Lung respiratory movement modeling method and related device

Abstract

The invention discloses a lung breathing movement modeling method and a related device. The method comprises the following steps: acquiring a 4D CT image data set of lung breathing movement; wherein the 4D CT image comprises 3D CT images at a plurality of moments, and the 3D CT images are composed of 2D CT images at different height positions; performing contrast standardization and de-noising processing on the 4D CT image to highlight a lung region with a specific geometric structure; segmenting the lung region to obtain the centroid of each segmented region, and extracting landmark points at corresponding positions; constructing a lung respiratory movement model according to the landmark points; and matching the landmark points extracted from the new CT image with the landmark points in the lung respiratory motion model to obtain motion state information of the landmark points in the new CT image. According to the invention, the global motion field and the local motion field of the lung can be processed, the calculation is lightweight, and the time for obtaining the final motion image is short.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to a lung respiratory motion modeling method and device, computing equipment, and computer program product. Background Art

[0002] In the field of medical imaging, particularly respiratory motion imaging, 4D CT (four-dimensional computed tomography) offers important new insights. By capturing multiple CT images at different time intervals during the respiratory cycle, 4D CT images can demonstrate the dynamic changes in the lungs during respiration, making them crucial for analyzing the impact of respiratory motion on lung diseases such as lung cancer. To more accurately describe respiratory motion, the concept of a deformable vector field (DVF) is introduced. The DVF contains important information about the local variations in the total lung displacement caused by respiratory motion. By deriving a respiratory motion model from 4D CT lung images and using the DVF as output, researchers can gain a deeper understanding of the characteristics of respiratory motion.

[0003] Most current DIRART (Direct Image Registration Technology) methods focus primarily on local changes while ignoring the continuity of global motion. This leads to errors in estimating lung displacement caused by respiratory motion, thus limiting the accuracy and reliability of respiratory motion models. Furthermore, motion artifacts caused by irregular patient breathing, equipment noise, or other factors can degrade imaging quality. These artifacts mask or distort the true lung motion during respiration, thus affecting the accurate assessment of respiratory motion.

[0004] Therefore, the present invention further improves the accuracy and reliability of 4D CT and DVF in respiratory motion imaging, including reducing the impact of motion artifacts, improving the accuracy and efficiency of image registration, and more deeply exploring the global motion pattern of the lungs during breathing to more accurately evaluate respiratory motion. Summary of the Invention

[0005] In view of the above problems, the present invention is proposed to provide a lung respiratory motion modeling method and apparatus, computing equipment, and computer program product that overcome the above problem of low respiratory motion assessment accuracy.

[0006] According to one aspect of the present invention, a method for modeling lung respiratory motion is provided, comprising: Acquire a 4D CT image dataset of lung respiratory motion; wherein the 4D CT image includes 3D CT images at multiple moments, and the 3D CT image is composed of 2D CT images at different height positions; performing contrast normalization and denoising on the 4D CT image to highlight lung regions with specific geometric structures; Segmenting the lung region to obtain the centroids of each segmented region and extracting landmark points at corresponding positions; constructing a lung respiratory motion model according to the landmark points; The landmark points extracted from the new CT image are matched with the landmark points in the lung respiratory motion model to obtain motion state information of the landmark points in the new CT image.

[0007] In an optional manner, the method further includes: Generate corresponding 3D point cloud data based on the landmark points at each breathing moment; The 3D point cloud data is subjected to correlation analysis, and only coordinate points with correlation within a preset range are retained to obtain sparse landmark points that exclude artifacts.

[0008] In an optional manner, the formula for calculating the displacement vector field of the lung respiratory motion model is: in, is the j component of the displacement vector field at position x, is the displacement vector between the landmark p and the position x, is the weight associated with landmark p, n is the number of landmarks, is the optimal change parameter.

[0009] In an optional manner, the optimal change parameter The optimization function is: in, For The relevant transformation function is is the regularization term.

[0010] In an optional manner, after extracting the landmark points at the corresponding positions, the method further includes: By constructing the affine transformation group Describes the rotation and translation changes of the lungs at different breathing moments; is the rotation matrix, is the translation vector, is a single parameter space, is a 3×3 reversible matrix, .

[0011] In an optional manner, the rotation matrix for: in, , , Respectively represent axis, Axis and The rotation angle of the axis.

