Method for generating virtual barium esophagogram based on digital reconstructed radiograph rendering of CT image segmentation

Abstract

The present application relates to the field of digital reconstruction and virtual angiography technology, and particularly relates to a virtual esophageal barium contrast generation method based on CT image segmentation and digital reconstruction X-ray image rendering, comprising the following steps: S1, collecting chest or upper abdominal CT images of a patient and preprocessing; S2, automatically segmenting an esophageal region from the original CT image; S3, dividing the entire esophageal region into three parts, namely, a lumen, a wall and a mucosa; S4, obtaining a segmentation result of a tumor; S5, mask calculation; S6, filling CT values for different parts of the esophagus to simulate the state of esophageal filling phase and mucosal phase under different barium doses; S7, fusing the original CT image with the filling phase image and the mucosal phase image respectively to obtain a fused three-dimensional CT image, modeling the propagation process of X-rays in the fused three-dimensional CT volume data by using a DRR technology, and generating a virtual esophageal barium meal image. The present application does not require the patient to swallow a barium meal, has no X-ray exposure risk, has low dependence on the experience of operators, and has stable and reliable image quality.

Description

Technical Field

[0001] This invention relates to the field of digital reconstruction and virtual imaging technology, specifically to a method for generating virtual esophageal barium contrast images based on digital reconstruction X-ray image rendering using CT image segmentation. Background Technology

[0002] Esophageal diseases are an important part of clinical digestive system diseases, encompassing structural, functional, and neoplastic lesions, including but not limited to esophagitis, esophageal stricture, esophageal diverticulum, esophageal motility disorders, and esophageal tumors. Different types of esophageal diseases exhibit significant differences in anatomical structure, cavity morphology, and functional status, placing higher demands on the coverage and visualization capabilities of imaging examinations.

[0003] Barium contrast esophagography, a traditional and well-established X-ray examination method, can display the overall outline of the esophagus, its patency, and mucosal morphological changes by filling the esophageal lumen with barium. It plays an irreplaceable role in the screening and preliminary assessment of esophageal diseases. In particular, barium contrast esophagography has the advantages of being intuitive, dynamic, and holistic in displaying esophageal morphological abnormalities, the degree of stenosis, and the continuity of the esophageal wall outline.

[0004] However, on the one hand, most patients with esophageal diseases are middle-aged and elderly, with slow movements and reactions, making them difficult to cooperate and posing potential risks such as aspiration, choking, and barium retention. Therefore, traditional esophageal barium swallow imaging is greatly affected by the patient's cooperation, swallowing movements, and body position, limiting its application in some esophageal diseases or specific patient populations. On the other hand, this examination method is highly dependent on examination conditions, operator experience, and patient cooperation, resulting in unstable application effects in the assessment of some functional esophageal diseases, complex anatomical variations, or early lesions. It is difficult to replicate and standardize, and the image presentation is uncertain, which is not conducive to quantitative analysis and subsequent research. It is also impossible to establish a direct correspondence with existing three-dimensional anatomical data, thus limiting the development of CT-based esophageal function simulation and virtual imaging. Summary of the Invention

[0005] The present invention aims to provide a virtual esophageal barium contrast imaging method based on CT image segmentation and digital reconstruction X-ray image rendering, which eliminates the need for patients to ingest barium meals and undergo X-ray irradiation, thereby reducing the risk of exposure.

[0006] A method for generating virtual esophageal barium contrast radiographs based on digital reconstruction X-ray image rendering using CT image segmentation, characterized by the following steps:

[0007] S1. Acquire CT images of the patient's chest or upper abdomen, and resample, standardize the orientation, and normalize the intensity of each frame of the CT image to form the original CT image.

[0008] S2. Automatically segment the original CT image to initially segment the esophageal region;

[0009] S3. The esophageal region of interest is segmented in one step, and the entire esophageal region is divided into three parts: lumen, wall and mucosa by a 3D slicer.

[0010] S4. Secondary segmentation of the esophageal region of interest to obtain the segmentation results of the tumor;

[0011] S5. Mask calculation: The masks of the lumen, wall and mucosa are obtained through step S3. The mask of the tumor is obtained through step S4. The lumen mask is subtracted from the tumor mask to obtain the real lumen. The real lumen area is subtracted from the entire esophageal area to obtain the real wall.

