Pre-calibration method for full-length multi-height multi-SID biplane ray system

Abstract

The invention relates to the field of medical equipment, and discloses a pre-calibration method of a full-length multi-height multi-SID biplane ray system, which comprises the following steps: setting a spiral steel ball die body and an arrangement method of steel balls on the spiral steel ball die body, setting any spiral steel ball die body matched with a self-calibration algorithm based on beam concentration adjustment, and matching with automatic exposure software for calibration. According to the arrangement method, large and small steel balls with two sizes are adopted, and the large and small steel balls are regularly arranged, so that the shot local steel balls with any length are uniquely positioned; a shooting exposure process is performed and controlled through the automatic exposure software, calibration image data are collected, image analysis and calibration calculation are performed through a self-calibration algorithm based on beam concentration adjustment, the calibration calculation comprises two stages of parameter estimation and parameter optimization, the self-calibration algorithm is used, the adaptability to different die bodies is expanded, and the calibration precision is improved. And meanwhile, a global optimization adjustment strategy of the biplane projection image under different SIDs, different heights and different die body placement angles is considered.

Description

Technical Field

[0001] The present invention relates to the field of medical equipment, and in particular to a pre-calibration method for a full-length multi-height multi-SID dual-plane ray system. Background Art

[0002] The geometric calibration of the X-ray system is a kind of geometric modeling of the spatial position relationship between the radiation source and the flat panel. It models the radiation source as a point light source and the flat panel as a spatial plane defined by four corner points. By shooting the calibration phantom, the geometric calibration algorithm can perform algorithmic analysis on the feature points on the calibration phantom in the X-ray image and calculate the spatial position of the radiation source and the flat panel. Many analysis and reconstruction algorithms based on X-ray images require the input of the spatial position of the radiation source and the flat panel. The accuracy of these positions directly affects the accuracy of these subsequent algorithms, so geometric calibration is an important prerequisite.

[0003] According to the requirements of subsequent algorithm scenarios, geometric calibration can be classified from simple to complex into single-height calibration and multi-height calibration. For example, the 6-DOF analysis technology of bone joint motion based on 2D-3D registration of dual-plane dynamic images only requires calibration results at a single height because it is only shot at a certain height; while for the two technologies of full-length X-ray of slit scanning and tomo reconstruction based on X-ray, since patients will be shot at any number of uncertain heights, calibration results at multiple heights are required. According to the calibration time, geometric calibration can be divided into pre-calibration and on-site calibration. Pre-calibration refers to the calibration work completed in advance by the algorithm engineer, and the results are stored in advance. When the subsequent algorithm needs the calibration results, they are extracted and used directly or after interpolation; while on-site calibration is to shoot the calibration model on-site before each image acquisition, and calculate the results on-site. Generally speaking, if the same accuracy can be guaranteed, pre-calibration is more efficient than on-site calibration because it simplifies the workflow. Because of its one-time calibration and multiple uses, it is more versatile. For example, for the above-mentioned dual-plane dynamic image 2D-3D registration bone joint motion 6-DOF analysis technology, if on-site calibration is used, the operator needs to place the calibration model and shoot before the patient takes the dynamic image that actually needs to be analyzed, and the results need to be calculated on-site; if the patient needs to analyze multiple parts at different heights, recalibration is required every time the height is switched. This will undoubtedly reduce the user experience and is contrary to the general perception of the convenience of scanning CT and MRI "just scan it when you come".

[0004] In the "Calibration of Stereo Radiography System for Radiostereometric Analysis Application", a water bottle with tantalum beads attached to the inner wall was used as a calibration object, and the calibration was performed again based on the design parameters of the French EOS company, but the improvement in the accuracy of 3D reconstruction of spatial points was limited. Tomo reconstruction also requires accurate calibration information. In order to expand the support of the calibration range, expanding the calibration phantom is a very direct and easy-to-think solution, but the existing technology simply elongates the commonly used cube calibration box design into a cuboid box design, which has two problems: the material and the number of holes will increase exponentially with the increase in height, and the cost of the cube calibration box is already expensive; the imaging range of the flat plate is limited, and it can only capture a part of the box at a time. How to determine which part of the box the captured image corresponds to is a key problem that needs to be solved urgently. All calibration algorithms, without exception, require a correct 3D-2D correspondence. If it is not possible to determine which steel beads project the 2D steel bead projection in the image, the calibration algorithm will not be able to proceed.

