Radiotherapy transport system and method based on digital twinning

Abstract

The application discloses a radiotherapy transfer system and method based on digital twinning, and the system comprises the following modules: a pose generation module generates a medical bed rigid body pose according to joint angles of a mechanical arm and lifting height of a bed surface; a deformation field calculation module calculates a bed surface deformation field according to bending moment readings and strain readings of the mechanical arm and the medical bed, in combination with a calibrated mapping matrix and a reserved deformation mode; a support area calculation module determines a patient support area according to installation position coordinates of a patient fixator and calculates equivalent vertical displacement and equivalent tilt angle; an offset generation module generates a target region position offset according to the above; a coordinate compensation module obtains compensated patient treatment coordinates; and a control processing module generates a control processing result according to a residual deviation between the compensated patient treatment coordinates and target positioning coordinates, and the bending moment readings and the strain readings. The application can compensate for the elastic deformation of the bed surface in the transfer process and improve the positioning accuracy of radiotherapy.

Description

Technical Field

[0001] This application relates to the field of radiotherapy equipment control technology, and more specifically, to a radiotherapy transport system and method based on digital twins. Background Technology

[0002] Radiotherapy demands high precision in patient positioning, as deviations in the target area relative to the treatment coordinate system directly impact dose delivery accuracy. In current radiotherapy procedures, patients typically undergo fixation and initial positioning during the preparation phase, before being transported to the treatment room for final positioning. Transport equipment often includes components such as a robotic arm, medical bed, and lifting mechanism. The robotic arm's posture, bed height, patient weight, body position, and fixation device placement all affect the stress on the bed and supporting structure.

[0003] During actual patient transport, even when the robotic arm and medical bed are within safe operating limits, they may still experience minor elastic deformations after bearing the patient and fixation devices, such as bed surface sinking, end deflection, or local tilting. These deformations are usually insufficient to trigger equipment safety alarms, but may cause a shift in the target area within the patient's body relative to the target positioning coordinates. Current technologies primarily focus on determining the rigid body posture of the equipment based on information such as the robotic arm joint angles and bed lifting height, or using sensor readings such as bending moment and strain to determine whether the equipment exceeds mechanical safety limits. However, it is difficult to establish an effective correspondence between stress deformation and the patient's treatment coordinates.

[0004] Meanwhile, mechanical safety limits and the allowable range of radiotherapy positioning accuracy belong to different judgment dimensions. Controlling solely based on mechanical safety status can easily overlook hidden positioning errors where the equipment posture is normal but the target area position has deviated. If simple image verification is used for correction, it will increase the complexity of the process and the positioning time. Summary of the Invention

[0005] This application provides a radiotherapy transport system and method based on digital twins, which at least solves some of the technical problems existing in the related technologies described above.

[0006] According to a first aspect of the embodiments of this application, a radiotherapy transport system based on digital twins is provided, comprising: The pose generation module is used to acquire the joint angles of the robotic arm, the lifting height of the bed surface, the installation position coordinates of the patient fixation device, the target positioning coordinates, and the relative position information of the target area in the treatment plan. The digital twin generates the rigid body pose of the medical bed based on the joint angles of the robotic arm and the lifting height of the bed surface. The deformation field calculation module is used to collect bending moment readings and strain readings on the robotic arm and medical bed, form an observation vector, input the observation vector into a calibrated mapping matrix, solve for the deformation mode participation coefficient, and calculate the bed surface deformation field based on the deformation mode participation coefficient and the retained deformation modes. The support area calculation module is used to determine the patient support area in the bed deformation field based on the installation position coordinates of the patient fixator, and to calculate the equivalent vertical displacement and equivalent tilt angle of the patient support area. The offset generation module is used to generate a target area position offset based on the equivalent vertical displacement, the equivalent tilt angle, and the spatial position of the target area relative to the patient support area. The coordinate compensation module is used to superimpose the target area position offset onto the nominal target area position under the rigid body pose of the medical bed to obtain the compensated patient treatment coordinates. The control processing module is used to compare the compensated patient treatment coordinates with the target positioning coordinates to obtain the residual deviation, and generate the control processing result of the transport equipment based on the residual deviation, bending moment reading and strain reading.

[0007] As an optional solution, a mapping matrix calibration module is also included. The mapping matrix calibration module is used to: establish a reduced-order structural model including a robotic arm, a medical bed, and a fixed support structure; extract retained deformation modes from the reduced-order structural model; calculate the theoretical bending moment value and theoretical strain value corresponding to each retained deformation mode at the installation positions of each bending moment sensor and strain sensor; arrange the theoretical bending moment value and the theoretical strain value into an initial mapping matrix; and update the parameters of the initial mapping matrix according to the bending moment reading, strain reading, and reference deformation data collected under the calibration load condition to obtain the calibrated mapping matrix that satisfies the full column rank condition.

[0008] As an optional approach, establishing the reduced-order structural model includes: dividing the robotic arm into beam units according to joint segments, dividing the medical bed into bed units according to the longitudinal direction of the bed surface and the coverage area of ​​the patient support region, and setting the fixed support structure as the constraint boundary; configuring bending stiffness and torsional stiffness for the beam units, and configuring bending stiffness and in-plane stiffness for the bed units; and using the bending stiffness, torsional stiffness, and in-plane stiffness as parameters to be calibrated in the parameter update.

[0009] As an optional approach, extracting retained deformation modes from the reduced-order structural model includes: applying load combinations of different positions and magnitudes to the reduced-order structural model, calculating the nodal displacements and nodal rotation responses under each load combination; performing principal component extraction on all nodal displacements and nodal rotation responses to obtain candidate deformation modes; selecting the retained deformation modes from the candidate deformation modes according to the cumulative variance contribution rate; and not selecting candidate deformation modes from the generated initial mapping matrix.

[0010] As an optional approach, updating the parameters of the initial mapping matrix includes: under each set of calibration load conditions, recording the mass, placement coordinates, robotic arm joint angle, and bed lifting height of the calibration load; collecting the corresponding bending moment readings, strain readings, and actual bed displacement values; generating reference deformation data based on the mass, placement coordinates, robotic arm joint angle, bed lifting height, and actual bed displacement values ​​of the calibration load, wherein the reference deformation data includes reference values ​​for the participation coefficients of each retained deformation mode; and updating the initial mapping matrix based on the bending moment readings, strain readings, and reference deformation data using a recursive least squares method.

[0011] As an optional approach, calculating the bed surface deformation field includes: multiplying the deformation mode participation coefficients by the corresponding retained deformation modes and summing them to obtain the nodal displacements and nodal rotations of each node on the bed surface; interpolating the nodal displacements and nodal rotations to obtain the vertical displacements and tilt angles at each position within a continuous area of ​​the bed surface; and using the vertical displacements and tilt angles at each position within the continuous area of ​​the bed surface as the bed surface deformation field.

