Three-dimensional image post-processing method for improving cbct low-contrast tissue resolution

Abstract

This invention provides a three-dimensional image post-processing method for improving the resolution of low-contrast tissues in CBCT, comprising: obtaining three-dimensional volumetric data of the object under test from CBCT through three-dimensional reconstruction; performing three-dimensional null value fusion filtering and noise reduction processing on the three-dimensional volumetric data, including: setting a three-dimensional neighborhood window and using the three-dimensional neighborhood window to traverse each specified voxel to obtain the three-dimensional effective neighborhood of the specified voxel; calculating the joint weight of the spatial domain and the value domain for each neighboring voxel within the three-dimensional effective neighborhood of the specified voxel, and obtaining the filtered gray value of the specified voxel based on the weighted average, which is used as the value of the filtered three-dimensional data at the corresponding position; and performing three-dimensional differential edge sharpening processing on the filtered three-dimensional data to obtain enhanced three-dimensional data. The method of this invention enhances the edges of low-contrast tissues while efficiently reducing noise, improving their distinguishability from surrounding tissues and meeting the needs of clinical diagnosis.

Description

Technical Field

[0001] This invention belongs to the field of medical image processing technology, specifically relating to a three-dimensional image post-processing method for improving the resolution of low-contrast tissues in CBCT images. It is applicable to clinical scenarios such as oral surgery, orthopedics, and oncology, and optimizes the imaging of low-contrast tissues such as soft tissues and small lesions in CBCT images. Background Technology

[0002] CBCT (cone-beam computed tomography) technology has been widely used in clinical diagnosis due to its advantages such as small size, low radiation dose, and fast imaging speed. However, due to the limitations of the imaging principle, CBCT images are prone to noise interference, and low-contrast tissues (such as oral soft tissues, bone trabeculae, and small tumor lesions) have small grayscale differences with surrounding tissues and blurred boundaries, making it difficult for clinicians to accurately identify them and affecting the accuracy of diagnosis.

[0003] In existing CBCT image post-processing methods, noise reduction algorithms (such as Gaussian filtering) tend to cause loss of edge details and blurry images; while edge enhancement algorithms (such as the Sobel operator) amplify noise and cause image artifacts. At the same time, traditional post-processing methods are mostly based on two-dimensional image operations, ignoring the inter-slice correlation of CBCT three-dimensional volume data, resulting in limited processing effects and failing to meet the clinical needs for accurate identification of low-contrast tissues.

[0004] Therefore, there is an urgent need for a post-processing method that balances noise reduction, edge preservation, and enhancement, meets clinical requirements for low-contrast resolution, and is compatible with 3D data. Summary of the Invention

[0005] This invention aims to solve the problems of insufficient low-contrast tissue resolution and the contradiction between noise reduction and detail preservation in existing CBCT post-processing methods. It provides a post-processing method based on the synergy of three-dimensional null value fusion filtering and three-dimensional differential edge sharpening, which enhances the edges of low-contrast tissues while efficiently reducing noise, thereby improving their differentiation from surrounding tissues and meeting the needs of clinical diagnosis.

[0006] To achieve the above objectives, the present invention provides a three-dimensional image post-processing method for improving the resolution of low-contrast tissues in CBCT, characterized in that it includes:

[0007] S1: The 3D volume data of the object under test obtained by CBCT is obtained through 3D reconstruction;

[0008] S2: Denoising processing of 3D volume data by performing 3D null value fusion filtering;

[0009] Step S2 specifically includes:

[0010] S21: Set a three-dimensional neighborhood window and use the three-dimensional neighborhood window for each specified voxel. Perform traversal processing to obtain the specified voxel. The three-dimensional effective neighborhood;

[0011] S22: Targeting a specific voxel Each neighborhood voxel within the three-dimensional effective neighborhood Calculate the joint weights of the spatial domain and the value domain, and obtain the specified voxel based on the weighted average. Filtered grayscale value , which serves as the value of the filtered 3D data at the corresponding position;

[0012] S3: Perform three-dimensional differential edge augmentation processing on the filtered three-dimensional data to obtain enhanced three-dimensional data.

[0013] The window size of the first 3D neighborhood window is 9×9×9, and the window radius of the first 3D neighborhood window is... The boundary of the first three-dimensional effective neighborhood. , , for:

[0014] ,

[0015] ,

[0016] ,

[0017] ,

[0018] ,

[0019] ,

[0020] in, These represent the lengths of the three-dimensional volume data in the transverse, coronal, and sagittal dimensions, respectively. These represent the value ranges of the first three-dimensional effective neighborhood in the three pixel dimensions of transverse, coronal, and sagittal planes.