[0012] S6. CT values ​​are filled for different parts of the esophagus to simulate the state of the esophageal filling phase and mucosal phase under different barium doses;

[0013] S7. The original CT images are fused with the filling phase image and the mucosal phase image respectively to obtain the fused three-dimensional CT image. The propagation process of X-rays in the fused three-dimensional CT volume data is modeled using DRR technology to generate virtual esophageal barium meal images.

[0014] The beneficial effects of this invention are as follows: This invention processes fused CT body data to generate virtual fluoroscopic images of esophageal barium swallow, eliminating the need for actual X-ray exposure for patients, significantly reducing radiation exposure and improving examination safety; this method utilizes image calculations to generate virtual filling and mucosal images, which is completely non-invasive and suitable for individuals who cannot tolerate swallowing contrast agents; by constructing a virtual projection model, this method can generate virtual esophageal barium swallow images from multiple angles, such as anteroposterior, lateral, and left and right anterior oblique views, without requiring rescanning of the patient, improving multi-view diagnostic capabilities; traditional barium swallow is easily affected by operator skill and patient swallowing status, while this invention automatically generates images using an algorithm, with low dependence on operator experience and stable, repeatable image quality. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the virtual esophageal barium contrast generation method based on CT image segmentation and digital reconstruction X-ray image rendering according to the present invention.

[0016] Figure 2a This invention provides a method for generating virtual esophageal barium contrast imaging based on CT image segmentation and digital reconstruction of X-ray images, specifically the esophageal delineation in the positive orientation.

[0017] Figure 2b The lateral view of esophageal delineation in the virtual esophageal barium contrast imaging method based on CT image segmentation of the present invention is used for digital reconstruction X-ray image rendering of CT image segmentation.

[0018] Figure 2c This invention relates to a method for generating virtual esophageal barium contrast imaging based on CT image segmentation and digital reconstruction of X-ray images, specifically the method for drawing transverse sections of the esophagus in a digital reconstruction X-ray image rendering system using CT image segmentation.

[0019] Figure 3a This is the esophageal segmentation image in the virtual esophageal barium contrast imaging method for digital reconstruction X-ray image rendering based on CT image segmentation in this invention;

[0020] Figure 3b This is a lumen separation image in the virtual esophageal barium contrast imaging method based on CT image segmentation and digital reconstruction X-ray image rendering of the present invention;

[0021] Figure 3c This is a 3D image of the lumen in the virtual esophageal barium contrast imaging method based on CT image segmentation and digital reconstruction of X-ray images in this invention.

[0022] Figure 4a This is a nested diagram of the esophagus and lumen in the virtual esophageal barium contrast imaging method for digital reconstruction X-ray image rendering based on CT image segmentation of the present invention.

[0023] Figure 4b This is a diagram of esophageal wall separation in the virtual esophageal barium contrast imaging method based on CT image segmentation and digital reconstruction of X-ray images according to the present invention.

[0024] Figure 4c This is a hollow esophageal wall image in the virtual esophageal barium contrast imaging method based on CT image segmentation and digital reconstruction X-ray image rendering of the present invention.

[0025] Figure 5a This is a map of the esophageal mucosa in the virtual esophageal barium contrast imaging method based on CT image segmentation and digital reconstruction of X-ray images according to the present invention.

[0026] Figure 5b This is a nested diagram of the esophagus and mucosa in the virtual esophageal barium contrast imaging method based on CT image segmentation and digital reconstruction X-ray image rendering according to the present invention.

[0027] Figure 5c This is a 3D mucosal image generated in the virtual esophageal barium contrast imaging method based on CT image segmentation and digital reconstruction of X-ray images in this invention.

[0028] Figure 6a This is a diagram showing the filling effect of the virtual esophageal barium contrast imaging generation method based on CT image segmentation and digital reconstruction X-ray image rendering according to the present invention.

[0029] Figure 6b This is a diagram showing the mucosal phase filling effect in the virtual esophageal barium contrast generation method based on CT image segmentation and digital reconstruction X-ray image rendering according to the present invention.

[0030] Figure 7a This invention provides a virtual barium meal image for the virtual esophageal barium contrast generation method based on digital reconstruction X-ray image rendering using CT image segmentation.