[0012] In an optional manner, the method further includes: By finding the parameters Minimize the image difference between adjacent respiratory moments; The optimal parameters are obtained by iterative solution based on the Gauss-Newton method, and the difference function is processed by linearization. Speed up the iteration process and use the Jacobi matrix Compute gradients and parameter updates; where, , the linearized function is , , Represents the residual between the true value and the simulated value of each point, is the actual point cloud coordinate; Generate the motion change field of the lungs according to the scaling method.

[0013] According to another aspect of the present invention, there is provided a lung respiratory motion modeling device, comprising: A data acquisition module, configured to acquire a 4D CT image dataset of lung respiratory motion; wherein the 4D CT image includes 3D CT images at multiple moments, and the 3D CT image is composed of 2D CT images at different height positions; a data preprocessing module, configured to perform contrast normalization and denoising on the 4D CT image to highlight a lung region with a specific geometric structure; A landmark acquisition module is used to segment the lung area to obtain the centroid of each segmented area and extract landmark points at corresponding positions; A model building module, used for building a lung respiratory motion model according to the landmark points; The matching module is used to match the landmark points extracted from the new CT image with the landmark points in the lung respiratory motion model to obtain motion state information of the landmark points in the new CT image.

[0014] According to another aspect of the present invention, there is provided a computing device, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus; According to yet another aspect of the present invention, a computer program product is provided, comprising at least one executable instruction, wherein the executable instruction enables the processor to execute operations corresponding to the above-mentioned lung respiratory motion modeling method.

[0015] According to the solution provided by the present invention, a 4D CT image dataset of lung respiratory motion is obtained; wherein the 4D CT image includes 3D CT images at multiple moments, and the 3D CT image is composed of 2D CT images at different height positions; the 4D CT image is subjected to contrast normalization and denoising to highlight lung regions with specific geometric structures; the lung regions are segmented to obtain the centroids of each segmented region, and landmark points at corresponding positions are extracted; a lung respiratory motion model is constructed based on the landmark points; and the landmark points extracted from the new CT image are matched with the landmark points in the lung respiratory motion model to obtain motion state information of the landmark points in the new CT image. The present invention can process global and local motion fields of the lungs, is computationally lightweight, and produces a final motion image in a short time.

[0016] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are specifically listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings: Figure 1 A schematic flow chart of a lung respiratory motion modeling method according to an embodiment of the present invention is shown; Figure 2 A schematic diagram of a motion image generation process according to an embodiment of the present invention is shown; Figure 3 A schematic diagram of a lung image containing artifacts according to an embodiment of the present invention is shown; Figure 4 A schematic diagram showing a segmented lung region according to an embodiment of the present invention is shown; Figure 5 A schematic diagram of a point cloud image according to an embodiment of the present invention is shown; Figure 6 A schematic diagram of a point cloud image (yellow points are corresponding landmark points) according to an embodiment of the present invention is shown; Figure 7 A schematic diagram of a scaling algorithm according to an embodiment of the present invention is shown; Figure 8 A schematic diagram showing a comparison of corresponding angles in an embodiment of the present invention is shown; Figure 9 A schematic diagram showing a comparison of lengths of corresponding vectors according to an embodiment of the present invention is shown; Figure 10 A schematic diagram of a vector field formula according to an embodiment of the present invention is shown; Figure 11 A schematic diagram of vector field visualization according to an embodiment of the present invention is shown; Figure 12 A schematic diagram showing a framework of a lung respiratory motion modeling device according to an embodiment of the present invention is shown; Figure 13 A schematic structural diagram of a computing device according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0018] Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0019] Figure 1 FIG. 4 is a flow chart showing a method for modeling lung respiratory motion according to an embodiment of the present invention. Figure 1 As shown, the following steps are included: Step S101 : Acquire a 4D CT image dataset of lung respiratory motion; wherein the 4D CT image includes 3D CT images at multiple moments, and the 3D CT image is composed of 2D CT images at different height positions.

[0020] like Figure 2 As shown in Figure 2, a 4D CT image dataset of lung respiratory motion (including breathing patterns such as free breathing, shallow breathing, and breath holding) was acquired from multiple subjects using CT scans. The 4D CT images consist of 3D CT images at multiple time points (each CT scan at a time point produces a complete 3D CT image). 3D CT images are typically composed of 2D CT images taken at different heights.

