Motion monitoring and imaging optimization system and method

Abstract

The invention discloses a motion monitoring and imaging optimization system and method. The system comprises a radar hardware system and a data processing system which are in electric signal connection, and the radar hardware system transmits a continuous wave signal to a monitored object, receives an echo signal and transmits the echo signal to the data processing system; after the data processing system obtains scanning data, the scanning data are aligned with echo signal time, feature data are obtained through preprocessing and feature extraction, physiological motion data are obtained through target detection and dynamic tracking, a mapping model of body surface displacement and in-vivo motion is constructed, and motion compensation parameters are generated; meanwhile, physiological parameters are extracted, and abnormal events and risk scoring results are evaluated. Non-contact high-precision motion monitoring is achieved, the problems that a traditional means is large in synchronization error, uncomfortable in monitored object and the like are solved, imaging artifacts are effectively eliminated, the imaging quality is improved, comprehensive data support is provided for medical diagnosis, and the method is suitable for medical scanning complex scenes.

Description

Technical Field

[0001] This application relates to the field of image imaging technology, and in particular to a motion monitoring and imaging optimization system and method. Background Technology

[0002] With the continuous development of medical imaging technology, high-precision imaging equipment such as CT and MRI are increasingly used in clinical diagnosis, leading to ever-increasing demands on image quality. However, the unavoidable physiological movements of the monitored subject during scanning, such as breathing and heartbeat, can cause motion artifacts like blurring and ghosting in the images, severely interfering with the accurate identification of lesions and affecting the reliability of diagnostic results. To solve this problem, it is necessary to accurately monitor the physiological movements of the monitored subject and optimize the imaging process based on the monitoring data.

[0003] Current methods for motion monitoring and imaging optimization mainly include respiratory gating technology, wearable sensor monitoring, and traditional image processing correction. Respiratory gating technology achieves imaging synchronization by capturing the respiratory phase, but it suffers from problems such as signal delay and phase recognition errors, resulting in limited synchronization accuracy. Wearable sensors require patients to wear the devices, which not only affects the naturalness and comfort of movement but may also produce soft tissue artifacts, limiting their applicability. Traditional image processing methods mostly rely on manual feature extraction and lack effective integration with motion monitoring data, making it difficult to achieve dynamic and accurate motion compensation and resulting in poor artifact elimination.

[0004] Therefore, how to achieve non-contact, high-precision physiological motion monitoring, ensure accurate time alignment between scanning data and motion monitoring data, eliminate artifacts, and simultaneously extract physiological parameters and complete risk assessment are urgent technical problems that need to be solved. Summary of the Invention

[0005] In view of this, the motion monitoring and imaging optimization system and method provided in this application can achieve non-contact high-precision motion monitoring, solving problems such as large synchronization errors and discomfort of the monitored object in traditional methods, effectively eliminating imaging artifacts, improving imaging quality, providing comprehensive data support for medical diagnosis, and adapting to complex medical scanning scenarios. The motion monitoring and imaging optimization system and method provided in this application are implemented as follows: This application provides a motion monitoring and imaging optimization system, which includes a radar hardware system and a data processing system connected by electrical signals. The radar hardware system is configured to: reflect a continuous wave signal to the monitored object and receive the echo, obtain the echo signal, and send the echo signal to the data processing system; The data processing system is configured to: acquire scan data, perform time alignment processing on the scan data and the echo signal, and obtain time-aligned scan data and echo signal; The echo signal is preprocessed and feature extracted to obtain feature data; The feature data is subjected to target detection and dynamic tracking processing to obtain physiological motion data; A mapping model is constructed, and motion compensation parameters are obtained based on the mapping model and the physiological motion data. The mapping model is used to characterize the mapping relationship between body surface displacement and internal motion. The physiological motion data is processed for parameter extraction and risk assessment to obtain physiological parameters, abnormal event types, and risk scores. The motion compensation parameters and the scan data are fused and optimized to obtain imaging data; The imaging data, physiological parameters, abnormal event types, and risk score results are output.

[0006] In some embodiments, the preprocessing and feature extraction of the echo signal to obtain feature data includes: The echo signal is subjected to clutter suppression processing to obtain a clutter-free signal; The denoised signal is subjected to signal optimization processing to obtain a signal with improved signal-to-noise ratio; The signal with improved signal-to-noise ratio is subjected to phase recovery processing to obtain an absolute phase signal; The absolute phase signal is subjected to feature acquisition processing to obtain multi-dimensional original features; The original multi-dimensional features are normalized and dimensionality reduced to obtain feature data.

[0007] In some embodiments, the step of performing target detection and dynamic tracking processing on the feature data to obtain physiological motion data includes: The feature data is processed for target detection to obtain the target confidence score and the coordinates of the moving area. Based on the coordinates of the motion region, the target confidence score is used to predict the target position, and the predicted position of the target in the next frame is obtained. The predicted position of the target in the next frame is matched with the actual detected position to obtain trajectory re-association data; The target detection results and the trajectory re-association data are integrated and processed to obtain physiological motion data.

[0008] In some embodiments, constructing a mapping model and obtaining motion compensation parameters based on the mapping model and the physiological motion data includes: The monitored object is subjected to synchronous calibration scanning to obtain multi-directional scan images of the monitored object and the surface displacement curves corresponding to the echo signals. The multi-directional scan images are registered to obtain in vivo motion data; The surface displacement curve and the internal motion data are subjected to model fitting to obtain a mapping model; Based on the mapping model and the physiological motion data, motion adjustment processing is performed on the imaging layer tracking direction to obtain the layer tracking compensation parameters. The imaging layer tracking direction is the motion direction of the monitored object in the front-back direction or the up-down direction. Based on the mapping model and the physiological motion data, phase shift correction is performed on the imaging artifact elimination direction to obtain artifact elimination correction parameters. The imaging artifact elimination direction is the movement direction of the monitored object in the left-right or front-back direction in the scanned image plane. The layer tracking compensation parameters and the artifact elimination correction parameters are integrated to obtain motion compensation parameters.

[0009] In some embodiments, the step of extracting parameters and conducting risk assessment on the physiological motion data to obtain physiological parameters, abnormal event types, and risk score results includes: The physiological motion data is filtered, feature-recognized, and statistically analyzed to obtain physiological parameters, including respiratory rate, heart rate, and exercise amplitude. Set an anomaly detection threshold, compare the physiological parameters with the anomaly detection threshold, and obtain the anomaly event type when the physiological parameters are greater than or equal to the anomaly detection threshold or when the parameter mutations within a preset time reach a preset proportion. The degree of deviation of the physiological parameters is classified into risk levels to obtain risk score results.