[0031] Figure 7b This is a genuine image of a barium meal. Detailed Implementation

[0032] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described below are only for explaining the present invention and do not limit the scope of protection of the present invention.

[0033] The present invention will be further described in detail below through preferred embodiments:

[0034] As attached Figure 1 The method for generating virtual esophageal barium contrast radiographs based on digital reconstruction X-ray image rendering using CT image segmentation includes the following steps:

[0035] S1. Acquire CT images of the patient's chest or upper abdomen, and resample, standardize the orientation, and normalize the intensity of each frame of the CT image to form the original CT image.

[0036] S2. Automatically segment the original CT image, initially segmenting the esophageal region, as shown in the attached image. Figure 2a , 2b 2c;

[0037] In this embodiment, specifically: the TotalSegmentator model, a deep learning-based automatic whole-body CT segmentation tool, is invoked. This model can automatically segment multiple anatomical structures in the human body in a single inference. In this invention, it is used to segment the esophagus and is an existing tool. The original CT image is imported into the AutoDL cloud algorithm platform. AutoDL (Auto DeepLearning) is a GPU computing power rental platform for researchers and AI developers, providing a pay-as-you-go deep learning server environment. On the AutoDL cloud algorithm platform, the TotalSegmentator model is used in Python to automatically segment the CT image, initially segmenting the esophageal region. The esophageal region includes all structures within the esophagus, including the esophageal wall, lumen, mucosa, and tumor locations. The segmentation result is stored in NIFTI format and imported into 3D Slicer. The Segmentation method is selected to check the esophageal region delineation effect. The delineation includes the entire esophagus, indicating a good segmentation effect. 3D Slicer is an open-source medical image 3D visualization and analysis platform, widely used in medical research, image analysis, surgical planning and AI model building. It is mainly used for medical image browsing and processing, segmentation, 3D reconstruction, image registration, quantitative analysis, etc. In this invention, it is used to accurately delineate the esophagus and tumor extent.

[0038] S3. The esophageal region of interest is segmented in one step, using a 3D slicer to divide the entire esophageal region into three parts: the lumen, the wall, and the mucosa; as shown in the attached figure. Figure 3a , 3b As shown in 3c, 4a, 4b, 4c, 5a, 5b, and 5c.

[0039] In actual esophagography, it is necessary to obtain the filling phase and mucosal phase of the esophagus, so the entire esophageal region needs to be subdivided again.

[0040] Specifically:

[0041] Lumen: In 3d-slicer, the smoothing tool is used to smooth the edges of the esophageal region. The Shrink tool in Margin is used to divide the lumen region. The margin tool can outline the lumen range from the entire esophagus. The 3d-slicer is then used to convert it into a 3D model.

[0042] Tube wall: The entire esophageal region is imported again, and the entire esophageal region is subtracted from the lumen region to obtain the tube wall. The effect is then checked by 3D processing.

[0043] Mucosa: Import the entire esophageal area again, add a new outline layer, use the threshold range tool to draw the esophageal mucosa range within the esophageal area, and perform 3D processing to check the effect.

[0044] S4. Secondary segmentation of the esophageal region of interest to obtain the segmentation results of the tumor;

[0045] Specifically:

[0046] S41. Acquire plain CT images and enhanced CT images of the same subject, and perform spatial registration processing on the plain CT images and enhanced CT images to align them in the same spatial coordinate system;

[0047] S42. Based on the registration, perform gray-level difference operation on the enhanced CT image and the plain CT image to obtain the gray-level difference volume before and after enhancement. The gray value of each voxel in the difference volume represents the degree of gray-level change of that voxel before and after enhancement.

[0048] S43. On the gray-level difference volume, voxels are screened according to a preset gray-level difference threshold, and voxel regions with gray-level difference volume values ​​higher than the gray-level difference threshold are extracted to obtain initial enhancement candidate regions.

[0049] S44. Limit the initial enhancement candidate region to the esophageal region and remove voxel regions located outside the esophageal region;

[0050] S45. Morphological processing is performed on the initial enhancement candidate region located in the esophagus, including connected domain analysis, hole filling and smoothing, to eliminate noise voxels and enhance the spatial coherence of the region, and finally obtain the initial volume segmentation result of the esophageal tumor.

