An elastic deformation modeling system for organ displacement in cyberknife treatment

Abstract

This invention discloses an elastic deformation modeling system for organ displacement during CyberKnife treatment, relating to the field of radiotherapy technology. It includes a data acquisition module for collecting static anatomical data, dynamic displacement data, and stiffness distribution data of the target organ; a dynamic registration module for constructing a heterogeneous elastic deformation field of the target organ through deformation modeling and a registration engine; a tracking and prediction module for acquiring real-time displacement of the target organ, predicting deformation trends, and obtaining predicted deformation data; a treatment planning integration module for acquiring 4D dose distribution data, calculating conformity and homogeneity indices, and evaluating and optimizing the dose; and a feedback control module for outputting motion compensation commands to the CyberKnife robotic arm, performing beam correction when the displacement exceeds a preset threshold, and adjusting the robotic arm's posture. This invention solves the problem of radiation deviation from the target area during CyberKnife treatment, improving the safety and therapeutic effect of radiotherapy.

Description

Technical Field

[0001] This invention relates to the field of radiotherapy technology, and more specifically, to an elastic deformation modeling system for organ displacement during CyberKnife treatment. Background Technology

[0002] CyberKnife, a high-precision stereotactic radiotherapy device, is widely used in tumor treatment due to its sub-millimeter-level treatment accuracy. However, during treatment, the patient's physiological movements such as breathing and heartbeat can easily cause displacement and elastic deformation of the organs in the target area. If this deformation process cannot be accurately depicted, the radiation will deviate from the target area, reducing the treatment effect and potentially increasing radiation damage to normal tissues. Organ displacement refers to the change in the position of an organ during treatment, mainly caused by physiological movements such as breathing and heartbeat. For example, when a lung cancer patient breathes, the lung tumor will move up and down with the lung tissue. Elastic deformation is the reversible shape change that occurs when an organ is subjected to respiratory pressure and traction from surrounding tissues. For example, the liver is compressed and stretched during breathing and can return to its original shape after treatment. The modeling system constructs a digital model simulating organ displacement and elastic deformation through mathematical formulas, image data, and mechanical principles. It is equivalent to "drawing a blueprint" for organ movement, allowing the system to predict its position and shape changes.

[0003] Shortcomings of existing technology: Elastic deformation modeling often employs the assumption of a homogeneous elastic body, constructing the deformation field through the Navier equation. However, it neglects the stiffness differences within organ tissues (such as blood vessels, lesions, and normal parenchyma), leading to insufficient accuracy in predicting the deformation of heterogeneous organs (such as the liver and lungs). Furthermore, traditional modeling methods rely on finite element analysis, resulting in high computational complexity and difficulty in meeting the millisecond-level response requirements of real-time intraoperative tracking. Simultaneously, their adaptive prediction capability for sudden changes in respiratory patterns (such as coughing and irregular breathing) is weak, easily leading to decreased tracking stability.

[0004] To address the above problems, this invention proposes a solution. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an elastic deformation modeling system for organ displacement during CyberKnife treatment. By predicting position and shape changes through the elastic deformation model, the system drives a robotic arm to automatically correct the beam direction, thereby solving the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: An elastic deformation modeling system for organ displacement during CyberKnife treatment includes a data acquisition module, a dynamic registration module, a tracking and prediction module, and a treatment planning integration module, with connections between the modules. The data acquisition module is used to collect static anatomical data, dynamic displacement data, and stiffness distribution data of the patient's target organ. The dynamic registration module constructs a heterogeneous elastic deformation field of the target organ based on the collected data through deformation modeling and registration engine to achieve dynamic registration; The tracking and prediction module obtains the real-time displacement of the target organ based on dynamic displacement data, and predicts the deformation trend by combining the respiratory-target area motion correlation model and the reduced-order model to obtain predicted deformation data. The treatment planning integration module acquires 4D dose distribution data based on a heterogeneous elastic deformation field, calculates conformity and homogeneity index based on the 4D dose distribution data, and evaluates the dose under deformation constraints in conjunction with preset organ-at-risk doses. If the evaluation results do not meet the requirements, the dose distribution is optimized. The feedback control module outputs sub-millimeter-level motion compensation commands to the SARK robotic arm based on the predicted deformation field data. When the displacement exceeds the preset threshold, beam correction is performed, and the PID control algorithm is used to adjust the robotic arm's posture.