[0005] In the prior art, the closest to the present invention is the spiral phantom proposed in "Geometrical calibration for 3D x-ray imaging", such as Figure 1 As shown, its height is about the height of a detector plate. By using a spiral line to arrange the steel balls, the problem of overlapping projections of the steel balls can be effectively solved. At the same time, the correspondence between the projection and the steel balls is also very easy to determine: the height of the steel ball and the vertical coordinate of its projection on the image are one-to-one corresponding. However, due to the periodicity of the spiral line, a simple image Figure 1 The same method uses steel balls of the same size, but it is still not possible to determine which segment the steel ball in the captured local image of the phantom corresponds to. Summary of the invention

[0006] The purpose of the present invention is to solve the shortcomings of the above-mentioned prior art and to provide a pre-calibration method for a full-length multi-height multi-SID dual-plane ray system, which can cope with single-height and multi-height calibration, adaptive motion alignment, slit scanning, tomo reconstruction, CBCT reconstruction and other algorithm requirements. Compared with the general calibration algorithm that requires the three coordinates of the feature points of the calibration object, this calibration scheme does not require the three coordinates and can complete high-precision calibration.

[0007] A pre-calibration method for a full-length multi-height multi-SID dual-plane ray system, comprising:

[0008] A spiral steel ball phantom and a method for arranging the steel balls thereon are provided, wherein the arrangement method uses large and small steel balls, and regularly arranges the large and small steel balls so that a unique positioning of the local steel balls of any length photographed is achieved, and the spiral steel ball phantom is placed at the imaging space center of a dual-plane ray system, and the dual-plane ray system includes an anterior and lateral X-ray source and an ante-lateral flat plate;

[0009] Any spiral steel ball phantom can be calibrated with the automatic exposure software through the self-calibration algorithm based on bundle adjustment, including:

[0010] S100: controlling the shooting exposure process through the automatic exposure software, collecting calibration image data, and arranging the collected calibration image data into a 4-dimensional multidimensional array;

[0011] S200: input the image of the 4-dimensional multidimensional array, use the image recognition algorithm and K means algorithm to identify and classify each image, and output the image of the multidimensional array, the content of the multidimensional array is a combination of the sequence ID of the steel ball and the two-dimensional coordinates of its projection point in the image;

[0012] S300: Calibrate and calculate the multidimensional array image outputted from step S200 by using the self-calibration algorithm, and output the spatial positions of several sets of X-ray sources and flat panels, all in the same global coordinate system.

[0013] Furthermore, in setting the spiral steel ball phantom and the arrangement method of the steel balls thereon, a virtual projection computer simulation program is constructed to determine whether the arrangement scheme of the spiral steel ball phantom and the steel balls thereon is feasible by checking the virtual simulation image, which specifically includes:

[0014] After receiving the key manufacturing parameters of the spiral steel ball phantom, the virtual projection computer simulation program generates a virtual phantom model in three-dimensional space and uses a ray projection GPU parallel algorithm to render virtual images of the phantom at different SIDs and different heights;

[0015] The key manufacturing parameters of the spiral steel ball mold include the cylindrical height of the mold, the radius of the cylindrical bottom surface, the number of steel balls, the height interval of the steel balls, the rotation angle of adjacent steel balls, and the radius of the large and small steel balls.

[0016] Further, the arbitrary spiral steel ball phantom provided with the self-calibration algorithm based on the bundle adjustment further comprises:

[0017] The cluster adjustment strategy simultaneously considers the global optimization adjustment of the dual-plane projection images under different SIDs, different heights, and different phantom placement angles, and freely adds constraints based on the mechanical and physical properties of the machine. At the same time, after the calibration is completed by the self-calibration algorithm, the 3D coordinates of the steel ball on the phantom are output to calibrate the phantom itself.