[0051] Step S3 yielded initial segmentation images of the three parts of the esophageal region of interest. However, this segmentation result has certain drawbacks. The automatic segmentation of the 3D-slicer will automatically connect a part of the esophageal wall into the lumen. If it is simply divided into three parts, the true information of the tumor will be lost. Therefore, step S4 is used to make up for this defect.

[0052] S5. Mask calculation: The masks for the lumen, wall and mucosa are obtained through step S3. The mask for the tumor is obtained through step S4. The lumen mask is subtracted from the tumor mask to obtain the real lumen. The real lumen area is subtracted from the entire esophageal area to obtain the real wall.

[0053] S6. CT value filling was performed on different parts of the esophagus to simulate the state of the esophageal filling phase and mucosal phase under different barium doses, as shown in the attached figure. Figure 6a , 6b As shown;

[0054] Specifically, it includes:

[0055] S61, Filling Phase: The actual lumen is introduced into the 3D slicer, and different CT values ​​are applied to fill the lumen and the tube wall respectively. In this embodiment, the lumen is filled with 6000 HU, and the tube wall is filled with 1000 HU.

[0056] Because the lumen is "filled" with barium in the filling phase, appearing as a continuous white columnar shadow, and the barium almost fills the lumen at this time, only the actual lumen area is filled when filling CT values.

[0057] S62, Mucosal phase filling: The actual tube wall and mucosa are introduced into the 3D-slicer, and CT values ​​are applied to fill the actual tube wall and mucosa respectively. In this embodiment, both the tube wall and mucosa are filled with 3000 HU.

[0058] Since the barium only adheres to the mucosal surface in the mucosal phase and no longer completely fills the lumen, the proportion of the mucosa is increased when filling the CT value, and the lumen is filled with a small amount of barium.

[0059] S7. The original CT images are fused with the filling phase image and the mucosal phase image respectively to obtain the fused three-dimensional CT image. The propagation process of X-rays in the fused three-dimensional CT volume data is modeled using DRR technology to generate virtual esophageal barium meal images.

[0060] Specifically, the following steps are included:

[0061] S71, Spatial Modeling and Attenuation Attribute Mapping, including:

[0062] S711. Load the fused 3D CT image into a regular 3D voxel grid to form 3D CT volume data. Each voxel contains spatial coordinate information and the corresponding CT gray value.

[0063] Specifically, the filling phase and mucosal phase filling images are imported into the autodl database, and the original CT images are imported into the Volume module, where they are resampled, coordinate system corrected, and HU-linear attenuation coefficient transformed. After preprocessing, three-dimensional CT volume data that meets the requirements of DRR rendering is obtained.

[0064] S712, Coordinate System 1, establishes the spatial mapping relationship between the voxel coordinate system and the world coordinate system to ensure the precise interaction between rays and voxels in three-dimensional space;

[0065] S713, Attenuation Attribute Mapping: Maps CT grayscale values ​​to equivalent X-ray attenuation coefficients, enabling each voxel to participate in X-ray energy attenuation calculations;

[0066] In this invention, to achieve accurate modeling of the physical process from CT images to X-ray imaging, a correspondence between CT grayscale values ​​and the equivalent X-ray linear attenuation coefficient is established. The grayscale values ​​in CT images are represented using Hounsfield Units (HU), defined as the ratio of the difference between the linear attenuation coefficient of the tissue being measured and the linear attenuation coefficient of water to the linear attenuation coefficient of water, expressed as a percentage (in thousands).

[0067] Where μ is the linear attenuation coefficient of the voxel being measured, and μ_water is the linear attenuation coefficient of water at a set effective photon energy. Therefore, the formula for calculating the voxel linear attenuation coefficient can be derived as: μ = μ_water × (1 + HU / 1000). In practical applications, the corresponding μ_water value is selected according to the preset effective X-ray energy, thereby converting the HU value in the CT volume data into an equivalent X-ray equivalent linear attenuation coefficient distribution for each voxel, which is then used for subsequent ray integration calculations based on Beer–Lambert's law. This is used to obtain a simulated X-ray projection image. Where I represents the intensity of the emitted rays. Let μ(x) represent the intensity of the incident ray, μ(x) represent the linear attenuation coefficient, x represent the spatial position, and L represent the ray path.

