Rapid orthogonal X-ray projection method for lung cancer CT / 4DCT tumor area

Abstract

The invention relates to the technical field of medical image processing, and particularly discloses a rapid orthogonal X-ray projection method for a lung cancer CT / 4DCT tumor area. The method comprises the following steps: acquiring a CT / 4DCT image of a lung cancer patient and preprocessing the CT / 4DCT image; bone structure interference and key spine and rib structures are removed from the preprocessed image; generating an orthogonal X-ray image on the basis of the CT / 4DCT image subjected to bone structure interference removal through Euler angle rotation control and an X-ray projection algorithm; and finally, performing auxiliary judgment on the respiratory movement tracking treatment feasibility based on the tumor area in the generated orthogonal X-ray image. Through an automatic image processing flow, subjective deviation of traditional manual sketching is eliminated, and a tumor orthogonal projection view without bone interference is quickly generated; in combination with 4DCT respiratory phase dynamic analysis, the accuracy of tumor area identification and the objectivity of respiratory tracking treatment feasibility judgment are improved, and efficient and reliable imaging auxiliary support is provided for lung cancer respiratory movement tracking radiotherapy.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, and in particular to a method for rapid orthogonal X-ray projection of tumor regions in lung cancer CT / 4DCT. Background Technology

[0002] In lung cancer radiotherapy, inoperable patients primarily rely on precise radiotherapy techniques to achieve therapeutic effects. However, tumor displacement caused by respiratory motion has long hindered improvements in treatment precision. During traditional radiotherapy, uneven dose distribution within the target area due to tumor movement during respiration poses significant clinical risks: areas receiving excessive doses may damage normal tissues such as the lung parenchyma and heart, increasing the incidence of radiation pneumonitis; areas receiving insufficient doses are prone to residual tumor cells, increasing the likelihood of local recurrence. To address this challenge, respiratory motion tracking radiotherapy technology is gradually being applied clinically. Its core lies in establishing a respiratory motion model of lung tumors, ensuring precise delivery of radiation dose to the tumor area through real-time tracking, while maximizing the protection of surrounding normal tissues.

[0003] However, existing technologies have systemic flaws: First, clinical practice relies entirely on doctors manually delineating tumors and their internal trackable areas. This process is limited by subjective experience differences—inconsistencies in the delineation results among different doctors lead to errors in the identification of trackable areas. Second, to achieve acceptable tracking accuracy, time-consuming iterative corrections are necessary: ​​doctors must manually delineate the contours based on localized CT scans, generate orthogonal images using projection algorithms, and then repeatedly adjust the contours and recalculate based on the reconstruction results—a time-consuming and labor-intensive process. This highly experience-dependent and inefficient process not only wastes medical resources but also leads to misjudgments of tracking feasibility, severely hindering the development of precision radiotherapy.

[0004] Therefore, there is an urgent need for a rapid orthogonal X-ray projection method for lung cancer CT / 4DCT tumor regions, which can eliminate the bottleneck of manual intervention through fully automatic orthogonal projection technology and improve the efficiency and accuracy of lung cancer tracking. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a method for rapid orthogonal X-ray projection of tumor regions in lung cancer CT / 4DCT, which can accurately and objectively identify trackable areas within the tumor and determine the feasibility of respiratory motion tracking, eliminating the influence of subjective experience differences and improving the positioning accuracy and treatment reliability of lung cancer tracking radiotherapy.

[0006] This invention provides a method for rapid orthogonal X-ray projection of tumor regions in lung cancer CT / 4DCT, comprising the following steps: S1. Acquire CT / 4DCT images of lung cancer patients and perform preprocessing; S2. Remove the interference from the spine and ribs from the preprocessed CT / 4DCT image to obtain a CT / 4DCT image with bone interference removed. S3. Based on CT / 4DCT images with bone interference removed, orthogonal X-ray images are generated through Euler angle rotation control and X-ray projection algorithm; S4. Based on the orthogonal X-ray images, assist in identifying the respiratory movement tracking area of ​​the tumor region and determine the feasibility of tracking treatment.

[0007] Furthermore, in S1, the preprocessing includes: S11. Truncate the voxel values ​​of the CT / 4DCT images of the lung cancer patient to a preset range; S12. Convert the pixel value range using a normalization formula; S13. Use the sliding window algorithm to locate the lung tissue feature region and perform image cropping.