[0021] Neighboring voxels grayscale difference for:

[0022] ,in, Neighboring voxels grayscale value, For the current voxel The grayscale value; the joint weight of the spatial domain and the value range. for:

[0023] ,

[0024] in, The standard deviation of the spatial domain is squared; The standard deviation of the range is the square of the range. Neighboring voxels coordinates For a specified voxel The coordinates; Neighboring voxels The difference in grayscale values;

[0025] The obtained specified voxel Filtered grayscale value for:

[0026] ,

[0027] in, For a specified voxel The filtered grayscale value, , , s represents the boundary of the first three-dimensional effective neighborhood. The joint weight of the spatial domain and the range, Neighboring voxels grayscale value, Neighboring voxels The coordinates.

[0028] The squared standard deviation of the spatial domain is 8, and the squared standard deviation of the value range is 100.

[0029] Step S3 specifically includes:

[0030] S31: Set a second three-dimensional neighborhood window and use the second three-dimensional neighborhood window to obtain the specified voxel. The second three-dimensional effective neighborhood;

[0031] S32: Traverse the specified voxels The second three-dimensional effective neighborhood is used to calculate the gray-level difference of the neighborhood. ;

[0032] S33: Sum the gray-level differences of the neighborhood With preset enhancement coefficient Multiply and add to the corresponding specified voxel. The filtered grayscale values ​​are used to obtain the enhanced 3D data.

[0033] The size of the second 3D neighborhood window is 3×3×3, and the window radius of the second 3D neighborhood window is... .

[0034] The three-dimensional image post-processing method of the present invention for improving the resolution of low-contrast tissues in CBCT has the following beneficial effects:

[0035] 1) Significant resolution improvement: Through the cascaded operation of "noise reduction + enhancement", the three-dimensional null value fusion filter can accurately suppress noise without blurring the edges, and the three-dimensional differential edge enhancement can strengthen the gray-scale difference between low-contrast tissues and surrounding tissues, taking into account both noise reduction and edge enhancement, and helping clinically identify low-contrast tissues such as soft tissues and small lesions.

[0036] 2) Meets CBCT registration image quality requirements: When testing image quality using Catphan700, the number of low-contrast targets identified before applying the method of this invention was 0, while at least 4 targets could be identified after application, and high-contrast and other indicators were not affected. This shows that the method of this invention can enhance the edges of low-contrast tissues while efficiently reducing noise, improving their differentiation from surrounding tissues, and taking into account other image quality requirements such as CBCT registration image quality requirements. It can accurately optimize the imaging effect of various low-contrast tissues, assist doctors in achieving early disease diagnosis, lesion localization, and treatment plan formulation, and effectively reduce the risk of misdiagnosis and missed diagnosis caused by insufficient image contrast, which meets clinical image quality standards and actual clinical application scenarios.

[0037] 3) Strong adaptability: The method of this invention is specifically designed for CBCT three-dimensional volumetric data, fully considering the inter-slice correlation of transverse, coronal, and sagittal views. It features proprietary neighborhood rules, parameter systems, and collaborative logic, making it more suitable for the three-dimensional data characteristics of CBCT compared to two-dimensional processing methods. It can be widely applied to CBCT images from different locations and with different scanning parameters, demonstrating strong versatility. Existing filters are mostly applied to two-dimensional scenes and require superposition with other filters. This invention designs proprietary three-dimensional neighborhood windows and boundary constraint rules, forming a unique spatial-intensity dual-weight calculation system, thereby enabling three-dimensional data processing without significant distortion.

[0038] 4) High flexibility: In the method of this invention, the core parameters (such as neighborhood window size, standard deviation, and enhancement coefficient) can be flexibly adjusted according to different clinical scenarios such as oral cavity, orthopedics, and oncology, so that clinicians can quickly adapt to different scenarios according to the needs of image reading, taking into account both universality and specificity, and meeting diverse diagnostic needs.

[0039] 5) Simple and efficient method: The method of this invention adopts a two-step cascade of three-dimensional null value fusion filtering and three-dimensional differential edge amplification. The logic is clear, no complex calculation support is required, it is easy to integrate into the post-processing system of existing CBCT equipment, no major hardware modification is required, and it has strong clinical applicability. Attached Figure Description

[0040] Figure 1This is a flowchart of the three-dimensional image post-processing method for improving the resolution of low-contrast tissues in CBCT according to the present invention.