[0021] Step S102 : performing contrast normalization and denoising processing on the 4D CT image to highlight the lung region with a specific geometric structure.

[0022] In this embodiment, contrast normalization is to improve the contrast between different tissues in the image, making the structure and features of the lung area more prominent. This can be achieved through methods such as histogram equalization or contrast stretching. In 4D CT images, noise may come from many aspects (such as patient movement, etc.). The image can be smoothed using methods such as Gaussian kernel function, median filtering, and wavelet denoising to reduce the impact of noise. Figure 3As shown, after contrast normalization and denoising are performed on each layer of the image, parts of the lungs with special geometric shapes, such as the trachea, are better highlighted. Figure 3 The unit of the coordinate axis is millimeter.

[0023] Step S103 , segmenting the lung region to obtain the centroids of each segmented region, and extracting landmark points at corresponding positions.

[0024] The purpose of lung region segmentation is to distinguish the lungs from other tissues (such as the chest wall, ribs, heart, etc.). The image can be segmented by methods such as threshold segmentation, region growing, and edge detection. After obtaining the segmentation results of the lung region, the centroid of each segmented region is calculated. The centroid is the average coordinate value of all pixels in the region. Each segmented region corresponds to a centroid. For each segmented region in a three-dimensional image, the centroid coordinates of the region are obtained by calculating the average coordinate value in the X, Y, and Z directions. Among them, the centroid can represent a landmark point, and the centroid coordinates can represent the coordinates of the landmark point. The landmark point describes a specific position or structure of the lung region (for example, the apex of the lung at the top of the lung, the bronchial bifurcation point at the intersection of the left and right main bronchi, etc.).

[0025] Figure 4 FIG. 4 shows a schematic diagram of a segmented lung region according to an embodiment of the present invention. Figure 4 As shown, Figure 4 The left side of the middle view may represent a cross-section of the chest. Figure 4 The view on the right side of the figure can represent the segmentation result. Figure 4 The units of the horizontal and vertical axes are both millimeters.

[0026] In an optional embodiment, the method further includes: Generate corresponding 3D point cloud data based on the landmark points at each breathing moment; The 3D point cloud data is subjected to correlation analysis, and only coordinate points with correlation within a preset range are retained to obtain sparse landmark points that exclude artifacts.

[0027] like Figure 4 、 Figure 5 As shown, the corresponding 3D point cloud data is generated according to the landmark points at each breathing moment, and the point cloud coordinates at each moment are subjected to point-to-point correlation analysis, and only the coordinate points with correlation within a preset range (such as 0.8 to 1) are retained. For example, the correlation between any two landmark points at adjacent moments can be analyzed. The remaining landmark points are not only sparse landmark points, but also landmark points that exclude the influence of artifacts, which more accurately reflects the morphology and position changes of the lungs during respiratory movement. Among them, Figure 5 The units of the coordinate axes are all millimeters.

[0028] Step S104: constructing a lung respiratory motion model based on the landmark points.

[0029] A lung respiratory motion model is built based on the motion trajectory of landmark points and the characteristics of lung respiratory motion, which can accurately calculate the motion of landmark points and explain the main characteristics of lung respiratory motion.

[0030] The temporal movement of the lungs during respiratory relaxation can be viewed as a single-parameter group with time as the variable. Consequently, the complete respiratory movement of the lungs from inhalation to exhalation can be viewed as a single-parameter Lie group variation. According to Lie group theory, a continuous motion field can be characterized by a generator that describes the changes in that field over a small time period. This minimum generator is the Lie algebra corresponding to the Lie group. In the lung respiratory motion described in the embodiments of this specification, the Lie algebra is specifically a three-dimensional displacement vector field.

[0031] As a specific implementation, the features of lung respiratory motion can be obtained based on landmark points in 4D CT images at adjacent moments, such as lung expansion features and lung contraction features.

[0032] Furthermore, lung respiratory motion is the rhythmic expansion and contraction of the thorax caused by the contraction and relaxation of respiratory muscles. This includes cyclical movements of the lungs, including 360-degree outward expansion, 360-degree inward contraction, slight translation, and up-and-down displacement. Therefore, a lung respiratory motion model requires determining the three-dimensional coordinates of each lung landmark. The set of differences between two landmark points is the displacement vector field.