[0010] In some embodiments, the fusion and optimization processing of motion compensation parameters and scan data to obtain imaging data includes: The scan data is aligned to obtain aligned scan data; Phase shift compensation is performed on the aligned scan data to obtain compensated scan data; The compensated scan data is then subjected to non-rigid registration optimization to obtain the optimized scan data. Image reconstruction processing is performed on the optimized scan data to obtain imaging data.

[0011] In some embodiments, the physiological motion data includes respiratory movement trajectory, heart rate fluctuation amplitude, and movement speed.

[0012] This application provides a motion monitoring and imaging optimization method applicable to a motion monitoring and imaging optimization system, comprising: Acquire scan data, and perform time alignment processing on the scan data and the echo signal received from the radar hardware system to obtain time-aligned scan data and echo signal; The echo signal is preprocessed and feature extracted to obtain feature data; The feature data is subjected to target detection and dynamic tracking processing to obtain physiological motion data; A mapping model is constructed, and motion compensation parameters are obtained based on the mapping model and the physiological motion data. The mapping model is used to characterize the mapping relationship between body surface displacement and internal motion. The physiological motion data is processed for parameter extraction and risk assessment to obtain physiological parameters, abnormal event types, and risk scores. The motion compensation parameters and the scan data are fused and optimized to obtain imaging data; The imaging data, physiological parameters, abnormal event types, and risk score results are output.

[0013] The computer device provided in this application includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the program, it implements the method described in this application.

[0014] The computer-readable storage medium provided in this application embodiment stores a computer program thereon, which, when executed by a processor, implements the method provided in this application embodiment.

[0015] This application provides a motion monitoring and imaging optimization system and method, comprising a radar hardware system and a data processing system electrically connected. The radar hardware system transmits continuous wave signals to the monitored object and receives echo signals, which are then transmitted to the data processing system. The data processing system acquires scan data, aligns it with the echo signals in time, and obtains feature data through preprocessing and feature extraction. Physiological motion data is acquired through target detection and dynamic tracking, a mapping model between surface displacement and internal motion is constructed, and motion compensation parameters are generated. Simultaneously, physiological parameters are extracted, abnormal events are assessed, and risk scoring results are obtained. This enables non-contact, high-precision motion monitoring, solving problems such as large synchronization errors and discomfort to the monitored object in traditional methods. It effectively eliminates imaging artifacts, improves imaging quality, provides comprehensive data support for medical diagnosis, adapts to complex medical scanning scenarios, and addresses the technical problems mentioned in the background art. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A schematic diagram illustrating the implementation process of a motion monitoring and imaging optimization method provided in this application embodiment; Figure 2 This is a schematic diagram illustrating an implementation process for acquiring feature data, provided in an embodiment of this application. Detailed Implementation

[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0019] The following description of some technologies involved in the embodiments of this application is provided to aid understanding and should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, some descriptions of well-known functions and structures are omitted in the following description.

[0020] Figure 1 This is a schematic diagram illustrating the implementation flow of a motion monitoring and imaging optimization method provided in an embodiment of this application, including steps 101 to 108. Wherein, Figure 1 This is merely one execution order shown in the embodiments of this application and does not represent the only execution order of a motion monitoring and imaging optimization method. Where the final result can be achieved, Figure 1 The steps shown can be performed in parallel or in reverse order.

[0021] Step 101: Reflect the continuous wave signal to the monitored object and receive the echo to obtain the echo signal, and send the echo signal to the data processing system.

[0022] In this embodiment, the radar hardware system serves as the core of signal acquisition and adopts a 60GHz band FMCW (Frequency-Modulated Continuous Wave) radar architecture.

[0023] The radar hardware system comprises four core components: the radar front-end, the antenna array, the signal processing module, and the synchronization interface. The radar front-end uses a 60GHz FMCW radar chip, supporting 4GHz ultra-wide bandwidth output with a transmit power controlled below 10dBm. The antenna array is an 8-channel MIMO (Multiple-Input Multiple-Output) array, employing a 4-transmitter plus 4-receiver layout. The antenna spacing is set to half the wavelength of the corresponding frequency band, using a horizontal and vertical dual-polarization design. It is fixed to the top of the CT (Computed Tomography) / MRI (Magnetic Resonance Imaging) scanning chamber via an adjustable bracket, 150-250mm away from the chest and abdomen of the monitored object, with an tilt angle adjusted to 15°-30° to fully cover the chest and abdomen detection area. Electromagnetic shielding is achieved using a copper foil and absorbing cotton composite structure. The signal processing module is equipped with a Xilinx Zynq-7000 series FPGA (Field-Programmable Gate Array). The system features a Field Programmable Gate Array (FPGA) chip and an ARM Cortex-A9 processor. The ADC (Analog-to-Digital Converter) has a 12-bit sampling accuracy and supports adaptive switching of the sampling rate from 20Hz to 50Hz. It automatically switches to 50Hz during intense motion and maintains 20Hz during stable motion. The synchronization interface connects to the main control unit of the CT / MRI equipment via a synchronization signal line to ensure that the synchronization error between radar data and scan data does not exceed 0.1ms.

[0024] During operation, the radar hardware system transmits a continuous wave linear frequency modulated (LFM) signal to the chest and abdominal region of the monitored object according to a preset program. The expression for the continuous wave LFM signal is: ,in, It is the time-domain signal of the continuous wave linear frequency modulated signal transmitted by the radar hardware system, that is, the function of the transmitted signal changing with time; The amplitude of the transmitted signal is denoted as , which is the amplitude parameter of the transmitted signal. The carrier frequency of the transmitted signal; The frequency modulation bandwidth of the transmitted signal; The frequency modulation period of the transmitted signal; It is a time variable used to describe the dynamic changes of the transmitted signal over time.

[0025] A continuous wave linear frequency modulated (LFM) signal is reflected from the surface of the monitored object to form an echo signal, which is received by the antenna array. The echo signal is converted into an intermediate frequency (IF) signal by a mixer. The expression for the IF signal is: ,in, The intermediate frequency echo signal obtained by the radar hardware system after receiving the echo reflected from the surface of the monitored object and performing frequency mixing is the baseband time domain function of the echo signal. This is the signal attenuation coefficient; The square of the radar transmitted signal amplitude; The linear frequency modulation of the transmitted signal; The real-time distance from the monitored object to the radar antenna array; For the time delay of the echo; The speed at which electromagnetic waves propagate in the air; The carrier frequency of the radar's transmitted signal; It is a time variable.

[0026] The signal is then amplified by a differential amplifier circuit, sampled and converted into a digital signal by an ADC chip, preliminarily processed by a signal processing module, and then sent to the data processing system.

