Methods and systems for monitoring and correcting body surface contours in radiotherapy patients

Abstract

This application relates to the field of medical image processing technology, and discloses a method and system for monitoring and correcting the body surface contour of radiotherapy patients. The method includes: acquiring raw point cloud data from the patient's body surface and generating a high-resolution body surface contour model by combining a 3D scanning device, a stereomicroscope, and an infrared sensor; dividing the body surface into multiple monitoring blocks based on preset anatomical region division rules; extracting local dynamic change features through time series analysis; generating correction guidance parameters by combining the positioning information of the offset region; and constructing a correction model to obtain a collaborative correction instruction set. Finally, correction is performed based on the collaborative correction instruction set, and the success of the correction is determined by non-rigid registration technology. If unsuccessful, adjustments are made until the correction is successful. This method can monitor and correct the dynamic changes of the patient's body surface contour in real time, ensuring accurate treatment positioning during radiotherapy and improving treatment efficacy and safety.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, and in particular to a method and system for monitoring and correcting the body surface contours of radiotherapy patients. Background Technology

[0002] During radiotherapy, the geometry of a patient's body surface undergoes dynamic changes due to respiration, positional changes, and other physiological factors, particularly in areas such as the chest and abdomen, where these changes are especially pronounced. Since the treatment area for radiotherapy is usually closely related to the patient's body surface geometry, any changes in the body surface can lead to deviations in the positioning of the treatment area, thereby affecting the treatment outcome and even causing side effects.

[0003] Traditional radiotherapy positioning methods rely on static image data or manual adjustment of the positioning system. However, these methods cannot track dynamic changes on the patient's body surface in real time, and therefore cannot effectively address changes in body surface morphology caused by physiological movements.

[0004] To address this issue, researchers have proposed various dynamic monitoring and correction methods in recent years, including using three-dimensional imaging technology and sensors to monitor changes on the patient's body surface in real time.

[0005] However, most existing methods are either limited to single devices and technologies or struggle to achieve high-precision and efficient real-time correction, and face significant technical challenges in dynamic change monitoring and physiological motion compensation. For example, how to capture minute changes on the body surface in real time during a patient's respiratory cycle, and how to dynamically adjust treatment parameters based on these changes to ensure radiotherapy accuracy and efficacy, remain pressing problems to be solved in the field of radiotherapy.

[0006] Furthermore, traditional methods suffer from insufficient accuracy and slow processing speed, failing to provide timely and effective correction solutions in rapidly changing treatment environments. To address these issues, this application provides a method combining three-dimensional scanning, stereomicroscopy, and infrared sensors to monitor and correct changes in the patient's body surface contours in real time. This method utilizes dynamic change characteristics for precise correction, thereby ensuring accurate execution of the radiotherapy process and improving treatment efficacy and patient safety. Summary of the Invention

[0007] This invention provides a method for monitoring and correcting the body surface contour of radiotherapy patients, mainly including:

[0008] S1. Obtain raw point cloud data from the patient's body surface, generate a high-resolution body surface contour model based on the point cloud data, divide the body surface into multiple monitoring blocks using preset anatomical region division rules, and determine the geometric boundary features of each monitoring block.

[0009] S2. For each monitoring block, acquire its corresponding anatomical structure characteristics and motion pattern data, extract the local dynamic change characteristics of the local area during the respiratory cycle, and if the local dynamic change characteristics exceed the preset deviation threshold, use the offset area positioning method based on geometric matching to determine the positioning information of the offset area.

[0010] S3. Based on the positioning information of the offset area, obtain the deviation vector and motion trajectory data, construct a weighted model to calculate the deviation data and obtain the correction guidance parameters, construct a correction model, and input the correction guidance parameters, geometric boundary features, anatomical structure characteristics and motion pattern data into the correction model to obtain a collaborative correction instruction set;

[0011] S4. Perform correction based on the collaborative correction instruction set, perform non-rigid registration between the corrected point cloud data and the high-resolution body surface contour model, obtain the registration result and determine whether the correction is successful. If the correction is unsuccessful, analyze the factors that caused the failure and perform different processing based on different factors until the correction is successful.

[0012] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects:

[0013] This invention overcomes the technical bottlenecks of insufficient accuracy in dynamic change monitoring and difficulties in real-time correction in traditional radiotherapy positioning methods by combining 3D scanning, stereomicroscopy, and infrared sensor technology. By acquiring raw point cloud data and generating a high-resolution body surface contour model, the geometric features of the patient's body surface can be accurately captured, especially during physiological processes such as breathing and changes in body position, where dynamic changes in the body surface contour are effectively monitored and corrected. Specifically, the combination of 3D scanning equipment and infrared sensors enables real-time acquisition of dynamic data of the patient's body surface during radiotherapy, greatly improving the accuracy and real-time nature of data acquisition. Simultaneously, through time series analysis and extraction of dynamic change features, the offset regions of the body surface under the influence of physiological movement can be accurately identified, thereby locating the offset information through geometric matching and ensuring the accuracy of correction of the offset regions. More importantly, by constructing a weighted model and correction guidance parameters, this invention can generate a collaborative correction instruction set for different monitoring blocks, ensuring that the motion patterns and physiological characteristics of each monitoring block are accurately compensated, greatly improving the accuracy and stability of the radiotherapy treatment process. Furthermore, this invention employs non-rigid registration technology, allowing the patient's body surface to adapt to deformations caused by physiological movement during correction, further optimizing the radiotherapy effect. This method provides real-time feedback on correction results and assesses the success of correction based on registration errors. If unsuccessful, it automatically analyzes the cause and re-corrects, ensuring high-precision treatment execution. Through this precise and intelligent monitoring and correction method, the patient's body surface contour remains consistent with the treatment plan, greatly improving the accuracy of radiotherapy, reducing the possibility of side effects, and providing a novel technical solution for the field of radiotherapy with broad application prospects and significant market value. Attached Figure Description

[0014] Figure 1 This is a flowchart of the radiotherapy patient body surface contour monitoring and correction method of the present invention;

[0015] Figure 2 This is a schematic diagram illustrating the specific process of determining the geometric boundary features of each block according to the present invention;

[0016] Figure 3 This is a comparison chart of the errors of the present invention using the traditional method and the LSTM model;

[0017] Figure 4 This is a schematic diagram of the structure of the radiotherapy patient surface contour monitoring and correction system of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0019] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 One embodiment of the method for monitoring and correcting the body surface contour of radiotherapy patients in this application includes:

[0020] S1. Obtain raw point cloud data from the patient's body surface, generate a high-resolution body surface contour model based on the point cloud data, divide the body surface into multiple monitoring blocks using preset anatomical region division rules, and determine the geometric boundary features of each monitoring block.