[0012] As an optional approach, determining the patient support area includes: determining the coverage area of ​​the patient fixation device in the bed coordinate system based on the installation position coordinates of the patient fixation device, and using the coverage area as the patient support area.

[0013] As an optional approach, calculating the equivalent vertical displacement and equivalent tilt angle of the patient support area includes: extracting the vertical displacement and tilt angle of each position within the patient support area from the bed deformation field; weighting and summing the vertical displacement according to the area weight corresponding to each position within the patient support area to obtain the equivalent vertical displacement; and weighting and summing the tilt angle according to the area weight to obtain the equivalent tilt angle.

[0014] As an optional approach, generating the target area position offset includes: when there are multiple patient support areas and the target area center is located within one of the patient support areas, determining that patient support area as the patient support area corresponding to the target area center; reading the spatial position of the target area center relative to the center of the patient support area, the spatial position including a planar position component and a vertical component; calculating the target area lateral offset based on the equivalent tilt angle and the vertical component in the spatial position, determining the target area vertical offset based on the equivalent vertical displacement, and combining the target area lateral offset and the target area vertical offset to form the target area position offset.

[0015] According to a second aspect of the embodiments of this application, a radiotherapy transport method based on digital twins is also provided, comprising: The system acquires information on the joint angles of the robotic arm, the lifting height of the bed surface, the coordinates of the patient fixation device installation position and the target positioning coordinates, as well as the relative position information of the target area in the treatment plan. The digital twin then generates the rigid body pose of the medical bed based on the joint angles of the robotic arm and the lifting height of the bed surface. The bending moment and strain readings of the robotic arm and the medical bed are collected to form an observation vector. The observation vector is input into the calibrated mapping matrix to solve for the deformation mode participation coefficient. The deformation field of the bed surface is calculated based on the deformation mode participation coefficient and the retained deformation mode. The patient support area is determined in the bed deformation field based on the coordinates of the patient fixation device installation position, and the equivalent vertical displacement and equivalent tilt angle of the patient support area are calculated. Based on the equivalent vertical displacement, the equivalent tilt angle, and the spatial position of the target area relative to the patient support area, a target area position offset is generated; The target area position offset is superimposed on the nominal target area position under the rigid body pose of the medical bed to obtain the compensated patient treatment coordinates. The control processing module is used to compare the compensated patient treatment coordinates with the target positioning coordinates to obtain the residual deviation, and generate the control processing result of the transport equipment based on the residual deviation, bending moment reading and strain reading.

[0016] This application incorporates the joint angles of the robotic arm, the lifting height of the bed, the installation position of the patient fixation device, and the relative position of the target area in the treatment plan into digital twin calculations. Based on the rigid body posture of the medical bed, it can further calculate the bed surface deformation field using bending moment and strain readings from the robotic arm and the medical bed. Combined with the equivalent vertical displacement and equivalent tilt angle of the patient support area, it generates the target area position offset, thus transforming the impact of equipment deformation on the patient's treatment coordinates into a quantifiable and compensable coordinate correction result. Compared to existing technologies that rely solely on mechanical safety limits for alarms or on rigid body posture for positioning, this application can identify hidden errors where the equipment remains within safe limits but the target area position has shifted. Based on the residual deviation, it generates processing results such as continued transport, positioning correction, positioning verification, or equipment protection. This improves the geometric positioning accuracy during radiotherapy transport and positioning, and reduces positioning errors caused by differences in patient weight, fixation device position, and robotic arm extension status.

[0017] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Furthermore, no embodiment in this disclosure is required to achieve all the effects described above. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0019] Figure 1 This is a flowchart illustrating a digital twin-based radiotherapy transport system and radiotherapy method provided in this disclosure.

[0020] Figure 2 A flowchart for calculating the deformation field of the bed surface provided in an embodiment of this disclosure.

[0021] Figure 3 A flowchart for generating target area position offset provided in the embodiments of this disclosure.

[0022] Figure 4 A flowchart for generating transfer control processing results provided in the embodiments of this disclosure.

[0023] Figure 5 A flowchart illustrating the process of obtaining the calibrated mapping matrix provided in this embodiment of the disclosure.

[0024] Figure 6 This is a schematic block diagram of a radiotherapy transport system and a radiotherapy system based on digital twins, provided as embodiments of this disclosure. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0026] The method provided in this disclosure is applicable to radiotherapy transport scenarios. The transport equipment includes a robotic arm, a medical bed, a bending moment sensor, a strain sensor, and a digital twin system. The robotic arm carries and moves the medical bed, which carries the patient and a patient immobilizer. The bending moment sensor and strain sensor are distributed across the force-bearing sections of the robotic arm and the medical bed. The digital twin system maintains a virtual model of the transport equipment and updates the virtual model's state based on sensor data. During radiotherapy transport, after the patient's immobilizer is installed and initial positioning is calibrated, the transport equipment moves the patient from the preparation room to the treatment room for final positioning. Due to variations in patient weight, position, immobilizer location, robotic arm extension length, and bed lifting height, the medical bed and robotic arm may still experience minor elastic deformation within safe force limits. Although this deformation may not trigger a safety alarm, it may cause a shift in the target area within the patient's body relative to the treatment coordinate system. This disclosure converts the bending moment and strain readings collected by the sensors into compensation values ​​for the patient's treatment coordinates, thereby identifying and correcting elastic deformations that, while not exceeding safety limits, have affected treatment positioning accuracy.

[0027] The implementation process of the method described in this application will be described in detail below with reference to specific embodiments. It should be noted that this embodiment is only used to explain this application and is not intended to limit the scope of protection of this application. Conventional adjustments or substitutions of each step by those skilled in the art without departing from the concept of this application should be included in the scope of protection of this application.

[0028] Please see Figure 1 , Figure 1 This is a flowchart of a digital twin-based radiotherapy transport system and radiotherapy method according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes: In box 10, the robot arm joint angle, bed lifting height, patient fixation device installation position coordinates, target positioning coordinates, and target area relative position information in the treatment plan are obtained. The digital twin generates the rigid body pose of the medical bed based on the robot arm joint angle and the bed lifting height.

[0029] In one embodiment, after the patient is secured to the medical bed and the fixation device is installed, the system acquires the robotic arm joint angles, bed lifting height, patient fixation device installation position coordinates, target positioning coordinates, and the relative position information of the target area in the treatment plan. The robotic arm joint angles are provided by encoders at each joint, the bed lifting height is provided by position feedback from the lifting mechanism, the patient fixation device installation position coordinates are entered by the operator during the treatment preparation phase or automatically collected by the positioning device, and the target positioning coordinates and the relative position information of the target area can be provided by the treatment plan management module.