[0068] This invention achieves a unified conversion of CT grayscale information into physical attenuation parameters through the above-mentioned linear mapping relationship, enabling image data to directly participate in the calculation of physical imaging models, ensuring numerical consistency and physical interpretability, and improving the authenticity and stability of simulated imaging results.

[0069] S72, Imaging Settings, including:

[0070] S721, X-ray source settings: Define the position and orientation of a virtual X-ray source in three-dimensional space to simulate the emission point of a real X-ray tube;

[0071] Specifically, after establishing a three-dimensional volumetric data coordinate system, a spatial Cartesian coordinate system (X, Y, Z) is constructed with the volumetric data center or a preset reference point as the origin. The position of the virtual X-ray source is represented by a three-dimensional coordinate vector S = (Sx, Sy, Sz).

[0072] S722, Detector settings: Define the position, size, pixel resolution, and spatial relationship between the virtual plane detector and the X-ray source; the center position of the detector is represented by a three-dimensional coordinate vector D = (Dx, Dy, Dz).

[0073] S723, Imaging geometry parameterization, sets the distance from the X-ray source to the detector, the direction of the main X-ray, and the angle rotation parameters to simulate the common anteroposterior, anterior oblique, and lateral projection conditions in clinical barium swallow examinations.

[0074] Specifically: In the positive phase, "60KV_AL35" is used to simulate the X-ray energy spectrum. 60KV_AL35 represents the X-ray multi-energy spectrum distribution model after passing through a 3.5 mm aluminum filter under a tube voltage of 60 kVp. In the lateral phase, since X-rays need to penetrate the entire left and right diameters of the thoracic cavity, the tissue is thicker, and X-ray attenuation is greater. To ensure image brightness, the X-ray penetration ability increases when the tube voltage is increased, and the image signal thickens when the tube current is increased. Therefore, the tube voltage is appropriately increased, using "90KV_AL40".

[0075] S73, Ray projection, including:

[0076] S731, Pixel-by-pixel ray generation: For each pixel on the virtual detector, calculate its corresponding ray direction vector and generate a ray from the virtual ray source;

[0077] S732, voxel-level ray sampling, each ray is sampled along its propagation path in a three-dimensional voxel grid with a fixed step size, passing through multiple voxels in sequence;

[0078] S733, Path integral calculation: At each voxel through which the ray passes, the product of the voxel's equivalent attenuation coefficient and the path length is accumulated to form the cumulative attenuation value of the ray. Based on the cumulative attenuation value of the ray, a virtual barium image is generated.

[0079] In this embodiment, the project() method of DeepDRR Projector is called to perform volume projection on the three-dimensional volume data, and the output result is a two-dimensional DRR floating-point image.

[0080] In this invention, the virtual barium meal image effect is not a post-processing superposition, but rather naturally formed during the X-ray projection calculation stage. The principle is as follows: voxels with high attenuation coefficients are placed within the lumen, absorbing X-rays during X-ray integration. When X-rays pass through the tumor region, the difference in attenuation coefficients between the voxels and the surrounding area causes localized morphological changes in the projected grayscale. These differences manifest in the two-dimensional projection image as filling defects, stenosis, or abnormal contours similar to those seen in real esophageal barium preparations. Therefore, the generation of the virtual barium meal image is the result of the combined effects of three-dimensional voxel space modeling, cumulative X-ray attenuation, and imaging geometric parameters.

[0081] S74, Image Transformation, including:

[0082] S741, Logarithmic Transformation: Convert the cumulative attenuation value of the ray into a grayscale image;

[0083] S742, intensity normalization and dynamic range adjustment, thereby ensuring that the generated virtual barium image still has stable contrast and visibility under different imaging parameters;

[0084] This involves observing the generated barium meal image and adjusting factors such as imaging distance, imaging range, and tube voltage to optimize the imaging effect. Since the directly generated image is relatively blurry, further image optimization is necessary; the quality of the initially generated image needs improvement.

[0085] The steps are as follows: light blur, first sharpening, light blur again, second sharpening, adaptive edge enhancement, and preservation of the original HU range.

[0086] S743, Two-dimensional image output, generating a virtual esophageal barium image for display, analysis, or storage.

[0087] As attached Figure 7a and 7b The images shown are the virtual barium meal image and the real barium meal image of the present invention, respectively.