[0007] In a preferred embodiment, the process of acquiring static anatomical data, dynamic displacement data, and stiffness distribution data of the patient's target organ is as follows: A full-range scan of the organs in the target area of ​​the patient was performed to obtain static anatomical images at the end of inspiration, end of expiration, and key time phases, generating a three-dimensional anatomical structure dataset to obtain static anatomical data. Stiffness scanning is performed on the target organ to obtain the original spatial distribution data of the elastic modulus, thus obtaining stiffness distribution data; Infrared markers were affixed to the patient's body surface in the area corresponding to respiratory movements. The infrared marker device and orthogonal X-ray imaging equipment were activated simultaneously. Data were continuously collected for the entire respiratory cycle while the patient was breathing naturally. The three-dimensional coordinate sequence of the infrared markers and the 6D pose data of the gold markers inside the target area were obtained at each respiratory phase. At the same time, dynamic image sequences were collected during the respiratory cycle to obtain data on the position and morphological changes of organs at different respiratory phases, thus obtaining dynamic displacement data.

[0008] In a preferred embodiment, the dynamic registration module is implemented as follows: Using images from the end-expiratory phase of static anatomical data as reference images, the target area and the outline of organs at risk are delineated on the reference images, and a reference coordinate system is established. Rigidly register dynamic images of other respiratory phases with reference images to eliminate global bias caused by overall patient positional movement; The target organ is regarded as a heterogeneous isotropic elastic body, and a set of heterogeneous elastic mechanical control equations is established based on stiffness distribution data. The displacement vector is obtained by using finite element discretization and numerical solution based on the equations, thereby generating a non-homogeneous elastic deformation field.

[0009] In a preferred embodiment, the displacement vector is obtained by finite element discretization and numerical solution based on the equation set, and the non-homogeneous elastic deformation field is generated as follows: The three-dimensional model of the target organ is divided into tiny tetrahedral elements, and the mechanical control equations are transformed into a discrete linear equation system that can be solved by a computer. The organ surface is set as a free boundary, and the actual displacement data of the collected gold standard is substituted as a constraint condition. The conjugate gradient method is used to solve the discretized linear equations to obtain the displacement vector components of each tetrahedral element. By integrating the displacement components of all units, the displacement vector of the entire organ at any time and any position within the entire respiratory cycle is obtained.

[0010] In a preferred embodiment, the process for acquiring predicted deformation data is as follows: Load the pre-trained respiratory-target motion correlation model, input the current body surface infrared signal into the model, output the initial value of target displacement prediction, fuse the initial value of target displacement prediction with real-time displacement data, correct the model parameters, and improve the correlation mapping accuracy. By calling a reduced-order model based on intrinsic orthogonal decomposition, the heterogeneous elastic deformation field is projected onto a low-dimensional intrinsic mode space, reducing computational complexity. The displacement data of the current and historical N frames are input into the reduced-order model to solve the temporal variation law of the modal coefficients. Based on the temporal variation law of the modal coefficients, the modal coefficients at the future time τ are predicted. The inhomogeneous elastic deformation field at the future time is reconstructed by inverse mapping of the reduced-order model to obtain the predicted deformation data.

[0011] In a preferred embodiment, the initial value of the target area displacement prediction is fused with the real-time displacement data to correct the model parameters. The process is as follows: Set the fusion weight coefficients for the initial predicted target displacement values ​​output by the model, and use a weighted average algorithm to calculate the fused target displacement values. The fused displacement value is correlated with the body surface infrared signal input to the model to construct a body surface signal-target area displacement sample pair at the current moment; The least squares iterative algorithm is used to correct the mapping parameters inside the model by minimizing the error between the displacement value output by the model based on the current body surface signal and the fused displacement value.

[0012] In a preferred embodiment, the process of acquiring 4D dose distribution data is as follows: Load the baseline treatment plan established before surgery, including beam parameters, target area and organ at risk contours, and initial dose distribution; Import heterogeneous elastic deformation field and predicted deformation data to establish a dynamic mapping relationship between deformation field and dose distribution; Based on the displacement vector in the deformation field, the dose distribution of the baseline plan is mapped from the reference phase to each phase of the entire respiratory cycle through the deformation vector field, so as to obtain the dose distribution of each phase. By traversing the entire respiratory cycle, the 4D dose distribution data is calculated.

[0013] In a preferred embodiment, if the evaluation results do not meet the requirements, the dose distribution is optimized as follows: Define the range of beam parameters that need to be adjusted, including beam incident angle and beam weight; The real-time displacement data and predicted deformation field data output by the tracking and prediction module are correlated with the dose distribution corresponding to each beam parameter. The influence of beam parameter changes on target area dose coverage and radiation dose to organs at risk under different respiratory phases is analyzed, and a mapping relationship between beam parameters, deformation state and dose distribution is established. Based on the mapping relationship, the beam parameters corresponding to the areas with insufficient or excessive dose coverage are initially adjusted. The adjusted beam parameters are then substituted into the 4D dose calculation model to recalculate the dose distribution throughout the respiratory cycle and then re-evaluate. If the reassessed indicators still fail to meet the target, record the deviation between the current dose distribution and the target, further fine-tune the beam parameters based on the direction of the deviation, and repeat the assessment and calculation process. When the requirements are met for three consecutive evaluations or the parameter adjustments have reached the preset maximum number of iterations, the optimization is terminated and the final beam parameter combination is confirmed. Dosage optimization is completed once all evaluation indicators meet the standards.