[0030] Based on the aforementioned robotic arm joint angles and bed lifting height, the digital twin system calculates the spatial position and attitude angles of each component in the treatment coordinate system using forward kinematics, generating the rigid body pose of the medical bed. This forward kinematics solution takes the robotic arm joint angles and link parameters as input and outputs the position and attitude of the medical bed in the treatment coordinate system. This rigid body pose reflects the ideal position under the assumption that the components of the device have no elastic deformation.

[0031] In box 20, bending moment and strain readings of the robotic arm and medical bed are collected to form an observation vector. The observation vector is input into the calibrated mapping matrix to solve for the deformation mode participation coefficient. The deformation field of the bed surface is calculated based on the deformation mode participation coefficient and the retained deformation modes.

[0032] Specifically, please refer to Figure 2 , Figure 2 A flowchart illustrating the calculation of the bed deformation field provided in an embodiment of this disclosure is shown. Figure 2 As shown, in box 201, the bending moment readings and strain readings of the robotic arm and the medical bed are collected to form an observation vector.

[0033] Specifically, during the transfer process, the system collects bending moment readings from all bending moment sensors distributed across the joints of the robotic arm and the medical bed in real time, as well as strain readings from all strain sensors. Bending moment sensors are typically installed at the connection points between the robotic arm segments and the bed, as well as at critical locations on the bed's main beam. Strain sensors are attached to the surfaces of the robotic arm links and areas of the bed subjected to significant stress. Each bending moment sensor outputs a bending moment reading, and each strain sensor outputs a strain reading. The system arranges all bending moment and strain readings collected simultaneously into a vector, i.e., the observation vector.

[0034] In box 202, the observation vector is input into the calibrated mapping matrix to solve for the deformation mode participation coefficient. The mapping matrix describes the linear correspondence between sensor readings and deformation mode participation coefficients. Deformation modes refer to the typical deformation shapes exhibited by the transfer equipment under different stress conditions, such as sinking of the middle of the bed, sinking of the front end, tilting left and right, and drooping of the end of the robotic arm. Each deformation mode corresponds to a mode participation coefficient, which reflects the degree of participation of this deformation shape under the current stress state.

[0035] Among them, the observation vector With the deformation mode participation coefficient vector The relationship between them can be represented as ,in For the calibrated mapping matrix, To measure noise, the system uses the calibrated mapping matrix as the coefficient matrix and solves the above equation using the least squares method to obtain the estimated values ​​of the deformation mode participation coefficients. Fully ranked and When reversible, This solution is used to minimize the sum of squared errors between the sensor observation vectors and the predicted observation vectors from the mapping matrix. When the number of sensors exceeds the number of retained deformation modes and When the column rank is full, this formula provides the optimal estimated solution to the overdetermined system of equations. The least squares method is a commonly used method in numerical computation for solving linear overdetermined systems of equations, obtaining the optimal parameter estimate by minimizing the sum of squared errors between observed and predicted values. The calibration process of the mapping matrix is ​​completed before the equipment is put into clinical use; the specific calibration method will be explained later and will not be repeated here.

[0036] In box 203, the bed surface deformation field is calculated based on the deformation mode participation coefficients and the retained deformation modes. Specifically, the retained deformation modes are several main deformation shapes extracted and retained from the reduced-order structural model during the model building phase. Each deformation mode is represented by a nodal degree-of-freedom vector, which includes the vertical displacement components and rotational components about the longitudinal and transverse axes of each node on the bed surface. The deformation mode participation coefficients obtained by the system solution are multiplied by the corresponding retained deformation modes and summed to obtain the nodal displacements and nodal rotations of each node on the bed surface under the current stress state. The principle is that the bed surface deformation under the actual stress state can be approximately decomposed into a linear combination of several basic deformation modes, and the mode participation coefficients are the weights of each basic mode in the combination.

[0037] Specifically, it is assumed that the following were retained. Deformation mode, the first The nodal displacement vector corresponding to the first deformation mode is , No. The participation coefficient of the first deformation mode is Then the nodal displacement vector under the current stress state is:

[0038] Nodal displacement vector It includes the vertical displacement components and the rotational components around the longitudinal and transverse axes of each discrete node on the bed surface. The vertical displacement reflects the subsidence of each node, and the rotational components reflect the inclination at each node.

[0039] Since the reduced-order structural model divides the bed surface into a finite number of nodes and elements, the output of modal superposition only covers these discrete nodes. To obtain the vertical displacement and tilt angle at any position within the continuous region of the bed surface, the system performs linear interpolation on the displacement and rotation values ​​between adjacent bed element nodes. The linear interpolation uses the displacement or rotation values ​​of adjacent element nodes as endpoints or vertices, and performs interpolation according to the distance ratio from the position to be determined to the adjacent element nodes. After the above interpolation, each position on the bed surface corresponds to a vertical displacement value and a set of tilt angle values, where the tilt angle includes components around the longitudinal axis and components around the transverse axis. These continuously distributed vertical displacements and tilt angles together constitute the bed surface deformation field.

[0040] In frame 30, the patient support area is determined in the bed deformation field according to the coordinates of the patient fixation device installation position, and the equivalent vertical displacement and equivalent tilt angle of the patient support area are calculated.

[0041] In some embodiments, patient fixation devices are mounted on the bed surface, with each fixation device occupying a certain contact area on the bed surface. The system determines the coverage area of ​​each fixation device in the bed surface coordinate system based on the installation position coordinates of the patient fixation devices, and uses this coverage area as the patient support area. The coverage area can be characterized by a rectangular region or the bottom contour of the fixation device, with its center position determined by the installation position coordinates and its size determined by the dimensional parameters corresponding to the fixation device model.

[0042] Specifically, the system pre-establishes a mapping table between fixator models and dimensional parameters. This mapping table uses the fixator model, fixator identification code, or fixator installation interface number as an index, recording the corresponding fixator's bottom length, bottom width, relative coordinates of mounting holes or positioning pins, bottom contour point set, and nominal height. When the system obtains the patient's fixator installation position coordinates, it queries the mapping table based on the fixator model to obtain the corresponding dimensional parameters and, combined with the installation position coordinates, determines its coverage area in the bed coordinate system. For fixators with irregular bottom contours, the system projects the bottom contour point set onto the bed coordinate system, determines the boundary coordinates of the projected contour along the longitudinal and transverse directions of the bed surface, and uses the smallest circumscribed rectangle enclosed by these boundary coordinates as the rectangular region; alternatively, when high calculation accuracy is required, the area inside the projected contour is directly used as the patient support area for area-weighted calculation.