[0021] Specifically, during radiotherapy, the patient's body contours dynamically change with time, position, and physiological processes such as respiration, especially in the abdominal and chest areas. A flexible pressure mattress is placed on top of a rigid carbon fiber bed within the radiotherapy space, directly contacting the patient. This mattress collects pressure signals from the patient's lower back in a supine position, with its output circuitry directly connected to a data acquisition module located on the side of the mattress. This module collects signals from pressure-sensitive points in each row and column of the mattress, processes them, and sends them to a data display module. The data acquisition module and the data display module are connected via a network cable. The data display module is located outside the radiotherapy room in a control room. Inside the treatment room, a two-dimensional image of the patient's back pressure signal is displayed in real time. The pressure sensing data, combined with a three-dimensional image, assists in adjusting positioning errors. Since the treatment area for radiotherapy is typically highly correlated with the geometry of the patient's body surface, accurate monitoring and timely correction of changes in the body surface contour are crucial to ensuring the accuracy and effectiveness of radiotherapy. If body surface contour deviation occurs during radiotherapy, it may lead to positioning errors in the treatment area, or even affect the patient's treatment outcome or cause side effects. Therefore, how to monitor and correct these changes in real time to ensure that the patient's body surface remains consistent with the treatment plan throughout the treatment process has become a pressing technical challenge in the field of radiotherapy. Therefore, this application provides a method for monitoring and correcting the body surface contour of radiotherapy patients.

[0022] First, raw point cloud data of the patient is acquired using a combination of 3D scanning equipment, stereomicroscope, and infrared sensors. A high-resolution body surface contour model is then generated based on the point cloud data. The specific method for generating the high-resolution body surface contour model will be explained in detail later. To facilitate subsequent monitoring and correction, the body surface is divided into multiple monitoring blocks according to anatomical region division rules (e.g., head and neck, chest, and abdomen). The divided monitoring blocks can accurately reflect the geometric features and movement patterns of each region. To facilitate subsequent monitoring and correction, the geometric boundary features of each monitoring block are also determined. The specific method for determining the geometric boundary features of each monitoring block will be explained in detail later.

[0023] S2. For each monitoring block, acquire its corresponding anatomical structure characteristics and motion pattern data, extract the local dynamic change characteristics of the local area during the respiratory cycle, and if the local dynamic change characteristics exceed the preset deviation threshold, use the offset area positioning method based on geometric matching to determine the positioning information of the offset area.

[0024] Specifically, during radiotherapy, the patient's body surface contours undergo dynamic changes with physiological activities such as breathing. To detect and correct these changes in a timely manner, it is first necessary to detect the local dynamic changes of each monitoring block through time series analysis. Dynamic change characteristics refer to the morphological changes that occur on the patient's body surface during physiological processes such as the respiratory cycle, especially the movement characteristics related to breathing. If the change in a certain area exceeds a preset deviation threshold, it indicates that the area has shifted. The shifted area is identified, and the shift information is further located. When the dynamic change characteristics exceed the threshold, a shifted area location method based on geometric matching is used. By matching the model and the current data, the location of the shifted area is determined. The specific method for determining the location information of the shifted area will be explained in detail later. The location information will provide key data for subsequent correction, ensuring that dynamic changes during radiotherapy do not affect the accuracy of treatment.

[0025] S3. Based on the positioning information of the offset area, obtain the deviation vector and motion trajectory data, construct a weighted model to calculate the deviation data and obtain the correction guidance parameters, construct a correction model, and input the correction guidance parameters, geometric boundary features, anatomical structure characteristics and motion pattern data into the correction model to obtain the collaborative correction instruction set.

[0026] Specifically, based on the determined offset area positioning information, the deviation vector (a vector representing the relative positional change of the offset area) and motion trajectory data are extracted. These data are a quantitative representation of the patient's surface motion changes. A weighted model is constructed using these data. Based on the physiological characteristics and motion patterns of different monitoring blocks, the deviation data of each monitoring block is calculated. Based on the deviation data calculated by the weighted model, correction guidance parameters are generated. The correction guidance parameters include translation, rotation angle, deformation, etc., indicating how to correct the position and posture of the monitoring blocks. The correction guidance parameters, geometric boundary features, anatomical structure characteristics, and motion pattern data are input into the correction model to obtain a set of collaborative correction instructions. The collaborative correction instruction set includes instructions for rigid monitoring blocks such as translation correction and rotation correction, as well as deformation correction instructions for deformable blocks.

[0027] S4. Based on the collaborative correction instruction set, perform correction, non-rigid registration of the corrected point cloud data and the high-resolution body surface contour model, obtain the registration result and determine whether the correction is successful. If unsuccessful, analyze the factors that caused the correction failure and perform different processing based on different factors until the correction is successful.

[0028] Specifically, based on the collaborative correction instruction set, the correction of the body surface contour is performed. Non-rigid registration technology is used to register the corrected point cloud data with the high-resolution body surface contour model. Non-rigid registration allows the body surface to make appropriate deformation adjustments during correction to adapt to physiological movement and offset. The registration result is evaluated, and the success of the correction is determined based on the registration error (such as residual vector and root mean square error). If the correction is unsuccessful, the factors causing the failure are analyzed, and necessary measures are taken to re-correct until it is successful.

[0029] The aforementioned method combines 3D scanning, stereomicroscopy, and infrared sensors to acquire precise point cloud data, generating a high-resolution body surface contour model divided into multiple monitoring blocks. Through time-series analysis and extraction of dynamic change features, it monitors the physiological movement effects on the patient's body surface and generates correction guidance parameters. Correction is performed based on a collaborative correction instruction set, and the results are evaluated to ensure precise alignment of the treatment area during radiotherapy, thereby improving treatment efficacy.

[0030] In one specific embodiment, generating a high-resolution body surface contour model based on point cloud data specifically includes the following steps:

[0031] S11. The initial point cloud data is aligned using a point cloud registration algorithm to obtain standard point cloud data in a unified coordinate system. The standard point cloud data is then denoised using a mean filtering algorithm to obtain smooth point cloud data.

[0032] S12. Based on the high-resolution image data acquired by the stereomicroscope, generate texture mapping data, and fuse the texture mapping data with the smooth point cloud data to obtain a high-resolution body surface contour model containing texture information.

[0033] S13. If there are local missing areas in the high-resolution body surface contour model, an interpolation algorithm is used to fill in the missing areas to obtain a complete body surface contour model.

