Accelerator-based multi-vision monitoring device and optical body surface positioning method
By installing a multi-vision monitoring device on the radiotherapy accelerator and using the visual monitoring device on the fixed support as a reference, the problems of limited field of view and time-consuming installation in the prior art have been solved, and stable monitoring of the treatment space and precise radiotherapy positioning have been achieved.
Patent Information
- Application Number
- CN202411038768.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-29
AI Technical Summary
In existing radiotherapy, fixed gantry have limited field of view, rotating gantry have blind spots, and installing a binocular vision system on the ceiling is time-consuming and labor-intensive. Furthermore, existing monitoring equipment cannot obtain complete point cloud information of the body surface, increasing the risk of secondary tumors.
An accelerometer-based multi-vision monitoring device is adopted, including vision monitoring devices mounted on a rotating support and a fixed support. Three-dimensional surface point cloud matching is performed through joint calibration and ICP algorithm. The vision monitoring device on the fixed support is used as a reference to achieve the acquisition of multi-view information and precise positioning.
It achieves stable monitoring of the treatment space, saves installation costs and time, obtains complete surface point cloud information, improves the accuracy of radiotherapy positioning, avoids additional radiation exposure, and reduces the risk of secondary tumors.
Smart Images

Figure CN119034118A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of radiotherapy technology, specifically relating to an accelerator-based multi-vision monitoring device and an optical body surface positioning method. Background Technology
[0002] A medical linear accelerator is a medical device that uses microwave electromagnetic fields to accelerate electrons and create linear motion paths. It is used for radiotherapy of tumors or other lesions in patients. It can produce high-energy X-rays and electron beams, and features high dose rate, short irradiation time, large irradiation field, dose uniformity, good stability, and small penumbra.
[0003] During the use of the accelerator, higher requirements are placed on the positioning of radiotherapy patients, the mutual interference of various components, and the accurate tracking of the tumor. To this end, the accelerator gantry is equipped with multiple positioning systems, including cone-beam CT positioning, MV-level flat panel positioning, laser lamp positioning, and binocular vision system positioning, which work together to locate the position of the tumor and the placement of the treatment head and treatment bed.
[0004] One or more binocular vision systems are installed on the existing rotating gantry to rotate with the treatment head; or a fixed binocular vision system is installed on the ceiling or treatment bed to track the posture of the treatment bed and gantry.
[0005] Medical imaging equipment such as electronic field imaging systems (EPID), kV-level planar imaging systems, 4DCT, and cone-beam computed tomography (CBCT) play an important role in radiotherapy. However, these devices have some limitations in real-time monitoring during treatment, and the additional dose effects that may result from multiple imaging sessions cannot be ignored, leading to unnecessary irradiation of normal tissues and increasing the risk of secondary tumors.
[0006] Therefore, the above method has the following problems:
[0007] First, the field of view of a fixed frame is fixed and limited, making it impossible to obtain point cloud information of the complete body surface.
[0008] Secondly, the single rotating gantry has blind spots, and the treatment bed surface cannot be seen when the gantry rotates downwards.
[0009] Third, the rotating frame has no fixed reference surface and cannot be used as a stable reference.
[0010] Fourth, installing a binocular vision system on the ceiling or treatment bed requires installation on the ceiling, as well as the installation of power and communication lines, which is time-consuming and labor-intensive.
[0011] Fifth, the ceiling mounting is separate from the rack mounting, and the reference calibrates need to be set separately. Summary of the Invention
[0012] To address the aforementioned technical problems, this invention proposes a multi-vision monitoring device based on an accelerometer and an optical body surface positioning method.
[0013] To achieve the above objectives, the technical solution of the present invention is as follows:
[0014] On one hand, the present invention discloses a multi-vision monitoring device based on an accelerator, comprising: one or more sets of first vision monitoring devices mounted on a rotating support of the accelerator and one or more sets of second vision monitoring devices mounted on a fixed support of the accelerator;
[0015] The first visual monitoring device can rotate with the rotating bracket;
[0016] The second visual monitoring device remains stationary and is used to monitor and track the posture of the treatment bed and rotating support.