[0043] When a patient uses multiple immobilizers, such as head and neck immobilizers and trunk immobilizers installed at different positions on the bed, the system determines the patient support area corresponding to each immobilizer and sends each patient support area to the subsequent equivalent deformation calculation step for independent processing.

[0044] The patient support area can be divided into head and neck support area, trunk support area and lower limb support area, or it can be divided according to the contact area of ​​the bottom surface of each fixator.

[0045] For each patient support area, the system extracts the vertical displacement and tilt angle at each location within the coverage area from the bed deformation field obtained in the preceding steps. Vertical displacement and tilt angle may differ at different locations within the same support area. The system performs a weighted summation of the vertical displacements according to the area weights corresponding to each location within the area to obtain the equivalent vertical displacement. The equivalent tilt angle is obtained by weighting the tilt angles according to the same area weight.

[0046] Specifically, the patient support area is divided into several micro-areas along the bed coordinate system. Each micro-element corresponds to a vertical displacement value and tilt angle value interpolated from the deformation field. The equivalent vertical displacement is equal to the sum of the products of the vertical displacement values ​​and their areas at each micro-element, divided by the total area of ​​the support area. The equivalent tilt angle is calculated in the same way. The area weight refers to the ratio of the area of ​​each micro-element to the total area of ​​the support area, with a value ranging from 0 to 1, and the sum of the area weights of all micro-elements equals 1. This area weighting method ensures that the portion with a larger area ratio within the support area has a greater impact on the equivalent value, and can more accurately reflect the influence of the contact state between the bottom surface of the fixation device and the bed surface on the overall posture of the patient.

[0047] In actual calculations, the deformation field of the bed surface is realized in the form of discrete nodal data and linear interpolation. The weighted summation mentioned above is realized numerically. The equivalent vertical displacement reflects the overall subsidence of the support area, and the equivalent tilt angle reflects the overall tilt of the support area. Together, they describe the rigid body equivalent motion of the support area.

[0048] In box 40, a target area position offset is generated based on the equivalent vertical displacement, the equivalent tilt angle, and the spatial position of the target area relative to the patient support area.

[0049] Specifically, please refer to Figure 3 , Figure 3 A flowchart illustrating the target area position offset generation process provided in an embodiment of this disclosure is shown. Figure 3 As shown in box 301, the spatial position of the target area center relative to the center of the patient's support area is read.

[0050] When multiple patient support areas exist, the system first determines the corresponding patient support area based on the projection position of the target area center in the bed coordinate system. When the projection position is located within a certain patient support area, that patient support area is used as the patient support area for calculating the target area position offset. The target area is the treatment target area specified in the treatment plan, and its position is represented by the target area center coordinates. The spatial position of the target area center relative to the center of the patient support area includes a planar position component and a vertical component.

[0051] The planar position component is used to determine the correspondence between the target area center and the patient support area, and to calculate the distance from the target area center to the center of the adjacent patient support area; the vertical component... This is the distance from the center of the target area to the bed surface. This value can be obtained from the positional parameters or fixator geometry parameters recorded in the treatment plan. When this value is not provided in the treatment plan, the nominal height of the fixator can be used as a substitute, because the design function of the radiotherapy fixator is to constrain the displacement of the patient's body surface relative to the fixation device, and the movement of the fixator surface has a high degree of consistency with the movement of the patient's body surface.

[0052] In box 302, the lateral offset of the target area is calculated based on the equivalent tilt angle and spatial position. When the patient support area tilts, the center of the target area, located at a certain distance above the bed surface, will experience a horizontal displacement relative to the center of the support area. This displacement depends on the equivalent tilt angle and the distance from the target area center to the bed surface. Since the tilt angle is small within the elastic deformation range of the bed, a small angle approximation can be used for calculation. The patient's longitudinal and transverse axes in the bed surface coordinate system are denoted as... direction and The direction, perpendicular to the bed surface and pointing from the bed surface towards the target area inside the patient's body, is denoted as the z-direction, and around the longitudinal axis is... The equivalent tilt angle of the direction is denoted as , around the horizontal axis The equivalent tilt angle of the direction is denoted as Then, under the small-angle approximation, the lateral offset of the target area can be expressed as:

[0053] The tilt of the support area causes the target area above it to translate in a direction perpendicular to the tilt axis. The amount of translation is proportional to the product of the tilt angle and the distance from the target area to the bed surface. The greater the distance from the target area to the bed surface, the greater the lateral offset caused by the same tilt angle. The sign direction is determined according to the coordinate system formed by the x, y, and z directions.

[0054] In box 303, the vertical offset of the target area is determined based on the equivalent vertical displacement. The vertical offset of the target area is directly equal to the equivalent vertical displacement of the patient support area, i.e. The amount of overall subsidence of the support area corresponds to the same vertical displacement of the target area above it. The horizontal and vertical displacements of the target area can be combined to obtain the target area position offset. .

[0055] In some embodiments, the target center may optionally be located between two adjacent patient support regions. In this case, the equivalent deformation of any single support region cannot fully reflect the actual offset at the target region. In this situation, the system calculates the offset of the two candidate target regions according to the aforementioned method, based on the equivalent vertical displacement, equivalent tilt angle, and spatial position of the target region relative to the center of each of the two adjacent patient support regions.

[0056] The offsets of the two candidate target areas are normalized and weighted based on the distances from the center of the target area to the centers of the two adjacent patient support regions. The weighting coefficients are determined by calculating the distances from the target area center to the centers of the two support regions, taking the reciprocal of each distance, and then normalizing the weights so that the sum of the two weights equals 1. Support regions that are closer to each other receive a larger weight, and those that are farther apart receive a smaller weight. This weighted weighting yields the final target area offset, which is then not individually used in subsequent processing.

[0057] In box 50, the target area position offset is superimposed on the nominal target area position under the rigid body pose of the medical bed to obtain the compensated patient treatment coordinates.

[0058] Specifically, the target area offset is superimposed onto the nominal target area position under the rigid body pose of the medical bed to obtain the compensated patient treatment coordinates. The compensated patient treatment coordinates are then compared with the target positioning coordinates to obtain the residual deviation. The nominal target area position under the rigid body pose of the medical bed refers to the coordinates of the target area in the treatment coordinate system, determined by the relative position information of the target area in the treatment plan and the medical bed rigid body pose output by the digital twin system, assuming no elastic deformation of the equipment. This is denoted as... The nominal position is calculated by the digital twin system while generating the rigid body pose of the medical bed.