[0018] Furthermore, in step S100, collecting calibration image data, and arranging the collected calibration image data into a 4-dimensional multidimensional array further includes:

[0019] S101: The spiral steel ball phantom is placed at the center of the imaging space. First, the SID is set to SID_1 through the automatic exposure software, the rack height is adjusted to 0, and a static frontal and lateral double shot is taken every fixed distance d. The final rack height is recorded as h. Then, the photography SID is adjusted to SID_2 and SID_3, and the above-mentioned static double shot image acquisition is performed at each distance d at SID_2 and SID_3 respectively;

[0020] S102: Rotate the spiral steel ball phantom three times, each time by 90°, and execute step S101 again after each rotation to collect images at four angles of 0°, 90°, 180° and 270° respectively;

[0021] S103: Arrange the collected image data into a 4-dimensional multidimensional array, which is 4×3×(h / d+1)×2, wherein 4 corresponds to 4 rotation angles, 3 corresponds to 3 SIDs, (h / d+1) is the number of exposure height positions, and 2 corresponds to two frontal and lateral images.

[0022] Furthermore, step S200 specifically includes:

[0023] Receive the collected 4-dimensional multidimensional array images as input, and use image recognition algorithm to identify the center and radius of the projected steel ball for each image;

[0024] The K means algorithm is used to perform binary classification on the steel balls identified in each image, and the classification results are arranged into a sequence, which is unique in the entire steel ball arrangement sequence. The shape of the output multidimensional array is the same as the input 4-dimensional multidimensional array, and the content of the array becomes a combination of the sequence ID of the steel ball and the two-dimensional coordinates of its projection point in the image.

[0025] Furthermore, in step S300, the image of the multidimensional array output in step S200 is calibrated by the self-calibration algorithm. The calibration calculation includes a parameter estimation stage and a parameter optimization stage. The final output result is 3×(h / d+1)×2 sets of spatial positions of X-ray sources and flat panels, where:

[0026] The parameter estimation stage includes estimating the three coordinates of the steel ball on the spiral steel ball phantom, the actual posture transformation of the spiral steel ball phantom at different placement angles, and the transformation in the SID direction, and using the above estimation as the initial value of the subsequent parameter optimization stage;

[0027] The parameter optimization stage uses an optimizer to optimize the estimation results of the parameter estimation stage.

[0028] Furthermore, in the parameter estimation stage, the coordinates of the reconstructed steel ball in each SID global coordinate system are first calculated, specifically including:

[0029] S301: Assuming that the lead screw is completely vertical, for each SID, using system design parameters, and according to the shooting height preset by the automatic exposure software, an ideal multi-SID, multi-height calibration environment is initialized;

[0030] S302: For each SID, the coordinates of the double-plane steel ball photographed at each height in the global coordinate system of the SID multi-height calibration environment are calculated using the triangulation method. The steel ball in the middle section of the spiral steel ball phantom will be photographed at multiple adjacent heights, so there are multiple coordinates of the steel ball in the calculation result. The average value of these coordinates is taken as the reconstructed coordinates of the steel ball. Since the spiral steel ball phantom is placed at a total of 4 rotation angles, a set of steel ball coordinates will be calculated for each angle and each SID, and the reconstructed steel ball coordinates in the global coordinate system of each SID are stored.

[0031] The parameter estimation stage further comprises:

[0032] S303: respectively calculating the transformation of each rotation angle of 90°, 180° and 270° of the spiral steel ball phantom to 0° under the same SID, and then averaging the same rotation angle of different SIDs to obtain an estimate of the transformation of the spiral steel ball phantom;

[0033] S304: Set the coordinates of the steel ball with an initial angle of 0° under SID_1 as local coordinates, align the coordinates of each other SID and rotation angle to the local coordinates, and take the average after alignment. The average result is considered to be an initial estimate of the local coordinate system;

[0034] S305: aligning the steel ball coordinates of different SIDs at the same rotation angle with the SID_1 steel ball coordinates at the rotation angle, taking the average and obtaining the relationship between the global coordinate system and the local coordinate system at each SID.

[0035] Furthermore, the parameter optimization stage specifically includes:

[0036] The optimized parameters include the aspect ratio and skew of the anteroposterior plate, the distance from the anteroposterior light source to the plate at different SIDs and heights, and the coordinates of the vertical projection of the light source under the plate, the transformation of the principal coordinate system of the dual-plane system with the local coordinate system at different SIDs and heights, the rotational posture transformation of the spiral steel ball phantom, and the globally unique anteroposterior plate transformation;

[0037] The optimization objective function is to minimize the square of the Euclidean distance between the two-dimensional point reprojected by the algorithm and the two-dimensional point detected on the real image at all angles, all SIDs, all heights and each viewing angle of the front and side positions.