[0008] Furthermore, in S13, the sliding window algorithm includes: S131. Scan the CT / 4DCT image according to the preset window size; S132. Calculate the sum of pixel values ​​in each window area; S133. Select the window with the largest sum as the lung tissue feature area; S134. Crop the image to the preset size with the lung tissue feature area as the center.

[0009] Furthermore, S2 specifically includes: S21. For the preprocessed CT / 4DCT images, generate a binary mask for the skeleton based on the voxel value threshold range of the spine and ribs. S22. Perform morphological operations on the binary bone mask to fill the holes and obtain a complete binary bone mask; S23. Perform connected component analysis on the complete binary mask of the skeleton, identify the largest connected component, and extract its corresponding structure to obtain a CT / 4DCT image with bone interference removed.

[0010] Furthermore, in S21, the binary mask generation for the skeleton employs a seed filling algorithm, including: S211. Randomly select initial seed points from pixels that meet the threshold range of voxel values ​​for the spine and ribs. S212. Using the initial seed point as the center, mark the spatially connected bony structure voxels using a region growing algorithm; S213. Assign a value of 1 to the voxels of pixels that meet the voxel value threshold range of the spine and ribs, and assign a value of 0 to the voxels of pixels that do not meet the voxel value threshold range of the spine and ribs, until all pixels of the CT / 4DCT image are traversed to generate a binary mask for the skeleton.

[0011] Furthermore, in S23, the largest connected region corresponds to the spinal structure, including: S231, calculating the area of ​​each connected region; S232, selecting the connected region with the largest area as the spine and ribs; S233, removing the voxels corresponding to the spine and ribs from the CT / 4DCT image.

[0012] Furthermore, S3 includes: S31. Establish the mapping relationship from the three-dimensional voxel space to the two-dimensional projection plane, and construct the projection matrix: ; in, The grayscale value of a point (x′, y′) in an orthogonal X-ray projection image; f represents the CT density function. The coordinates of three-dimensional points in CT / 4DCT images of lung cancer patients; L represents the path of orthogonal projection X-rays; The projection matrix; S32. Based on the mapping relationship from the three-dimensional voxel space to the two-dimensional projection plane, with the geometric center of the CT / 4DCT image as the rotation center, the rotation angles around the z-axis are fixed at 45° and -45° to generate the rotated three-dimensional voxel space. S33. Based on each detector pixel in the rotated three-dimensional voxel space, calculate the sum of the CT values ​​of all voxels traversed by the X-ray penetration path, normalize them to generate projected grayscale values, and generate orthogonal X-ray images.

[0013] Furthermore, when processing 4DCT images: S34, repeat S31-S33 for all phase CT sequences within the respiratory cycle; S35, generate an orthogonal X-ray image dataset containing 10 respiratory phases to fully capture the morphological changes of tumor respiratory motion.

[0014] Furthermore, S4 includes: S41. Automatically delineate the tumor projection boundary based on the orthogonal X-ray image; S42. Compare the displacement and deformation of the projected boundary under different breathing phases; S43. Assess the feasibility of respiratory motion tracking based on displacement and deformation patterns.

[0015] The present invention has the following technical effects: This invention acquires complete lung imaging baseline data through preprocessing, and effectively avoids the obscuring of tumor areas by bony artifacts through efficient elimination of spinal interference. Relying on an orthogonal projection control mechanism with a preset fixed rotation angle, it ensures the rapid generation of tumor projection views with clear image boundaries at the optimal viewing angle. This invention is the first in the field of radiotherapy planning to achieve a two-dimensional, intuitive presentation of the dynamic features of tumors in three-dimensional space. The complete removal of spinal structures effectively eliminates the obscuring effect of high-density bony tissue on the tumor area, allowing the tumor outline in the hilar region to be fully revealed in the projection view. The symmetrical fixed-angle projection control strategy ensures the best viewing angle for both tumors while fundamentally avoiding the resource consumption of traditional multi-angle trial-and-error reconstruction. Finally, the traceable region identification and feasibility assessment based on this projection transform the three-dimensional decision-making process, which originally relied on the physician's spatial imagination, into a quantifiable two-dimensional image evaluation process. This completely eliminates the reliance on subjective experience from traditional manual drawing, fundamentally improving the objective accuracy of respiratory motion tracking assessment. This invention provides highly reliable imaging support for radiotherapy planning while significantly reducing time consumption and decision uncertainty in traditional processes, fundamentally optimizing the clinical implementation quality of lung cancer tracking radiotherapy. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a flowchart of a method for rapid orthogonal X-ray projection of the tumor area in lung cancer CT / 4DC provided in an embodiment of the present invention.

