss is represented by the scaling factor β. i To balance the regression losses of position and orientation with different dimensions, the two losses are defined as follows: Among them, ||·|| δ Representing the δ-norm, R(·) transforms the attitude Euler angle mapping into a rotation matrix, ||·|| F Represents the Frobenius norm, symbol This represents the output of the pose estimation network PoseNet.
11. The digestive tract transport trajectory acquisition system according to claim 10, characterized in that, Uniformly distributed sampling is performed within the workspace of the positioning system to generate a dataset. The position sampling interval is 1-50mm, and the angle interval is 1-50°.
12. The digestive tract transport trajectory acquisition system according to claim 11, characterized in that, The position sampling interval is 10mm, and the angle interval is 10°.
13. The digestive tract transport trajectory acquisition system according to claim 11, characterized in that, During training, noise is randomly added to the sensor measurements in the training dataset to improve the robustness of the network.
14. The digestive tract transport trajectory acquisition system according to claim 1, characterized in that, The motion artifact interference elimination method includes: using the edible positioning marker (2) as a single target point and the auxiliary positioning markers (P1, P2, P3) as multiple reference points, and performing the following steps: A1. Data Acquisition and Processing Steps: Acquire motion data sequences including single target points and multiple reference points; A2. Analyze the correlation between a single target point and multiple reference points: Calculate the correlation index between each reference point and the single target point, and determine the time delay matching point and the degree of correlation; A3. Construction of multi-reference point weighted model: Construct a weighted model based on correlation, introduce a stability index to correct the weighting coefficients, and construct a single target point data estimation model based on multiple reference points; A4. Model Optimization and Artifact Removal: The weighted model is optimized using the least squares method to obtain an optimized single-target point data sequence after eliminating motion artifact interference.
15. The digestive tract transport trajectory acquisition system according to claim 14, characterized in that, The method for eliminating motion artifact interference also includes: A5. Data Validation and Quality Assessment: The root mean square error and correlation coefficient are used to assess the quality of the single target point data before and after processing, verify the effectiveness of the algorithm, and adjust the parameters or model as needed.
16. The digestive tract transport trajectory acquisition system according to claim 14, characterized in that, Step A1: Collect motion data sequences containing a single target point and multiple reference points, wherein the collected single target point data sequence is S(t)={s1,s2,...,s n }, where t represents the time series, s i The data value of the drivable positioning marker (2) at time i; simultaneously, data sequences of m reference points are collected, and the data sequence of the j-th reference point is denoted as R. j (t)={r j1 ,r j2 ,...,r jn }, j = 1, 2, ..., m; perform preliminary processing on the collected data to remove obviously abnormal data points.
17. The digestive tract transport trajectory acquisition system according to claim 16, characterized in that, The obvious abnormal data points are judged and removed by setting a data threshold range.
18. The digestive tract transport trajectory acquisition system according to claim 14, characterized in that, In step A2: The cross-correlation function is used to measure the correlation. For the j-th reference point and the single target point, the cross-correlation function C is... s,Rj The formula for calculating (τ) is as follows: in The mean of a single target point data sequence. Let τ be the mean of the data sequence at the j-th reference point, and τ be the time delay parameter, with a value ranging from [-τ]. max ,τ max By analyzing the peak position and magnitude of the cross-correlation function, the optimal matching point and correlation degree between each reference point and the single target point in terms of time delay can be determined.
19. The digestive tract transport trajectory acquisition system according to claim 14, characterized in that, In step A3: Let the weighting coefficient be ω j And introduce the stability index St j Stability indices can be calculated by measuring the variance of the reference point data sequence. Measured by the reciprocal, that is Furthermore, the weighting coefficients are corrected, and the corrected weighting coefficients ω' j The calculation formula is: Construct a single-target point data estimation model based on multiple reference points: Where Δt j This is the time delay compensation value between the j-th reference point and the single target point, determined based on the cross-correlation function.
20. The digestive tract transport trajectory acquisition system according to claim 19, characterized in that, Its initial values are initially allocated based on the degree of correlation; the higher the degree of correlation, the higher the weighting coefficient ω of the reference point. j The larger.
21. The digestive tract transport trajectory acquisition system according to claim 19, characterized in that, In step A4: The weighted model is optimized using the least squares method to minimize the error between the estimated value S′(t) and the original single-target point data S(t), where the error function is set as follows: By analyzing E with respect to the weighting coefficient ω' j Taking the partial derivatives and setting them equal to zero yields a set of equations. Solving this set of equations yields the optimized weighting coefficients ω. j Substituting the optimized weighting coefficients into the data estimation model, we obtain the single-target point data sequence S"(t) after eliminating motion artifact interference.