[0027] Step 102: Acquire scan data, perform time alignment processing on the scan data and echo signal to obtain time-aligned scan data and echo signal.

[0028] In this embodiment, the data processing system is electrically connected to the radar hardware system and is responsible for scanning data acquisition, signal synchronization processing, feature extraction, target tracking, motion compensation, parameter evaluation, and imaging optimization.

[0029] The data processing system scans data and simultaneously receives echo signals transmitted from the radar hardware system. It performs time alignment processing on the scan data and echo signals to ensure that each set of echo signals is precisely matched with its corresponding scan data in the time dimension, ultimately obtaining time-aligned scan data and echo signals.

[0030] Step 103: Perform preprocessing and feature extraction on the echo signal to obtain feature data.

[0031] In this embodiment, the time-aligned echo signal is first subjected to clutter suppression processing. A static clutter filtering algorithm is used, which generates a clutter template through background modeling and performs point-by-point subtraction on the echo signal to eliminate clutter interference from the fixed structure inside the scanning chamber, thus obtaining a de-cluttered signal. The de-cluttered signal is then input into a nonlinear programming optimizer based on a preset objective function: ,in, Let the objective function of the nonlinear programming optimizer be... This refers to the radar echo signal collected at the current moment. The radar echo signal at the reference time, , To adjust the parameters (p=1, q=2).

[0032] Iteratively update and optimize the weights, and finally pass Output optimized signal, where, For the optimized radar echo signal, These are the optimal weight coefficients obtained after iterative optimization. This is the radar data matrix corresponding to the region of interest (ROI) (only the signal of the target area is retained). Weak motion signal components are enhanced to obtain a signal with improved signal-to-noise ratio (SNR). Fourier transform is performed on the enhanced SNR signal to extract phase information. A phase unwrapping algorithm based on adjacent sampling points is used to eliminate the phase entanglement limitation from -π to π, restoring the absolute phase signal corresponding to the true displacement. Next, a dual-sliding-window multimodal feature extraction method is used to acquire features from the absolute phase signal. A fixed time window of 500ms (sliding step size of 100ms) is set to extract time-domain and frequency-domain features. The point cloud data is divided into windows according to a 5cm × 5cm spatial grid to extract spatial features, obtaining multi-dimensional original features. Finally, Z-score normalization is performed on the multi-dimensional original features to eliminate dimensional differences, and PCA dimensionality reduction is used to retain principal components with a cumulative contribution rate ≥ 95%, resulting in highly robust feature data.

[0033] Step 104: Perform target detection and dynamic tracking processing on the feature data to obtain physiological motion data.

[0034] In this embodiment, feature data is input into a lightweight CNN (Convolutional Neural Network) + Transformer hybrid detection model for target detection. The CNN backbone network of the model adopts the MobileNetV3 lightweight architecture, extracting local features through depthwise separable convolutions. The Transformer attention module introduces a 2-layer encoder + 1-layer decoder structure, using an 8-head self-attention mechanism to capture global feature associations. Finally, the target confidence score (threshold set to 0.7) and motion region coordinates are output to form the target detection result. Based on the motion region coordinates in the target detection result, Kalman filtering is used to establish the target motion state equation. ,in, For the first The target motion state vector at time t. Here is the state transition matrix. For the first The target motion state vector at time t. This represents process noise. (Compared to the observation equation) ,in, For the first The observed value at time, For the observation matrix, For observation noise, state vector To address motion uncertainty, the system predicts the target's position in the next frame. It then performs trajectory matching between the predicted position and the actual detected motion area coordinates, employing the Hungarian algorithm with a cost function of position distance plus velocity similarity (threshold set to 0.3) to achieve trajectory re-association in occluded scenarios (occlusion time ≤ 3s), obtaining re-association data. Finally, it integrates the target detection results with the re-association data to extract information such as the monitored object's respiratory trajectory, heart rate amplitude, and movement speed, forming physiological motion data. The respiratory trajectory is represented by a three-dimensional coordinate sequence of the center point of the chest and abdomen, with a motion amplitude resolution of 0.1mm.

[0035] Step 105: Construct a mapping model and obtain motion compensation parameters based on the mapping model and physiological motion data.

[0036] In this embodiment, the data processing system first executes a synchronous calibration scan procedure. Before the formal CT / MRI scan, a 30-second calibration scan is performed on the monitored object, simultaneously acquiring sagittal and coronal CT / MRI images and the corresponding surface displacement curves of the radar echo signals. For the acquired multi-planar scan images, a diastolic cardiac image is selected as the reference image, and a normalized two-dimensional cross-correlation registration algorithm is used for registration processing. The registration algorithm formula is as follows: ,in, The scanned image to be registered (measuring the image) With template In displacement (the degree of matching at the location) Template image (pixel coordinates) (grayscale value at the location) For image In displacement The average gray level within the corresponding local area. template image The average value.

[0037] The internal motion data of the scanned area in three directions—head-to-foot, anterior-posterior, and lateral—were obtained. The surface displacement curves and the internal motion data in these three directions were fitted using fractional multinomial regression. The least squares method was used to solve for the fitting parameters, constructing a mapping model characterizing the relationship between surface displacement and internal motion. The specific mapping formula is as follows:

[0038]

[0039]

[0040] in, The displacement of the body surface as measured by radar. , , , , , , , as well as For different fitting parameters.

[0041] Based on the mapping model and physiological motion data, the motion displacement in the tracking direction (head-to-foot or anteroposterior direction) is determined at the imaging plane. Real-time tracking at the imaging plane is achieved by adjusting the center frequency of the radio frequency pulse. The frequency adjustment formula is as follows: ,in =42.58MHz / T (gyrodynamic ratio), This is a layer-selected gradient magnetic field.

[0042] To ensure the imaging plane is relatively stationary with respect to the scanned area, plane tracking compensation parameters are obtained. Simultaneously, the motion displacement in the phase encoding direction and frequency encoding direction within the scanned image plane is determined. The phase shift correction amount in the corresponding directions is calculated using a phase shift correction algorithm to eliminate motion-induced imaging artifacts. The formula for the phase shift correction amount in the frequency encoding direction is:

[0043] in, For frequency coding direction, The displacement of the monitored object in the frequency coding direction. This represents the number of sampling points in the frequency coding direction. This is the index for the sampling points.

[0044] The formula for phase shift correction in the phase encoding direction is:

[0045] in, For phase encoding direction, The displacement of the monitored object in the phase encoding direction. The number of sampling points in the phase encoding direction. This is the index for the sampling points.