[0059] The system will output the target area offset from the aforementioned steps. Superimposed onto the nominal target area, the compensated patient treatment coordinates are obtained. This coordinate reflects the actual estimated position of the target area under the action of elastic deformation of the bed.

[0060] In box 60, the compensated patient treatment coordinates are compared with the target positioning coordinates to obtain the residual deviation. Based on the residual deviation, bending moment reading, and strain reading, the control processing result of the transport equipment is generated.

[0061] Specifically, please refer to Figure 4 , Figure 4 A flowchart illustrating the generation of transfer control processing results provided in an embodiment of this disclosure is shown. Figure 4 As shown in box 401, the residual deviation in each direction is obtained component by subtracting the compensated patient treatment coordinates from the target positioning coordinates.

[0062] The residual deviation represents the correction vector from the compensated patient treatment coordinates to the target positioning coordinates. A residual deviation of zero indicates that the actual estimated position of the target area is completely consistent with the target position. The larger the residual deviation, the greater the impact of bed elastic deformation on treatment positioning accuracy. Therefore, the system can detect target area position shifts caused by elastic deformation even when the equipment posture is correct, solving the problem of hidden shifts being missed due to the conflation of mechanical safety thresholds and radiotherapy positioning accuracy thresholds in existing solutions.

[0063] In box 402, the current bending moment and strain readings are compared with their corresponding mechanical safety limits to determine the mechanical safety status. The mechanical safety limits are determined by the equipment manufacturer based on structural strength and fatigue life, and are written into the system parameters at the time of equipment delivery. If any bending moment reading exceeds its corresponding bending moment safety limit, or any strain reading exceeds its corresponding strain safety limit, the mechanical safety status is determined to be abnormal; if all bending moment and strain readings are within their respective safety limits, the mechanical safety status is determined to be normal.

[0064] In box 403, the obtained residual deviation is compared with the permissible range of positioning accuracy to determine the radiotherapy geometric accuracy status. The permissible range of positioning accuracy is set by the treatment plan or clinical procedure, and its value is independent of the mechanical safety limit. When the residual deviation in each direction is within the permissible range of positioning accuracy, the radiotherapy geometric accuracy status is determined to be satisfactory; when the residual deviation in any direction exceeds the permissible range of positioning accuracy, the radiotherapy geometric accuracy status is determined to be insufficient.

[0065] The two state dimensions mentioned above are determined independently, and the data sources and threshold meanings used for their determination are different. Among them, the mechanical safety state is based on the comparison between the original sensor readings and the structural safety limit, while the radiotherapy geometric accuracy state is based on the comparison between the residual deviation of the patient coordinates after bed deformation compensation and the permissible range of positioning.

[0066] In box 404, control processing results are generated based on two state dimensions. Specifically, when the mechanical safety state is abnormal, regardless of the radiotherapy geometric accuracy state, the system generates equipment protection processing results, including pausing robotic arm movement and locking the bed position. Mechanical safety abnormalities are the highest priority events; under this condition, no positioning correction is performed.

[0067] When the mechanical safety status is normal and the radiotherapy geometric accuracy status is met, the system generates the results for continued transport and positioning, and the transport equipment continues to execute according to the planned process.

[0068] When the mechanical safety condition is normal but the radiotherapy geometric accuracy condition is insufficient, the equipment itself is in a safe operating state. However, the elastic deformation of the bed has caused the patient's coordinates to deviate beyond the permissible range for radiotherapy. This situation is easily overlooked in existing solutions because they typically only focus on whether sensor readings exceed safety limits, without involving the conversion calculation from sensor readings to patient treatment coordinates. In this embodiment, the system generates a positioning correction result based on the residual deviation in this situation, specifically as follows.

[0069] The system compares the residual deviation with the fine-tuning stroke of the transfer equipment, which includes the bed height fine-tuning range and the end-effector horizontal adjustable range converted from the robotic arm angle fine-tuning range.

[0070] When the residual deviation is within the fine-tuning stroke, the system converts the vertical component of the residual deviation into a bed height correction amount and the horizontal component into a robotic arm angle fine-tuning amount. The vertical component directly corresponds to the adjustment amount of the bed lifting mechanism, and the adjustment direction is opposite to the direction of the vertical residual deviation; that is, the bed is raised when the target area is too low and lowered when the target area is too high. The horizontal component is converted into the angle fine-tuning increment of each joint through the differential relationship between the displacement of the robotic arm end and the joint angle. This differential relationship is represented by the Jacobian matrix corresponding to the current robotic arm posture. The system packages the bed height correction amount and the robotic arm angle fine-tuning amount to generate a positioning correction processing result containing the above two items, which is then used by the device actuator for adjustment. After the adjustment is completed, the system re-acquires sensor data and repeats the aforementioned processing process to verify whether the adjusted residual deviation has entered the allowable range of positioning accuracy.

[0071] When the residual deviation exceeds the fine-tuning stroke, bed height adjustment and robotic arm angle fine-tuning cannot completely eliminate the deviation. The system generates a positioning verification result, prompting the operator to verify the patient's positioning through image guidance or other methods. This positioning verification result is marked as a positioning accuracy prompt on the system interface, distinguishing it from mechanical safety alarms in terms of display method and priority.

[0072] In some embodiments, the calibration of the mapping matrix is ​​performed before the device is put into clinical use, or it can be re-performed during the periodic maintenance of the device. The calibration process is divided into establishing a reduced-order structural model, extracting and retaining deformable modes, calculating the initial mapping matrix, and updating parameters.

[0073] Specifically, please refer to Figure 5 , Figure 5 A flowchart illustrating the process of obtaining the calibrated mapping matrix provided in an embodiment of this disclosure is shown. Figure 5 As shown, a reduced-order structure model is established in box 501.

[0074] Specifically, the reduced-order structural model is a simplified deformation calculation model of the robotic arm, medical bed, and fixed support structure of the transport equipment. Its degrees of freedom are much lower than the full-size finite element model, resulting in less computation and making it suitable for embedding in the equipment control system. When building this model, the robotic arm is divided into beam elements according to its joint segments. Each beam element corresponds to a link segment or rotatable joint segment of the robotic arm, and the length of the beam element is consistent with the actual length of the link segment. The medical bed is divided into bed body elements based on the longitudinal direction of the bed surface and the coverage area of ​​the patient support region. Bed body elements can be plate elements or beam elements, and the spacing between them is determined based on the length of the bed surface and the sensor placement. Typically, at least one node is set between adjacent sensors to ensure the resolution of the deformation field. The fixed support structure is set as a constraint boundary, and displacement and rotation constraints are applied to it.