[0033] As a specific implementation, an affine transformation group may be constructed based on the characteristics of the lung respiratory motion to obtain the motion trajectory of the landmark point.

[0034] In the embodiment of this specification, the lung breathing motion model may be a model that describes the motion trajectory of a landmark point.

[0035] In an optional embodiment, the formula for calculating the displacement vector field of the lung respiratory motion model is: in, is the j component of the displacement vector field at position x, is the displacement vector between the landmark p and the position x, is the weight associated with landmark p, n is the number of landmarks, is the optimal change parameter.

[0036] In an optional embodiment, the formula for calculating the displacement vector field using the lung respiratory motion model may be: in, is the displacement vector field describing the general motion of the lungs, is the displacement vector field of the jth experimental patient object, is the evaluation weight of the jth experimental patient (for example, the proportion of the vital capacity of the jth experimental patient in the total vital capacity of the n experimental patients), n is the total number of experimental patients, is the optimal change parameter for the jth experimental patient; x is axis, axis, The axes form three-dimensional coordinate points on the three-dimensional coordinates, The positive axis points to the left side of the experimental patient subject. The positive axis points to the back of the experimental patient. The positive direction of the axis points to the coordinate point on the head of the experimental patient subject.

[0037] Different landmark points have different contributions to the displacement vector field (DVF) due to their different positions, importance or data quality in the lung structure. In this embodiment, the optimal change parameter μ (also known as the global scaling factor) is introduced to reduce the influence of the landmark points on the overall displacement vector field. and the optimal variation parameter μ provide an intuitive understanding of the extent to which different landmark points contribute to the displacement vector field.

[0038] Assume that the entire respiratory time of the patient during the 4D CT scan is T, ignore some small time interval errors that may be caused by the CT machine, and assume that the respiratory motion of the lung represented by the differential transformation is completed in a continuous time interval. Then, the entire motion transformation of the lung in the time interval T can be decomposed into 9 sub-time intervals between 10 3D CT images. The homeomorphic transformation can be expressed as Then the continuous change in time T can be broken down into the accumulation of changes in 9 specific sub-time intervals. It can represent the optimal change parameter μ.

[0039] Specifically, the diffeomorphic transformation can be expressed as: .

[0040] Where Ai and bi are the affine homeomorphic variation parameters at the i-th moment. Since 4D CT scans patient 1 in a preset equal time, it is assumed that all gap times are equal and equal to Then the transformation can be expressed as ,in , . This transformation can represent the motion trajectory of the landmark point. Where A and b are μ. Therefore, describing the changes in lung breathing throughout the respiratory cycle can be broken down into the optimal problem of affine changes between each specific CT image. That is, at each extremely small time interval, how to make the point at time ti become the point at time ti+1 under affine change, and achieve optimization by comparing it with the actual point cloud at time ti+1, and then obtain the optimal change parameter μ with the smallest gap between the actual point cloud and the estimated point cloud.

[0041] In an optional embodiment, the optimal change parameter The optimization function is: in, For The relevant transformation function is is the regularization term.

[0042] When optimizing the optimal change parameter μ, the above optimization function (Objective function, i.e. the square of L2 norm) finds the value of μ that minimizes the error. The error is the sum of the data point x and the transformation function The difference between the predicted values obtained is measured, and the regularization term Prevent overfitting. Data point x can be expressed as coordinate point x, represents the regularization function.

[0043] In practical applications, the lungs can be considered as a submanifold embedded in three-dimensional space, with each lung tissue considered a collection of points within this submanifold. Lung respiratory motion primarily involves contraction and expansion, a deformation type that can be viewed as the rotation and translation of each lung point over time. Therefore, lung respiratory motion can be specifically described as the rotation and translation of each lung point, and can be characterized using affine group transformations involving rotation and translation.

[0044] In an optional embodiment, after extracting the landmark points at the corresponding positions, the method further includes: By constructing the affine transformation group Describes the rotation and translation changes of the lungs at different breathing moments; is the rotation matrix, is the translation vector, is a single parameter space, It can also be expressed as an affine transformation group, which can represent the motion trajectory of landmark points, such as the rotation and translation changes of the lungs at different times. For example, it can represent the relative position change of a landmark point at a certain moment and the landmark point at the next moment. is a 3×3 reversible matrix, It can represent linear space, When describing the coordinates of landmark points, a coordinate system is introduced for the spatial location of the lungs. In the field of medical imaging, this can be a specific patient reference coordinate system (where the positive axis points to the patient's left side, the positive axis points to the patient's back, and the positive axis points to the patient's head). In this introduced coordinate system, a three-dimensional coordinate description can be found for each specific point in the lungs, which is recorded as x.