[0038] Compared with the prior art, the present invention has the following beneficial effects:

[0039] (1) The method of the present invention is combined with automatic exposure software, and engineers do not need to perform any other operations except placing the calibration phantom, thereby improving image acquisition efficiency;

[0040] (2) The present invention can cope with all algorithm requirements such as single-height and multi-height calibration, adaptive motion registration, slit scanning, tomo reconstruction, CBCT reconstruction, etc. through the bundle adjustment strategy, and has wide applicability;

[0041] (3) By setting a self-calibration algorithm, the present invention does not require the three coordinates of the characteristic points of the calibration object, compared with the general calibration algorithm. On the contrary, the three coordinates of the steel ball on the calibration model are output, thereby completing high-precision calibration and expanding the adaptability to different models.

[0042] (4) The algorithm needs to input the three coordinates of the calibration object. Since it relies on the three coordinates of the object, if the thermal stability and rigidity of the material of the object are not good, its true coordinates will deviate from the three coordinates measured at a certain moment, which will cause the accuracy of the calibration algorithm to decrease. Whether using more stable materials or regularly measuring the three coordinates, it will increase manufacturing or later costs. This algorithm will calculate the three coordinates itself, completely eliminating the input of the three coordinate state of the object, greatly reducing R&D costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0044] In the attached picture:

[0045] Figure 1 It is a schematic diagram of a spiral phantom calibration structure in the prior art;

[0046] Figure 2 It is a flow chart of a pre-calibration method of a full-length multi-height multi-SID dual-plane ray system of the present invention;

[0047] Figure 3 This is a schematic diagram of the structure of a spiral steel ball mold body of the present invention. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0049] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an", "said" and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0050] The specific implementation of the present invention is described below with reference to the accompanying drawings and embodiments.

[0051] Example 1

[0052] like Figure 2 As shown, this embodiment provides a pre-calibration method for a full-length multi-height multi-SID dual-plane ray system, and the technical solution includes:

[0053] First, a spiral steel ball phantom and a method for arranging the steel balls thereon are provided. The arrangement method uses steel balls of two sizes, large and small, and arranges the large and small steel balls regularly so that a unique positioning of the local steel balls of any length captured is achieved. The spiral steel ball phantom is placed at the center of the imaging space of a dual-plane ray system, and the dual-plane ray system includes an X-ray source in the front and side positions and two flat plates in the front and side positions.

[0054] At the same time, the self-calibration algorithm based on bundle adjustment can be used to calibrate any spiral steel ball phantom that is set up with automatic exposure software, including:

[0055] S100: controlling the shooting exposure process through the automatic exposure software, collecting calibration image data, and arranging the collected calibration image data into a 4-dimensional multidimensional array;

[0056] S200: input the image of the 4-dimensional multidimensional array, use the image recognition algorithm and K means algorithm to identify and classify each image, and output the image of the multidimensional array, the content of the multidimensional array is a combination of the sequence ID of the steel ball and the two-dimensional coordinates of its projection point in the image;

[0057] S300: Calibrate and calculate the multidimensional array image outputted from step S200 by using the self-calibration algorithm, and output the spatial positions of several sets of X-ray sources and flat panels, all in the same global coordinate system.

[0058] Among them, in setting the spiral steel ball phantom and the arrangement method of the steel balls thereon, a virtual projection computer simulation program is constructed to quickly determine whether the arrangement scheme of the spiral steel ball phantom and the steel balls thereon is feasible by checking the virtual simulation image, specifically including:

[0059] After receiving the key manufacturing parameters of the spiral steel ball phantom, the virtual projection computer simulation program generates a virtual phantom model in three-dimensional space, and uses a ray projection GPU parallel algorithm to quickly render virtual images of the phantom at different SIDs and different heights;

[0060] The key manufacturing parameters of the spiral steel ball mold include the cylindrical height of the mold, the radius of the cylindrical bottom surface, the number of steel balls, the height interval of the steel balls, the rotation angle of adjacent steel balls, and the radius of the large and small steel balls.

[0061] By checking the virtual simulation images, you can quickly determine whether a design solution is feasible or compare the advantages and disadvantages of different solutions.