[0014] In a preferred embodiment, the beam correction step when the displacement exceeds a preset threshold is as follows: The 6D pose deviation between the target center at the current and predicted time and the target center in the optimized treatment plan is calculated based on the predicted deformation data and real-time displacement data. If the deviation does not exceed the threshold, maintain the current robotic arm posture and beam parameters; If the current or predicted pose deviation exceeds the preset threshold, a beam interruption command is immediately triggered to pause beam irradiation; at the same time, a correction process is initiated, inputting real-time displacement data, predicted deformation field data, and target area position information of the optimized treatment plan into the robotic arm control unit.

[0015] In a preferred embodiment, the process of adjusting the robotic arm's posture using a PID control algorithm is as follows: First, integrate the previously calculated 6D pose deviations into a unified deviation variable, and then load the preset PID control parameters, including proportional coefficient, integral coefficient and derivative coefficient, and set the initial cumulative value of the integral term and the initial difference value of the derivative term to 0. The three core components—proportional, integral, and differential—are calculated separately. The calculated proportional, integral, and differential components are then added together to obtain the required angle change compensation for each joint of the robotic arm. The synthesized joint angle compensation is converted into electrical signal control commands that can be recognized by the robotic arm drive unit. These commands are then sent to the drive motors of each joint of the robotic arm via the system bus. The drive motors rotate the joints according to the commands, thus completing the initial correction of the robotic arm's position and posture.

[0016] The technical effects and advantages of the elastic deformation modeling system for organ displacement during CyberKnife treatment of the present invention are as follows: 1. This invention constructs a heterogeneous elastic deformation field of the target organ by fusing multimodal data, breaking through the limitations of traditional homogeneous elastic body modeling. Combined with individualized mechanical parameter calibration and precise displacement vector solution, it can accurately characterize the elastic deformation and displacement of the organ caused by physiological movements such as breathing, significantly reducing deformation prediction errors. This provides a reliable deformation basis for subsequent dose calculation and robotic arm compensation, effectively avoiding the problem of radiation deviating from the target area during CyberKnife treatment and improving the accuracy of target area irradiation.

[0017] 2. This invention achieves real-time target displacement capture and accurate prediction of deformation trends through a tracking and prediction module. Combined with the 4D dose optimization of the treatment planning integration module and the PID closed-loop control of the feedback control module, a complete treatment chain of "modeling-prediction-optimization-control" is formed. This can reduce the target area's external boundary, lower the radiation dose to organs at risk, reduce the side effects of radiotherapy, and adapt to patients who cannot hold their breath or have irregular breathing, expanding the indications for CyberKnife treatment. At the same time, it ensures the real-time performance and stability of the entire treatment process, improving the safety and therapeutic effect of radiotherapy. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the elastic deformation modeling system for organ displacement during CyberKnife treatment according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1, Figure 1 This invention presents an elastic deformation modeling system for organ displacement during CyberKnife treatment.

[0021] The data acquisition module is used to collect static anatomical data, dynamic displacement data, and stiffness distribution data of the patient's target organs; providing data support for the subsequent construction of a heterogeneous elastic deformation field. The specific acquisition process is as follows: The 4DCT / MRI equipment is activated to perform a full-range scan of the organs where the patient's target area is located, and static anatomical images of key phases such as end of inspiration and end of expiration are obtained to generate a three-dimensional anatomical structure dataset and obtain static anatomical data. Simultaneously, the MRE device is activated to perform a stiffness scan on the target organ, obtain the original spatial distribution data of the elastic modulus μ(x,y,z), and obtain stiffness distribution data. This data is used to characterize the stiffness differences of different tissues (blood vessels, lesions, normal parenchyma) within the organ. Six to eight infrared markers are affixed to the patient's body surface in areas associated with respiratory movements (such as the abdomen and chest) to ensure a strong correlation between the markers and the target area movement. The infrared markers and orthogonal X-ray imaging equipment are activated simultaneously to continuously collect data for 10 to 15 complete respiratory cycles while the patient is breathing naturally. The three-dimensional coordinate sequence of the infrared markers and the 6D pose data (3D linear displacement + 3D rotation) of the gold markers (preoperatively implanted, 3-5) inside the target area are obtained at each respiratory phase. At the same time, a 4D CT / MRI device is activated to collect dynamic image sequences within the respiratory cycle (0-5s) at a sampling frequency of 10Hz to obtain data on the position and morphological changes of organs at different respiratory phases, thus obtaining dynamic displacement data.