[0041] Figure 2 This is a low-contrast test layer image of Catphan700, which was not processed using the method of this invention (before optimization).

[0042] Figure 3 This is a low-contrast test layer image of Catphan700, which uses the method of this invention (optimized).

[0043] Figure 4 This is a high-contrast test layer image of Catphan700, which was not processed using the method of this invention (before optimization).

[0044] Figure 5 This is a high-contrast test layer image of Catphan700, which uses the method of this invention (optimized). Detailed Implementation

[0045] The preferred embodiments of the present invention are given below with reference to the accompanying drawings and described in detail.

[0046] like Figure 1 As shown, according to an embodiment of the present invention, a three-dimensional image post-processing method for improving the resolution of low-contrast tissues in CBCT specifically includes:

[0047] Step S1: 3D reconstruction yields 3D volumetric data of the object under test obtained from CBCT (cone-beam computed tomography).

[0048] Step S1 specifically includes: acquiring the raw CBCT projection data of the object to be tested, obtaining the 3D volume data of the CBCT through a 3D reconstruction algorithm, denoted as the 3D volume data matrix InputData[X×Y×Z], where X, Y, and Z are the lengths of the 3D volume data in the transverse, coronal, and sagittal dimensions, respectively, and the data format is floating-point, preparing for subsequent 3D post-processing. The object to be tested is usually the patient's oral cavity, bones, or other parts of the body.

[0049] Three-dimensional reconstruction algorithms are used to reconstruct three-dimensional volume data, and these algorithms can be FDK, SIRT, SART, T-FDK, etc.

[0050] Step S2: Perform 3D null value fusion filtering on the 3D volume data to reduce noise while preserving tissue edge details;

[0051] Step S2 specifically includes:

[0052] Step S21: Set a first three-dimensional neighborhood window, and use the first three-dimensional neighborhood window for each specified voxel. Perform traversal processing to obtain the specified voxel. The first three-dimensional effective neighborhood;

[0053] To achieve noise reduction, the window size of the first 3D neighborhood window is 9×9×9 (window side length). (9), that is, the window radius (The relationship between the window radius r and the window side length W is W = 2 × r + 1). Using too low a value will weaken the low-contrast rendering effect; using too high a value may result in over-filtering. Boundary constraint rules determine the effective neighborhood range to avoid distortion in boundary pixel calculations.

[0054] For a given voxel The boundary of its first three-dimensional effective neighborhood , , for:

[0055] ,

[0056] ,

[0057] ,

[0058] ,

[0059] ,

[0060] ,

[0061] in, These represent the lengths of the three-dimensional volume data in the transverse, coronal, and sagittal dimensions, respectively. These represent the value ranges of the first three-dimensional effective neighborhood in the three pixel dimensions of transverse, coronal, and sagittal planes.

[0062] The purpose of the first three-dimensional effective neighborhood is to reduce computational load and avoid distortion of boundary values. Image data has boundaries; without neighborhood checks, the program will crash when it exceeds these boundaries. Pre-setting the boundaries of the first three-dimensional effective neighborhood can reduce unnecessary computation to some extent. Furthermore, conventional solutions often use zero values / NAN / other constant values ​​as substitutes when values ​​exceed the boundaries, since no values ​​actually exist outside the boundaries. However, this causes abrupt changes in the boundary values, resulting in distortion.

[0063] Step S22: Two-dimensional weight calculation and filtering output: for a specified voxel Each neighborhood voxel within the first three-dimensional effective neighborhood Calculate the joint weights of the spatial domain and the value domain, and obtain the specified voxel based on the weighted average. Filtered grayscale value As filtered 3D data The value taken at the corresponding position.

[0064] Neighboring voxels grayscale difference for:

[0065] ,

[0066] in, Neighboring voxels grayscale value, For the current voxel The grayscale value.

[0067] The grayscale values ​​here do not need to be normalized. This is because the reconstruction results output earlier have already undergone HU correction, and the values ​​already conform to the medical physical attenuation model. Therefore, the standard deviation is set to a fixed value of 100.