[0017] Based on the above technical solution, the following improvements can be made:
[0018] As a preferred option, the second visual monitoring device is connected to the fixed bracket via a mounting bracket.
[0019] As a preferred embodiment, the data cable or power cable of the second visual monitoring device is integrated and installed on the mounting bracket and / or fixed bracket.
[0020] Furthermore, this invention also discloses an optical body surface positioning method, which utilizes any of the aforementioned multi-vision monitoring devices for positioning, including:
[0021] Step S1: Perform joint calibration of the first visual monitoring device and the second visual monitoring device;
[0022] Step S2: Capture the patient's three-dimensional body surface point cloud using the first visual monitoring device and the second visual monitoring device;
[0023] Step S3: Match and compare the captured 3D surface point cloud with the reference surface point cloud to obtain real-time position error information and make adjustments.
[0024] As a preferred embodiment, step S1 includes:
[0025] Step S1.1: Place the calibration board vertically under the multi-vision monitoring device, and each first vision monitoring device and second vision monitoring device can capture images of the corresponding calibration board;
[0026] Step S1.2: Select a camera from either the first or second visual monitoring device as the main camera, and the remaining cameras as secondary cameras;
[0027] Step S1.3: Perform perspective deviation correction on each camera, extract the corresponding marker points on the calibration board after correction, and calculate the mapping relationship between the sub-camera pixel coordinate system and the main camera pixel coordinate system.
[0028] Step S1.4: After all the secondary cameras have acquired images, convert them to the main camera's pixel coordinate system using a mapping relationship;
[0029] Step S1.5: After coordinate system one, perform several high-precision calibrations on the main camera, and obtain the optimal calibration parameters through nonlinear optimization.
[0030] As a preferred option, one of the cameras of the second visual monitoring device is selected as the main camera, and the cameras of the remaining first and second visual monitoring devices are used as secondary cameras.
[0031] As a preferred approach, in step S2, the SIFT algorithm is used to extract features from the patient's three-dimensional body surface point cloud.
[0032] As a preferred option, in step S3, the point cloud stitching and registration are performed using the point-to-plane ICP algorithm.
[0033] This invention discloses an accelerometer-based multi-vision monitoring device and an optical body surface positioning method, which have the following beneficial effects:
[0034] First, this invention discloses a multi-vision detection device, in which one or more sets of second vision monitoring devices are installed on a fixed bracket. Using the camera of the second vision monitoring device as a reference allows for pre-calibration, resulting in more stable monitoring of the treatment space; it also saves on ceiling installation costs; and since the second vision monitoring device moves with the accelerometer to a new machine room, it can be directly installed onto the rack mounting interface without further adjustment, saving installation time.
[0035] Second, multi-vision monitoring devices can add multi-scale and multi-view information to the body surface point cloud, enabling the ICP algorithm to escape local optima and achieve more accurate radiotherapy positioning. Attached Figure Description
[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.
[0037] Figure 1 A side view of an accelerometer-based multi-vision monitoring device provided in an embodiment of the present invention.
[0038] Figure 2The front view of the accelerometer-based multi-vision monitoring device provided in an embodiment of the present invention.
[0039] Figure 3 This is a top view of an accelerometer-based multi-vision monitoring device provided in an embodiment of the present invention.
[0040] Wherein: 1-rotating bracket, 2-fixed bracket, 3-mounting bracket, 41-first visual monitoring device, 42-second visual monitoring device. Detailed Implementation
[0041] The preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings.
[0042] 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 skilled in the art without creative effort are within the scope of protection of the present invention.
[0043] Using ordinal numbers such as “first,” “second,” “third,” etc. to describe ordinary objects merely indicates different instances of similar objects and is not intended to imply that the objects being described must have a given order in time, space, sequence, or any other way.
[0044] Furthermore, the expression "includes" is an "open-ended" expression, which only means that the corresponding component exists and should not be interpreted as excluding additional components.