[0022] The acquired raw data were preprocessed, and a median filtering algorithm (3×3 window size) was used to remove noise from the MRE elastic modulus data. A grayscale-based image registration method was used to perform inter-frame alignment of the 4DCT / MRI dynamic image sequences to eliminate errors caused by slight patient positional movements during scanning. Outlier removal was performed on the surface infrared marker data and gold standard 6D pose data. The 3σ criterion was used to identify and delete data points exceeding the normal fluctuation range, and then linear interpolation was used to complete the missing data. Finally, the spatial consistency of each modality data was verified. The maximum mutual information method was used to register the MRE data, 4DCT / MRI data, and orthogonal X-ray image data to the same coordinate system (world coordinate system), with a registration error ≤0.3mm.

[0023] The dynamic registration module, based on the acquired data, constructs a heterogeneous elastic deformation field of the target organ through deformation modeling and a registration engine, achieving accurate registration between dynamic images and reference images. This provides deformation basis for dose calculation and robotic arm compensation. The process is as follows: Using 4DCT / MRI images of the end-expiratory phase in static anatomical data as reference images, the target area and organ at risk (OAR) contours are delineated on the reference images to establish a reference coordinate system; dynamic images of other respiratory phases are rigidly registered with the reference images to eliminate global deviations caused by overall patient positional movement. The target organ is treated as a heterogeneous isotropic elastic body. Based on stiffness distribution data, a set of non-homogeneous elasticity governing equations is established. This set of equations is based on the fundamental principles of elasticity and is constructed around the logical chain of "force equilibrium - stress-strain relationship - strain-displacement relationship". The construction process is as follows: Establishing force balance equations: When an organ undergoes elastic deformation, all points inside must satisfy mechanical equilibrium. Therefore, the Navier equilibrium equations are used to describe the "spatial variation of stress and the equilibrium relationship of external forces" as the core constraint of the equation set. Establishing the stress-strain relationship: In order to correlate the "stress" in the equilibrium equation with the deformation, the generalized Hooke's law is introduced to establish a quantitative relationship between the stress tensor, the strain tensor, and the material mechanical parameters (elastic modulus, Lamé constant), so as to realize the transformation from "force" to "degree of deformation". Establishing the strain-displacement relationship: Strain is an intermediate parameter describing the degree of deformation. It needs to be further related to the final "displacement" to be determined. Therefore, the strain tensor is transformed into the differential form of the displacement vector through geometric equations (strain-displacement relationship), and finally a complete closed system of equations is formed. The displacement parameters can be obtained by solving the equations.

[0024] The displacement vector u(x,y,z,t) is obtained by using the equation system through "finite element discretization + numerical solution", and then a non-homogeneous elastic deformation field is generated. The specific process is as follows: The three-dimensional model of the target organ is divided into tiny tetrahedral elements, and the mechanical control equations are transformed into a discrete linear equation system that can be solved by a computer. The organ surface is set as a free boundary (without external constraints), and the actual displacement data of the collected gold standard is substituted as a constraint condition (to ensure that the gold standard displacement calculated by the model is consistent with the measured data and improve the solution accuracy). The conjugate gradient method is used to solve the discretized linear equations to obtain the displacement vector components of each tetrahedral element. By integrating the displacement components of all units, the displacement vector u(x,y,z,t) at any time t and any position (x,y,z) within the entire organ and the entire respiratory cycle is obtained, laying the foundation for the subsequent construction of a non-homogeneous elastic deformation field.

[0025] The elastic modulus μ(x,y,z) (spatial distribution data, characterizing the stiffness of tissues at different locations) collected by the MRE device was extracted from the preprocessed data; the initial value of Poisson's ratio ν(x,y,z) was set (selected according to organ type: liver 0.35-0.45, lung 0.30-0.40), and subsequently optimized through calibration.

[0026] The tracking and prediction module obtains the real-time displacement of the target organ based on dynamic displacement data, and predicts the deformation trend by combining the respiratory-target area motion correlation model and the reduced-order model to obtain predicted deformation data. The infrared marker device on the body surface and the orthogonal X-ray imaging equipment are activated to synchronously acquire respiratory motion signals (three-dimensional coordinate sequence of infrared marker points) and dynamic images of the target area (6D pose data of gold standard) at a sampling frequency of 30Hz. The system clock synchronization mechanism ensures that the timestamps of the two types of data are consistent, and the data transmission delay is controlled within 10ms. The acquired orthogonal X-ray images are identified and located using gold standard identification. The 6D pose of the target area at the current moment is calculated using the image registration algorithm (3D linear displacement: Δx, Δy, Δz; 3D rotation: Δα, Δβ, Δγ) to obtain the real-time displacement data of the target organ. At the same time, the infrared signal on the body surface is filtered and preprocessed (using Kalman filtering algorithm) to eliminate motion noise interference. Load the preoperatively trained respiratory-target motion correlation model (this model is trained based on preoperative 4DCT dynamic sequences and surface infrared signals to establish a mapping relationship between surface motion and target area displacement); input the current surface infrared signal into the model, output the initial predicted value of target area displacement, fuse it with real-time displacement data, correct the model parameters, and improve the correlation mapping accuracy; the process of fusing with real-time displacement data and correcting model parameters is as follows: First, the fusion weight coefficients for the initial predicted target displacement output by the model are set. This weight allocation is determined based on the accuracy characteristics of the two types of data—orthogonal X-ray gold standard localization directly reflects the true position of the target area and has higher accuracy, so it is given a higher weight. Second, a weighted average algorithm is used to calculate the fused target displacement value. Next, the fused displacement value is correlated with the body surface infrared signal input to the model to construct a "body surface signal-target displacement" sample pair at the current moment. Then, a least squares iterative algorithm is used to correct the mapping parameters inside the model (including body surface signal feature extraction weights, displacement mapping coefficients, etc.) with the goal of "minimizing the error between the displacement value output by the model based on the current body surface signal and the fused displacement value". Finally, after 3-5 iterations, the model parameters are corrected in real time to ensure that the model can adapt to the subtle changes in the patient's breathing pattern during surgery and further improve the correlation mapping accuracy between body surface movement and target area displacement.