[0068] Joint weight of spatial domain and range for:

[0069] ,

[0070] in, This represents the squared standard deviation of the spatial domain, with a default value of 8. This represents the squared standard deviation of the range, with a default value of 100. Neighboring voxels coordinates For a specified voxel The coordinates; Neighboring voxels The grayscale difference. This joint weight. It balances spatial location correlation with grayscale consistency, achieving a balance between noise suppression and edge preservation.

[0071] In this embodiment, the squared standard deviation of the spatial domain A value of 8 indicates that neighboring voxels and the central voxel have relatively higher weights if they are within 8 voxel units of each other; otherwise, their weights are negligible. (Value range, standard deviation, squared) A value of 100 means that if the grayscale value of neighboring voxels and the central voxel is less than 100, the weight will be higher for smoothing. If it is greater than 100, it may represent other unrelated tissues or edges, so the weight will be lowered to protect the edges. 8 and 100 are suitable values ​​verified by actual data acquisition from the Aipuqiang proton radiotherapy CBCT system. They are compatible with most scenarios and are set based on practical experience. The values ​​can be fine-tuned for different treatment scenarios.

[0072] The specified voxel is obtained based on the weighted average. Filtered grayscale value for:

[0073] ,

[0074] in, For a specified voxel The filtered grayscale value, , , s represents the boundary of the first three-dimensional effective neighborhood. The joint weight of the spatial domain and the range, Neighboring voxels grayscale value, Neighboring voxels The coordinates.

[0075] Step S3: Process the filtered 3D data Three-dimensional differential edge augmentation is performed to obtain enhanced three-dimensional data. .

[0076] Therefore, the filtered 3D data By using three-dimensional differential edge sharpening to enhance the edges of low-contrast tissues, tissue differentiation is improved, making the results more obvious to the naked eye.

[0077] Step S3 specifically includes:

[0078] Step S31: Set a second three-dimensional neighborhood window and use the second three-dimensional neighborhood window to obtain the specified voxel. The second three-dimensional effective neighborhood; thus, distortion in boundary voxel calculation is avoided;

[0079] The second three-dimensional neighborhood window has a window size of 3×3×3, which is the window radius. This allows for coverage of adjacent voxels in three dimensions: transverse, coronal, and sagittal.

[0080] For a given voxel The boundary of the second three-dimensional effective neighborhood , , for:

[0081] ,

[0082] ,

[0083] ,

[0084] ,

[0085] ,

[0086] ,

[0087] in, These represent the lengths of the three-dimensional volume data in the transverse, coronal, and sagittal dimensions, respectively. These represent the value ranges of the second three-dimensional effective neighborhood in the transverse, coronal, and sagittal dimensions, respectively.

[0088] Step S32: Solving the neighborhood gray-level difference sum: Traversing the specified voxels The second three-dimensional effective neighborhood is used to calculate the gray-level difference of the neighborhood. The gray-level difference in this neighborhood It can accurately characterize the gray level difference between the current voxel and its neighbors, with significant gray level differences in the edge region and obvious gray level differences and features in the neighborhood.

[0089] Calculate the gray-level difference of the neighborhood Specifically, this includes: statistical analysis of specified voxels. Total number of voxels in the second three-dimensional effective neighborhood Accumulate all neighborhood voxels within the second three-dimensional effective neighborhood. The gray value is used to calculate the gray-level difference of the neighborhood. .

[0090] Neighborhood gray-level difference sum for:

[0091] ,

[0092] in, For a specified voxel The total number of voxels in the second three-dimensional effective neighborhood. For a specified voxel The filtered grayscale value, i.e. Corresponding values; , , This represents the boundary of the second three-dimensional effective neighborhood. This represents the filtered grayscale value of the voxels within the neighborhood.

[0093] Step S33: Edge enhancement operation: sum the gray-level differences of the neighborhood With preset enhancement coefficient Multiply and add to the corresponding specified voxel. The filtered grayscale values ​​are used to obtain the enhanced 3D data. This enhances the grayscale difference in the edge areas, achieving clearer boundaries for low-contrast tissues.

[0094] Preset enhancement coefficient The default value is 0.01, which can be flexibly adjusted according to the clinical tissue type. 0.01 is a value that has been tested and verified in practice. Too low a value will have no effect, while too high a value will make the edges too obvious and thus distort the image, affecting the clinical judgment of the client (doctor).

[0095] Enhanced 3D data In the specified voxel The value at this location is:

[0096] ,

[0097] in, To enhance the post-3D data In the specified voxel The value at that location, For a specified voxel The filtered grayscale value, For the preset enhancement coefficient, It represents the sum of gray-level differences in the neighborhood.