[0046] To eliminate motion-induced imaging artifacts, artifact elimination correction parameters are obtained. By integrating the layer tracking compensation parameters with the artifact elimination correction parameters, motion compensation parameters covering multiple directions are obtained.

[0047] Step 106: Extract parameters and perform risk assessment on the physiological motion data to obtain physiological parameters, abnormal event types, and risk score results.

[0048] In this embodiment, the data processing system performs targeted analysis on physiological motion data, extracts key physiological parameters, and completes risk assessment. Low-pass filtering is applied to the respiratory motion trajectory; peak and trough points are identified using a peak detection algorithm; the reciprocal mean of the intervals between adjacent peaks is calculated to obtain the respiratory rate; polynomial fitting is used to eliminate respiratory harmonic interference; matched filtering is used to enhance the heartbeat signal; and the R-wave peak interval is extracted to calculate the heartbeat rate; peak-to-peak values ​​of the three-dimensional coordinates of the center point of the chest and abdomen are statistically analyzed to obtain the motion amplitude parameter; based on statistical results of normal human data, normal range thresholds for physiological parameters are set, including a respiratory rate of 12-20 breaths / minute, a heart rate of 60-100 beats / minute, and a motion amplitude ≤5mm. When physiological parameters exceed this range or a parameter mutation of ≥30% within 3 seconds occurs, it is judged as an abnormal event, and the type of abnormal event is determined; a weighted scoring method is used to quantitatively assess the deviation of physiological parameters, and the scoring formula is: ,in, , , The deviations of respiration, heart rate, and exercise parameters are normalized to 0-1. In the scoring formula, the deviation of respiration parameter has a weight of 0.3, the deviation of heart rate parameter has a weight of 0.5, and the deviation of exercise parameter has a weight of 0.2. The deviations are normalized to the 0-1 range. The final score is ≥0.6 for high risk, 0.3-0.6 for medium risk, and <0.3 for low risk, thus obtaining the risk score result.

[0049] Step 107: The motion compensation parameters and scanning data are fused and optimized to obtain imaging data.

[0050] In this embodiment, the data processing system fuses and optimizes the motion compensation parameters with the time-aligned scan data. First, it uses layer tracking compensation parameters to perform real-time imaging layer alignment processing on the scan data, eliminating the relative displacement interference between the scan layer and the target part caused by forward / backward or up / down movement. Then, it uses artifact elimination correction parameters to perform coded direction phase shift compensation processing on the layer-aligned scan data, suppressing blurring and ghosting artifacts caused by in-plane motion. Subsequently, it performs non-rigid registration optimization on the compensated scan data to further correct the imaging deviation caused by residual motion. Finally, it performs image reconstruction processing on the multi-dimensional correction and optimization scan data to obtain high-definition, low-artifact imaging data.

[0051] Step 108: Output imaging data, physiological parameters, abnormal event types, and risk score results.

[0052] In this embodiment, the data processing system outputs imaging data, physiological parameters, abnormal event types, and risk score results. High-risk events trigger audible and visual alarms and simultaneously output a pause scan signal to the CT / MRI equipment. Medium- and low-risk events and various data are displayed in real time on the medical staff's console, and all data can be exported via an Ethernet interface, providing comprehensive support for medical diagnosis and treatment.

[0053] This application embodiment transmits continuous wave signals and receives echoes through a radar hardware system, capturing physiological movements such as breathing and heartbeat without contact with the monitored object. This avoids the patient discomfort caused by traditional contact monitoring devices (such as breathing straps and electrodes), making it suitable for special medical scenarios such as burns and infectious disease cases, and solving the problem of limited use of traditional contact methods. Time alignment processing eliminates data time differences, addressing the large synchronization errors between traditional non-contact monitoring (such as optical cameras and lidar) and medical scanning equipment, providing a time consistency basis for subsequent motion compensation and imaging optimization. It not only eliminates imaging artifacts by fusing motion compensation parameters with scanning data, but also simultaneously outputs physiological parameters, abnormal event types, and risk score results, breaking through the limitations of traditional methods that only focus on imaging correction, and providing comprehensive data support for medical diagnosis.

[0054] In the above Figure 1 Based on the above, this application embodiment also provides a schematic diagram of the implementation process for obtaining feature data. For example... Figure 2 As shown, steps 201 to 205 are included: Step 201: Perform clutter suppression processing on the echo signal to obtain a clutter-free signal.

[0055] In this embodiment, clutter suppression processing is performed on the time-aligned echo signal to eliminate interference from the fixed structure inside the scanning chamber. Specifically, a static clutter filtering algorithm is adopted. First, environmental signals in the scanning scene without a patient are collected through background modeling to generate a corresponding clutter template. Then, the echo signal containing the target signal is subtracted from the clutter template point by point to accurately remove the clutter components generated by the fixed structure such as the inner wall and support of the scanning chamber, and finally obtain a clutter-free signal that retains only the patient's surface reflection signal.

[0056] Step 202: Perform signal optimization processing on the denoised signal to obtain a signal with improved signal-to-noise ratio.

[0057] In this embodiment, the clutter-removed signal undergoes signal optimization processing to improve the signal-to-noise ratio (SNR) of weak motion signals. The radar data matrix corresponding to the clutter-removed signal is input into a nonlinear programming optimizer, which iteratively updates the optimization weights based on a preset objective function. The optimizer focuses on enhancing signal components corresponding to weak motions such as breathing and heartbeat, while suppressing remaining random noise interference, making the motion signal characteristics more prominent, and finally outputting a signal with improved SNR.

[0058] Step 203: Perform phase recovery processing on the signal after improving the signal-to-noise ratio to obtain the absolute phase signal.

[0059] In this embodiment, the signal after signal-to-noise ratio enhancement is processed to obtain an absolute phase signal. First, a Fourier transform is performed on the optimized signal to extract phase information related to the target displacement. Since the original phase information is limited by phase entanglement from -π to π, it cannot directly reflect the true displacement. Therefore, a phase unwinding algorithm based on adjacent sampling points is adopted. By verifying the phase difference between adjacent sampling points point by point, the distortion caused by phase entanglement is eliminated, and the absolute phase signal corresponding one-to-one with the true displacement of the patient's body surface is recovered.

[0060] Step 204: Perform feature acquisition processing on the absolute phase signal to obtain multi-dimensional original features.

[0061] In this embodiment, the absolute phase signal is processed to obtain multi-dimensional original features. A dual-sliding-window multimodal feature extraction method is employed. In the time dimension, a fixed time window of 500ms is set with a sliding step of 100ms. The temporal features (including mean, variance, peak factor, and kurtosis) and frequency domain features (including respiratory and heartbeat frequency band energy, dominant frequency, and harmonic components) of the signal within the window are extracted. In the spatial dimension, the point cloud data corresponding to the absolute phase signal is divided into spatial grid windows of 5cm × 5cm. Spatial features (including point cloud density, distance standard deviation, and target duty cycle) within each window are extracted to distinguish the motion characteristics of different regions of the chest and abdomen. The extraction results from both the time and spatial dimensions are combined to form multi-dimensional original features.