[0088] This invention has the following advantages:

[0089] 1. Avoiding the radiation risks of real imaging: This invention processes the fused CT body data to generate a virtual fluoroscopic image of esophageal barium contrast, eliminating the need for actual X-ray exposure for patients, significantly reducing radiation exposure and improving examination safety.

[0090] 2. No need to actually swallow contrast agents, reducing clinical risks: Traditional barium meals require patients to swallow barium, which may lead to risks such as aspiration and choking. This method uses image calculation to generate virtual filling and mucosal images, which is completely non-invasive and suitable for groups that cannot tolerate swallowing contrast agents.

[0091] 3. Reproducible multi-view and multi-position images: By constructing a virtual projection model, this method can generate virtual esophageal barium images from multiple angles, such as anteroposterior, lateral, and left and right anterior oblique views, without the need to rescan the patient, thus improving multi-view diagnostic capabilities.

[0092] 4. Stable and consistent examination results, avoiding operational dependence: Traditional barium meals are easily affected by the operator's skills and the patient's swallowing status. This invention is automatically generated by an algorithm, which has low dependence on the operator's experience and the image quality is stable and repeatable.

[0093] 5. Quantitative analysis of lumen morphology: Utilizing CT segmentation results, this method can output quantitative parameters such as lumen diameter, degree of stenosis, and volume distribution, providing an objective basis for disease severity assessment, which is something that traditional barium meals cannot achieve.

[0094] 6. Improved lesion localization accuracy through fusion with original CT images: Virtual barium meal images can be registered and displayed with original CT, reconstructed MPR, and 3D models, which helps to accurately locate lesions such as esophageal tumors and strictures and to plan for surgery.

[0095] 7. Controllable image contrast, density, and perspective: By simulating X-ray energy spectrum, scattering, and attenuation models, parameters such as contrast agent density, irradiation voltage, and distance can be adjusted to achieve a realistic barium meal visual effect and adjust consistency.

[0096] 8. Can be used as a source of AI-assisted diagnosis and training data: Virtual barium meal images have controllable quality and unlimited quantity, which can be used as training data for deep learning models to automatically detect esophageal lesions and improve the performance of AI diagnosis.

[0097] 9. Suitable for special populations who cannot undergo barium meal examination: including those with swallowing disorders, elderly patients, critically ill patients, postoperative patients, etc. This invention provides a non-invasive and safe alternative.

[0098] 10. Significantly reduced costs and resource consumption: No more barium reagent, fluoroscopy equipment, or special operating environment are needed. The generation process can be completed on a regular workstation or server, which is conducive to its widespread use in primary hospitals and in scenarios without fluoroscopy equipment.

[0099] 11. Enhanced ability to differentiate imaging features of different types of esophageal diseases: By setting differentiated virtual barium attenuation models for the lumen and lesion-related areas, virtual barium contrast imaging can simultaneously reflect multiple imaging features in the projected images, such as changes in lumen morphology, abnormal wall contours, and local filling defects. This feature allows virtual barium contrast imaging not only to display luminal stenosis caused by space-occupying lesions, but also to help identify morphological changes caused by esophageal and gastric varices, esophageal compression, inflammation, edema, or functional abnormalities, thereby improving the comprehensive ability to identify the imaging manifestations of different types of esophageal diseases.

[0100] While existing technologies disclose techniques for esophageal barium swallow imaging, such as patent application number 202510346289.0 entitled "A System for In Vitro Detection of Esophageal Cancer," this approach also involves scanning patients to obtain CT images and then extracting esophageal features from them. However, it does not disclose the specific methods for extracting these features (thus failing to provide the technical inspiration for the three-stage extraction of esophageal features in this invention). Furthermore, it relies on the extracted esophageal features (such as diameter, length, and curvature) to create a three-dimensional model of the esophagus and surrounding tissues. A virtual barium swallow solution is then injected into the three-dimensional model to simulate the flow and distribution of the barium swallow solution within the esophagus and predict the effects of the barium swallow imaging. However, this existing document does not actually disclose the specific technical methods for performing three-dimensional modeling or predicting the effects of the barium swallow imaging. The present invention only provides an idea, and the final result, based on the prediction, is mainly used to determine which body position is suitable for the patient. In other words, the patient still needs to take a barium meal and then be irradiated with X-rays. The present invention first converts the three-dimensional CT image into two dimensions, and then fills the two-dimensional esophageal wall, lumen and mucosa with CT values ​​to simulate the filling phase and mucosal phase. Then, the filling phase and mucosal phase are fused with the original three-dimensional CT image to generate a new three-dimensional fused image. The irradiation of X-rays is simulated in the new three-dimensional fused image. It can be seen that the existing technology is completely different from the overall technical concept of the present invention, and it does not disclose the key technical means to achieve the technical solution. Moreover, the patient still needs to take a barium meal in the end. It can be seen that the technical problem it solves is different from the technical means of the present invention, and it does not achieve the various effects of the present invention.