[0027] A reduced-order model based on intrinsic orthogonal decomposition (POD) is invoked to project the heterogeneous elastic deformation field onto a low-dimensional intrinsic mode space (extracting the first 50 intrinsic modes with a cumulative energy percentage ≥ 95%), thereby reducing computational complexity; and displacement data from the current and historical N frames (N=10, corresponding to 0.33s) are input into the reduced-order model to solve for the temporal variation of the modal coefficients. Based on the temporal variation law of modal coefficients, the modal coefficients at the future time τ (τ=3 frames, corresponding to 0.1s) are predicted. The non-homogeneous elastic deformation field at the future time is reconstructed by inverse mapping of the reduced-order model to obtain the predicted deformation data.

[0028] The treatment planning integration module acquires 4D dose distribution data based on a heterogeneous elastic deformation field, calculates conformity and homogeneity index based on the 4D dose distribution data, and evaluates the dose under deformation constraints in combination with preset organ-at-risk doses. If the evaluation results do not meet the requirements, the dose distribution is optimized by combining real-time displacement data and predicted deformation field data. Load the pre-established baseline treatment plan (including beam parameters, target area and organ at risk contours, and initial dose distribution); import the heterogeneous elastic deformation field and predicted deformation data to establish a dynamic mapping relationship between the deformation field and the dose distribution; Based on the displacement vector u(x,y,z,t) in the deformation field, the dose distribution of the baseline plan is mapped from the reference phase (end-expiration) to each phase of the entire respiratory cycle through the deformation vector field (DVF) to obtain the dose distribution of each phase. By traversing the entire respiratory cycle, the 4D dose distribution calculation is completed. The dose conformity index (CI) and homogeneity index (HI) of the target area during the entire respiratory cycle are calculated as follows: First, target contour data and corresponding 4D dose distribution data for each phase of the entire respiratory cycle are extracted to determine the prescribed dose Dpres (e.g., 60 Gy). For the dose conformity index (CI), the volume of the target area covered by the prescribed dose (i.e., the volume within the target area where the dose is ≥ Dpres, denoted as Vt,Dpres) is calculated first, and then the total volume covered by the prescribed dose (i.e., the volume within the entire imaging range where the dose is ≥ Dpres, denoted as Vall,Dpres) is calculated. The dose conformity index (CI) is obtained by dividing the volume of the target area covered by the prescribed dose by the total volume covered by the prescribed dose. The closer the index is to 1, the more accurate the coverage of the target area by the prescribed dose, and the less it affects the surrounding normal tissues. For the homogeneity index (HI), the highest dose Dmax within the target area is obtained first, and then divided by the prescribed dose Dpres to obtain the homogeneity index (HI). The closer the index is to 1, the more uniform the dose distribution within the target area, avoiding excessively high or low local doses. The evaluation criteria are: CI ≥ 0.9 (the degree of matching between the actual irradiated volume of the target area and the prescribed dose volume) and HI ≤ 1.1 (the ratio of the highest dose in the target area to the prescribed dose); at the same time, the actual irradiated dose of the organ at risk (OAR) is calculated and compared with the preset dose limit thresholds (such as the maximum dose of the spinal cord ≤ 45 Gy and the lung V20 ≤ 30%). If the assessment results do not meet the requirements (CI < 0.9, HI > 1.1, or the dose to organs at risk exceeds the limit), the dose optimization process is initiated. Using "target area CI ≥ 0.9, HI ≤ 1.1 throughout the respiratory cycle" as the objective function and "the dose to organs at risk does not exceed the preset threshold" as the constraint, parameters such as beam angle and weights are adjusted based on real-time / predicted deformation data. The dose distribution is optimized through iterative calculations until all assessment indicators meet the standards. The process is as follows: First, parameter initialization and constraint setting: The range of beam parameters to be adjusted is defined, including the beam incident angle (adjustment step size of 0.5°), beam weights, etc. The first step involves adjusting the initial weight (range 0-1.2 times) and converting the "dose on organs at risk does not exceed a preset threshold" into a quantitative constraint (e.g., maximum spinal cord dose ≤ 45 Gy, lung V20 ≤ 30%). The core optimization objective is determined as "target area CI ≥ 0.9, HI ≤ 1.1 throughout the respiratory cycle". The second step involves deformation data correlation and dose impact analysis. This involves correlating the real-time displacement data and predicted deformation field data output from the tracking and prediction modules with the dose distribution corresponding to each beam parameter. The analysis examines the impact of beam parameter changes at different respiratory phases on target area dose coverage and the radiation dose to organs at risk. The first step involves establishing a mapping relationship between "beam parameters - deformation state - dose distribution". The second step involves initial parameter adjustment, based on the above mapping relationship, prioritizing the initial adjustment of beam parameters corresponding to areas with insufficient or excessive dose coverage. For example, for areas with low dose at the edge of the target area, the weight of the beam at the corresponding incident angle is appropriately increased. For beams near organs at risk, the angle is finely adjusted to avoid highly sensitive areas. The third step involves iterative calculation and dose assessment, substituting the adjusted beam parameters into the 4D dose calculation model to recalculate the dose distribution throughout the respiratory cycle, and then assessing CI, HI, and the dose to organs at risk according to the above standards. The fifth step is convergence judgment and parameter fine-tuning. If the re-evaluated indicators still do not meet the standards, record the deviation between the current dose distribution and the target, and further fine-tune the beam parameters based on the direction of the deviation. Repeat the calculation and evaluation process of the third and fourth steps. The sixth step is optimization termination and result confirmation. When the CI is ≥0.9 and HI is ≤1.1 for three consecutive evaluations in the iterative calculation and the dose to the organs at risk is not exceeded, or the parameter adjustment has reached the preset maximum number of iterations (usually 20 times) and the indicators are closest to the target value, the optimization is terminated and the final beam parameter combination is confirmed. The dose optimization is completed until all evaluation indicators meet the standards.