[0045] In this embodiment, based on the landmark points, a diffeomorphic registration method is used to treat the lung respiratory movement process as a single parameter group change acting on a three-dimensional subspace: This formula proves that the affine transformation group in the embodiment of this specification is a single-parameter group. For this group of changes, is a single parameter space, The 3D space where the lungs are located is essentially The corresponding time The invariant corresponding to the change of the single parameter group is the lung itself, so this change can be regarded as an affine transformation group. This embodiment describes the rotation and translation changes of the lungs at different breathing moments through an affine transformation group, which can accurately simulate the three-dimensional movement of the lungs during breathing. Among them, the affine transformation group consists of linear transformations (including rotation and scaling) and translations. The linearization of this radial change group can obtain a corresponding Lie algebra: The lung respiratory motion model can be constructed through this Lie algebra. are the corresponding 3D point coordinates, is the corresponding initial state, The coordinates of all points in the initial state can be expressed, and then the Lie algebra can be restored by processing the point cloud data to obtain the corresponding lung breathing motion change field. The radial change can be further deconstructed into rotation and translation motion. Represents the corresponding rotation change, for example, A can represent the rotation matrix. The rotation change in three-dimensional space can be reflected in the rotation relative to , , The rotation of three axes. Based on this, the rotation matrix can be expressed as: in, , , Respectively represent axis, Axis and The rotation angle of the axis.

[0046] In an optional embodiment, the method further includes: By finding the parameters Minimize the image difference between adjacent respiratory moments, such as the image difference between simulated values and actual values; The optimal parameters are obtained by iterative solution based on the Gauss-Newton method, and the difference function is processed by linearization. Speed up the iteration process and use the Jacobi matrix Compute gradients and parameter updates; where, , Indicates the coordinate value of the actual landmark point at a certain moment. Represents the coordinate value of the actual landmark point at the next moment of a certain moment. The linearized function is , , Represents the residual between the true value and the simulated value of each point, is the actual point cloud coordinate, is the Jacobian ratio matrix, is the parameter value at each iteration; Generate the motion change field of the lungs according to the scaling method.

[0047] In this embodiment, lung image registration is essentially to find the parameters Make the lungs three-dimensional image arrive The image difference is the smallest. That is, the actual Values and obtained under radiometric changes Minimize the difference between the values: , The Gauss-Newton method is used to help solve the specific parameter values. First, the difference is linearized as follows: In the continuous loop iteration, the parameters are finally generated Finally, the corresponding radiation changes are obtained by averaging the parameter values in all time domains. It can represent the parameter value of a certain moment in all the moments corresponding to the 4D CT image, and the parameter values of all time domains can represent the parameter values of all moments of the 4D CT image. Figure 7 After the scaling method shown in FIG. 3 is used, a corresponding motion transformation field is generated, where the motion transformation field can represent the motion trajectory of the landmark point.

[0048] Step S105 : matching the landmark points extracted from the new CT image with the landmark points in the lung respiratory motion model to obtain motion state information of the landmark points in the new CT image.

[0049] In the embodiments of this specification, the lung respiratory motion model may include motion state information corresponding to each landmark point. Optionally, landmark points in the new CT image may be matched with landmark points in the lung respiratory motion model to obtain landmark points that match the lung respiratory motion model, and the motion state information of the matched landmark points may be used as the motion state information of the landmark points in the new CT image.