[0034] Specifically, to obtain an accurate and complete body surface contour model, firstly, a point cloud registration algorithm is used to align the original point cloud data obtained from different viewpoints or devices. Commonly used point cloud registration algorithms include iterative nearest neighbor (ICP) algorithm, coarse alignment, and fine alignment methods. This step transforms all point cloud data into a unified coordinate system for further processing. The aligned point cloud data is then denoised, typically using a mean filtering algorithm (or other denoising techniques) to smooth the data and eliminate noise caused by equipment errors or external environmental interference, resulting in smoothed point cloud data. High-resolution image data of the patient's body surface is then acquired using a stereomicroscope (or other high-resolution imaging equipment). This image data provides detailed information about the body surface, such as skin texture, tumor markers, and surgical areas. Based on the acquired data... High-resolution image data is processed using a texture mapping algorithm to map the image's texture information onto smoothed point cloud data. Texture mapping technology maps the color information of image pixels to each point in the point cloud model, making the point cloud model not only a representation of geometric shapes but also containing specific texture details. The generated texture mapping data is then fused with the smoothed point cloud data. Through the combination of color and geometric information, the resulting body surface contour model not only accurately describes the shape but also displays realistic body surface texture. In the generated high-resolution body surface contour model, the presence of local missing regions is checked. Missing regions may be caused by blind spots during scanning, equipment precision limitations, or patient posture. These regions are detected and labeled as missing regions. For the detected missing regions, interpolation algorithms are used to fill in these regions. Common interpolation methods include nearest neighbor interpolation, bilinear interpolation, or surface-based higher-order interpolation methods. The interpolation process uses data from surrounding known points to estimate the shape of the missing region, ensuring the continuity and smoothness of the model. The missing region is filled in by the interpolation algorithm, generating a complete body surface contour model. The complete body surface contour model has smooth boundaries and complete texture information, providing accurate data for subsequent body surface monitoring and correction.

[0035] In one specific embodiment, determining the geometric boundary features of each block includes the following steps:

[0036] S21. Based on the spatial distribution data of each monitoring block, determine the corresponding initial boundary range. If the initial boundary ranges overlap, use a grid segmentation algorithm to spatially divide the overlapping areas to obtain the boundary data of non-overlapping monitoring blocks.

[0037] S22. Based on the boundary data, extract the geometric boundary features of each monitoring block and generate a spatial geometric description of each monitoring block. The spatial geometric description refers to the expression of the boundary of the monitoring block using a mathematical model.

[0038] S23. Combining spatial geometric description, a surface fitting algorithm is used to smooth the boundaries of each monitoring block to obtain optimized boundary surface data. An interpolation algorithm is used to repair the discontinuous areas in the boundary surface data to obtain a continuous boundary surface model.

[0039] S24. Based on the boundary surface model, generate three-dimensional visualization data for each monitoring block and determine the final geometric features of the block.

[0040] Specifically, to ensure that the geometric boundaries of each monitoring block are accurately defined and described, such as Figure 2 The diagram illustrates the specific process of determining the geometric boundary features of each monitoring block. Based on the spatial distribution data of each monitoring block, the corresponding initial boundary range is determined. Spatial distribution data refers to the geometric information of the position and distribution of each monitoring block in three-dimensional space. For example, the spatial position and distribution of the head and neck, chest, and abdomen on the patient's body surface. Spatial distribution data is usually obtained through 3D scanning or imaging technology and includes information such as the size, shape, position, and relative relationships of each region. Based on the spatial distribution data, the initial boundary range of each monitoring block (such as the head and neck, chest, and abdomen) is determined. The initial boundary range is roughly delineated based on anatomical structures and the patient's body surface scan data. If the initial boundary ranges of multiple monitoring blocks overlap, they need to be processed using a grid segmentation algorithm. The grid segmentation algorithm divides the overlapping area into multiple sub-regions and assigns an independent monitoring block boundary to each sub-region, avoiding inaccuracies caused by region overlap. The grid segmentation method creates a grid model to divide the overlapping area into non-overlapping small regions. These small regions are used as the boundary data of new monitoring blocks, ensuring that there is no overlap between each monitoring block and that they can be accurately located.

[0041] Based on the initially defined boundary data of the monitoring blocks, specific geometric boundary features of the monitoring blocks are extracted using a geometric feature extraction algorithm. These geometric boundary features include the edge lines, surfaces, or spatial distribution characteristics of the monitoring blocks. For example, the boundary of the head and neck may be represented as a curved surface, while the boundary of the abdomen may present a certain geometric shape, such as an ellipse or a rectangle. Based on the geometric boundary features, a spatial geometric description of each monitoring block is generated. This spatial geometric description refers to expressing the boundary of the monitoring block using a mathematical model. This process represents the geometric shape of the monitoring block through formulas, surfaces, or mathematical curves. For example, if the boundary of the monitoring block is a curved surface, a parametric equation can be used to represent the surface, or a B-spline curve can be used to represent the boundary line.

[0042] The boundaries of the monitoring blocks are smoothed using surface fitting algorithms. These algorithms fit a smooth surface based on known point data, ensuring the smoothness and continuity of the boundaries. Common surface fitting algorithms include least squares, spline curve fitting, and B-spline fitting. If discontinuous areas exist in the fitted boundary surface, it indicates potential cracks or gaps caused by data noise or measurement errors. In such cases, interpolation algorithms can be used to repair these discontinuous areas. Based on the repaired boundary surface model, computer graphics technology is used to generate 3D visualization data of the monitoring blocks, transforming the geometry of each monitoring block into a 3D image that can be displayed on a computer screen, facilitating further monitoring and correction. By extracting the geometric features of each monitoring block (such as area, volume, curvature, and boundary smoothness), the spatial attributes of each monitoring block are ultimately determined. These geometric features provide a quantitative basis for subsequent radiotherapy monitoring and correction. For example, the curvature of the head and neck region is relatively large, while the curvature of the abdominal region is relatively small. These geometric features can help to accurately locate the radiotherapy area.

[0043] In one specific embodiment, extracting the local dynamic change features of a local region during the respiratory cycle specifically includes the following steps:

[0044] S31. Use a stereomicroscope and motion capture system to acquire data on the corresponding anatomical structural characteristics and motion patterns;

[0045] S32. The collected motion pattern data is preprocessed by time series analysis to remove noise and standardize it. Fourier transform is used to perform frequency domain analysis on the standardized motion pattern data to extract the frequency characteristics of the block within the respiratory cycle.

[0046] S33. Use a low-pass filter to smooth the high-frequency noise in the frequency domain features to obtain stable time series data. Use a long short-term memory network model to extract the local dynamic change features under the influence of respiratory motion from the time series data to obtain dynamic feature vectors.