[0018] Figure 2 This is a schematic diagram of the method for rapidly generating orthogonal X-ray projections from lung cancer CT / 4DCT images provided in this embodiment of the invention.

[0019] Figure 3 This is a dataset of 10-phase sequence orthogonal X-ray projections of 4DCT images of lung cancer patients provided in this embodiment of the invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0021] In lung cancer radiotherapy, tumor displacement caused by respiratory motion severely restricts radiotherapy accuracy. While respiratory motion tracking technology can achieve dynamic tumor tracking, it has long been hampered by two major technical bottlenecks: first, it relies entirely on doctors manually delineating the tumor and trackable areas, leading to significant delineation errors due to individual experience differences and a high degree of subjectivity in judging tracking feasibility; second, it requires multiple time-consuming iterative cycles of "delineation-projection reconstruction-manual correction," resulting in low efficiency and insufficient result stability. This invention utilizes a precise spinal artifact removal algorithm to remove the obscuring of the tumor by bony structures, rapidly generating high-contrast orthogonal images at the optimal viewing angle based on fixed dual-angle projection control, and automatically delineating the projection area using tumor boundary adaptive recognition technology. Furthermore, a multi-phase dynamic displacement analysis module quantifies the tumor deformation patterns within the respiratory cycle, generating an objective tracking feasibility decision report. This invention eliminates the subjective bias of manual delineation and the time consumption of iterative reconstruction at the root, shifting respiratory motion tracking assessment from experience-dependent to data-driven, establishing a standardized and reproducible implementation path for high-precision tracking radiotherapy in lung cancer, and comprehensively improving the reliability of radiotherapy.

[0022] Figure 1 This is a flowchart of a method for rapid orthogonal X-ray projection of lung cancer tumor areas using CT / 4DC, provided in an embodiment of the present invention. See also... Figure 1 A rapid orthogonal X-ray projection method for lung cancer CT / 4DCT tumor regions includes the following steps: S1. Acquire CT / 4DCT images of lung cancer patients and perform preprocessing.

[0023] In some embodiments, preprocessing includes: S11. Truncate the voxel values ​​of the CT / 4DCT images of the lung cancer patient to a preset range; S12. Convert the pixel value range using a normalization formula; S13. Use the sliding window algorithm to locate the lung tissue feature region and perform image cropping.

[0024] Furthermore, in S13, the sliding window algorithm includes: S131. Scan the CT / 4DCT image according to the preset window size; S132. Calculate the sum of pixel values ​​in each window area; S133. Select the window with the largest sum as the lung tissue feature area; S134. Crop the image to the preset size with the lung tissue feature area as the center.

[0025] Specifically, in the preprocessing stage of lung cancer CT / 4DCT images, the core design focuses on eliminating redundant image information and enhancing the expression of lung features. CT / 4DCT images of lung cancer patients are stored in 512×512 Dicom format, represented using Hounsfield units (HU). Medical imaging data shows that CT values ​​of lung soft tissue are concentrated in the 0-1300 HU range. Voxel values ​​outside this range, from -2000 HU (representing air at standard pressure and temperature) to over +3000 HU (representing metal or bone), interfere with tumor identification efficiency. Therefore, this invention truncates voxel values ​​to the range of 0 to 1300 HU.

[0026] Then, normalization is performed to dynamically map the voxel values ​​to a grayscale range of 0-255, effectively eliminating contrast fluctuations caused by differences in scanning parameters from different devices. The normalization formula is: Where D represents the pixel matrix of CT / 4DCT images of lung cancer patients whose voxel values ​​are truncated to a preset range; min(D) represents the minimum pixel value in matrix D; max(D) represents the maximum pixel value in matrix D; and image represents the normalized pixel matrix.