[0050] The method in this embodiment is compared with the existing DIRART method on MATLAB. Since the elements in the vector field are three-dimensional vectors, the vectors of the marker motion corresponding to the corresponding lung tissue position are compared. Specifically, the motion vectors obtained by the method based on the embodiment of this specification and the existing DIRART method are compared. The first is the comparison of angles. Figure 8 It can be seen that the distribution of most vectors is between -20 degrees and 20 degrees, that is, the movement directions of the vector fields obtained by the two different methods are basically the same. Figure 8 In the figure, P related to the horizontal axis represents the angle, and the vertical axis represents the number of vectors. The second is the comparison of the lengths of the corresponding vectors. Figure 9 It can be seen that the distribution of the x, y, and z lengths of the two vector fields are basically the same. Figure 9 In the formula, Empirical Value can be the abbreviation of the existing method, NLLS can be the abbreviation of the method of the embodiment of this specification, mean can be the average coordinate, and std can be the coordinate variance. The specific vector field formula and its visualization are as follows: Figure 10 and Figure 11 As shown. Figure 10 The vector field formula of the embodiment method of this specification can be expressed as follows: Figure 10 Where X can represent a vector field, and V can represent vital capacity. Figure 11 It can represent the vector field at a certain moment obtained based on the method of the embodiment of this specification, where Figure 11 The units of each coordinate axis are millimeters. The yellow portion of the vector field may represent a landmark point with a larger motion trend, and the blue portion may represent a landmark point with a smaller motion trend. In the embodiment of this specification, the motion trajectory of the landmark point can be obtained based on the vector field at multiple moments.

[0051] According to the solution provided by the present invention, a 4D CT image dataset of lung respiratory motion is obtained; wherein the 4D CT image includes 3D CT images at multiple moments, and the 3D CT image is composed of 2D CT images at different height positions; the 4D CT image is subjected to contrast normalization and denoising to highlight lung regions with specific geometric structures; the lung regions are segmented to obtain the centroids of each segmented region, and landmark points at corresponding positions are extracted; a lung respiratory motion model is constructed based on the landmark points; the landmark points extracted from the new CT image are matched with the landmark points in the lung respiratory motion model to obtain motion state information of the landmark points in the new CT image. The present invention can process global motion fields and local motion fields of the lungs, such as the motion field of the entire lung and the motion field of a local area of the lungs, and the calculation is lightweight and the time to obtain the final motion image is short.

[0052] Specifically, the embodiments of this specification can not only accurately obtain the motion state information of landmark points in CT images, achieving the accuracy of the existing DIRART method, but also because the motion state information of landmark points in CT images can be obtained through the constructed lung respiratory motion model, the calculation is more lightweight than the existing methods, making it possible to quickly obtain the motion state information of landmark points.

[0053] Based on the same idea, the embodiments of this specification also provide a device corresponding to the above method.

[0054] Figure 12 FIG2 is a schematic diagram showing the structure of a lung respiratory motion modeling device according to an embodiment of the present invention. The lung respiratory motion modeling device includes: a data acquisition module 1210 , a data preprocessing module 1220 , a landmark acquisition module 1230 , a model construction module 1240 , and a matching module 1250 .

[0055] The data acquisition module 1210 is configured to acquire a 4D CT image dataset of lung respiratory motion; wherein the 4D CT image includes 3D CT images at multiple moments, and the 3D CT image is composed of 2D CT images at different height positions; The data preprocessing module 1220 is configured to perform contrast normalization and denoising on the 4D CT image to highlight the lung region with a specific geometric structure; The landmark acquisition module 1230 is configured to segment the lung region to obtain the centroid of each segmented region and extract landmark points at corresponding positions, including: calculating, for any segmented region, the average coordinates of all pixels in the segmented region, the average coordinates being the coordinates of the centroid of the segmented region, and the coordinates of the centroid of the segmented region being the coordinates of the landmark points of the segmented region; The model construction module 1240 is configured to construct a lung respiratory motion model based on the landmark points, including: obtaining lung respiratory motion features based on the landmark points in the 4D CT images at adjacent moments, constructing an affine transformation group based on the lung respiratory motion features, and obtaining motion trajectories of the landmark points, wherein the lung respiratory motion model is a model that describes the motion trajectories of the landmark points; The matching module 1250 is configured to match the landmark points extracted from the new CT image with the landmark points in the lung respiratory motion model to obtain motion state information of the landmark points in the new CT image.

[0056] In an optional manner, the device further includes: Generate corresponding 3D point cloud data based on the landmark points at each breathing moment; The 3D point cloud data is subjected to correlation analysis, and only coordinate points with correlation within a preset range are retained to obtain sparse landmark points that exclude artifacts.