[0047] S34. Based on the dynamic feature vector, principal component analysis is used to reduce the dimensionality of high-dimensional features to obtain low-dimensional dynamic change features. If the variance contribution rate of the low-dimensional dynamic change features is lower than the preset threshold, the parameters of principal component analysis are adjusted and recalculated to obtain optimized local dynamic change features.

[0048] Specifically, to accurately extract the dynamic changes of the patient's body surface during the respiratory cycle during radiotherapy, a stereomicroscope and a motion capture system are used to acquire corresponding anatomical structural characteristics and motion pattern data. The stereomicroscope is used to capture high-resolution images of the monitored area and obtain its anatomical structural characteristics, including important parameters such as the area's geometry, size, and relative position. The motion capture system is used to collect real-time motion trajectory data of the monitored area. By tracking the movement of the monitored area during the respiratory cycle, motion pattern data of the area is obtained. Motion pattern data refers to the movement trajectory of the monitored area during the respiratory cycle. The obtained motion pattern data undergoes time series analysis preprocessing. Noise reduction algorithms are used to remove noise caused by equipment errors, environmental interference, and other factors, ensuring data smoothness and continuity. The denoised data is then standardized to eliminate differences between different data sources or measurement scales, ensuring all motion pattern data fall within the same dimensional range. Fourier transform is applied to the standardized time series data, converting the time-domain data into frequency-domain data. Fourier transform can extract features of different frequency components in the motion pattern data, identifying periodic patterns within the respiratory cycle. Frequency features are extracted from the frequency-domain data, focusing primarily on frequency components representing periodic fluctuations. These components reflect the regular dynamic changes of the monitored blocks within the respiratory cycle. High-frequency noise in the frequency domain is filtered out, typically using a low-pass filter to smooth the signal. Low-pass filters effectively remove high-frequency noise while retaining lower-frequency dynamic features, ensuring smoother data. The smoothed time-series data is then input into a Long Short-Term Memory (LSTM) network model. LSTM is a special type of recurrent neural network that effectively processes and predicts long-term dependencies in time-series data, extracting local dynamic change features influenced by respiratory motion, such as changes in block shape and location, to derive dynamic features representing these changes. Principal Component Analysis (PCA) maps high-dimensional data to a low-dimensional space through linear transformation, preserving the main information in the data while removing redundant and unnecessary features. The dynamic change features after dimensionality reduction can better represent the local motion patterns of blocks, reducing noise and computational burden. To ensure that the dimensionality-reduced features can accurately reflect the main change patterns of the data and avoid information loss, it is also determined whether the variance contribution rate of the dimensionality-reduced low-dimensional dynamic change features is lower than a preset threshold. If so, the PCA parameters are adjusted and the dimensionality reduction calculation is performed again, so that the optimized low-dimensional features can provide a more accurate description of dynamic changes and ensure that the features play a role in subsequent correction and monitoring processes.

[0049] Traditional methods such as edge detection, motion estimation, and optical flow may suffer from significant errors in dynamic feature extraction, especially when dealing with complex or non-periodic changes. Our proposed method, using an LSTM model, exhibits smaller errors in dynamic feature extraction and can better track changing trends, particularly under the influence of complex physiological movements. Figure 3 The figure shown is a comparison of the errors using the traditional method and the LSTM model.

[0050] In one specific embodiment, determining the positioning information of the offset region includes the following steps:

[0051] S41. Obtain the preliminary offset range of the target monitoring block. Based on the geometric characteristics of the offset range, initially locate the offset area as a surface fitting model and generate preliminary offset area positioning data.

[0052] S42. Using a shape matching algorithm based on surface fitting, the geometric features of the target monitoring block are compared with the preset standard surface. The position of the offset area is determined by calculating the surface similarity metric, and a preliminary positioning result is obtained.

[0053] S43. Based on the preliminary positioning results, a spatial registration method based on geometric projection is used to spatially calibrate the offset areas at different angles and positions. The positioning error is optimized by using a projection alignment algorithm to further improve the positioning accuracy.

[0054] S44. Based on the geometric features of the offset region, an adaptive weighted distance calculation method is used to assign a unique weight coefficient to each offset region, and accurate positioning information is obtained through weighted calculation.

[0055] Specifically, to ensure the precise positioning of the offset area and thus provide an accurate basis for subsequent radiotherapy correction, during the monitoring process, the preliminary offset range of the target monitoring block (such as the head and neck, chest, or abdomen) is first determined using the aforementioned dynamic change feature extraction method and monitoring results. The offset range refers to the displacement area of ​​the monitoring block during the respiratory cycle or other physiological activities. It can be estimated by comparing the surface contour model of the initial state and the current state, or by analyzing the patient's surface movement trajectory. By analyzing the geometric features of the target monitoring block, especially the shape and curvature of its boundaries, more accurate positioning data is provided for the preliminary offset range. The offset range is usually represented by a region in three-dimensional space, covering the boundary area where offset may occur, and needs to be compared with the preset geometric model. A surface fitting algorithm is used to perform preliminary modeling of the target monitoring block to generate a preliminary surface fitting model. This model represents the initial geometric features of the target block. The initial location of the offset region is determined based on this surface fitting model, and its range is the offset region, involving the geometric deformation and spatial position of the region. Using a shape matching algorithm based on surface fitting, the geometric features of the target monitoring block are compared with a preset standard surface. The standard surface is a trained and optimized standard anatomical model or reference model that represents the normal body surface contour. Through the shape matching algorithm, the similarity between the surface of the target monitoring block and the standard surface is calculated. The similarity measure usually includes geometric indicators such as the distance and angle error between the surfaces. The similarity calculated by shape matching generates the initial offset region location result. This result provides the spatial position and boundary of the offset region. Based on the initial location result, the offset region of the target monitoring block is further optimized using a spatial registration method based on geometric projection. The spatial registration method adjusts the positioning of the target block by calculating the transformation relationship between the target monitoring block and the reference model (such as the standard anatomical model) to ensure that the two positions are accurately aligned. A projection alignment algorithm is used to optimize the positioning accuracy of the offset region at different angles and positions. The projection alignment algorithm projects the target monitoring block onto the geometric space of the reference model, compares the projection differences, and further optimizes the positioning accuracy. The optimized spatial registration result yields the final accurate offset region positioning information. Based on the geometric characteristics of the offset region, an adaptive weighted distance calculation method assigns a unique weight coefficient to each offset region. The weight coefficient is typically calculated based on the importance, shape complexity, and importance of each region in radiotherapy; the more critical the region, the greater its weight. The weight coefficients of each offset region are weighted and calculated to obtain the precise positioning information for each region. This weighted calculation yields the most accurate offset region positioning information. This information guides the subsequent radiotherapy correction process, ensuring accurate treatment of the treatment area.