[0027] Next, to reduce redundant information and unnecessary computational burden, the image size was uniformly cropped to 256×256 pixels, effectively eliminating irrelevant areas at the image edges and further improving data processing efficiency and accuracy. During the size optimization stage, the system employed a sliding window algorithm to locate lung tissue feature regions and perform image cropping. Since the lung region exhibits significant high-pixel value aggregation in CT images, a 256×256 pixel window was used for step-by-step scanning, calculating the sum of voxel values ​​within each window in real time. When the program captured the region with the maximum cumulative value, it was identified as the core lung feature region, and a 256×256 pixel sub-image was precisely cropped from this center. This density-feature-based localization mechanism completely avoids the risk of tumor region truncation caused by traditional center cropping.

[0028] Optionally, data augmentation operations can be performed after image cropping, including: image scaling, which simulates the observation effect at different viewing distances by adjusting the image size; aspect ratio perturbation, which breaks the original proportion of the image and forces the model to learn more flexible size transformation features; horizontal or vertical flipping, which increases the diversity of images in the dataset and reduces the model's dependence on orientation; and color gamut transformation, which involves adjusting hue, saturation, and brightness, aiming to simulate color variations under different lighting conditions and shooting devices, thereby improving the model's ability to recognize and adapt to image features in different environments.

[0029] S2. Remove the spine and ribs from the preprocessed CT / 4DCT image to obtain a CT / 4DCT image with the interference of bony structures removed.

[0030] In the removal of bony interference in lung cancer CT / 4DCT images, the core design focuses on the accurate identification and elimination of artifacts in the spine and ribs. Medical anatomy shows that the spine and ribs, as continuous bone structures running the entire length of the thoracic cavity, have significantly higher CT values ​​than soft tissues (typically ranging from 300 to 2000 HU). However, traditional threshold segmentation is prone to causing fractures in the spinal region due to intervertebral disc gaps, resulting in discrete bone fragments that interfere with subsequent analysis.

[0031] In some embodiments, S2 specifically includes: S21. For the preprocessed CT / 4DCT image, generate a binary mask for the skeleton based on the voxel value threshold range of the spine and ribs.

[0032] Furthermore, in S21, the binary mask generation for the skeleton employs a seed filling algorithm, including: S211. Randomly select initial seed points from pixels that meet the threshold range of voxel values ​​for the spine and ribs. S212. Using the initial seed point as the center, mark the spatially connected bony structure voxels using a region growing algorithm; S213. Assign a value of 1 to the voxels of pixels that meet the voxel value threshold range of the spine and ribs, and assign a value of 0 to the voxels of pixels that do not meet the voxel value threshold range of the spine and ribs, until all pixels of the CT / 4DCT image are traversed to generate a binary mask for the skeleton.

[0033] Specifically, the preprocessed CT / 4DCT images are first preliminarily screened based on a preset bone density threshold range: all voxels in the image are traversed, and voxels that meet the threshold conditions are marked as candidate bone regions. At this time, the spine and rib regions present a discrete island-like distribution. To connect the fractured vertebrae to form a complete spine and rib structure, the system uses the Seed Filling algorithm to perform region growth optimization—starting from any bone voxel that meets the conditions, spatially connected bone structures are dynamically marked through recursive scanning of eight neighborhoods, and finally a complete binary mask of the skeleton is generated. In this process, the algorithm automatically assigns a value of 1 to the marked bone regions and a value of 0 to the non-bone regions. The generated binary mask completely avoids the risk of misidentifying the intervertebral space as a tumor region while preserving the spatial continuity of the entire bone structure.

[0034] S22. Perform morphological operations on the binary bone mask to fill the holes, and obtain a complete binary bone mask.

[0035] Furthermore, the morphological operation in S22 is a closing operation, including: S221, performing a dilation operation on the binary mask of the skeleton to fill the tiny gaps; S222, performing an erosion operation on the dilated mask to restore the original contour of the skeleton.

[0036] Optionally, after obtaining the binary mask of the skeleton, further optimization of morphological features is needed. Since the spine and ribs present a ring-shaped structure in cross-sectional images, their central region often contains a low-density medullary cavity, resulting in holes within the binary mask. The implementation process employs morphological closing operations for precise repair: first, a circular kernel dilation operation with a radius of 3 pixels is performed on the mask to fully fill the tiny gaps between adjacent bone regions; then, an erosion operation is performed using the same kernel to restore the original contour boundaries of the bones. This combined strategy of dilation followed by erosion completely closes the holes while avoiding the risk of contour deformation, establishing a geometrically complete topological structure for spine and rib recognition.