[0045] To achieve the objectives of this invention, some embodiments of an accelerometer-based multi-vision monitoring device and optical body surface positioning method are provided, such as... Figure 1-3 As shown, the multi-vision monitoring device includes: one or more sets of first vision monitoring devices 41 mounted on the rotating support 1 of the accelerator and one or more sets of second vision monitoring devices 42 mounted on the fixed support 2 of the accelerator.
[0046] The first visual monitoring device 41 can rotate with the rotating bracket 1;
[0047] The second visual monitoring device 42 is fixed in place and is used to monitor and track the posture of the treatment bed and the rotating support 1.
[0048] The installation angle and number of the second visual monitoring device 42 can be adjusted according to the actual situation, and are not limited in this invention.
[0049] The aforementioned first visual monitoring device 41 and second visual monitoring device 42 are composed of a grating projector, a camera, a power supply component, a communication device, etc.
[0050] This invention can monitor the treatment space and correct the positioning of the treatment bed and the patient; it allows direct observation of the patient's condition, especially the condition of the patient's body surface, without being disturbed by the rotation of the rotating frame, laying a foundation for subsequent development of body surface condition tracking; it can optimize the movement trajectory within the treatment range and avoid collisions between the patient, doctor, and other personnel on the treatment bed and the various components on the frame.
[0051] In order to further optimize the implementation effect of the present invention, in some other embodiments, the remaining features are the same, except that the second visual monitoring device 42 is connected to the fixed bracket 2 through the mounting bracket 3.
[0052] The mounting bracket 3 can be any form, such as fixed, manually adjustable, or electrically adjustable, and has a large degree of freedom of movement. It is not limited in this invention.
[0053] To further optimize the implementation effect of the present invention, in some other embodiments, the remaining features are the same, except that the data line or power line of the second visual monitoring device 42 is integrated and installed on the mounting bracket 3 and / or the fixed bracket 2.
[0054] The above installation method has a high degree of integration. The installation and power supply wiring of the second visual monitoring device 42 are integrated on a fixed frame or a rotating frame, saving installation and calibration time as well as decoration costs.
[0055] This invention also discloses an optical body surface positioning method, which utilizes the aforementioned multi-vision monitoring device for positioning, including:
[0056] Step S1: Perform joint calibration of the first visual monitoring device and the second visual monitoring device;
[0057] Step S2: Capture the patient's three-dimensional body surface point cloud using the first visual monitoring device and the second visual monitoring device;
[0058] Step S3: Match and compare the captured 3D surface point cloud with the reference surface point cloud to obtain real-time position error information and make adjustments.
[0059] By using a fixed rig to acquire more comprehensive surface point clouds, the problem of incomplete 3D point cloud data due to occlusion in single-view 3D surface reconstruction can be solved. It has the advantages of comprehensive imaging with a wide field of view, and can also acquire complete point cloud information for each field of view by moving and rotating the rig. However, the point cloud data acquired by the camera at different poses of the same object is based on the current camera coordinate system. Excessive differences in pose information between the point cloud data will lead to a decrease in subsequent registration efficiency.
[0060] This invention transforms point clouds obtained from various viewpoints into a fixed camera coordinate system, making the poses of each point cloud basically similar.
[0061] Specifically, step S1 includes:
[0062] Step S1.1: Place the calibration board vertically under the multi-vision monitoring device, and each first vision monitoring device and second vision monitoring device can capture images of the corresponding calibration board;
[0063] Step S1.2: Select one of the cameras of the second visual monitoring device as the main camera, and the cameras of the remaining first visual monitoring device and second visual monitoring device as secondary cameras;
[0064] Step S1.3: Perform perspective deviation correction on each camera, extract the corresponding marker points on the calibration board after correction, and calculate the mapping relationship between the sub-camera pixel coordinate system and the main camera pixel coordinate system.
[0065] Step S1.4: After all the secondary cameras have acquired images, they are converted to the pixel coordinate system of the main camera through mapping, which is equivalent to infinitely expanding the field of view of the main camera.