[0056] In one specific embodiment, obtaining correction guidance parameters based on deviation data includes the following steps:

[0057] S51. Based on the positioning information of the offset area, extract the deviation vector, and use the motion trajectory data collected by the infrared sensor and motion capture system to perform time series analysis on the deviation vector and motion trajectory data to obtain the deviation data.

[0058] S52. Based on the degree of influence of each monitoring block on the overall radiotherapy effect, the importance of the region and its correlation with the patient's physiological movement, a weighted model is constructed. The weighted result of the deviation vector of each monitoring block is calculated based on the weighted model. The weighted result represents the weighted deviation data of each monitoring block at each time point.

[0059] S53. Use optimization algorithms to optimize the weighted results of multiple time points in each monitoring block to obtain optimized deviation data, and generate correction guidance parameters based on the optimized deviation data.

[0060] Specifically, the deviation vector refers to the displacement vector of the offset region relative to its initial or standard position. It represents the direction and magnitude of the displacement of the region in space. Using infrared sensors and motion capture systems (such as optical motion capture, infrared sensors, or other motion tracking devices), motion trajectory data of the patient's body surface at different time points are collected. This motion trajectory data includes the positional information of various points on the body surface. Time series analysis methods can be used to extract the motion patterns of each monitoring block and combine them with the offset region positioning information to obtain deviation data. Time series analysis is performed on the collected motion trajectory data to determine the deviation vector and dynamic changes of the offset region during the motion process. Techniques such as Fourier transform are used to process the time series data, identifying periodic changes and possible high-frequency noise, further extracting effective data reflecting deviation information. This deviation data reflects the degree of deviation of the patient's body surface relative to its initial state at different time periods. Based on the physiological importance of each monitoring block and its impact on the overall radiotherapy effect, the data is further analyzed. A weighted model is constructed to assess the impact of the disease on radiotherapy and its correlation with the patient's physiological movements. For different monitoring blocks (such as the head and neck, chest, and abdomen), weights are assigned based on their impact on radiotherapy and physiological characteristics (such as respiration and range of motion). For example, the head and neck may have higher requirements for radiotherapy accuracy, while the abdomen experiences more dramatic dynamic changes. Therefore, these areas may be assigned different weight values. Based on the weighted model, the deviation vector of each monitoring block is weighted to obtain the weighted deviation data for each monitoring block. Each weighted deviation result corresponds to the dynamic deviation information at each time point. These weighted deviation data comprehensively consider the physiological impact of the monitoring block and its contribution to the overall radiotherapy effect, providing accurate deviation data to guide the subsequent correction process. The weighted deviation data for each monitoring block is usually represented in the form of a matrix or vector, where each element represents the degree of deviation of the monitoring block at a specific time point. For the weighted results of multiple time points for each monitoring block, an optimization algorithm is used for optimization processing. Commonly used optimization algorithms include least squares, genetic algorithms, and particle swarm optimization (PSO). The goal of optimization algorithms is to eliminate redundant information or unnecessary errors by adjusting weighted bias data, and to minimize the deviation between the corrected data and the target position. Optimization algorithms dynamically adjust the weighted results based on the deviation of different monitoring blocks to ensure accurate correction of each monitoring block during radiotherapy. With the help of optimization algorithms, the weighted bias data of each monitoring block is optimized into optimized bias data. This data can accurately reflect the deviation that occurs on the patient's body surface during radiotherapy and provide guidance for subsequent correction operations. Based on the optimized bias data, correction guidance parameters are generated. The correction guidance parameters include the specific adjustment amount for each monitoring block, such as translation, rotation, or deformation, to ensure that the target area is accurately aligned with the radiotherapy plan.

[0061] In one specific embodiment, extracting the local dynamic change features of a local region during the respiratory cycle specifically includes the following steps:

[0062] In one specific embodiment, obtaining the cooperative correction instruction set specifically includes the following steps:

[0063] S61. For rigid monitoring blocks, calculate the three-dimensional translation correction amount and rotation correction angle based on the corresponding deviation data;

[0064] S62. For deformable monitoring blocks, based on the corresponding deviation data and motion pattern data, and based on the biomechanical deformation model, generate a set of driving point displacement instructions through the reverse solution method.

[0065] S63. The translation correction amount and rotation correction angle of the rigid monitoring block are combined with the displacement command set of the driving point of the deformable monitoring block for conflict detection and weighted fusion to obtain the collaborative correction command set.

[0066] Specifically, to ensure accurate calibration of each monitoring block during radiotherapy, for rigid monitoring blocks (such as the head or other less deformable areas), the three-dimensional translational correction and rotational correction angles of the block are first calculated based on deviation data. The three-dimensional translational correction refers to the translational amount along the X, Y, and Z axes calculated by analyzing the deviation vector of the target block. Typically, the translational correction is determined by analyzing the position of the patient's body surface offset in three-dimensional space. The rotational correction is usually calculated based on the angular offset in the deviation data. Matrix transformation methods can be used to calculate the rotational angle of the monitoring block in space (e.g., the rotational angle around the X, Y, and Z axes). When calculating the translational correction, the least squares method or other optimization algorithms are used to obtain the translational amount that minimizes the deviation. The rotational angle can be calculated using geometric matching methods, combining the initial position and current state of the target area, and using angle transformation matrices or quaternion methods to obtain the accurate rotational amount. The calculated three-dimensional translational correction and rotational correction angles serve as calibration instructions for the rigid monitoring blocks and are used for subsequent adjustments of the physical therapy equipment.

[0067] For deformable monitoring areas (such as the abdomen, which are significantly affected by physiological movements like respiration), the analysis is first performed based on relevant deviation data and motion pattern data (such as dynamic changes within the respiratory cycle). Deformable monitoring areas typically exhibit significant geometric changes, so relying solely on deviation vectors for correction may be insufficient. It is necessary to combine motion pattern data to obtain more accurate deformation information. A biomechanical deformation model is then used to model the deformable monitoring area. This model can help predict the deformation pattern of the area by simulating the deformation behavior of biological tissues under external forces. Through a reverse engineering method, a suitable set of driving point displacement instructions is generated based on the motion pattern data and deviation data. These driving point displacement instructions can guide the radiotherapy equipment to dynamically adjust the deformable area to compensate for offsets caused by respiration, tumors, or other physiological movements. The generated driving point displacement instruction set provides position correction instructions for key points of each monitoring area. These instructions can guide the treatment equipment to precisely adjust the patient's body surface area, ensuring that the target area remains in the accurate position throughout the treatment process.