[0062] Specifically, in this embodiment, a spiral steel ball phantom is shown, such as Figure 3 As shown, the manufacturing parameters corresponding to the phantom are as follows:

[0063] Calibration column height: 1760;

[0064] Calibration column bottom radius: 150;

[0065] Total number of calibration column steel balls: 89;

[0066] Radius of small steel ball of calibration column: 2;

[0067] The radius of the large steel ball of the calibration column: 3.25;

[0068] Ball radius sequence (from low to high): 2 2 2 2 2 2 2 3.25 2 2 2 2 2 3.25 3.25 2 2 2 2 3.25 2 3.25 2 2 2 2 3.25 3.25 3.25 2 2 2 3.25 2 2 3.25 2 2 2 3.25 2 3.25 3.25 2 2 2 3.25 3.25 2 3.25 2 2 2 3.25 3.25 3.25 3.25 2 2 3.25 2 2 3.25 3.25 2 2 3.25 2 3.25 2 3.25 2 2 3.25 2 3.25 3.25 3.25 2 2 3.25 3.25 2 3.25 3.25 2 2 3.25 3.25;

[0075] The difference between the adjacent steel ball angles (°): 33.4884;

[0076] The distance between adjacent steel balls in the z direction: 20;

[0077] The calibration column guarantees the shortest steel ball sequence length that can be uniquely positioned: 7; Theoretical coordinates of the steel balls (from low to high):

[0078]

[0079]

[0080]

[0081]

[0082] Further, the arbitrary spiral steel ball phantom provided with the self-calibration algorithm based on the bundle adjustment further comprises:

[0083] The bundle adjustment strategy (Bundle Adjustment) simultaneously considers the global optimization adjustment of the dual-plane projection images under different SIDs, different heights, and different phantom placement angles, fully considers the mechanical and physical characteristics of the machine, and freely adds constraints. At the same time, after the calibration is completed by the self-calibration algorithm, the 3D coordinates of the steel ball on the phantom are output to calibrate the phantom itself.

[0084] In this embodiment, the algorithm does not require the accurate 3D coordinates of the calibration model as input. Instead, after the calibration is completed, it will output the 3D coordinates of the steel ball on the calibration model. Therefore, it is called a "self-calibration" algorithm, because it also calibrates the model itself. Compared with many calibration algorithms in the prior art that require accurate 3D coordinates of steel balls, self-calibration algorithms have huge advantages. This is because as the calibration model increases, the spatial range of the steel balls distributed on it will also increase, and the measurable space of a general 3D coordinate measurement machine is limited. Therefore, it is very difficult to measure the 3D coordinates of steel balls on a large model, and the cost is also very high. Using a self-calibration algorithm can undoubtedly expand the adaptability to different models.

[0085] The bundle adjustment strategy is essentially a strategy that considers the dual-plane projection images at different SIDs, different heights, and different phantom placement angles and performs global optimization and adjustment. This strategy can also fully consider the mechanical and physical characteristics of the machine and freely add constraints to make the final calibration result more accurate.

[0086] In step S100, collecting calibration image data, and arranging the collected calibration image data into a 4-dimensional multidimensional array further includes:

[0087] S101: placing the spiral steel ball phantom at the center of the imaging space, first setting the SID to SID_1 through the automatic exposure software, adjusting the rack height to h, taking a static frontal and lateral double shot every fixed distance d, then adjusting the photographic SID to SID_2 and SID_3, and performing the aforementioned static double shot image acquisition every distance d at SID_2 and SID_3 respectively;

[0088] S102: Rotate the spiral steel ball phantom three times, each time by 90°, and execute step S101 again after each rotation to collect images at four angles of 0°, 90°, 180° and 270° respectively;

[0089] S103: Arrange the collected image data into a 4-dimensional multidimensional array, which is 4×3×(h / d+1)×2, wherein 4 corresponds to 4 rotation angles, 3 corresponds to 3 SIDs, (h / d+1) is the number of exposure height positions, and 2 corresponds to two frontal and lateral images.

[0090] Specifically, in this embodiment, the spiral steel ball phantom is placed at the center of the imaging space, and the automatic exposure software first controls the SID to 1350mm, then the rack is raised to 1760mm, and every 40mm, a static front and side double shot is taken; then the SID is adjusted to 1500mm and 1800mm, and the aforementioned static double shot image acquisition every 40mm is performed at the two SIDs. Then the spiral column is rotated 90°, and the aforementioned 3 SIDs are executed, and the image exposure is 40mm apart under each SID. A total of 3 rotations are performed, and a total of four angles of 0°, 90°, 180° and 270° are taken. The collected image data is organized into a 4-dimensional multidimensional array, 4x3x45x2, where 4 corresponds to 4 angles, 3 corresponds to 3 SIDs, 45 corresponds to (1760 / 40+1) exposure height positions, and 2 corresponds to two front and side images.