[0066] Step S1.5: After coordinate system one, perform several high-precision calibrations on the main camera, and obtain the optimal calibration parameters through nonlinear optimization.
[0067] Select one of the cameras of the second visual monitoring device as the main camera, and the cameras of the remaining first and second visual monitoring devices as secondary cameras.
[0068] Using the camera of the second vision monitoring device on the fixed frame as a constant reference, the conditions on the rotating frame are calibrated, such as the angle of the rotating frame and the camera status of the first vision monitoring device on the rotating frame.
[0069] Using a fixed frame to install a second visual monitoring device as a benchmark facilitates advance calibration and provides more stable monitoring of the treatment space.
[0070] Point clouds acquired by multi-vision inspection devices contain hundreds of thousands of points, while ICP searches the entire point cloud when matching point pairs. Furthermore, the presence of noise in the point cloud can easily lead to matching errors; therefore, an excessive number of point clouds significantly reduces registration efficiency. While commonly used voxel filtering methods can effectively reduce the number of point clouds, they do not fully preserve the features of the point cloud itself, thus reducing registration efficiency and accuracy.
[0071] The feature points extracted by the SIFT algorithm have good invariance to noise, rotation, and translation. Its main principle is to search for feature points that are invariant to scale and rotation in different scale spaces of the target point set.
[0072] For a point cloud dataset P with N points, the coordinates of any point pi are (xi, yi, zi), i = 1, 2, ..., N.
[0073] The steps of the SIFT feature point extraction algorithm are as follows:
[0074] (1) Construct scale space: Convolve the point cloud with a 3D Gaussian function to calculate the scale space L(x,y,z,σ).
[0075]
[0076] Where G(x,y,z,kσ) is the Gaussian convolution kernel function, p(x,y,z) is the point cloud coordinates, σ is the spatial scale parameter, and k is the scale space size adjustment parameter.
[0077] The Gaussian convolution kernel function in three-dimensional space is as follows:
[0078]
[0079] (2) Calculate the Gaussian difference model: Let s be the number of layers in each group of the Gaussian pyramid. Subtract the point clouds of two adjacent layers in the Gaussian pyramid to form the Gaussian difference map DoG of the point cloud, as shown in the following formula:
[0080] DoG i =L i (x,y,z,kσ)-L i-1 (x,y,z,kσ)
[0081] Where DoGi is the difference model between adjacent scale layers, i∈(0,s+2).
[0082] (3) Feature point detection: Calculate DoG i If a point is an extreme point within a given layer and is still considered an extreme point when compared with the extreme points in the two layers above and below it, then that point is considered a feature point.
[0083] When dealing with locally curved structures on point cloud surfaces, the ICP algorithm is prone to numerous mismatches during the matching process, causing the objective function to get trapped in local optima. A point-to-plane matching method, which represents the correspondences using the distance from a point to a tangent plane, improves this problem, better reflects the spatial structure of the point cloud, is more resistant to erroneous point pairs, and achieves faster iterative convergence. The standard ICP algorithm uses the distance error between points as the objective function for iterative solution.
[0084] The point-to-plane ICP algorithm uses minimizing the distance from a point in the source point cloud to the plane containing the corresponding point in the target point cloud as the registration criterion, as shown in the following equation:
[0085]
[0086] Where, n i For the target point cloud q i The unit normal vector at that location;
[0087] n p n q These represent the number of points in the source point cloud and the target point cloud, respectively.
[0088] R is the point cloud rotation transformation matrix; T is the point cloud translation vector;
[0089] p i and q i These are points in the source point cloud and the target point cloud, respectively.
[0090] After precise stitching using the point-to-surface ICP algorithm, an accurate registration matrix can be obtained, which transforms each point cloud to be stitched into the coordinate system of the target point cloud (i.e., the fixed camera coordinate system) and completes the point cloud stitching.