[0029] The optimized 4D treatment plan (including dose distribution and beam parameters at each time phase) is output to the feedback control module as the dose basis for robotic arm compensation and beam control.

[0030] The feedback control module outputs sub-millimeter-level motion compensation commands to the CyberKnife robotic arm based on the predicted deformation data. When the displacement exceeds the preset threshold, a beam correction process is performed, and a PID control algorithm is used to adjust the robotic arm's posture.

[0031] The 6D pose deviation between the target center at the current and predicted time and the target center in the optimized treatment plan is calculated based on the predicted deformation data and real-time displacement data. The preset thresholds are: linear displacement deviation > 1.5mm and rotational deviation > 1.5°. If the deviation does not exceed the threshold, the current robotic arm posture and beam parameters are maintained. If the current or predicted pose deviation exceeds the preset threshold, the beam interruption command is immediately triggered to pause beam irradiation; at the same time, the correction process is started, and the real-time displacement data, predicted deformation field data and target area position information of the optimized treatment plan are input into the robotic arm control unit. The required compensation amount (joint angle change Δθ) for the robotic arm is calculated based on the pose deviation. A PID control algorithm is then used to adjust the robotic arm's posture. The process is as follows: First, the previously calculated 6D pose deviations (including linear displacement deviations in the x, y, and z directions and rotational deviations in the α, β, and γ directions) are integrated into a unified deviation variable e(t) for easier subsequent calculations. Simultaneously, preset PID control parameters, including proportional, integral, and derivative coefficients, are loaded, and the initial cumulative value of the integral term and the initial difference value of the derivative term are set to 0 to prepare for subsequent calculations. The three core components—proportional, integral, and derivative—are calculated separately: The proportional term is obtained by multiplying the proportional coefficient by the current deviation e(t). Its function is to quickly respond to the current deviation and directly output the basic compensation amount, allowing the robotic arm to initially adjust its posture to reduce the deviation. The integral term is obtained by multiplying the integral coefficient by the cumulative integral value of the deviation e(t) from the start of treatment to the current moment. It is mainly used to eliminate long-term static deviations, avoid residual deviations caused by system inertia and other factors, and improve the stability of posture adjustment. The accuracy of the state is as follows: The differential term is obtained by multiplying the differential coefficient by the ratio of the difference between the current deviation e(t) and the deviation e(t-1) at the previous moment. This is used to predict the trend of deviation change, suppress deviation expansion in advance, and avoid overshoot caused by excessive adjustment of the robot arm posture. The calculated proportional term, integral term, and differential term are added together to obtain the angle change compensation amount Δθ(t) required for each joint of the robot arm, which clarifies the specific angle that each joint needs to be adjusted. The synthesized joint angle compensation amount Δθ(t) is converted into an electrical signal control command that can be recognized by the robot arm drive unit. It is sent to the drive motor of each joint of the robot arm through the system bus. The drive motor drives the joint to rotate according to the command, completing the initial correction of the robot arm position and posture. During the robot arm posture adjustment process, the 6D posture data of the target area is re-acquired by the orthogonal X-ray imaging device every 10ms to calculate the new real-time deviation e(t). With the new deviation as input, the calculation, compensation and execution process is repeated to continuously fine-tune the robot arm posture until the deviation initially converges to a small range, laying the foundation for subsequent accuracy verification.