[0101] The preferred embodiments of this application have been described in detail above with reference to the accompanying drawings. Typical known structures and common knowledge techniques in the preferred embodiments have not been described in detail here. Those skilled in the art can improve and implement the technical solutions of this invention based on the guidance provided in these embodiments and their own capabilities. Some typical known structures, known methods or common knowledge techniques should not be obstacles for those skilled in the art to implement this application.

[0102] The scope of protection claimed in this application shall be determined by the contents of its claims, and the contents described in the invention description, specific embodiments and drawings shall be used to interpret the claims.

[0103] Within the scope of the technical concept of this application, several modifications can be made to the specific implementation of this application, and these modified implementations should also be considered within the protection scope of this application.

Claims

1. A method for generating virtual esophageal barium contrast radiographs based on digital reconstruction X-ray image rendering using CT image segmentation, characterized in that, Includes the following steps: S1. Acquire CT images of the patient's chest or upper abdomen, and resample, standardize the orientation, and normalize the intensity of each frame of the CT image to form the original CT image. S2. Automatically segment the original CT image to initially segment the esophageal region; S3. The esophageal region of interest is segmented in one step, and the entire esophageal region is divided into three parts: lumen, wall and mucosa by a 3D slicer. S4. Secondary segmentation of the esophageal region of interest to obtain the segmentation results of the tumor; S5. Mask calculation: The masks of the lumen, wall and mucosa are obtained through step S3. The mask of the tumor is obtained through step S4. The lumen mask is subtracted from the tumor mask to obtain the real lumen. The real lumen area is subtracted from the entire esophageal area to obtain the real wall. S6. CT values ​​are filled for different parts of the esophagus to simulate the state of the esophageal filling phase and mucosal phase under different barium doses; S7. The original CT images are fused with the filling phase image and the mucosal phase image respectively to obtain the fused three-dimensional CT image. The propagation process of X-rays in the fused three-dimensional CT volume data is modeled using DRR technology to generate virtual esophageal barium meal images.

2. The method for generating virtual esophageal barium contrast radiographs based on digital reconstruction X-ray image rendering using CT image segmentation according to claim 1, characterized in that, Step S4 specifically includes the following steps: S41. Acquire plain CT images and enhanced CT images of the same subject, and perform spatial registration processing on the plain CT images and enhanced CT images to align them in the same spatial coordinate system; S42. Based on the registration, perform gray-level difference operation on the enhanced CT image and the plain CT image to obtain the gray-level difference volume before and after enhancement. The gray value of each voxel in the difference volume represents the degree of gray-level change of that voxel before and after enhancement. S43. On the gray-level difference volume, voxels are screened according to a preset gray-level difference threshold, and voxel regions with gray-level difference volume values ​​higher than the gray-level difference threshold are extracted to obtain initial enhancement candidate regions. S44. Limit the initial enhancement candidate region to the esophageal region and remove voxel regions located outside the esophageal region; S45. Morphological processing is performed on the initial enhancement candidate region located in the esophagus, including connected domain analysis, hole filling and smoothing, to eliminate noise voxels and enhance the spatial coherence of the region, and finally obtain the initial volume segmentation result of the esophageal tumor.

3. The method for generating virtual esophageal barium contrast radiographs based on digital reconstruction X-ray image rendering using CT image segmentation according to claim 1, characterized in that, Step S6 specifically includes the following steps: S61, Filling phase filling: The real lumen is introduced into the 3D-slicer, and different CT values ​​are applied to fill the lumen and the tube wall respectively. S62, Mucosal phase filling: The real tube wall and mucosa are introduced into the 3D-slicer, and CT values ​​are filled into the real tube wall and mucosa respectively.