[0091] Step S200 specifically includes:

[0092] Receive the collected 4-dimensional multidimensional array images as input, and use image recognition algorithm to identify the center and radius of the projected steel ball for each image;

[0093] The K means algorithm is used to perform binary classification on the steel balls identified in each image, and the classification results are arranged into a sequence, which is unique in the entire steel ball arrangement sequence. The shape of the output multidimensional array is the same as the input 4-dimensional multidimensional array, and the content of the array becomes a combination of the sequence ID of the steel ball and the two-dimensional coordinates of its projection point in the image.

[0094] Further, in step S300, the image of the multidimensional array output in step S200 is calibrated by the self-calibration algorithm. The calibration calculation includes a parameter estimation stage and a parameter optimization stage. The final output result is 3×(h / d+1)×2 sets of X-ray system light sources and spatial positions of the flat panel. In this embodiment, the result is 3×45×2 sets of X-ray system light sources and spatial positions of the flat panel, where 3 corresponds to 3 SIDs, 45 corresponds to 45 heights, and 2 corresponds to two sets of X-ray systems in the frontal and lateral positions.

[0095] The parameter estimation stage includes estimating the three coordinates of the steel ball on the spiral steel ball phantom, the actual posture transformation of the spiral steel ball phantom at different placement angles, and the transformation in the SID direction, and using the above estimation as the initial value of the subsequent parameter optimization stage;

[0096] The parameter optimization stage uses an optimizer to optimize the estimation results of the parameter estimation stage.

[0097] In the parameter estimation stage, the coordinates of the reconstructed steel ball in each SID global coordinate system are first calculated, specifically including:

[0098] S301: Assuming that the lead screw of the rack is completely vertical, for each SID, using system design parameters, and according to the shooting height preset by the automatic exposure software, an ideal multi-SID, multi-height calibration environment is initialized;

[0099] S302: For each SID, the coordinates of the double-plane steel ball photographed at each height in the global coordinate system of the SID multi-height calibration environment are calculated using the triangulation method. The steel ball in the middle section of the spiral steel ball phantom will be photographed at multiple adjacent heights, so there are multiple coordinates of the steel ball in the calculation result. The average value of these coordinates is taken as the reconstructed coordinates of the steel ball. Since the spiral steel ball phantom is placed at a total of 4 rotation angles, a set of steel ball coordinates will be calculated for each angle and each SID, and the reconstructed steel ball coordinates in the global coordinate system of each SID are stored.

[0100] Specifically, we organize the calculation results of S302 into the following two-dimensional table to facilitate the subsequent algorithm description and writing:

[0101] Deg_1(0°) Deg_2(90°) Deg_3(180°) Deg_4(270°) SID_1(1350) SID_2(1500) SID_3(1800)

[0102] Each cell in the table stores the SID and the steel ball coordinates reconstructed in the SID global coordinate system under the rotation angle combination.

[0103] The parameter estimation stage further comprises:

[0104] S303: Calculate the transformation of each rotation angle of 90° (Deg_2), 180° (Deg_3) and 270° (Deg_4) of the spiral steel ball phantom to 0° under the same SID, and then average the same rotation angles of different SIDs to obtain an estimate of the transformation of the spiral steel ball phantom;

[0105] S304: Set the coordinates of the steel ball with an initial angle of 0° under SID_1 (SID_1×Deg_1) as local coordinates, align the coordinates of each other SID and rotation angle to the local coordinates, and take the average after alignment. The average result is considered to be an initial estimate of the local coordinate system;

[0106] S305: aligning the steel ball coordinates of different SIDs at the same rotation angle with the SID_1 steel ball coordinates at the rotation angle, taking the average and obtaining the relationship between the global coordinate system and the local coordinate system at each SID.