[0037] S23. Perform connected component analysis on the complete binary mask of the skeleton, identify the largest connected component, and extract its corresponding structure to obtain a CT / 4DCT image with the interference of bone structure removed.

[0038] In some embodiments, S23, where the largest connected region corresponds to the spine and ribs, includes: S231, calculating the area of ​​each connected region; S232, selecting the connected region with the largest area as the spine and ribs; S233, removing the voxels corresponding to the spine and ribs from the CT / 4DCT image.

[0039] Specifically, the spine and ribs are precisely located through connected component analysis of the mask. The pixel area of ​​all connected regions is calculated. Since the spine and ribs possess the largest continuous skeletal volume in the human thoracic cavity, the algorithm automatically selects the largest connected region and labels it as the spine and ribs. Then, using spatial coordinate mapping technology, the corresponding voxels are precisely removed from the original CT / 4DCT images. The output image then completely eliminates the obscuring of the tumor area by high-density spine and rib artifacts, and the tumor boundary in the hilar region is clearly revealed against the low-density background.

[0040] The entire debonding process described above is executed in full automation, providing a clean image foundation for subsequent orthogonal projection by removing interference from bony structures, thus fundamentally solving the problem of subjective bias in traditional manual drawing.

[0041] S3. Based on CT / 4DCT images with bone structure interference removed, orthogonal X-ray images are generated through Euler angle rotation control and X-ray projection algorithm.

[0042] Based on image processing that removes interference from bony structures, this invention achieves efficient orthogonal X-ray image generation through the fusion design of Euler angle space transformation and X-ray projection algorithm.

[0043] In some embodiments, such as Figure 2 As shown, S3 includes: S31. Establish the mapping relationship from the three-dimensional voxel space to the two-dimensional projection plane, and construct the projection matrix: ; in, The grayscale value of a point (x′, y′) in an orthogonal X-ray projection image; f represents the CT density function. The coordinates of three-dimensional points in CT / 4DCT images of lung cancer patients; L represents the path of orthogonal projection X-rays; The projection matrix; S32. Based on the mapping relationship from the three-dimensional voxel space to the two-dimensional projection plane, with the geometric center of the CT / 4DCT image as the rotation center, the rotation angles around the z-axis are fixed at 45° and -45° to generate the rotated three-dimensional voxel space. S33. Based on each detector pixel in the rotated three-dimensional voxel space, calculate the sum of the CT values ​​of all voxels traversed by the X-ray penetration path, normalize them to generate projected grayscale values, and generate orthogonal X-ray images.

[0044] Furthermore, when processing 4DCT images: S34, repeat S31-S33 for all phase CT sequences within the respiratory cycle; S35, generate an orthogonal X-ray image dataset containing 10 respiratory phases to fully capture the morphological changes of tumor respiratory motion, such as... Figure 3 As shown, Figure 3 The middle (a) image shows orthogonal X-ray images of 10 respiratory phases at a 45-degree angle to the left (i.e., a rotation angle of 45°); Figure 3 The middle (b) image shows the orthogonal X-ray images of 10 respiratory phases at a 45-degree angle to the right (i.e., a rotation angle of -45°).

[0045] In practical implementation, a three-dimensional coordinate system is established using the geometric center as the reference point, and voxel data of CT / 4DCT images, after removing interference from bony structures, are loaded into the image analysis development software. The core design concept originates from anatomical characteristics: when CT images are rotated ±45° around the long axis of the human body (i.e., the z-axis), the optimal observation angle for the bilateral hilar regions can be formed, maximizing the exposure of the tumor area. When the software performs projection matrix calculations, it constructs an accurate mapping from the spatial point p(x, y, z) to the projection point P(x′, y′), where the integration process of the penetration path L is optimized using the Bresenham algorithm. The rotation angle r is set. z The rotation parameters are fixed angle values ​​of 45° and -45°, and the rotation operation is achieved through Euler matrix transformation.

[0046] In the three-dimensional space after rotation transformation, a penetration path tracing is performed for each detector pixel (x′, y′). The straight path L from the focal point of the X-ray source through the rotation space to the detector point is calculated in reverse. All voxels are traversed along this path and their CT values ​​(Hounsfield units) are accumulated. Finally, the values ​​are converted into standard grayscale values ​​through normalization.