[0057] In an optional manner, the formula for calculating the displacement vector field of the lung respiratory motion model is: in, is the j component of the displacement vector field at position x, is the displacement vector between the landmark p and the position x, is the weight associated with landmark p, n is the number of landmarks, is the optimal change parameter.

[0058] In an optional manner, the optimal change parameter The optimization function is: in, For The relevant transformation function is is the regularization term.

[0059] In an optional manner, after extracting the landmark points at the corresponding positions, the apparatus further includes: By constructing the affine transformation group Describes the rotation and translation changes of the lungs at different breathing moments; is the rotation matrix, is the translation vector, is a single parameter space, is a 3×3 reversible matrix, .

[0060] In an optional manner, the rotation matrix for: in, , , Respectively represent axis, Axis and The rotation angle of the axis.

[0061] In an optional manner, the device further includes: By finding the parameters Minimize the image difference between adjacent respiratory moments; The optimal parameters are obtained by iterative solution based on the Gauss-Newton method, and the difference function is processed by linearization. Speed up the iteration process and use the Jacobi matrix Compute gradients and parameter updates; where, , the linearized function is , , Represents the residual between the true value and the simulated value of each point, is the actual point cloud coordinate; Generate the motion change field of the lungs according to the scaling method.

[0062] Figure 13 The schematic diagram of the structure of the computing device embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computing device.

[0063] like Figure 13 As shown, the computing device may include: a processor (processor) 1302, a communication interface (Communications Interface) 1304, a memory (memory) 1306, and a communication bus 1308.

[0064] Processor 1302, communication interface 1304, and memory 1306 communicate with each other via communication bus 1308. Communication interface 1304 is used to communicate with other devices, such as client devices or other server network elements. Processor 1302 is used to execute program 1310, which may specifically perform the steps described in the aforementioned embodiment of the lung respiratory motion modeling method.

[0065] Specifically, the program 130 may include program codes, which include computer operating instructions.

[0066] Processor 1302 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The one or more processors included in a computing device may be of the same type, such as one or more CPUs, or may be of different types, such as one or more CPUs and one or more ASICs.

[0067] The memory 1306 is used to store the program 1310. The memory 1306 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage.

[0068] According to the solution provided by the present invention, a 4D CT image dataset of lung respiratory motion is obtained; wherein the 4D CT image includes 3D CT images at multiple moments, and the 3D CT image is composed of 2D CT images at different height positions; the 4D CT image is subjected to contrast normalization and denoising to highlight lung regions with specific geometric structures; the lung regions are segmented to obtain the centroids of each segmented region, and landmark points at corresponding positions are extracted; a lung respiratory motion model is constructed based on the landmark points; and the landmark points extracted from the new CT image are matched with the landmark points in the lung respiratory motion model to obtain motion state information of the landmark points in the new CT image. The present invention can process global and local motion fields of the lungs, is computationally lightweight, and produces a final motion image in a short time.

[0069] Those skilled in the art will appreciate that the modules in the devices of the embodiments can be adaptively modified and deployed in one or more devices different from the embodiments. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and furthermore, they can be divided into multiple sub-modules, sub-units, or sub-components. All features disclosed in this specification (including the accompanying claims, abstract, and drawings), as well as all processes or units of any method or device disclosed therein, can be combined in any combination, unless at least some of such features and / or processes or units are mutually exclusive. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature serving the same, equivalent, or similar purpose. Furthermore, those skilled in the art will appreciate that although some embodiments herein include certain features and not others included in other embodiments, combinations of features from different embodiments are intended to fall within the scope of the present invention and form different embodiments. For example, in the claims below, any of the claimed embodiments may be used in any combination. The present invention may be implemented by means of hardware comprising several distinct elements and by means of a suitably programmed computer. In a unit claim that lists several means, several of these means may be embodied by the same hardware item. Unless otherwise specified, the steps in the above embodiments should not be understood as limiting the order of execution.