[0098] Step S4: Output the enhanced 3D data : For the enhanced 3D data The format is converted and output to a clinically compatible format, which can then be used for subsequent operations such as clinical image reading and lesion annotation, completing the entire post-processing workflow.

[0099] The default values ​​and functions of the core parameters mentioned above are shown in Table 1.

[0100] Table 1: Default values ​​and functions of core parameters

[0101]

[0102] The adjustment logic for different scenarios is as follows:

[0103] Oral / Orthopedic: For hard tissues with clear margins, the range should be appropriately expanded (spatial domain standard deviation squared 8-10) to avoid losing details. For mild noise (such as dental scans), the spatial domain standard deviation squared should be 6-8. For tumors: Lesions are mostly low-contrast soft lesions, so the spatial neighborhood needs to be more refined (spatial domain standard deviation squared 4-8) to avoid blurring the spatial location relationship of small lesions.

[0104] Oral / Orthopedic: Large grayscale difference between soft and hard tissues (teeth / bone vs. mucosa), expanding the value range (value range standard deviation squared 100-120) can preserve grayscale differences. For mild noise in periodontal soft tissues, the value range standard deviation squared 80-100 is used. Tumor: Small grayscale difference between lesions and surrounding tissues, narrowing the value range (value range standard deviation squared 50-100) can accurately smooth noise while preserving slight grayscale differences in lesions.

[0105] Oral / Orthopedic: Hard tissue edges have some grayscale differences, so slight enhancement is sufficient (enhancement coefficient 0.008-0.01). Low-contrast soft tissues such as periodontal tissues and mucosa can be moderately enhanced (enhancement coefficient 0.01-0.015). Tumors: Small lesions / infiltrates have extremely blurred boundaries with surrounding tissues, so the enhancement intensity needs to be moderately increased (enhancement coefficient 0.01-0.02) to enhance the slight grayscale differences. Since the noise has been suppressed after filtering, a slight enhancement will not amplify the noise.

[0106] Experimental results:

[0107] In step S1, the three-dimensional volume data of CBCT is obtained: FDK reconstruction is performed on the oral CBCT scan data to obtain 512×512×512 dimension floating-point three-dimensional volume data, which is used as the processing input data.

[0108] In step S2, the three-dimensional null value fusion filtering configuration is as follows: the size of the three-dimensional neighborhood window is set to 9×9×9, the spatial domain standard deviation is 8, and the value domain standard deviation is 100. The input data is subjected to three-dimensional null value fusion filtering to obtain the filtered data.

[0109] In step S3, the three-dimensional differential edge amplification configuration is as follows: the size of the three-dimensional neighborhood window is set to 3×3×3, the enhancement coefficient is set to 0.01, and three-dimensional differential edge amplification processing is performed based on the filtered data to enhance the boundary grayscale differences of different tissues and obtain the enhanced data.

[0110] Output results: Save the enhanced data in clinical data format.

[0111] Comparison of the effects of the method of the present invention with those of the prior art Figures 2-5 As shown.

[0112] Figure 2 This is a low-contrast test layer image of Catphan700, which does not use the method of this invention. Figure 3 This is a low-contrast test layer image of Catphan700, which uses the method of this invention. Figure 3 In the middle, circles of different sizes are faintly visible on the ring, with the largest at the bottom and decreasing in size as you move left (observe from a distance). The registration review requires that only the bottom circle be visible to pass. Figure 2 If only a blurry, indistinct image can be seen at the same location, this situation will not pass the registration review, as the low contrast index is considered unacceptable.

[0113] Figure 4 This is a high-contrast test layer image of Catphan700, which does not use the method of this invention. Figure 5 This is a test layer image of Catphan700 with high contrast, which utilizes the method of this invention. Figure 4 , Figure 5 The upper part is the high-contrast index test chart. A line pair refers to a black line plus a white line in the chart. The top line is the thickest and easiest to identify. Rotating clockwise downwards reveals different recognition levels. Figure 4 , Figure 5 The levels of the midline pairs are 1, 2, 3, 4, 5, 6, 7, 8, 9, 10... in clockwise order. Level 7 meets the minimum requirements for registration review, and levels above that reflect the superior performance of the algorithm. Figure 4 , Figure 5 The lower half of the image is a waveform diagram used to assist in observation to determine whether the line pairs can be identified. Draw a line on the corresponding area of ​​the image. Since the visual observation is already very blurry at this point, as long as the numerical fluctuations on this line can be clearly seen to rise or fall, it is considered that this level can also be identified. Figure 4 , Figure 5 This demonstrates that using this patented technology improves the recognizability of line pairs by one level. Conventionally, with common image post-processing methods, such as noise reduction, images often become blurred, and high contrast usually results in poorer recognition. However, this patent balances noise reduction and enhancement effects, achieving not only no difference but actually stronger results, showcasing the excellent performance of the algorithm.