[0062] Step 205: Normalize and reduce the dimensionality of the original multi-dimensional features to obtain feature data.

[0063] In this embodiment, the multi-dimensional original features are normalized and dimensionality reduced to obtain feature data. First, the Z-score normalization method is used to process the multi-dimensional original features. By calculating the mean and standard deviation of the feature data, the dimensional differences between different features are eliminated, making each feature of the same order of magnitude. Then, the PCA (Principal Component Analysis) algorithm is used for dimensionality reduction. Principal components with a cumulative contribution rate ≥95% are selected and retained, and redundant features are removed. This reduces the computational complexity of subsequent data processing while ensuring that core information is not lost, and finally, highly robust feature data is obtained.

[0064] This application's embodiments precisely remove interference signals from the fixed structure within the scanning chamber through clutter suppression processing, solving the problem that traditional radar signals are easily affected by environmental clutter and target signals are submerged. This ensures that subsequent processing focuses only on the motion signals of the monitored object, improving signal purity. Signal optimization and phase recovery processing enhance the signal-to-noise ratio of weak motion signals, eliminate displacement distortion caused by phase entanglement, and address the shortcomings of traditional signal processing, such as difficulty in identifying weak motion signals and inaccurate phase information, achieving precise extraction of millimeter-level displacement signals. Dual sliding window multimodal feature acquisition, normalization, and dimensionality reduction processing integrate core features in the temporal and spatial dimensions, while eliminating dimensional differences and redundant information, resulting in highly robust feature data. This reduces the computational complexity of subsequent target detection and improves the adaptability of features to different motion states, solving the problem of insufficient robustness in traditional single-dimensional feature extraction.

[0065] In some embodiments, target detection and dynamic tracking processing are performed on feature data to obtain physiological motion data, including: target detection processing of feature data to obtain target detection results with target confidence and motion region coordinates.

[0066] Specifically, the feature data is processed for target detection to obtain target detection results including target confidence and motion region coordinates. The data processing system uses a lightweight CNN+Transformer hybrid detection model to perform target detection. The CNN backbone network uses the lightweight MobileNetV3 architecture, with an input feature map size of 32×32×16. Local motion features are extracted through depthwise separable convolution, and a 64-dimensional feature vector is finally output. The model parameter size is controlled within 1.2M to ensure real-time computing efficiency. The Transformer attention module introduces a 2-layer encoder and 1-layer decoder structure. The encoder uses an 8-head self-attention mechanism to capture global feature associations. The decoder outputs target confidence and motion region coordinates based on the extracted feature vector. The target confidence threshold is set to 0.7. Regions with a confidence level higher than the target confidence threshold are considered valid motion regions, and their bounding box coordinates are the motion region coordinates. Together, they constitute the target detection result, achieving accurate localization of breathing and heartbeat-related motion regions.

[0067] Furthermore, based on the target confidence score of the motion region coordinates, target position prediction processing is performed to obtain the target's predicted position in the next frame.

[0068] Specifically, the target's position is predicted based on the coordinates of the motion region to obtain the predicted position of the target in the next frame. A Kalman filter algorithm is used to construct a target motion model, establishing state and observation equations. The state vector contains the target's three-dimensional position and three-dimensional velocity. By analyzing historical data of the motion region coordinates, the trend and pattern of the target's motion are deduced, thereby predicting the target's position coordinates in the next frame. This addresses uncertainties during motion and provides a predictive basis for subsequent trajectory tracking.

[0069] Furthermore, trajectory matching processing is performed on the predicted position of the target in the next frame and the actual detection position to obtain trajectory re-association data.

[0070] Specifically, trajectory matching is performed between the predicted and actual detected positions of the target in the next frame to obtain trajectory re-association data. The Hungarian algorithm is used as the core algorithm for trajectory matching, with positional distance acceleration similarity as the cost function and a cost threshold of 0.3. When the actual detected position deviates from the predicted position due to occlusion or other factors, the algorithm calculates the matching degree between the predicted and actual detected positions. If the matching degree is higher than the threshold, trajectory association is achieved. For scenarios where the occlusion time is ≤3 seconds, trajectory re-association after occlusion can be achieved through comprehensive matching of historical trajectory data with the current predicted and actual detected positions, ensuring trajectory continuity and ultimately obtaining trajectory re-association data.

[0071] Furthermore, the target detection results and trajectory re-correlation data are integrated and processed to obtain physiological motion data.

[0072] Specifically, the target detection results and trajectory re-association data are integrated to obtain physiological motion data. The effective motion area coordinates from the target detection results are combined with the continuous trajectory information from the trajectory re-association data to extract the three-dimensional coordinate sequence of the patient's chest and abdomen center point, forming a respiratory motion trajectory. The amplitude of heart rate fluctuations is obtained by calculating the peak-to-peak value of coordinate changes in the trajectory. Based on the rate of change of coordinates over time, the motion velocity is calculated. The respiratory motion trajectory, heart rate fluctuation amplitude, and motion velocity together constitute the physiological motion data.

[0073] This application's embodiments accurately output target confidence and motion region coordinates through target detection processing. Combined with a lightweight detection model design, it achieves precise localization of breathing and heartbeat motion regions while ensuring real-time computational efficiency, solving the problems of fuzzy localization and insufficient accuracy in medical micro-motion scenarios caused by traditional target detection. Based on Kalman filter-based position prediction and Hungarian algorithm-based trajectory matching, it effectively addresses motion uncertainty and occlusion scenarios, achieving trajectory re-association and overcoming the shortcomings of traditional tracking algorithms such as trajectory breakage and tracking failure after target occlusion, ensuring the continuity and integrity of physiological motion trajectories.

[0074] In some embodiments, a mapping model is constructed, and motion compensation parameters are obtained based on the mapping model and physiological motion data, including: performing synchronous calibration scanning processing on the monitored object to obtain the surface displacement curves corresponding to the multi-directional scan images and echo signals of the monitored object.