22. The digestive tract transport trajectory acquisition system according to claim 15, characterized in that, In step A5: The root mean square error (RMSE) and correlation coefficient (CC) are used to assess the quality of single-target point data before and after eliminating motion artifact interference. The formula for calculating RMSE is as follows: The formula for calculating the relevant CC coefficient is: The effectiveness of this algorithm is verified based on the evaluation results. If the results do not meet the expected standards, return to step A2 or step A3 to adjust the parameters or rebuild the model.
23. A method for eliminating motion artifact interference in a single-target motion data acquisition system, the acquisition system comprising: A single target, multiple reference points, and a terminal device, wherein the terminal device is used to receive and process the acquired motion data of the single target and multiple reference points, characterized in that the method includes the following steps: A1. Data Acquisition and Processing Steps: Acquire motion data sequences including single target points and multiple reference points; A2. Analyze the correlation between a single target point and multiple reference points: Calculate the correlation index between each reference point and the single target point, and determine the time delay matching point and the degree of correlation; A3. Construction of multi-reference point weighted model: Construct a weighted model based on correlation, introduce a stability index to correct the weighting coefficients, and construct a single target point data estimation model based on multiple reference points; A4. Model Optimization and Artifact Removal: The weighted model is optimized using the least squares method to obtain an optimized single-target point data sequence after eliminating motion artifact interference.
24. The method for eliminating motion artifact interference according to claim 23, characterized in that, Also includes: A5. Data Validation and Quality Assessment: The root mean square error and correlation coefficient are used to assess the quality of the single target point data before and after processing, verify the effectiveness of the algorithm, and adjust the parameters or model as needed.
25. The method for eliminating motion artifact interference according to claim 23 or 24, characterized in that, Step A1: Collect motion data sequences containing a single target point and multiple reference points, wherein the collected single target point data sequence is S(t)={s1,s2,...,s n }, where t represents the time series, s i The data value of the drivable positioning marker (2) at time i; simultaneously, data sequences of m reference points are collected, and the data sequence of the j-th reference point is denoted as R. j (t)={r j1 ,r j2 ,...,r jn }, j = 1, 2, ..., m; perform preliminary processing on the collected data to remove obviously abnormal data points.
26. The method for eliminating motion artifact interference according to claim 25, characterized in that, The obvious abnormal data points are judged and removed by setting a data threshold range.
27. The method for eliminating motion artifact interference according to claim 23, characterized in that, In step A2: The cross-correlation function is used to measure the correlation. For the j-th reference point and the single target point, the cross-correlation function C is... s,Rj The formula for calculating (τ) is as follows: in The mean of a single target point data sequence. Let τ be the mean of the data sequence at the j-th reference point, and τ be the time delay parameter, with a value ranging from [-τ]. max ,τ max By analyzing the peak position and magnitude of the cross-correlation function, the optimal matching point and correlation degree between each reference point and the single target point in terms of time delay can be determined.
28. The method for eliminating motion artifact interference according to claim 23, characterized in that, τ max It is 5-15% of the total data sequence length.
29. The method for eliminating motion artifact interference according to claim 23, characterized in that, In step A3: Let the weighting coefficient be ω j And introduce the stability index St j Stability indices can be calculated by measuring the variance of the reference point data sequence. Measured by the reciprocal, that is Furthermore, the weighting coefficients are corrected, and the corrected weighting coefficients ω' j The calculation formula is: Construct a single-target point data estimation model based on multiple reference points: Where Δt j This is the time delay compensation value between the j-th reference point and the single target point, determined based on the cross-correlation function.
30. The method for eliminating motion artifact interference according to claim 29, characterized in that, Its initial values are initially allocated based on the degree of correlation; the higher the degree of correlation, the higher the weighting coefficient ω of the reference point. j The larger.
31. The method for eliminating motion artifact interference according to claim 23, characterized in that, In step A4: The weighted model is optimized using the least squares method to minimize the error between the estimated value S′(t) and the original single-target point data S(t), where the error function is set as follows: By analyzing E with respect to the weighting coefficient ω' j Taking the partial derivatives and setting them equal to zero yields a set of equations. Solving this set of equations yields the optimized weighting coefficients ω. j Substituting the optimized weighting coefficients into the data estimation model, we obtain the single-target point data sequence S"(t) after eliminating motion artifact interference.
32. The method for eliminating motion artifact interference according to claim 24, characterized in that, In step A5: The root mean square error (RMSE) and correlation coefficient (CC) are used to assess the quality of single-target point data before and after eliminating motion artifact interference. The formula for calculating RMSE is as follows: The formula for calculating the relevant CC coefficient is: Verify the validity of the calculation results based on the evaluation results. If the results do not meet the expected standards, return to step A2 or step A3 to adjust the parameters or rebuild the model.
33. The method for eliminating motion artifact interference according to claim 23, characterized in that, The single target and the multiple reference points are magnetic points containing magnetic material, and the motion data are magnetic field strength or position and direction parameters determined based on the magnetic field strength.