[0075] Each beam element is configured with bending and torsional stiffness, and each bed element is configured with bending and in-plane stiffness. Bending stiffness determines the degree of bending deformation of the element under lateral load, torsional stiffness determines the torsional angle of the element under torque, and in-plane stiffness determines the deformation characteristics of the bed element in the plane. The initial values ​​of these stiffness parameters can be estimated based on material mechanics parameters and cross-sectional geometry, but there are usually deviations between these values ​​and those of the actual equipment. Therefore, these parameters are used as parameters to be calibrated in subsequent parameter updates.

[0076] In box 502, extract the preserved deformation modes from the reduced-order structural model. Apply load combinations of different positions and magnitudes to the reduced-order structural model, and calculate the nodal displacement and nodal rotation responses under each load combination. The load combinations should cover common patient weight ranges and body position distributions in clinical use, and may optionally include various working conditions such as concentrated loads at the head end of the bed, concentrated loads in the middle, concentrated loads at the foot end, and distributed loads. After summing all nodal displacement responses, perform principal component analysis (PCA) on them. That is, use PCA to perform eigenvalue decomposition on the nodal displacement and nodal rotation response data matrix, and extract principal components sorted by variance contribution rate. Each principal component corresponds to a candidate deformation mode. PCA is a commonly used method in multivariate statistics for dimensionality reduction and feature extraction. Its core is to project high-dimensional data onto several orthogonal directions with the largest variance, and approximate the main features of the original data with a small number of principal components.

[0077] Optionally, retained deformation modes are selected from candidate deformation modes based on their cumulative variance contribution rate. The calculation of the cumulative variance contribution rate involves sorting each candidate deformation mode by variance contribution rate from largest to smallest, sequentially accumulating the contribution rates of each order, and stopping selection when the accumulated value reaches a preset threshold. This threshold can be configured to be no less than 95%. Unselected candidate deformation modes do not participate in the generation of the initial mapping matrix; that is, their corresponding deformation shapes are not considered in subsequent calculations. The retained deformation modes typically include several typical shapes such as mid-bed sinking, front-end sinking, left and right tilting, and robotic arm end-effector drooping, with orders generally ranging from several to over ten.

[0078] In box 503, calculate the initial mapping matrix. At each installation location of the moment sensor and strain sensor, calculate the theoretical bending moment and theoretical strain value corresponding to each retained deformation mode according to the reduced-order structural model. The theoretical bending moment value is the bending moment value calculated from the structural model at the sensor location, assuming that the participation factor of a certain deformation mode is a unit value and the participation factors of the other modes are zero. The theoretical strain value has a similar meaning. Arrange all theoretical bending moment values ​​and theoretical strain values ​​into a matrix according to the sensor number and mode number, which is the initial mapping matrix. The number of rows in the initial mapping matrix is ​​equal to the total number of sensors, and the number of columns is equal to the order of the retained deformation modes.

[0079] In box 504, the initial mapping matrix is ​​updated. The initial mapping matrix is ​​generated based on theoretical calculations from the model and differs from the actual device due to manufacturing errors, assembly deviations, and material dispersion. The initial mapping matrix is ​​updated through a calibration process to ensure that it accurately reflects the sensor response characteristics of the actual device.

[0080] During specific calibration, under each set of calibration load conditions, the mass, placement coordinates, robotic arm joint angles, and bed lifting height of the calibration load are recorded. Corresponding bending moment readings, strain readings, and actual bed displacement values ​​are collected. The actual bed displacement value is obtained through an external measuring tool, optionally a displacement gauge or laser rangefinder. Based on the mass, placement coordinates, robotic arm joint angles, bed lifting height, and actual bed displacement value of the calibration load, the system uses a reduced-order structural model to back-calculate the reference values ​​of the participation coefficients of each deformation mode under the calibration conditions, thereby generating reference deformation data.

[0081] Specifically, the system applies the calibration load as an external force condition to the reduced-order structural model, determines the boundary conditions of the model by combining the current robotic arm posture and bed height, and obtains the theoretical nodal displacements by solving the model. The theoretical nodal displacements are compared with the measured actual displacements of the bed to determine the reference values ​​of the participation coefficients of each deformation mode. The reference deformation data is the set of reference values ​​of the participation coefficients of each retained deformation mode.

[0082] For example, the system employs the recursive least squares method to progressively update the initial mapping matrix based on the bending moment readings, strain readings, and reference deformation data collected under each set of calibration conditions. Upon obtaining a new set of calibration data, the system corrects the mapping matrix parameters based on the predicted residuals and updates the covariance matrix. The recursive least squares method is an online parameter estimation algorithm. Its basic principle is: based on existing parameter estimates, upon obtaining a new set of observation data, the system calculates the predicted residual of the current parameter estimate for that observation, corrects the parameter estimate based on the predicted residuals and the gain matrix, and recursively updates the covariance matrix. This algorithm does not require storing and reprocessing all historical data, and the computational cost of each update step is constant, making it suitable for operation in equipment control systems.

[0083] Optionally, a forgetting factor can be introduced into the recursive least squares method. Forgetting factor The range of values ​​is Its function is to assign higher weights to recent data and lower weights to earlier data when updating parameters. The closer The higher the degree to which the algorithm retains historical data, the smoother the parameter updates. The smaller the value, the faster the algorithm tracks new data. This forgetting factor can be preset during equipment debugging based on the long-term stability of the sensor and the aging rate of the equipment structure, or it can be adjusted during subsequent use based on the statistical characteristics of the parameter update residuals.

[0084] During the parameter update process described above, the bending and torsional stiffness of each beam element, and the bending and in-plane stiffness of each bed element in the reduced-order structural model are also included as parameters to be calibrated. Specifically, these stiffness parameters are incorporated into the parameter vector of the recursive least squares method. Under calibration load conditions, the stiffness values ​​are corrected by the residual between the measured displacement and the model's predicted displacement. Based on the corrected stiffness values, the theoretical bending moment and theoretical strain values ​​corresponding to each retained deformation mode are recalculated to simultaneously correct the initial mapping matrix, making the calibrated reduced-order structural model closer to the deformation characteristics of the actual equipment.

[0085] Optionally, during long-term use of the equipment, the system uses the sensor data and images accumulated from each transfer to guide the verification results and continue to perform recursive least squares updates on the mapping matrix to adapt to the slow changes in the equipment structure caused by wear, fatigue, and other factors. When the cumulative change in the mapping matrix parameters exceeds the preset maintenance threshold, the system prompts for a complete recalibration. The maintenance threshold can be set according to the sensor accuracy level and placement accuracy requirements.