[0068] Conflict detection refers to ensuring that there are no contradictions between the correction commands of rigid and deformable monitoring blocks when simultaneously correcting them. Rigid and deformable blocks may overlap at certain boundaries, or their correction commands may conflict due to physical limitations. Conflict detection ensures that translation or rotation of the rigid block will not affect the displacement of the driving point of the deformable block. Conflict detection can be performed using geometric constraints and optimization algorithms. After conflict detection, the correction commands for the rigid and deformable blocks need to be weighted and fused to integrate them into a final collaborative correction command set. The weighted fusion takes into account… The impact of each block's correction instructions on the overall treatment effect is considered during the fusion process. Different weights are assigned to the correction instructions for rigid and deformable blocks. The weights are usually calculated based on the block's physiological importance, its position in radiotherapy, and the degree of influence of physiological motion. After weighted fusion, the final collaborative correction instruction set will take into account the correction needs of both rigid and deformable monitoring blocks, ensuring accurate positioning on the patient's body surface and improving radiotherapy accuracy. The final collaborative correction instruction set contains correction data for all monitoring blocks, including translational correction and rotational correction angles for rigid monitoring blocks, as well as drive point displacement instructions for deformable monitoring blocks.

[0069] In one specific embodiment, determining whether the correction was successful based on the registration result includes the following steps:

[0070] S71. The registration results include the residual vector and root mean square error of each monitoring block. The real-time dynamic error is obtained based on the residual vector. The feature similarity within the respiratory cycle is calculated based on the real-time dynamic error and dynamic change characteristics.

[0071] S72. If all monitoring blocks simultaneously meet the preset conditions, the preset conditions are that the root mean square error is less than or equal to the preset geometric accuracy threshold and the feature similarity is greater than or equal to the dynamic consistency threshold, the correction is judged to be successful; otherwise, the correction is judged to be unsuccessful.

[0072] Specifically, to ensure precise alignment of the body surface contour during radiotherapy and guarantee treatment effectiveness and patient safety, the residual vector represents the magnitude and direction of deviations in each monitoring block during registration. The root mean square error (RMSE) is the standard for measuring registration accuracy; it represents the average error between the target and the registration result. During radiotherapy, the residual vector represents the offset of the monitoring block on the body surface, typically a vector indicating the difference between the target area and its actual location. RMSE is a standard method for quantifying model accuracy; it is obtained by calculating the mean of the sum of squared errors of all monitoring blocks, providing an indicator of correction accuracy. The smaller the RMSE, the higher the accuracy. The higher the accuracy of the correction, the better the dynamic change characteristics are obtained. These characteristics are mainly obtained by analyzing the motion trajectory of the patient's body surface during the respiratory cycle. These characteristics reflect the dynamic change pattern of the body surface, such as the deformation or displacement of the monitoring block caused by the respiratory cycle or other physiological movements. Combining the residual vector and dynamic change characteristics, the feature similarity within the respiratory cycle is calculated. Feature similarity is measured by comparing the similarity between the dynamic change pattern of the target area at different time points and the standard dynamic pattern. Euclidean distance or cosine similarity methods can be used for calculation. The higher the feature similarity, the better the match between the dynamic change of the patient's body surface and the preset model throughout the radiotherapy process. The better, the better. Feature similarity can be calculated by comparing the registration results with the dynamic response of the preset standard model. During the calibration process, preset conditions are set to determine whether the calibration is successful. The preset conditions include a root mean square error less than or equal to a preset geometric accuracy threshold and a feature similarity greater than or equal to a dynamic consistency threshold. The first condition ensures that the registration error of each monitoring block is within an acceptable range during the calibration process, ensuring that the geometric matching accuracy of the body surface contour is high enough. The second condition ensures that the movement of the monitoring block matches the preset model well during the dynamic changes of the patient's body surface, especially under the influence of respiratory cycle and physiological movement, ensuring dynamic consistency. If all monitoring blocks meet the preset conditions, the calibration is considered successful. Successful calibration means that the radiotherapy target area has been accurately aligned and the dynamic changes of the patient's body surface have been effectively considered, ensuring high-precision execution of the treatment. If any monitoring block does not meet the preset conditions, i.e., the root mean square error exceeds the geometric accuracy threshold or the feature similarity is lower than the dynamic consistency threshold, the calibration is considered to have failed. If the calibration fails, the cause of failure will be analyzed and further analysis will be conducted based on the cause of failure. It may be necessary to adjust the calibration parameters, re-acquire data, or consider other factors (such as changes in patient posture, equipment calibration issues, etc.) until the calibration meets the preset conditions.

[0073] In one specific embodiment, different processes are applied based on different factors, specifically including the following steps:

[0074] If the root mean square error of all monitoring blocks is greater than the preset geometric accuracy threshold or the similarity of all features is less than the dynamic consistency threshold, it is judged as a systematic offset. The geometric accuracy threshold or dynamic consistency threshold is adjusted according to the offset direction. Otherwise, the deviation data of the monitoring blocks that do not meet the preset conditions is extracted. Based on the deviation data, the elastic parameters of the biomechanical deformation model are adjusted. Based on the updated elastic parameters, S3-S4 are re-executed until the correction is judged to be successful.

[0075] Specifically, to ensure accurate surface monitoring and correction during radiotherapy, when correction fails, the root mean square error (RMSE) and feature similarity of all monitoring blocks are first checked. If the RMSE of all monitoring blocks is greater than the preset geometric accuracy threshold, or the feature similarity of all monitoring blocks is less than the dynamic consistency threshold, it indicates that there is a systematic shift during radiotherapy. Systematic shift may be caused by factors such as equipment problems, model errors, and changes in patient posture, resulting in correction deviations of all blocks in similar directions and magnitudes. If a systematic shift is confirmed, the geometric accuracy threshold or dynamic consistency threshold will be adjusted according to the shift direction. The shift direction is usually determined by analyzing the residual vector and dynamic change characteristics. The shift direction indicates the spatial direction and trend of the deviation, which may be due to the positional shift of the radiotherapy equipment, posture errors, or other systematic factors. For example, if the systematic shift is manifested as a uniform geometric change, the geometric accuracy threshold will be relaxed to tolerate a certain degree of error until correction is successful. If the systematic shift occurs during dynamic changes (such as the influence of the respiratory cycle), the dynamic consistency threshold will be adjusted accordingly to allow for greater dynamic changes.