[0047] When processing 4DCT dynamic image sequences, the system automatically scans all 10 phases of data within the respiratory cycle. The aforementioned rotational projection process is fully executed for each independent phase, resulting in a time-series dataset that maintains ±45° orthogonal projection characteristics in the spatial dimension and fully records the deformation trajectory of the tumor during respiration in the temporal dimension. This spatiotemporal dynamic coupling allows clinicians to directly observe the tumor displacement and deformation patterns within the respiratory cycle, providing multidimensional quantitative evidence for feasibility assessment of tracking.

[0048] This implementation approach, firstly, avoids the computational burden of traditional multi-angle trial-and-error methods by using a fixed rotation angle, reducing the time required for a single projection generation to the second level; secondly, the projection algorithm directly accumulates the original voxel values, eliminating the boundary blurring effect caused by interpolation. While ensuring the accuracy of anatomical projection, it creates an efficient and reliable image analysis foundation for lung cancer tracking radiotherapy.

[0049] S4. Based on the orthogonal X-ray images, assist in identifying the respiratory movement tracking area of ​​the tumor region and determine the feasibility of tracking treatment.

[0050] In some embodiments, S4 includes: S41. Automatically delineate the tumor projection boundary based on the orthogonal X-ray image; S42. Compare the displacement and deformation of the projected boundary under different breathing phases; S43. Assess the feasibility of respiratory motion tracking based on displacement and deformation patterns.

[0051] After acquiring orthogonal X-ray images to remove interference from bony structures, tumor boundaries are automatically delineated using density gradient analysis. Since the tumor region exhibits a higher density characteristic than the lung parenchyma in CT images (typically 80-150 HU), the process first performs edge enhancement processing on ±45° orthogonal projection images. The Sobel operator is used to calculate density gradient changes along the projection direction, and gradient abrupt changes are automatically identified as potential boundaries using a preset threshold. Subsequently, closed contour lines are generated through radial basis function interpolation, automatically avoiding interfering structures such as blood vessels and bronchial cross-sections (which exhibit linear characteristics in orthogonal projection). The final output is a continuous and smooth tumor projection boundary curve. This workflow completely replaces traditional manual delineation, eliminating differences in contour drawing between different physicians.

[0052] After single-phase delineation, orthogonal X-ray projection sequences of all 10 phases within a 4DCT respiratory cycle are automatically loaded. Boundary extraction is repeated for each phase to form a spatiotemporal dynamic contour set. Using the 0% phase (end of inspiration) as a baseline, mutual information registration technology is used to spatiotemporally synchronize the projection boundaries of other phases, ensuring anatomical consistency in displacement measurements. The three core parameters at this stage are: ① boundary center point displacement vector (quantitative tumor spatial movement within a respiratory cycle); ② contour deformation index (contour area change rate); ③ boundary irregularity (change in the ratio of contour perimeter to area). These parameters are presented in real-time on the 3D visualization interface, forming a dynamic respiratory motion trajectory atlas.

[0053] Based on the above dynamic analysis results, a medical decision-making rule base was established to assess the feasibility of tracking. When the displacement vector direction is constant and the deformation index is less than a threshold, it is determined to be a linearly trackable region. If the displacement exhibits non-linear fluctuations or the deformation index exceeds the limit, a high-risk region is marked and requires physician review. All assessment results are displayed on a baseline phase projection image via heatmap overlay. Red-marked areas represent areas of significant displacement, while blue areas indicate controllable deformation. A structured report is finally generated, containing displacement vector field data for each phase and feasibility grading conclusions, providing clinicians with objective and quantitative basis for tracking and treatment decisions. The entire process transforms the traditional experience-based judgment process into a data-driven standardized analysis, making the feasibility assessment of respiratory motion tracking verifiable and repeatable.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims

1. A method for rapid orthogonal X-ray projection of tumor regions in lung cancer CT / 4DCT, characterized in that, Includes the following steps: S1. Acquire CT / 4DCT images of lung cancer patients and perform preprocessing; S2. Remove the spine and ribs from the preprocessed CT / 4DCT image to obtain a CT / 4DCT image with the interference of bony structures removed. S3. Based on CT / 4DCT images with bone structure interference removed, orthogonal X-ray images are generated through Euler angle rotation control and X-ray projection algorithm; S4. Based on the orthogonal X-ray images, assist in identifying the respiratory movement tracking area of ​​the tumor region and determine the feasibility of tracking treatment.