[0075] Specifically, the monitored subjects undergo synchronous calibration scanning to acquire multi-directional scan images and body surface displacement curves. Before the formal CT / MRI scan, a synchronous calibration process is initiated, with a calibration duration set at 30 seconds. During this period, two data acquisitions are completed simultaneously: First, sagittal and coronal scan images of the monitored subjects are acquired using CT / MRI equipment, with each frame acquired at a 50ms interval, for a total of 600 frames, ensuring complete coverage of 1-2 respiratory cycles; second, echo signals are synchronously acquired using the radar hardware system, and after preprocessing, a body surface displacement curve aligned with the scan image time is extracted. This curve accurately reflects the real-time displacement changes of the monitored subject's chest and abdominal surface.

[0076] Furthermore, the multi-directional scan images are registered to obtain in vivo motion data.

[0077] Specifically, multi-planar scan images are registered to obtain in vivo motion data. The diastolic image of the heart is selected as the baseline template from the acquired multi-planar scan images. A normalized two-dimensional cross-correlation registration algorithm is used to register each of the remaining scan images with this baseline template frame by frame. By calculating the correlation differences between image pixels, the relative displacement of the scanned area in the head-to-foot, anterior-posterior, and lateral spatial directions is accurately determined. These displacements are then integrated to form complete in vivo motion data, intuitively presenting the actual motion state of internal organs.

[0078] Furthermore, the surface displacement curve and the internal motion data are subjected to model fitting to obtain a mapping model.

[0079] Specifically, a model fitting process is performed on the surface displacement curve and the internal motion data to obtain a mapping model. A fractional multinomial regression method is used to construct the fitting relationship, with the surface displacement data acquired by the radar hardware system as the independent variable and the registered internal motion data in the head-to-foot, front-back, and left-right directions as the dependent variables, respectively, establishing a three-dimensional mapping equation. The fitting parameters in each equation are solved using the least squares method to determine the quantitative correspondence between surface displacement and internal motion in each direction, ultimately forming a mapping model that accurately represents the mapping relationship between surface displacement and internal motion.

[0080] Furthermore, based on the mapping model and physiological motion data, motion adjustment processing is performed on the tracking direction of the imaging layer to obtain the layer tracking compensation parameters. The tracking direction of the imaging layer is the motion direction of the monitored object in the front-back or up-down direction.

[0081] Specifically, based on the mapping model and physiological motion data, motion adjustment processing is performed on the imaging layer tracking direction to obtain layer tracking compensation parameters. The imaging layer tracking direction is specifically the anterior-posterior direction or the superior-inferior (head-to-foot) direction of the monitored object. Movement in this direction easily causes relative displacement between the imaging layer and the target scanning area. Using the constructed mapping model, the surface displacement data in this direction from the physiological motion data is converted into the corresponding in vivo motion displacement. The radiofrequency pulse center frequency of the CT / MRI equipment is adjusted according to this displacement, based on the product relationship between the gyromagnetic ratio and the slice gradient magnetic field. Through real-time adjustment, the imaging layer synchronously follows the movement of the scanning area, ensuring that the two remain relatively stationary, thereby obtaining the layer tracking compensation parameters in this direction.

[0082] Furthermore, based on the mapping model and physiological motion data, phase shift correction is performed on the imaging artifact elimination direction to obtain artifact elimination correction parameters. The imaging artifact elimination direction is the direction of motion of the monitored object in the left-right or front-back direction within the scanned image plane.

[0083] Specifically, based on a mapping model and physiological motion data, phase shift correction is applied to the imaging artifact elimination direction to obtain artifact elimination correction parameters. The imaging artifact elimination direction refers to the left-right or front-back direction of the monitored object within the scanned image plane. Motion in these directions is the main cause of artifacts such as blurring and ghosting in the image, and corresponds to the phase encoding direction and frequency encoding direction of the imaging device. The mapping model converts the body surface displacement within this plane in the physiological motion data into phase encoding direction displacement and frequency encoding direction displacement. Based on the displacement amounts in these two directions and the corresponding number of sampling points in the encoding direction, the phase shift correction amount for each sampling point is calculated. Phase shift compensation is used to offset the signal phase deviation caused by motion, thereby obtaining the artifact elimination correction parameters.

[0084] Furthermore, the layer tracking compensation parameters and artifact elimination correction parameters are integrated to obtain motion compensation parameters.

[0085] Specifically, the layer tracking compensation parameters and artifact elimination correction parameters are integrated to obtain motion compensation parameters. These parameters are then combined to form a complete set of motion compensation parameters covering three spatial dimensions and taking into account both imaging layer alignment and artifact elimination.

[0086] This application's embodiments accurately obtain the correspondence between surface displacement and internal movement by synchronously calibrating scanning and image registration processing. Combined with a mapping model fitted by fractional multinomial regression, it achieves precise quantitative mapping between the two, solving the problems of poor adaptability and inaccurate reflection of internal organ movement in traditional surface displacement mapping techniques. Corrections are made separately for the tracking direction (front-back / up-down direction) and the imaging artifact elimination direction (left-right / front-back direction within the image plane) to achieve comprehensive motion compensation. This overcomes the limitations of traditional motion compensation, which only targets a single direction and has incomplete artifact elimination, effectively offsetting imaging deviations caused by motion in different dimensions.

[0087] In some embodiments, physiological motion data is subjected to parameter extraction and risk assessment processing to obtain physiological parameters, abnormal event types and risk scores, including filtering, feature recognition and statistical analysis of physiological motion data to obtain physiological parameters, including respiratory rate, heart rate and exercise amplitude.

[0088] Specifically, physiological motion data is filtered, feature-recognized, and statistically analyzed to obtain physiological parameters including respiratory rate, heart rate, and exercise amplitude. For respiratory rate extraction, the respiratory trajectory in the physiological motion data is first low-pass filtered with a cutoff frequency of 1Hz to remove high-frequency noise interference. Then, a peak detection algorithm is used to identify the peaks and troughs of the respiratory signal, and the time interval between adjacent peaks is calculated. The reciprocal mean of all intervals is taken as the respiratory rate. For heart rate extraction, a polynomial fitting method is used to process the physiological motion data to eliminate interference from respiratory harmonics. Subsequently, a matched filtering algorithm is used to enhance the heart rate signal. A template of 200ms is selected to extract the R-wave peak intervals in the signal, and the heart rate is calculated based on the interval time. For exercise amplitude extraction, the three-dimensional coordinate sequence of the center point of the chest and abdomen in the physiological motion data is statistically analyzed. The peak-to-peak value (difference between the maximum and minimum values) of each coordinate within the exercise cycle is calculated, and this peak-to-peak value is used as the exercise amplitude parameter, with a resolution of up to 0.1mm.

[0089] Furthermore, an anomaly detection threshold is set, and physiological parameters are compared with the anomaly detection threshold. When the physiological parameters are greater than or equal to the anomaly detection threshold, or when the parameter mutation reaches a preset proportion within a preset time, the abnormal event type is obtained.