[0086] Therefore, this invention converts the readings of the moment and strain sensors on the robotic arm and medical bed into a bed surface deformation field through a reduced-order structural model and deformation mode superposition. This, combined with the patient fixation device's support area division and rigid body kinematic geometry transformation, yields the compensation amount for the patient's treatment coordinates. This allows sensor data, originally used only for equipment safety monitoring, to simultaneously perform treatment positioning accuracy assessment and compensation. By independently determining the residual deviation between the compensated patient treatment coordinates and the target positioning coordinates, and by independently judging the original sensor readings and mechanical safety limits, it can identify hidden positioning deviations where the equipment posture is correct but the patient coordinates are off-target. This reduces positioning errors caused by differences in patient weight, fixation device position, and robotic arm extension status, while avoiding misjudging insufficient positioning accuracy as mechanical failure or overlooking positioning accuracy deviations due to a lack of mechanical safety alarms.

[0087] Please see Figure 6 , Figure 6 This is a block diagram of a radiotherapy transport system and a radiotherapy system based on digital twins, provided in an embodiment of this application. Figure 6 As shown, the system includes: The pose generation module 601 is used to acquire the joint angle of the robotic arm, the lifting height of the bed surface, the installation position coordinates of the patient fixation device, the target positioning coordinates, and the relative position information of the target area in the treatment plan. The digital twin generates the rigid body pose of the medical bed based on the joint angle of the robotic arm and the lifting height of the bed surface. The deformation field calculation module 602 is used to collect bending moment readings and strain readings on the robotic arm and medical bed, form an observation vector, input the observation vector into a calibrated mapping matrix, solve for the deformation mode participation coefficient, and calculate the bed surface deformation field based on the deformation mode participation coefficient and the retained deformation mode. The support area calculation module 603 is used to determine the patient support area in the bed deformation field according to the installation position coordinates of the patient fixator, and to calculate the equivalent vertical displacement and equivalent tilt angle of the patient support area; The offset generation module 604 is used to generate a target area position offset based on the equivalent vertical displacement, the equivalent tilt angle, and the spatial position of the target area relative to the patient support area. The coordinate compensation module 605 is used to superimpose the target area position offset onto the nominal target area position under the rigid body pose of the medical bed to obtain the compensated patient treatment coordinates. The control processing module 606 is used to compare the compensated patient treatment coordinates with the target positioning coordinates to obtain the residual deviation, and generate the control processing result of the transport device based on the residual deviation, bending moment reading and strain reading.

[0088] Optionally, the system further includes a mapping matrix calibration module, which is used to: establish a reduced-order structural model including a robotic arm, a medical bed, and a fixed support structure; extract retained deformation modes from the reduced-order structural model; calculate the theoretical bending moment value and theoretical strain value corresponding to each retained deformation mode at each installation position of the bending moment sensor and strain sensor; arrange the theoretical bending moment value and the theoretical strain value into an initial mapping matrix; and update the parameters of the initial mapping matrix according to the bending moment reading, strain reading, and reference deformation data collected under the calibration load condition to obtain the calibrated mapping matrix that satisfies the full column rank condition.

[0089] Optionally, establishing the reduced-order structural model includes: dividing the robotic arm into beam units according to joint segments, dividing the medical bed into bed units according to the longitudinal direction of the bed surface and the coverage area of ​​the patient support region, and setting the fixed support structure as a constraint boundary; configuring bending stiffness and torsional stiffness for the beam units, and configuring bending stiffness and in-plane stiffness for the bed units; and using the bending stiffness, torsional stiffness, and in-plane stiffness as parameters to be calibrated in the parameter update.

[0090] Optionally, extracting the retained deformation modes from the reduced-order structural model includes: applying load combinations of different positions and magnitudes to the reduced-order structural model, calculating the nodal displacements and nodal rotation responses under each load combination; performing principal component extraction on all nodal displacements and nodal rotation responses to obtain candidate deformation modes; selecting the retained deformation modes from the candidate deformation modes according to the cumulative variance contribution rate; and not selecting candidate deformation modes, which do not participate in the generation of the initial mapping matrix.

[0091] Optionally, updating the parameters of the initial mapping matrix includes: under each set of calibration load conditions, recording the mass, placement coordinates, robotic arm joint angle, and bed lifting height of the calibration load; collecting the corresponding bending moment readings, strain readings, and actual bed displacement values; generating reference deformation data based on the mass, placement coordinates, robotic arm joint angle, bed lifting height, and actual bed displacement values ​​of the calibration load, wherein the reference deformation data includes reference values ​​for the participation coefficients of each retained deformation mode; and updating the initial mapping matrix using the recursive least squares method based on the bending moment readings, strain readings, and reference deformation data.

[0092] Optionally, calculating the bed surface deformation field includes: multiplying the deformation mode participation coefficients by the corresponding retained deformation modes and summing them to obtain the nodal displacements and nodal rotations of each node on the bed surface; interpolating the nodal displacements and nodal rotations to obtain the vertical displacements and tilt angles at each position within the continuous area of ​​the bed surface; and using the vertical displacements and tilt angles at each position within the continuous area of ​​the bed surface as the bed surface deformation field.

[0093] Optionally, determining the patient support area includes: determining the coverage area of ​​the patient fixation device in the bed coordinate system based on the installation position coordinates of the patient fixation device, and using the coverage area as the patient support area.

[0094] Optionally, calculating the equivalent vertical displacement and equivalent tilt angle of the patient support area includes: extracting the vertical displacement and tilt angle of each position within the patient support area from the bed deformation field; weighting and summing the vertical displacement according to the area weight corresponding to each position within the patient support area to obtain the equivalent vertical displacement; and weighting and summing the tilt angle according to the area weight to obtain the equivalent tilt angle.

[0095] Optionally, generating the target area position offset includes: when there are multiple patient support areas and the target area center is located within one of the patient support areas, determining the patient support area as the patient support area corresponding to the target area center; reading the spatial position of the target area center relative to the center of the patient support area, the spatial position including a planar position component and a vertical component; calculating the target area lateral offset based on the equivalent tilt angle and the vertical component in the spatial position, determining the target area vertical offset based on the equivalent vertical displacement, and combining the target area lateral offset and the target area vertical offset into the target area position offset.

[0096] Each processing unit and / or module in the embodiments of this application can be implemented by an analog circuit that implements the functions described in the embodiments of this application, or by software that executes the functions described in the embodiments of this application.

[0097] In the above embodiments, the descriptions of each embodiment have different focuses. Parts not described in detail in a certain embodiment can be referred to in the relevant descriptions of other embodiments. The above descriptions are merely preferred embodiments of this application and explanations of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to the technical solutions formed by specific combinations of the above technical features, but should also cover other technical solutions formed by arbitrary combinations of the above technical features or their equivalent features without departing from the inventive concept.