Claims

1. A lung respiratory motion modeling method, characterized in that: include: Acquire a 4D CT image dataset of lung respiratory motion; wherein the 4D CT image includes 3D CT images at multiple moments, and the 3D CT image is composed of 2D CT images at different height positions; performing contrast normalization and denoising on the 4D CT image to highlight lung regions with specific geometric structures; Segmenting the lung region to obtain the centroid of each segmented region and extracting landmark points at corresponding positions, including: calculating, for any segmented region, average coordinates of all pixels in the segmented region, the average coordinates being the coordinates of the centroid of the segmented region, and the coordinates of the centroid of the segmented region being the coordinates of the landmark points of the segmented region; Constructing a lung respiratory motion model based on the landmark points, comprising: obtaining lung respiratory motion features based on the landmark points in the 4D CT images at adjacent moments, constructing an affine transformation group based on the lung respiratory motion features, and obtaining a motion trajectory of the landmark points, wherein the lung respiratory motion model is a model that describes the motion trajectory of the landmark points; The landmark points extracted from the new CT image are matched with the landmark points in the lung respiratory motion model to obtain motion state information of the landmark points in the new CT image.

2. The lung respiratory motion modeling method according to claim 1, characterized in that: The method further comprises: Generate corresponding 3D point cloud data based on the landmark points at each breathing moment; The 3D point cloud data is subjected to correlation analysis, and only coordinate points with correlation within a preset range are retained to obtain sparse landmark points that exclude artifacts.

3. The lung respiratory motion modeling method according to claim 1 or 2, characterized in that: The formula for calculating the displacement vector field of the lung respiratory motion model is: in, is the displacement vector field describing the lung motion, is the displacement vector field of the jth experimental patient object, is the evaluation weight of the jth experimental patient, n is the total number of experimental patients, is the optimal change parameter for the jth experimental patient, and x is axis, axis, The axes form three-dimensional coordinate points on the three-dimensional coordinates, The positive axis points to the left side of the experimental patient subject. The positive axis points to the back of the experimental patient. The positive axis pointed toward the head of the experimental patient subject.

4. The lung respiratory motion modeling method according to claim 3, characterized in that: The optimal variation parameter The optimization function is: in, For The relevant transformation function is is the regularization term.

5. The lung respiratory motion modeling method according to claim 1, characterized in that: After extracting the landmark points at the corresponding positions, the method further includes: By constructing the affine transformation group Describes the rotation and translation changes of the lungs at different breathing moments; is the rotation matrix, is the translation vector, is a single parameter space, is a 3×3 reversible matrix, .

6. The lung respiratory motion modeling method according to claim 5, characterized in that: The rotation matrix for: in, , , Respectively represent axis, Axis and The rotation angle of the axis.

7. The lung respiratory motion modeling method according to claim 1, characterized in that: The method further comprises: By finding the parameters Minimize the image difference between adjacent breathing moments; The optimal parameters are obtained by iterative solution based on the Gauss-Newton method, and the difference function is processed by linearization. Speed up the iteration process and use the Jacobi matrix Compute gradients and parameter updates; where, , the linearized function is , , Represents the residual between the true value and the simulated value of each point, is the actual point cloud coordinate; Generate the motion change field of the lungs according to the scaling method.

8. A lung respiratory motion modeling device, characterized in that: include: A data acquisition module, configured to acquire a 4D CT image dataset of lung respiratory motion; wherein the 4D CT image includes 3D CT images at multiple moments, and the 3D CT image is composed of 2D CT images at different height positions; a data preprocessing module, configured to perform contrast normalization and denoising on the 4D CT image to highlight a lung region with a specific geometric structure; A landmark acquisition module is used to segment the lung area to obtain the centroid of each segmented area and extract the landmark points at the corresponding positions, including: for any of the segmented areas, calculating the average coordinates of all pixels in the segmented area, the average coordinates being the coordinates of the centroid of any of the segmented areas, and the coordinates of the centroid of any of the segmented areas being the coordinates of the landmark points of any of the segmented areas; a model construction module, configured to construct a lung respiratory motion model based on the landmark points, comprising: obtaining lung respiratory motion features based on the landmark points in 4DCT images at adjacent moments, constructing an affine transformation group based on the lung respiratory motion features, and obtaining motion trajectories of the landmark points, wherein the lung respiratory motion model is a model describing the motion trajectories of the landmark points; The matching module is used to match the landmark points extracted from the new CT image with the landmark points in the lung respiratory motion model to obtain motion state information of the landmark points in the new CT image.

9. A computing device comprising: A processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned lung respiratory motion modeling method.

10. A computer program product comprising at least one executable instruction, wherein the executable instruction enables a processor to execute operations corresponding to the lung respiratory motion modeling method according to any one of claims 1 to 7.

Images