[0032] After calibration, the target area image is captured by an orthogonal X-ray imaging device, the current 6D pose of the target area is calculated, and the pose deviation after calibration is verified to be ≤0.5mm. If the deviation meets the standard, a beam recovery command is output and treatment continues. If the deviation does not meet the standard, calibration continues until the deviation meets the requirements. The entire calibration process takes less than 4 seconds.

[0033] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0034] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0035] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0036] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0037] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0038] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A system for modeling the elastic deformation of organ displacement during CyberKnife treatment, characterized in that, It includes a data acquisition module, a dynamic registration module, a tracking and prediction module, and a treatment plan integration module. These modules are interconnected. The data acquisition module is used to collect static anatomical data, dynamic displacement data, and stiffness distribution data of the patient's target organ. The dynamic registration module constructs a heterogeneous elastic deformation field of the target organ based on the collected data through deformation modeling and registration engine to achieve dynamic registration; The tracking and prediction module obtains the real-time displacement of the target organ based on dynamic displacement data, and predicts the deformation trend by combining the respiratory-target area motion correlation model and the reduced-order model to obtain predicted deformation data. The treatment planning integration module acquires 4D dose distribution data based on a heterogeneous elastic deformation field, calculates conformity and homogeneity index based on the 4D dose distribution data, and evaluates the dose under deformation constraints in conjunction with preset organ-at-risk doses. If the evaluation results do not meet the requirements, the dose distribution is optimized. The feedback control module outputs sub-millimeter-level motion compensation commands to the SARK robotic arm based on the predicted deformation field data. When the displacement exceeds the preset threshold, beam correction is performed, and the PID control algorithm is used to adjust the robotic arm's posture.

2. The elastic deformation modeling system for organ displacement during CyberKnife treatment according to claim 1, characterized in that, The process of collecting static anatomical data, dynamic displacement data, and stiffness distribution data of the patient's target organ is as follows: A full-range scan of the organs in the target area of ​​the patient was performed to obtain static anatomical images at the end of inspiration, end of expiration, and key time phases, generating a three-dimensional anatomical structure dataset to obtain static anatomical data. Stiffness scanning is performed on the target organ to obtain the original spatial distribution data of the elastic modulus, thus obtaining stiffness distribution data; Infrared markers were affixed to the patient's body surface in the area corresponding to respiratory movements. The infrared marker device and orthogonal X-ray imaging equipment were activated simultaneously. Data were continuously collected for the entire respiratory cycle while the patient was breathing naturally. The three-dimensional coordinate sequence of the infrared markers and the 6D pose data of the gold markers inside the target area were obtained at each respiratory phase. At the same time, dynamic image sequences were collected during the respiratory cycle to obtain data on the position and morphological changes of organs at different respiratory phases, thus obtaining dynamic displacement data.

3. The elastic deformation modeling system for organ displacement during CyberKnife treatment according to claim 2, characterized in that, The implementation process of the dynamic registration module is as follows: Using images from the end-expiratory phase of static anatomical data as reference images, the target area and the outline of organs at risk are delineated on the reference images, and a reference coordinate system is established. Rigidly register dynamic images of other respiratory phases with reference images to eliminate global bias caused by overall patient positional movement; The target organ is regarded as a heterogeneous isotropic elastic body, and a set of heterogeneous elastic mechanical control equations is established based on stiffness distribution data. The displacement vector is obtained by using finite element discretization and numerical solution based on the equations, thereby generating a non-homogeneous elastic deformation field.

4. The elastic deformation modeling system for organ displacement during CyberKnife treatment according to claim 3, characterized in that, The displacement vector is obtained by using finite element discretization and numerical solution based on the equations, and the process of generating a non-homogeneous elastic deformation field is as follows: The three-dimensional model of the target organ is divided into tiny tetrahedral elements, and the mechanical control equations are transformed into a discrete linear equation system that can be solved by a computer. The organ surface is set as a free boundary, and the actual displacement data of the collected gold standard is substituted as a constraint condition. The conjugate gradient method is used to solve the discretized linear equations to obtain the displacement vector components of each tetrahedral element. By integrating the displacement components of all units, the displacement vector of the entire organ at any time and any position within the entire respiratory cycle is obtained.