2. The method for rapid orthogonal X-ray projection of lung cancer CT / 4DCT tumor regions according to claim 1, characterized in that, In step S1, the preprocessing includes: S11. Truncate the voxel values ​​of the CT / 4DCT images of the lung cancer patient to a preset range; S12. Convert the pixel value range using a normalization formula; S13. Use the sliding window algorithm to locate the lung tissue feature region and perform image cropping.

3. The method for rapid orthogonal X-ray projection of lung cancer CT / 4DCT tumor regions according to claim 1, characterized in that, In S13, the sliding window algorithm includes: S131. Scan the CT / 4DCT image according to the preset window size; S132. Calculate the sum of pixel values ​​in each window area; S133. Select the window with the largest sum as the lung tissue feature area; S134. Crop the image to the preset size with the lung tissue feature area as the center.

4. The method for rapid orthogonal X-ray projection of lung cancer CT / 4DCT tumor regions according to claim 1, characterized in that, S2 specifically includes: S21. For the preprocessed CT / 4DCT image, generate a binary mask for the skeleton based on the voxel value threshold range of the spine and ribs. S22. Perform morphological operations on the binary bone mask to fill the holes and obtain a complete binary bone mask; S23. Perform connected component analysis on the complete binary mask of the skeleton, identify the largest connected component, and extract its corresponding structure to obtain a CT / 4DCT image with the interference of bone structure removed.

5. The method for rapid orthogonal X-ray projection of lung cancer CT / 4DCT tumor regions according to claim 4, characterized in that, In step S21, the binary mask generation for the skeleton employs the Seed Filling algorithm, including: S211. Randomly select initial seed points from pixels that meet the threshold range of voxel values ​​for the spine and ribs. S212. Using the initial seed point as the center, mark the spatially connected bony structure voxels using a region growing algorithm; S213. Assign a value of 1 to the voxels of pixels that meet the voxel value threshold range of the spine and ribs, and assign a value of 0 to the voxels of pixels that do not meet the voxel value threshold range of the spine and ribs, until all pixels of the CT / 4DCT image are traversed to generate a binary mask for the skeleton.

6. The method for rapid orthogonal X-ray projection of lung cancer CT / 4DCT tumor regions according to claim 4, characterized in that, In step S23, the largest connected region corresponds to the spinal structure, including: S231, calculating the area of ​​each connected region; S232, selecting the connected region with the largest area as the spine and ribs; S233, removing the voxels corresponding to the spine and ribs from the CT / 4DCT image.

7. The method for rapid orthogonal X-ray projection of lung cancer CT / 4DCT tumor regions according to claim 1, characterized in that, S3 includes: S31. Establish the mapping relationship from the three-dimensional voxel space to the two-dimensional projection plane, and construct the projection matrix: ; in, represents the gray value of the coordinate point (x′, y′) in an orthogonal X-ray projection image; f represents the CT density function. The coordinates of a three-dimensional point in the CT / 4DCT image of a lung cancer patient are represented; L represents the path of the orthogonal projection X-ray. Represents coordinate points The projection matrix; S32. Based on the mapping relationship from the three-dimensional voxel space to the two-dimensional projection plane, with the geometric center of the CT / 4DCT image as the rotation center, the rotation angles around the z-axis are fixed at 45° and -45° to generate the rotated three-dimensional voxel space. S33. Based on each detector pixel in the rotated three-dimensional voxel space, calculate the sum of the CT values ​​of all voxels traversed by the X-ray penetration path, normalize them to generate projected grayscale values, and generate orthogonal X-ray images.

8. The method for rapid orthogonal X-ray projection of lung cancer CT / 4DCT tumor regions according to claim 7, characterized in that, When processing 4DCT images: S34, repeat S31-S33 for all phase CT sequences within the respiratory cycle; S35, generate an orthogonal X-ray image dataset containing 10 respiratory phases to fully capture the morphological changes of tumor respiratory motion.

9. The method for rapid orthogonal X-ray projection of lung cancer CT / 4DCT tumor regions according to claim 1, characterized in that, S4 includes: S41. Automatically delineate the tumor projection boundary based on the orthogonal X-ray image; S42. Compare the displacement and deformation of the projected boundary under different breathing phases; S43. Assess the feasibility of respiratory motion tracking based on displacement and deformation patterns.

Images