[0090] Specifically, anomaly detection thresholds are set, and physiological parameters are compared with these thresholds to determine the type of abnormal event. These thresholds are based on statistical results of normal human physiological data, specifically: normal respiratory rate is 12-20 breaths / minute, normal heart rate is 60-100 beats / minute, and normal range of motion is ≤5mm. The extracted physiological parameters are compared with their corresponding thresholds. If the respiratory rate is below 12 breaths / minute or above 20 breaths / minute, it is considered an abnormal respiratory rate; if the heart rate is below 60 beats / minute or above 100 beats / minute, it is considered an abnormal heart rate; if the motion amplitude is greater than 5mm, it is considered an abnormal motion amplitude. Simultaneously, the dynamic changes of the physiological parameters are monitored. If the abrupt change in any parameter within a preset 3-second timeframe reaches a preset proportion of 30%, it is considered an abnormal parameter change. The above determination results are the corresponding abnormal event type.

[0091] Furthermore, the degree of deviation of physiological parameters is classified into risk levels to obtain risk score results.

[0092] Specifically, the degree of deviation of physiological parameters is classified into risk levels to obtain risk scores. First, the deviation of each physiological parameter is calculated: the difference between parameter values ​​exceeding the normal range and the threshold boundary is normalized to ensure the deviation falls within the 0-1 range; parameters within the normal range have a deviation of 0. A weighted scoring method is used to calculate the comprehensive risk score, with the following weights: respiratory parameter deviation weight 0.3, heart rate parameter deviation weight 0.5, and exercise parameter deviation weight 0.2. The comprehensive score is calculated using the formula: Risk Score = 0.3 × Respiratory Deviation + 0.5 × Heart Rate Deviation + 0.2 × Exercise Deviation. Risk levels are then classified based on the comprehensive score: a score ≥ 0.6 indicates high risk, 0.3 ≤ score < 0.6 indicates medium risk, and a score < 0.3 indicates low risk. This risk level, along with the comprehensive score, constitutes the overall risk score.

[0093] This application employs targeted filtering and feature recognition algorithms to accurately extract respiratory rate, heart rate, and amplitude of movement, addressing the problems of low accuracy and susceptibility to interference in traditional physiological parameter extraction. This provides reliable vital sign data for medical assessment. Based on a large amount of normal human data, anomaly detection thresholds are set, combined with parameter mutation monitoring, enabling rapid identification and type determination of abnormal events. This overcomes the shortcomings of traditional monitoring, which only outputs raw data and cannot proactively warn of anomalies, providing support for timely intervention by medical personnel. A weighted scoring method enables quantitative risk assessment, clearly defining high, medium, and low risk levels. This addresses the problems of vagueness and lack of quantitative standards in traditional anomaly assessment, allowing medical personnel to quickly grasp the physiological risk level of the monitored subject and improve the efficiency of medical decision-making.

[0094] In some embodiments, the motion compensation parameters and the scanning data are fused and optimized to obtain imaging data, including: aligning the scanning data to obtain aligned scanning data.

[0095] Specifically, the scan data is aligned to obtain aligned scan data. Alignment is performed using slice tracking compensation parameters. For the movement of the monitored object in the front-back or up-down (head-to-foot) direction, based on the product of the magnetogyroscope ratio and the slice gradient magnetic field, the radiofrequency pulse center frequency of the CT / MRI equipment is adjusted to ensure that the imaging slice follows the movement trajectory of the scanned area in real time. This eliminates the relative displacement interference between the scanning slice and the target area caused by movement in that direction, ensuring that the scan data accurately matches the target area in spatial dimension, ultimately yielding aligned scan data.

[0096] Furthermore, phase shift compensation processing is performed on the aligned scan data to obtain compensated scan data.

[0097] Specifically, phase shift compensation is performed on the aligned scan data to obtain compensated scan data. Based on artifact elimination correction parameters, for motion in the left-right or front-back directions within the scanned image plane, the phase shift correction amount is calculated for each sampling point, taking into account the number of sampling points in the frequency encoding direction and the number of sampling points in the phase encoding direction. This phase shift correction amount is then applied to the scan data to offset the signal phase deviation caused by motion, suppressing blurring and ghosting artifacts, thus obtaining compensated scan data.

[0098] Furthermore, non-rigid registration optimization processing is performed on the compensated scan data to obtain optimized scan data.

[0099] Specifically, non-rigid registration optimization is performed on the compensated scan data to obtain optimized scan data. A normalized two-dimensional cross-correlation registration algorithm is employed. Based on the compensated scan data, the scan data frame corresponding to the diastolic phase of the heart is selected as the reference template. Each of the remaining frames is then registered frame-by-frame with the reference template. By calculating the correlation differences between pixels, residual motion deviations caused by inconsistent motion amplitudes in different tissue areas are corrected, further improving the spatial consistency of the scan data, ultimately yielding the optimized scan data.

[0100] Furthermore, image reconstruction processing is performed on the optimized scan data to obtain imaging data.

[0101] Specifically, the optimized scan data undergoes image reconstruction to obtain imaging data. Depending on the imaging type of the CT or MRI equipment, the corresponding reconstruction algorithm is used to reconstruct the optimized scan data: CT imaging employs a filtered back-projection algorithm, which filters the projection data and then projects it back into the image space; MRI imaging uses an iterative reconstruction algorithm, combining precise data from previous motion compensation to iteratively optimize image quality. During the reconstruction process, synchronously aligned motion monitoring data is integrated to form an imaging result adapted to the physiological motion cycle, ultimately outputting high-resolution, low-artifact imaging data.

[0102] This application's embodiments eliminate misalignment between the scanning layer and the target area caused by front-to-back / up-to-down movement through alignment processing based on layer tracking compensation parameters. This solves the image offset problem caused by layer selection direction movement in traditional imaging, ensuring the spatial accuracy of the scan data. Phase shift compensation processing specifically counteracts phase deviations caused by in-plane movement, effectively suppressing artifacts such as blurring and ghosting, solving the artifact residue problem caused by coding direction movement in traditional imaging, and improving image detail recognition. Non-rigid registration optimization further corrects residual motion deviations, especially imaging differences caused by inconsistent motion amplitudes in different tissue areas, overcoming the shortcomings of traditional rigid registration in adapting to complex movements, and achieving high-precision spatial consistency calibration of scan data.

[0103] While this application provides method operation steps as shown in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-inventive labor. The order of steps listed in this embodiment is merely one possible execution order among many and does not represent the only execution order. In actual device or client product execution, the method can be executed sequentially according to this embodiment or the accompanying drawings, or in parallel (e.g., in a parallel processor or multi-threaded processing environment).