[0114] The core technology of this invention is a cascaded processing flow of "three-dimensional null value fusion filtering and noise reduction → three-dimensional differential edge enhancement". First, three-dimensional null value fusion filtering and noise reduction is performed to enhance low-contrast tissues while protecting the edge morphology of each tissue. Then, three-dimensional differential edge enhancement is used to enhance the edges, making the results more obvious to the naked eye. Thus, for the three-dimensional volume data reconstructed by CBCT, a two-step collaborative operation is designed based on the inter-layer correlation characteristics of three-dimensional data to improve the resolution of low-contrast tissues.

[0115] The three-dimensional image post-processing method of the present invention for improving the resolution of low-contrast tissues in CBCT has the following beneficial effects:

[0116] 1) Significant resolution improvement: Through the cascaded operation of "noise reduction + enhancement", the three-dimensional null value fusion filter can accurately suppress noise without blurring the edges, and the three-dimensional differential edge enhancement can strengthen the gray-scale difference between low-contrast tissues and surrounding tissues, taking into account both noise reduction and edge enhancement, and helping clinically identify low-contrast tissues such as soft tissues and small lesions.

[0117] 2) Meets CBCT registration image quality requirements: When testing image quality using Catphan700, the number of low-contrast targets identified before applying the method of this invention was 0, while at least 4 targets could be identified after application, and high-contrast and other indicators were not affected. This shows that the method of this invention can enhance the edges of low-contrast tissues while efficiently reducing noise, improving their differentiation from surrounding tissues, and taking into account other image quality requirements such as CBCT registration image quality requirements. It can accurately optimize the imaging effect of various low-contrast tissues, assist doctors in achieving early disease diagnosis, lesion localization, and treatment plan formulation, and effectively reduce the risk of misdiagnosis and missed diagnosis caused by insufficient image contrast, which meets clinical image quality standards and actual clinical application scenarios.

[0118] 3) Strong adaptability: The method of this invention is specifically designed for CBCT three-dimensional volumetric data, fully considering the inter-slice correlation of transverse, coronal, and sagittal views. It features proprietary neighborhood rules, parameter systems, and collaborative logic, making it more suitable for the three-dimensional data characteristics of CBCT compared to two-dimensional processing methods. It can be widely applied to CBCT images from different locations and with different scanning parameters, demonstrating strong versatility. Existing filters are mostly applied to two-dimensional scenes and require superposition with other filters. This invention designs proprietary three-dimensional neighborhood windows and boundary constraint rules, forming a unique spatial-intensity dual-weight calculation system, thereby enabling three-dimensional data processing without significant distortion.

[0119] 4) High flexibility: In the method of this invention, the core parameters (such as neighborhood window size, standard deviation, and enhancement coefficient) can be flexibly adjusted according to different clinical scenarios such as dentistry and orthopedics, so that clinicians can quickly adapt to different scenarios according to the needs of image reading, taking into account both universality and specificity, and meeting diverse diagnostic needs.

[0120] 5) Simple and efficient method: The method of this invention adopts a two-step cascade of three-dimensional null value fusion filtering and three-dimensional differential edge amplification. The logic is clear, no complex calculation support is required, it is easy to integrate into the post-processing system of existing CBCT equipment, no major hardware modification is required, and it has strong clinical applicability.

[0121] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of the invention. Various variations can be made to the above embodiments of the present invention. All simple and equivalent changes and modifications made in accordance with the claims and description of this application fall within the protection scope of the claims of this patent. All aspects not described in detail in this invention are conventional technical content.