Claims

1. A radiotherapy transport system based on digital twins, characterized in that, include: The pose generation module is used to acquire the joint angles of the robotic arm, the lifting height of the bed surface, the installation position coordinates of the patient fixation device, the target positioning coordinates, and the relative position information of the target area in the treatment plan. The digital twin generates the rigid body pose of the medical bed based on the joint angles of the robotic arm and the lifting height of the bed surface. The deformation field calculation module is used to collect bending moment readings and strain readings on the robotic arm and medical bed, form an observation vector, input the observation vector into a calibrated mapping matrix, solve for the deformation mode participation coefficient, and calculate the bed surface deformation field based on the deformation mode participation coefficient and the retained deformation modes. The support area calculation module is used to determine the patient support area in the bed deformation field based on the installation position coordinates of the patient fixator, and to calculate the equivalent vertical displacement and equivalent tilt angle of the patient support area. The offset generation module is used to generate a target area position offset based on the equivalent vertical displacement, the equivalent tilt angle, and the spatial position of the target area relative to the patient support area. The coordinate compensation module is used to superimpose the target area position offset onto the nominal target area position under the rigid body pose of the medical bed to obtain the compensated patient treatment coordinates. The control processing module is used to compare the compensated patient treatment coordinates with the target positioning coordinates to obtain the residual deviation, and generate the control processing result of the transport equipment based on the residual deviation, bending moment reading and strain reading.

2. The system according to claim 1, characterized in that, It also includes a mapping matrix calibration module, which is used to: establish a reduced-order structural model including a robotic arm, a medical bed, and a fixed support structure; extract retained deformation modes from the reduced-order structural model; calculate the theoretical bending moment value and theoretical strain value corresponding to each retained deformation mode at the installation positions of each bending moment sensor and strain sensor; arrange the theoretical bending moment value and the theoretical strain value into an initial mapping matrix; and update the parameters of the initial mapping matrix according to the bending moment reading, strain reading, and reference deformation data collected under the calibration load condition to obtain the calibrated mapping matrix that satisfies the full column rank condition.

3. The system according to claim 2, characterized in that, The process of establishing the reduced-order structural model includes: dividing the robotic arm into beam units according to joint segments, dividing the medical bed into bed units according to the longitudinal direction of the bed surface and the coverage area of ​​the patient support region, and setting the fixed support structure as the constraint boundary; configuring bending stiffness and torsional stiffness for the beam units, and configuring bending stiffness and in-plane stiffness for the bed units; and using the bending stiffness, torsional stiffness, and in-plane stiffness as parameters to be calibrated in the parameter update.

4. The system according to claim 2, characterized in that, Extracting retained deformation modes from the reduced-order structural model includes: applying load combinations of different positions and magnitudes to the reduced-order structural model, calculating the nodal displacements and nodal rotation responses under each load combination; performing principal component extraction on all nodal displacements and nodal rotation responses to obtain candidate deformation modes; selecting the retained deformation modes from the candidate deformation modes according to the cumulative variance contribution rate; unselected candidate deformation modes do not participate in the generation of the initial mapping matrix.

5. The system according to claim 2, characterized in that, The parameter update of the initial mapping matrix includes: under each set of calibration load conditions, recording the mass, placement coordinates, robotic arm joint angle, and bed lifting height of the calibration load; collecting the corresponding bending moment readings, strain readings, and actual bed displacement values; generating reference deformation data based on the mass, placement coordinates, robotic arm joint angle, bed lifting height, and actual bed displacement values ​​of the calibration load, wherein the reference deformation data includes reference values ​​for the participation coefficients of each retained deformation mode; and updating the initial mapping matrix based on the bending moment readings, strain readings, and reference deformation data using the recursive least squares method.

6. The system according to claim 1, characterized in that, The calculation of the bed surface deformation field includes: multiplying the deformation mode participation coefficients by the corresponding retained deformation modes and summing them to obtain the nodal displacements and nodal rotations of each node on the bed surface; interpolating the nodal displacements and nodal rotations to obtain the vertical displacements and tilt angles of each position within the continuous area of ​​the bed surface; and using the vertical displacements and tilt angles of each position within the continuous area of ​​the bed surface as the bed surface deformation field.

7. The system according to claim 1, characterized in that, Determining the patient support area includes: determining the coverage area of ​​the patient fixation device in the bed coordinate system based on the installation position coordinates of the patient fixation device, and using the coverage area as the patient support area.

8. The system according to claim 7, characterized in that, Calculating the equivalent vertical displacement and equivalent tilt angle of the patient support area includes: extracting the vertical displacement and tilt angle of each position within the patient support area from the bed deformation field; weighting and summing the vertical displacement according to the area weight of each position within the patient support area to obtain the equivalent vertical displacement; and weighting and summing the tilt angle according to the area weight to obtain the equivalent tilt angle.

9. The system according to claim 1, characterized in that, The generation of target area position offset includes: when there are multiple patient support areas and the target area center is located within one of the patient support areas, determining the patient support area as the patient support area corresponding to the target area center; reading the spatial position of the target area center relative to the center of the patient support area, the spatial position including a planar position component and a vertical component; calculating the target area lateral offset based on the equivalent tilt angle and the vertical component in the spatial position, determining the target area vertical offset based on the equivalent vertical displacement, and combining the target area lateral offset and the target area vertical offset into the target area position offset.

10. A radiotherapy transport method based on digital twins, characterized in that, include: The system acquires information on the joint angles of the robotic arm, the lifting height of the bed surface, the coordinates of the patient fixation device installation position and the target positioning coordinates, as well as the relative position information of the target area in the treatment plan. The digital twin then generates the rigid body pose of the medical bed based on the joint angles of the robotic arm and the lifting height of the bed surface. The bending moment and strain readings of the robotic arm and the medical bed are collected to form an observation vector. The observation vector is input into the calibrated mapping matrix to solve for the deformation mode participation coefficient. The deformation field of the bed surface is calculated based on the deformation mode participation coefficient and the retained deformation mode. The patient support area is determined in the bed deformation field based on the coordinates of the patient fixation device installation position, and the equivalent vertical displacement and equivalent tilt angle of the patient support area are calculated. Based on the equivalent vertical displacement, the equivalent tilt angle, and the spatial position of the target area relative to the patient support area, a target area position offset is generated; The target area position offset is superimposed on the nominal target area position under the rigid body pose of the medical bed to obtain the compensated patient treatment coordinates. The compensated patient treatment coordinates are compared with the target positioning coordinates to obtain the residual deviation. The control processing results of the transport equipment are generated based on the residual deviation, bending moment reading, and strain reading.

Images