4. The method for generating virtual esophageal barium contrast radiographs based on digital reconstruction X-ray image rendering using CT image segmentation according to claim 1, characterized in that, Step S7 specifically includes the following steps: S71, Spatial Modeling and Attenuation Attribute Mapping; S72, Imaging setup of the X-ray source and detector; S73, X-ray projection, generating virtual barium image.

5. The method for generating virtual esophageal barium contrast radiographs based on digital reconstruction X-ray image rendering using CT image segmentation according to claim 4, characterized in that, Step S71 specifically includes the following steps: S711. Load the fused 3D CT image into a regular 3D voxel grid to form 3D CT volume data. Each voxel contains spatial coordinate information and the corresponding CT gray value. S712, Coordinate System 1: Establish the spatial mapping relationship between the voxel coordinate system and the world coordinate system; S713, Attenuation Attribute Mapping: Maps CT grayscale values ​​to equivalent X-ray attenuation coefficients, enabling each voxel to participate in X-ray energy attenuation calculations.

6. The method for generating virtual esophageal barium contrast radiographs based on digital reconstruction X-ray image rendering using CT image segmentation according to claim 4, characterized in that, Step S72 specifically includes the following steps: S721, X-ray source settings: Define the position and orientation of a virtual X-ray source in three-dimensional space to simulate the emission point of a real X-ray tube; S722, Detector settings: Define the position, size, pixel resolution, and spatial relationship between the virtual plane detector and the radiation source; S723, Imaging geometry parameterization, sets the distance from the X-ray source to the detector, the direction of the main X-ray, and the angle rotation parameters to simulate the common anteroposterior, anterior oblique, and lateral projection conditions in clinical barium swallow examinations.

7. The method for generating virtual esophageal barium contrast radiographs based on digital reconstruction X-ray image rendering using CT image segmentation according to claim 4, characterized in that, Step S73 specifically includes the following steps: S731, Pixel-by-pixel ray generation: For each pixel on the virtual detector, calculate its corresponding ray direction vector and generate a ray from the virtual ray source; S732, voxel-level ray sampling, each ray is sampled along its propagation path in a three-dimensional voxel grid with a fixed step size, passing through multiple voxels in sequence; S733, Path integral calculation: At each voxel through which the ray passes, the product of the voxel's equivalent attenuation coefficient and the path length is accumulated to form the cumulative attenuation value of the ray. Based on the cumulative attenuation value of the ray, a virtual barium image is generated.

8. The method for generating virtual esophageal barium contrast radiographs based on digital reconstruction X-ray image rendering using CT image segmentation according to claim 4, characterized in that, S7 also includes the following steps: S74, Image Transformation, including: S741, Logarithmic Transformation: Convert the cumulative attenuation value of the ray into a grayscale image; S742, Intensity Normalization and Dynamic Range Adjustment, including light blur, first sharpening, then light blur again, second sharpening, and adaptive edge enhancement; S743, Two-dimensional image output.

9. The method for generating virtual esophageal barium contrast radiographs based on digital reconstruction X-ray image rendering using CT image segmentation according to claim 8, characterized in that, Step S713 specifically includes the following: In CT images, the grayscale value is defined as the difference between the linear attenuation coefficient of the tissue being measured and the linear attenuation coefficient of water, expressed as a percentage of the linear attenuation coefficient of water. Where μ is the linear attenuation coefficient of the voxel to be measured, and μ_water is the linear attenuation coefficient of water under a set effective photon energy, the formula for calculating the linear attenuation coefficient of the voxel can be obtained as: μ = μ_water × (1 + HU / 1000); the HU value in the CT volume data is converted into the equivalent X-ray linear attenuation coefficient distribution for each voxel.

10. The method for generating virtual esophageal barium contrast radiographs based on digital reconstruction X-ray image rendering using CT image segmentation according to claim 1, characterized in that, S3 specifically includes the following: In 3d-slicer, smooth the edges of the esophageal region to divide the lumen region. Import the entire esophageal region again, subtract the lumen region from the entire esophageal region to obtain the esophageal wall, import the entire esophageal region again, add a new outline layer, and use the threshold range tool to draw the esophageal mucosa within the esophageal region.

Images