[0091] A key step in the positioning algorithm is the ICP iterative solution to the registration error. ICP (Iterative Closest Point) is a common technique in point cloud registration. Point cloud registration typically refers to aligning multiple point cloud datasets into a common coordinate system, making them spatially coincident. The ICP algorithm iteratively optimizes transformation parameters (including rotation and translation) to achieve optimal alignment between point clouds. However, the ICP algorithm is affected by matching noise and outliers, and may converge to a local optimum. This invention can obtain more accurate and comprehensive surface point cloud information, thus solving the problem of the ICP algorithm easily getting trapped in local optima.
[0092] This invention discloses an accelerometer-based multi-vision monitoring device and an optical body surface positioning method, which have the following beneficial effects:
[0093] First, this invention discloses a multi-vision detection device, in which one or more sets of second vision monitoring devices are installed on a fixed bracket. Using the camera of the second vision monitoring device as a reference allows for pre-calibration, resulting in more stable monitoring of the treatment space; it also saves on ceiling installation costs; and since the second vision monitoring device moves with the accelerometer to a new machine room, it can be directly installed onto the rack mounting interface without further adjustment, saving installation time.
[0094] Second, multi-vision monitoring devices can add multi-scale and multi-view information to the body surface point cloud, enabling the ICP algorithm to escape local optima and achieve more accurate radiotherapy positioning.
[0095] The above embodiments are only for illustrating the technical concept and features of the present invention, and are intended to enable those skilled in the art to understand the content of the present invention and implement it. They should not be used to limit the scope of protection of the present invention. All equivalent changes or modifications made in accordance with the spirit and essence of the present invention should be covered within the scope of protection of the present invention.
Claims
1. An accelerometer-based multi-vision monitoring device, characterized in that: include: One or more sets of first visual monitoring devices mounted on a rotating support of the accelerator and one or more sets of second visual monitoring devices mounted on a fixed support of the accelerator; The first visual monitoring device is capable of rotating with the rotating bracket; The second visual monitoring device is fixed and is used to monitor and track the posture of the treatment bed and the rotating bracket.
2. The multi-vision monitoring device according to claim 1, characterized in that: The second visual monitoring device is connected to the fixing bracket via a mounting bracket.
3. The multi-vision monitoring device according to claim 2, characterized in that: The data line or power line of the second visual monitoring device is integrated and installed on the mounting bracket and / or the fixing bracket.
4. An optical body surface positioning method, characterized in that: Positioning using the multi-vision monitoring device according to any one of claims 1 to 3 comprises: Step S1: jointly calibrating the first visual monitoring device and the second visual monitoring device; Step S2: using the first visual monitoring device and the second visual monitoring device to capture the three-dimensional body surface point cloud of the patient; Step S3: Match and compare the captured three-dimensional body surface point cloud with the reference body surface point cloud to obtain real-time position error information and make adjustments.
5. The optical body surface positioning method according to claim 4, characterized in that: The step S1 comprises: Step S1.1: vertically place the calibration plate under the multi-vision monitoring device, and each first vision monitoring device and second vision monitoring device can photograph the corresponding calibration plate; Step S1.2: selecting a camera of the first visual monitoring device or the second visual monitoring device as a main camera, and the remaining cameras as auxiliary cameras; Step S1.3: correct the perspective deviation of each camera, extract the identification points on the calibration plate corresponding to the correction, and obtain the mapping relationship between the pixel coordinate system of the secondary camera and the pixel coordinate system of the main camera; Step S1.4: After all the secondary cameras have collected the images, they are converted to the pixel coordinate system of the main camera through mapping relationships; Step S1.5: After the coordinate system is established, the main camera is calibrated several times with high precision, and the optimal calibration parameters are obtained through nonlinear optimization.
6. The optical body surface positioning method according to claim 5, characterized in that: The camera of one of the second visual monitoring devices is selected as the main camera, and the cameras of the remaining first visual monitoring devices and second visual monitoring devices are selected as auxiliary cameras.
7. The optical body surface positioning method according to claim 4, characterized in that: In step S2, the SIFT algorithm is used to extract features from the patient's three-dimensional body surface point cloud.
8. The optical body surface positioning method according to claim 4, characterized in that: In step S3, the point cloud is stitched and registered using the point-to-plane ICP algorithm.