5. The elastic deformation modeling system for organ displacement during CyberKnife treatment according to claim 4, characterized in that, The process of obtaining predicted deformation data is as follows: Load the pre-trained respiratory-target motion correlation model, input the current body surface infrared signal into the model, output the initial value of target displacement prediction, fuse the initial value of target displacement prediction with real-time displacement data, correct the model parameters, and improve the correlation mapping accuracy. By calling a reduced-order model based on intrinsic orthogonal decomposition, the heterogeneous elastic deformation field is projected onto a low-dimensional intrinsic mode space, reducing computational complexity. The displacement data of the current and historical N frames are input into the reduced-order model to solve the temporal variation law of the modal coefficients. Based on the temporal variation law of the modal coefficients, the modal coefficients at the future time τ are predicted. The inhomogeneous elastic deformation field at the future time is reconstructed by inverse mapping of the reduced-order model to obtain the predicted deformation data.

6. The elastic deformation modeling system for organ displacement during CyberKnife treatment according to claim 5, characterized in that, The initial predicted displacement value of the target area is fused with the real-time displacement data to correct the model parameters. The process is as follows: Set the fusion weight coefficients for the initial predicted target displacement values ​​output by the model, and use a weighted average algorithm to calculate the fused target displacement values. The fused displacement value is correlated with the body surface infrared signal input to the model to construct a body surface signal-target area displacement sample pair at the current moment; The least squares iterative algorithm is used to correct the mapping parameters inside the model by minimizing the error between the displacement value output by the model based on the current body surface signal and the fused displacement value.

7. The elastic deformation modeling system for organ displacement during CyberKnife treatment according to claim 6, characterized in that, The process of obtaining 4D dose distribution data is as follows: Load the baseline treatment plan established before surgery, including beam parameters, target area and organ at risk contours, and initial dose distribution; Import heterogeneous elastic deformation field and predicted deformation data to establish a dynamic mapping relationship between deformation field and dose distribution; Based on the displacement vector in the deformation field, the dose distribution of the baseline plan is mapped from the reference phase to each phase of the entire respiratory cycle through the deformation vector field, so as to obtain the dose distribution of each phase. By traversing the entire respiratory cycle, the 4D dose distribution data is calculated.

8. The elastic deformation modeling system for organ displacement during CyberKnife treatment according to claim 7, characterized in that, If the evaluation results do not meet the requirements, the dose distribution optimization process is as follows: Define the range of beam parameters that need to be adjusted, including beam incident angle and beam weight; The real-time displacement data and predicted deformation field data output by the tracking and prediction module are correlated with the dose distribution corresponding to each beam parameter. The influence of beam parameter changes on target area dose coverage and radiation dose to organs at risk under different respiratory phases is analyzed, and a mapping relationship between beam parameters, deformation state and dose distribution is established. Based on the mapping relationship, the beam parameters corresponding to the areas with insufficient or excessive dose coverage are initially adjusted. The adjusted beam parameters are then substituted into the 4D dose calculation model to recalculate the dose distribution throughout the respiratory cycle and then re-evaluate. If the reassessed indicators still fail to meet the target, record the deviation between the current dose distribution and the target, further fine-tune the beam parameters based on the direction of the deviation, and repeat the assessment and calculation process. When the requirements are met for three consecutive evaluations or the parameter adjustments have reached the preset maximum number of iterations, the optimization is terminated and the final beam parameter combination is confirmed. Dosage optimization is completed once all evaluation indicators meet the standards.

9. The elastic deformation modeling system for organ displacement during CyberKnife treatment according to claim 8, characterized in that, When the displacement exceeds a preset threshold, the beam correction step is as follows: The 6D pose deviation between the target center at the current and predicted time and the target center in the optimized treatment plan is calculated based on the predicted deformation data and real-time displacement data. If the deviation does not exceed the threshold, maintain the current robotic arm posture and beam parameters; If the current or predicted pose deviation exceeds the preset threshold, a beam interruption command is immediately triggered to pause beam irradiation; at the same time, a correction process is initiated, inputting real-time displacement data, predicted deformation field data, and target area position information of the optimized treatment plan into the robotic arm control unit.

10. The elastic deformation modeling system for organ displacement during CyberKnife treatment according to claim 9, characterized in that, The process of adjusting the robot arm's posture using the PID control algorithm is as follows: First, integrate the previously calculated 6D pose deviations into a unified deviation variable, and then load the preset PID control parameters, including proportional coefficient, integral coefficient and derivative coefficient, and set the initial cumulative value of the integral term and the initial difference value of the derivative term to 0. The three core components—proportional, integral, and differential—are calculated separately. The calculated proportional, integral, and differential components are then added together to obtain the required angle change compensation for each joint of the robotic arm. The synthesized joint angle compensation is converted into electrical signal control commands that can be recognized by the robotic arm drive unit. These commands are then sent to the drive motors of each joint of the robotic arm via the system bus. The drive motors rotate the joints according to the commands, thus completing the initial correction of the robotic arm's position and posture.

Images