[0076] If a non-systematic shift is detected, monitoring blocks that do not meet the preset conditions are identified. These blocks may fail to meet the preset conditions of root mean square error and feature similarity due to local geometric errors or abnormal physiological movements. Deviation data is extracted from the monitoring blocks that do not meet the conditions. Deviation data usually includes information such as position changes, shape deformation, and dynamic movement to help determine the cause of the correction failure. The extracted deviation data is analyzed to determine whether the error is caused by biomechanical deformation. Biomechanical deformation typically involves localized deformation of the patient's body surface or organs, such as deformation caused by physiological factors like respiration and changes in body posture. If localized deformation is considered the cause of calibration failure, the biomechanical deformation model will be adjusted. The elastic parameters of the biomechanical deformation model usually determine the degree of response of the human body or organ to external forces. By adjusting these parameters, deformation caused by physiological movements (such as respiration) can be simulated and compensated more accurately. The adjusted elastic parameters will help to better match the dynamic changes of the patient's body surface and improve the accuracy of calibration. Using the updated biomechanical deformation model and elastic parameters, steps S3 and S4 will be re-executed, i.e., a new calibration will be performed based on the new deformation model and calibration guidance parameters. During the re-execution of registration and calibration, the new elastic parameters will be considered to ensure that the calibration data of each monitoring block is more consistent with the actual dynamic changes.

[0077] The process continues iteratively until the correction results of all monitoring blocks meet the preset conditions, namely, the root mean square error is less than or equal to the geometric accuracy threshold and the feature similarity is greater than or equal to the dynamic consistency threshold. The correction is then determined to be successful. Through multiple adjustments and optimizations, the accurate correction of the patient's body surface is finally ensured, so as to facilitate high-precision radiotherapy.

[0078] The above describes a method for monitoring and correcting the body surface contour of a radiotherapy patient according to an embodiment of this application. The following describes a system for monitoring and correcting the body surface contour of a radiotherapy patient according to an embodiment of this application. Please refer to [link to relevant documentation]. Figure 4 One embodiment of a radiotherapy patient surface contour monitoring and correction system in this application includes:

[0079] The data acquisition module obtains raw point cloud data from the patient's body surface, generates a high-resolution body surface contour model based on the point cloud data, divides the body surface into multiple monitoring blocks using preset anatomical region division rules, and determines the geometric boundary features of each monitoring block.

[0080] The dynamic monitoring module acquires the corresponding anatomical structural characteristics and motion pattern data for each monitoring block, extracts the local dynamic change characteristics of the local area during the respiratory cycle, and if the local dynamic change characteristics exceed the preset deviation threshold, the offset area positioning method based on geometric matching is used to determine the positioning information of the offset area.

[0081] The correction calculation module obtains the deviation vector and motion trajectory data based on the positioning information of the offset area, constructs a weighted model to calculate the deviation data and obtain correction guidance parameters, constructs a correction model, and inputs the correction guidance parameters, geometric boundary features, anatomical structure characteristics and motion pattern data into the correction model to obtain a collaborative correction instruction set;

[0082] The calibration execution module performs calibration based on the collaborative calibration instruction set. It performs non-rigid registration between the calibrated point cloud data and the high-resolution body surface contour model, obtains the registration result, and determines whether the calibration is successful. If the calibration is unsuccessful, it analyzes the factors that caused the failure and performs different processing based on different factors until the calibration is successful.

[0083] The above description is merely a specific implementation of this specification. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the scope of protection of this specification is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this specification, and these modifications or substitutions should all be covered within the scope of protection of this specification.

Claims

1. A method of monitoring and correcting a patient's body surface contour for radiotherapy, characterized by, The method comprises: S1, obtaining original point cloud data from the patient's body surface, generating a high-resolution body surface contour model based on the point cloud data, dividing the body surface into multiple monitoring blocks using a preset anatomical region division rule, and determining the geometric boundary features of each monitoring block; S2, for each monitoring block, obtain its corresponding anatomical structure characteristics and motion mode data, extract the local dynamic change characteristics of the local area in the respiratory cycle, if the local dynamic change characteristics exceed the preset deviation threshold, use a geometric matching-based offset region positioning method to determine the positioning information of the offset region, wherein the anatomical structure characteristics include the geometric shape, size, and relative position of the block, and the motion mode data refers to the motion trajectory of the monitoring block in the respiratory cycle; S3, according to the positioning information of the offset region, obtain the deviation vector and motion trajectory data, construct a weighted model to calculate the deviation data and obtain the correction guidance parameters, construct a correction model, input the correction guidance parameters, geometric boundary features, anatomical structure characteristics and motion mode data into the correction model to obtain a collaborative correction instruction set, wherein the collaborative correction instruction set contains the correction data of all monitoring blocks, including the translation correction amount, rotation correction angle of rigid monitoring blocks, and driving point displacement instruction of deformable monitoring blocks; for different monitoring blocks, according to their influence degree on radiotherapy and physiological characteristics, the weighted model is constructed; based on the weighted model, the deviation vector of each monitoring block is weighted calculated, the deviation data is obtained, and the correction guidance parameters are generated according to the deviation data, the correction guidance parameters include the translation adjustment amount, rotation adjustment amount and deformation adjustment amount of each monitoring block; geometric transformation calculation is performed on rigid monitoring blocks, and inverse solution based on biomechanical deformation model is performed on deformable monitoring blocks, and a correction model is constructed; S4, based on the collaborative correction instruction set, the corrected point cloud data and the high-resolution body surface contour model are registered, the registration result is obtained, and it is judged whether the correction is successful, if not, analyze the factors of unsuccessful correction, and process different factors based on different factors until the correction is successful.

2. The method of radiotherapy patient body surface contour monitoring and correction of claim 1, wherein, Generating a high-resolution body surface contour model based on point cloud data comprises: S11, aligning the initial point cloud data using a point cloud registration algorithm to obtain standard point cloud data in a unified coordinate system, and denoising the standard point cloud data using a mean filtering algorithm to obtain smooth point cloud data; S12, based on the high-resolution image data collected by the stereomicroscope, generate texture mapping data, fuse the texture mapping data with the smooth point cloud data, and obtain a high-resolution body surface contour model containing texture information; S13, if the high-resolution body surface contour model has a local missing area, use an interpolation algorithm to fill the missing area to obtain a complete body surface contour model.

3. The method of radiotherapy patient body surface contour monitoring and correction of claim 1, wherein, Determining the geometric boundary features of each block comprises: S21, determine the initial boundary range corresponding to each monitoring block based on the spatial distribution data of each monitoring block, if the initial boundary range overlaps, use the grid segmentation algorithm to divide the space of the overlapping area, and obtain the boundary data of the non-overlapping monitoring block; S22, according to the boundary data, extract the geometric boundary features of each monitoring block, and generate the spatial geometric description of each monitoring block, the geometric boundary features include the edge line, surface or spatial distribution characteristics of the monitoring block, and the spatial geometric description refers to using mathematical model to express the boundary of the monitoring block; S23, through the spatial geometric description, the curved surface fitting algorithm is used to smooth the boundary of each monitoring block, and the optimized boundary curved surface data is obtained, and the interpolation algorithm is used to repair the discontinuous area of the boundary curved surface data, and the continuous boundary curved surface model is obtained; S24, according to the boundary curved surface model, generate three-dimensional visualization data of each monitoring block, and determine the final block geometric feature.