[0107] Furthermore, the parameter optimization stage specifically includes:

[0108] The optimized parameters include the aspect ratio and skew of the anteroposterior plate, the distance from the anteroposterior light source to the plate at different SIDs and heights, and the coordinates of the vertical projection of the light source under the plate, the transformation of the principal coordinate system of the dual-plane system with the local coordinate system at different SIDs and heights, the rotational position transformation of the spiral steel ball phantom, and the globally unique anteroposterior plate transformation;

[0109] The optimization objective function is to minimize the square of the Euclidean distance between the two-dimensional points reprojected by the algorithm and the two-dimensional points detected on the real image at all angles, all SIDs, all heights and all viewing angles (front and side).

[0110] Obviously, those skilled in the art can make various changes and modifications to the invention without departing from the spirit and scope of the invention. Thus, if these modifications and variations of the invention fall within the scope of the claims of the invention and their equivalents, the invention is also intended to include these modifications and variations.

[0111] Finally, it should be noted that the above are only preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A pre-calibration method for a full-length multi-height multi-SID dual-plane ray system, characterized in that: include: A spiral steel ball phantom and a method for arranging the steel balls thereon are provided, wherein the arrangement method uses large and small steel balls, and regularly arranges the large and small steel balls so that a unique positioning of the local steel balls of any length photographed is achieved, and the spiral steel ball phantom is placed at the imaging space center of a dual-plane ray system, and the dual-plane ray system includes an anterior and lateral X-ray source and an ante-lateral flat plate; Any spiral steel ball phantom can be calibrated with the automatic exposure software through the self-calibration algorithm based on bundle adjustment, including: S100: controlling the shooting exposure process through the automatic exposure software, collecting calibration image data, and arranging the collected calibration image data into a 4-dimensional multidimensional array; S200: input the image of the 4-dimensional multidimensional array, use the image recognition algorithm and K means algorithm to identify and classify each image, and output the image of the multidimensional array, the content of the multidimensional array is a combination of the sequence ID of the steel ball and the two-dimensional coordinates of its projection point in the image; S300: Calibrate and calculate the multidimensional array image outputted from step S200 by using the self-calibration algorithm, and output the spatial positions of several sets of X-ray sources and flat panels, all in the same global coordinate system.

2. The pre-calibration method of the full-length multi-height multi-SID dual-plane ray system according to claim 1 is characterized in that: In the method for setting the spiral steel ball phantom and the arrangement of the steel balls thereon, a virtual projection computer simulation program is constructed, and the arrangement scheme of the spiral steel ball phantom and the steel balls thereon is judged to be feasible by checking the virtual simulation image, which specifically includes: After receiving the key manufacturing parameters of the spiral steel ball phantom, the virtual projection computer simulation program generates a virtual phantom model in three-dimensional space and uses a ray projection GPU parallel algorithm to render virtual images of the phantom at different SIDs and different heights; The key manufacturing parameters of the spiral steel ball mold include the cylindrical height of the mold, the radius of the cylindrical bottom surface, the number of steel balls, the height interval of the steel balls, the rotation angle of adjacent steel balls, and the radius of the large and small steel balls.

3. The pre-calibration method of the full-length multi-height multi-SID dual-plane ray system according to claim 1 is characterized in that: The arbitrary spiral steel ball phantom provided with the self-calibration algorithm based on the bundle adjustment further comprises: The cluster adjustment strategy simultaneously considers the global optimization adjustment of the dual-plane projection images under different SIDs, different heights, and different phantom placement angles, and freely adds constraints based on the mechanical and physical properties of the machine. At the same time, after the calibration is completed by the self-calibration algorithm, the 3D coordinates of the steel ball on the phantom are output to calibrate the phantom itself.

4. The pre-calibration method of the full-length multi-height multi-SID dual-plane ray system according to claim 1 is characterized in that: In step S100, collecting calibration image data, and arranging the collected calibration image data into a 4-dimensional multidimensional array further includes: S101: The spiral steel ball phantom is placed at the center of the imaging space. First, the SID is set to SID_1 through the automatic exposure software, the rack height is adjusted to 0, and a static frontal and lateral double shot is taken every fixed distance d. The final rack height is recorded as h. Then, the photography SID is adjusted to SID_2 and SID_3, and the above-mentioned static double shot image acquisition is performed at each distance d at SID_2 and SID_3 respectively; S102: Rotate the spiral steel ball phantom three times, each time by 90°, and execute step S101 again after each rotation to collect images at four angles of 0°, 90°, 180° and 270° respectively; S103: Arrange the collected image data into a 4-dimensional multidimensional array, which is 4×3×(h / d+1)×2, wherein 4 corresponds to 4 rotation angles, 3 corresponds to 3 SIDs, (h / d+1) is the number of exposure height positions, and 2 corresponds to two frontal and lateral images.