Claims

1. A three-dimensional image post-processing method for improving the resolution of low-contrast tissues in CBCT, characterized in that, include: Step S1: 3D reconstruction obtains the 3D volumetric data of the object under test using CBCT; Step S2: Perform noise reduction processing on the 3D volume data using 3D null value fusion filtering; Step S2 specifically includes: Step S21: Set up a three-dimensional neighborhood window and use the three-dimensional neighborhood window for each specified voxel. Perform traversal processing to obtain the specified voxel. The three-dimensional effective neighborhood; Step S22: Targeting the specified voxel Each neighborhood voxel within the three-dimensional effective neighborhood Calculate the joint weights of the spatial domain and the value domain, and obtain the specified voxel based on the weighted average. Filtered grayscale value , which serves as the value of the filtered 3D data at the corresponding position; Step S3: Perform three-dimensional differential edge augmentation processing on the filtered three-dimensional data to obtain the enhanced three-dimensional data.

2. The three-dimensional image post-processing method for improving the resolution of low-contrast tissues in CBCT according to claim 1, characterized in that, The dimensions of the CBCT 3D volumetric data of the object under test are 512×512×512; the window size of the first 3D neighborhood window is 9×9×9, and the window radius of the first 3D neighborhood window is... ; Boundary of the first three-dimensional effective neighborhood , , for: ， ， ， ， ， ， in, These represent the lengths of the three-dimensional volume data in the transverse, coronal, and sagittal dimensions, respectively. These represent the value ranges of the first three-dimensional effective neighborhood in the three pixel dimensions of transverse, coronal, and sagittal planes.

3. The three-dimensional image post-processing method for improving the resolution of low-contrast tissues in CBCT according to claim 1, characterized in that, Neighboring voxels grayscale difference for: ， in, Neighboring voxels grayscale value, For the current voxel grayscale value; Joint weight of spatial domain and range for: ， in, The standard deviation of the spatial domain is squared; The standard deviation of the range is the square of the range. Neighboring voxels coordinates For a specified voxel The coordinates; Neighboring voxels The difference in grayscale values; The obtained specified voxel Filtered grayscale value for: ， in, For a specified voxel The filtered grayscale value, , , s represents the boundary of the first three-dimensional effective neighborhood. The joint weight of the spatial domain and the range, Neighboring voxels grayscale value, Neighboring voxels The coordinates.

4. A three-dimensional image post-processing method for improving the resolution of low-contrast tissues in CBCT according to claim 3, characterized in that, The squared standard deviation of the spatial domain is 8, and the squared standard deviation of the value range is 100.

5. A three-dimensional image post-processing method for improving the resolution of low-contrast tissues in CBCT according to claim 1, characterized in that, Step S3 specifically includes: Step S31: Set a second three-dimensional neighborhood window and use the second three-dimensional neighborhood window to obtain the specified voxel. The second three-dimensional effective neighborhood; Step S32: Traverse the specified voxels The second three-dimensional effective neighborhood is used to calculate the gray-level difference of the neighborhood. ; Step S33: Sum the grayscale differences of the neighborhood With preset enhancement coefficient Multiply and add to the corresponding specified voxel. The filtered grayscale values ​​are used to obtain the enhanced 3D data.

6. A three-dimensional image post-processing method for improving the resolution of low-contrast tissues in CBCT according to claim 5, characterized in that, The size of the second 3D neighborhood window is 3×3×3, and the window radius of the second 3D neighborhood window is... .

7. A three-dimensional image post-processing method for improving the resolution of low-contrast tissues in CBCT according to claim 5, characterized in that, Calculate the gray-level difference of the neighborhood Specifically, this includes: statistical analysis of specified voxels. Total number of voxels in the second three-dimensional effective neighborhood Accumulate all neighborhood voxels within the second three-dimensional effective neighborhood. The gray value is used to calculate the gray-level difference of the neighborhood. .

8. A three-dimensional image post-processing method for improving the resolution of low-contrast tissues in CBCT according to claim 7, characterized in that, Neighborhood gray-level difference sum for: ， in, For a specified voxel The total number of voxels in the second three-dimensional effective neighborhood. For a specified voxel The filtered grayscale value; , , This represents the boundary of the second three-dimensional effective neighborhood. The filtered grayscale value of the voxels in the neighborhood.

9. A three-dimensional image post-processing method for improving the resolution of low-contrast tissues in CBCT according to claim 5, characterized in that, Preset enhancement coefficient It is 0.

01.

10. A three-dimensional image post-processing method for improving the resolution of low-contrast tissues in CBCT according to claim 1, characterized in that, Also includes: For the enhanced 3D data Perform format conversion and output in a clinically compatible format.

Images