[0104] The methods, apparatus, or modules described in this application can be implemented in a computer-readable program code manner. The controller can be implemented in any suitable manner, such as a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of a memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code manner, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included within it for implementing various functions can also be considered as structures within the hardware component. Alternatively, the device used to implement various functions can be viewed as either a software module that implements the method or a structure within a hardware component.

[0105] This application also provides an apparatus, the apparatus comprising: a processor; a memory for storing processor-executable instructions; wherein, when the processor executes the executable instructions, it implements the method described in this application.

[0106] This application also provides a non-volatile computer-readable storage medium storing a computer program or instructions thereon, which, when executed, enables the method described in this application embodiment to be implemented.

[0107] Furthermore, in the various embodiments of the present invention, each functional module can be integrated into a processing module, or each module can exist independently, or two or more modules can be integrated into a single module.

[0108] The aforementioned storage media include, but are not limited to, Random Access Memory (RAM), Read-Only Memory (ROM), Cache, Hard Disk Drive (HDD), or Memory Card. The memory can be used to store computer program instructions.

[0109] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary hardware. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product, or it can be embodied in the process of data migration. The computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, mobile terminal, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0110] The various embodiments described in this specification are presented in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. All or part of this application can be used in numerous general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, mobile communication terminals, multiprocessor systems, microprocessor-based systems, programmable electronic devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices, etc.

[0111] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of this application.

Claims

1. A motion monitoring and imaging optimization system, characterized in that, This includes the radar hardware system and data processing system with electrical signal connections; The radar hardware system is configured to: reflect a continuous wave signal to the monitored object and receive the echo, obtain the echo signal, and send the echo signal to the data processing system; The data processing system is configured to: acquire scan data, perform time alignment processing on the scan data and the echo signal, and obtain time-aligned scan data and echo signal; The echo signal is preprocessed and feature extracted to obtain feature data; The feature data is subjected to target detection and dynamic tracking processing to obtain physiological motion data; A mapping model is constructed, and motion compensation parameters are obtained based on the mapping model and the physiological motion data. The mapping model is used to characterize the mapping relationship between body surface displacement and internal motion. The physiological motion data is processed for parameter extraction and risk assessment to obtain physiological parameters, abnormal event types, and risk scores. The motion compensation parameters and the scan data are fused and optimized to obtain imaging data; The imaging data, physiological parameters, abnormal event types, and risk score results are output.

2. The system according to claim 1, characterized in that, The preprocessing and feature extraction of the echo signal to obtain feature data includes: The echo signal is subjected to clutter suppression processing to obtain a clutter-free signal; The denoised signal is subjected to signal optimization processing to obtain a signal with improved signal-to-noise ratio; The signal with improved signal-to-noise ratio is subjected to phase recovery processing to obtain an absolute phase signal; The absolute phase signal is subjected to feature acquisition processing to obtain multi-dimensional original features; The original multi-dimensional features are normalized and dimensionality reduced to obtain feature data.

3. The system according to claim 1, characterized in that, The process of performing target detection and dynamic tracking on the feature data to obtain physiological motion data includes: The feature data is processed for target detection to obtain the target confidence score and the coordinates of the moving area. Based on the coordinates of the motion region, the target confidence score is used to predict the target position, and the predicted position of the target in the next frame is obtained. The predicted position of the target in the next frame is matched with the actual detected position to obtain trajectory re-association data; The target detection results and the trajectory re-association data are integrated and processed to obtain physiological motion data.

4. The system according to claim 1, characterized in that, The construction of the mapping model, and the obtaining of motion compensation parameters based on the mapping model and the physiological motion data, includes: The monitored object is simultaneously calibrated and scanned to obtain multi-directional scan images of the monitored object and the surface displacement curves corresponding to the echo signals. The multi-directional scan images are registered to obtain in vivo motion data; The surface displacement curve and the internal motion data are subjected to model fitting to obtain a mapping model; Based on the mapping model and the physiological motion data, motion adjustment processing is performed on the imaging layer tracking direction to obtain the layer tracking compensation parameters. The imaging layer tracking direction is the motion direction of the monitored object in the front-back or up-down direction. Based on the mapping model and the physiological motion data, phase shift correction is performed on the imaging artifact elimination direction to obtain artifact elimination correction parameters. The imaging artifact elimination direction is the movement direction of the monitored object in the left-right or front-back direction in the scanned image plane. The layer tracking compensation parameters and the artifact elimination correction parameters are integrated to obtain motion compensation parameters.

5. The system according to claim 1, characterized in that, The process of extracting parameters and assessing risks from the physiological motion data yields physiological parameters, abnormal event types, and risk scores, including: The physiological motion data is filtered, feature-recognized, and statistically analyzed to obtain physiological parameters, including respiratory rate, heart rate, and exercise amplitude. Set an anomaly detection threshold, compare the physiological parameters with the anomaly detection threshold, and obtain the anomaly event type when the physiological parameters are greater than or equal to the anomaly detection threshold or when the parameter mutations within a preset time reach a preset proportion. The degree of deviation of the physiological parameters is classified into risk levels to obtain risk score results.

6. The system according to claim 1, characterized in that, The process of fusing and optimizing the motion compensation parameters with the scan data to obtain imaging data includes: The scan data is aligned to obtain aligned scan data; Phase shift compensation is performed on the aligned scan data to obtain compensated scan data; The compensated scan data is then subjected to non-rigid registration optimization to obtain the optimized scan data. Image reconstruction processing is performed on the optimized scan data to obtain imaging data.

7. The system according to claim 3, characterized in that, The physiological motion data includes respiratory movement trajectory, heart rate fluctuation amplitude, and movement speed.

8. A motion monitoring and imaging optimization method applicable to the system described in any one of claims 1-7, characterized in that, include: Acquire scan data, and perform time alignment processing on the scan data and the echo signal received from the radar hardware system to obtain time-aligned scan data and echo signal; The echo signal is preprocessed and feature extracted to obtain feature data; The feature data is subjected to target detection and dynamic tracking processing to obtain physiological motion data; A mapping model is constructed, and motion compensation parameters are obtained based on the mapping model and the physiological motion data. The mapping model is used to characterize the mapping relationship between body surface displacement and internal motion. The physiological motion data is processed for parameter extraction and risk assessment to obtain physiological parameters, abnormal event types, and risk scores. The motion compensation parameters and the scan data are fused and optimized to obtain imaging data; The imaging data, physiological parameters, abnormal event types, and risk score results are output.

9. A computer device comprising a memory and a processor, the memory storing a computer program executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method of claim 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method as described in claim 8.

Images