5. The pre-calibration method of the full-length multi-height multi-SID dual-plane ray system according to claim 4 is characterized in that: Step S200 specifically includes: Receive the collected 4-dimensional multidimensional array images as input, and use image recognition algorithm to identify the center and radius of the projected steel ball for each image; The K means algorithm is used to perform binary classification on the steel balls identified in each image, and the classification results are arranged into a sequence, which is unique in the entire steel ball arrangement sequence. The shape of the output multidimensional array is the same as the input 4-dimensional multidimensional array, and the content of the array becomes a combination of the sequence ID of the steel ball and the two-dimensional coordinates of its projection point in the image.

6. The pre-calibration method of the full-length multi-height multi-SID dual-plane ray system according to claim 5 is characterized in that: In step S300, the image of the multidimensional array output in step S200 is calibrated by the self-calibration algorithm. The calibration calculation includes a parameter estimation stage and a parameter optimization stage. The final output result is 3×(h / d+1)×2 sets of spatial positions of X-ray sources and flat panels, where: The parameter estimation stage includes estimating the three coordinates of the steel ball on the spiral steel ball phantom, the actual posture transformation of the spiral steel ball phantom at different placement angles, and the transformation in the SID direction, and using the above estimation as the initial value of the subsequent parameter optimization stage; The parameter optimization stage uses an optimizer to optimize the estimation results of the parameter estimation stage.

7. The pre-calibration method of the full-length multi-height multi-SID dual-plane ray system according to claim 6 is characterized in that: In the parameter estimation stage, the coordinates of the reconstructed steel ball in each SID global coordinate system are first calculated, specifically including: S301: Assuming that the lead screw is completely vertical, for each SID, using system design parameters, and according to the shooting height preset by the automatic exposure software, an ideal multi-SID, multi-height calibration environment is initialized; S302: For each SID, the coordinates of the double-plane steel ball photographed at each height in the global coordinate system of the SID multi-height calibration environment are calculated using the triangulation method. The steel ball in the middle section of the spiral steel ball phantom will be photographed at multiple adjacent heights, so there are multiple coordinates of the steel ball in the calculation result. The average value of these coordinates is taken as the reconstructed coordinates of the steel ball. Since the spiral steel ball phantom is placed at a total of 4 rotation angles, a set of steel ball coordinates will be calculated for each angle and each SID, and the reconstructed steel ball coordinates in the global coordinate system of each SID are stored.

8. The pre-calibration method of the full-length multi-height multi-SID dual-plane ray system according to claim 7 is characterized in that: The parameter estimation stage further comprises: S303: respectively calculating the transformation of each rotation angle of 90°, 180° and 270° of the spiral steel ball phantom to 0° under the same SID, and then averaging the same rotation angle of different SIDs to obtain an estimate of the transformation of the spiral steel ball phantom; S304: Set the coordinates of the steel ball with an initial angle of 0° under SID_1 as local coordinates, align the coordinates of each other SID and rotation angle to the local coordinates, and take the average after alignment. The average result is considered to be an initial estimate of the local coordinate system; S305: aligning the steel ball coordinates of different S IDs at the same rotation angle with the SID_1 steel ball coordinates at the rotation angle, taking the average and obtaining the relationship between the global coordinate system and the local coordinate system at each S ID.

9. The pre-calibration method of the full-length multi-height multi-SID dual-plane ray system according to claim 8, characterized in that: The parameter optimization stage specifically includes: The optimized parameters include the aspect ratio and skew of the anteroposterior plate, the distance from the anteroposterior light source to the plate at different SIDs and heights, and the coordinates of the vertical projection of the light source under the plate, the transformation of the principal coordinate system of the dual-plane system with the local coordinate system at different SIDs and heights, the position transformation of the spiral steel ball phantom rotation, and the globally unique anteroposterior plate transformation; The optimization objective function is to minimize the square of the Euclidean distance between the two-dimensional point reprojected by the algorithm and the two-dimensional point detected on the real image at all angles, all SIDs, all heights and each viewing angle of the front and side positions.

Images