4. The method of radiotherapy patient body surface contour monitoring and correction of claim 1, wherein, Extract the local dynamic change characteristics of the local area in the respiratory cycle, including: S31, use a stereomicroscope and a motion capture system to obtain the corresponding anatomical structure characteristics and motion mode data; S32, through time series analysis method, the collected motion mode data is preprocessed, noise is removed and standardized, and fourier transform is used to analyze the frequency domain of the standardized motion mode data, and the frequency characteristics of the block in the respiratory cycle are extracted; S33, using low-pass filter to smooth the high-frequency noise in the frequency domain characteristics, and obtaining the smooth time series data, using long short-term memory network model to extract the local dynamic change characteristics under the influence of respiratory motion from the time series data, and obtaining the dynamic characteristic vector; S34, according to the dynamic characteristic vector, using principal component analysis method to reduce the high-dimensional feature, obtaining the low-dimensional dynamic change characteristic, if the variance contribution rate of the low-dimensional dynamic change characteristic is lower than the preset threshold, adjusting the parameter of principal component analysis and recalculating, obtaining the optimized local dynamic change characteristic.

5. The method of radiotherapy patient body surface contour monitoring and correction of claim 1, wherein, Determine the positioning information of the offset area, including: S41, obtain the preliminary offset range of the target monitoring block, according to the geometric characteristics of the offset range, preliminarily locate the offset area as a curved surface fitting model, and generate the preliminary offset area positioning data; S42, using shape matching algorithm based on curved surface fitting, compare the geometric characteristics of the target monitoring block with the preset standard curved surface, judge the position of the offset area by calculating the curved surface similarity measure, and obtain the preliminary positioning result; S43, combined with the preliminary positioning result, using the space registration method based on geometric projection, the offset areas at different angles and positions are calibrated, and the projection alignment algorithm is used to optimize the positioning error; S44, according to the geometric characteristics of the offset area, using adaptive weighted distance calculation method, assigning a unique weight coefficient to each offset area, and obtaining accurate positioning information through weighted calculation.

6. The method of radiotherapy patient body surface contour monitoring and correction of claim 1, wherein, According to the positioning information of the offset area, obtain the deviation vector and the motion trajectory data, construct the weighted model to calculate the deviation data and obtain the correction guidance parameter, including: S51, extract a deviation vector according to the positioning information of the offset region, and perform time series analysis on the deviation vector and motion trajectory data collected by the infrared sensor and the motion capture system to obtain deviation data; S52, construct a weighted model according to the influence degree of each monitoring block on the overall radiotherapy effect, the importance of the region, and the relevance to the physiological movement of the patient, calculate the weighted results of the deviation vector of each monitoring block based on the weighted model, and the weighted results represent the weighted deviation data of each monitoring block at each time point; S53, use an optimization algorithm to optimize the weighted results of each monitoring block at multiple time points to obtain optimized deviation data, and generate correction guidance parameters based on the optimized deviation data.

7. The method of radiotherapy patient body surface contour monitoring and correction of claim 1, wherein, The cooperative correction instruction set is obtained, including: S61, for the rigid monitoring block, calculate the three-dimensional translation correction amount and the rotation correction angle based on the corresponding deviation data; S62, for the deformable monitoring block, based on the corresponding deviation data and the motion mode data, based on the biomechanical deformation model, generate a driving point displacement instruction set through an inverse solving method; S63, perform conflict detection and weighted fusion on the translation correction amount and the rotation correction angle of the rigid monitoring block and the driving point displacement instruction set of the deformable monitoring block to obtain the cooperative correction instruction set.

8. The method of radiotherapy patient body surface contour monitoring and correction of claim 1, wherein, Determine whether the correction is successful based on the registration result, including: S71, the registration result includes residual vectors and root mean square errors of each monitoring block, and the feature similarity in the respiratory cycle is calculated based on the residual vectors and dynamic change characteristics; S72, if all monitoring blocks simultaneously meet the preset conditions, the preset conditions are that the root mean square error is less than or equal to the preset geometric precision threshold and the feature similarity is greater than or equal to the dynamic consistency threshold, it is judged that the correction is successful, otherwise it is judged that the correction fails.

9. The method of monitoring and correcting patient body surface contours for radiotherapy treatment of claim 1, wherein, Analyze the factors that the correction is unsuccessful, and perform different processing based on different factors, including: If the root mean square errors of all monitoring blocks are all greater than the preset geometric precision threshold or all the feature similarities are less than the dynamic consistency threshold, it is judged that it is a systematic offset, the geometric precision threshold or the dynamic consistency threshold is adjusted according to the offset direction, otherwise, the deviation data of the monitoring block that does not meet the preset condition is extracted, the elastic parameter of the biomechanical deformation model is adjusted based on the deviation data, and S3-S4 is re-executed based on the updated elastic parameter until it is judged that the correction is successful.

10. A radiotherapy patient body surface contour monitoring and correction system for implementing a radiotherapy patient body surface contour monitoring and correction method as claimed in any one of claims 1-9, characterized by, The system includes: A data acquisition module acquires original point cloud data from the patient's body surface, generates a high-resolution body surface contour model based on the point cloud data, divides the body surface into multiple monitoring blocks using a preset anatomical region division rule, and determines the geometric boundary features of each monitoring block; A dynamic monitoring module acquires the corresponding anatomical structure characteristics and motion mode data for each monitoring block, extracts the local dynamic change characteristics of the local region within the respiratory cycle, and if the local dynamic change characteristics exceed the preset deviation threshold, uses a geometric matching-based offset region positioning method to determine the positioning information of the offset region. The correction calculation module obtains a deviation vector and motion trajectory data according to the positioning information of the offset region, constructs a weighting model to calculate deviation data and obtain correction guidance parameters, constructs a correction model, inputs the correction guidance parameters, geometric boundary features, anatomical structure characteristics and motion mode data into the correction model to obtain a cooperative correction instruction set; The correction execution module performs correction based on the cooperative correction instruction set, performs non-rigid registration on the corrected point cloud data and the high-resolution body surface contour model to obtain a registration result and determine whether the correction is successful, analyzes factors for unsuccessful correction in the case of unsuccessful correction, performs different processing based on different factors until the correction is successful.

Images