Flat-scanning CT skeleton segmentation and three-dimensional reconstruction method based on KNN multi-dimensional data

By employing a multi-dimensional data processing method based on the KNN algorithm, combined with image processing technology, CT images are preprocessed and reconstructed in three dimensions. This addresses the issues of insufficient accuracy and automation in 3D bone reconstruction, achieving efficient and accurate bone segmentation and 3D reconstruction, thus enhancing its auxiliary role in clinical diagnosis and treatment.

CN119048537APending Publication Date: 2024-11-29ZHUHAI HENGQIN ALL-STAR MEDICAL TECHNOLOGY CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411248833.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-06
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

Current medical imaging technologies lack the precision and automation required for 3D bone reconstruction, failing to meet the high standards of clinical diagnosis.

Method used

A multi-dimensional data processing method based on the K-Nearest Neighbor (KNN) algorithm is adopted, combined with image processing technology to preprocess CT images, automatically identify and segment bone structures, and generate an optimized bone model through a three-dimensional reconstruction system.

Benefits of technology

It significantly improves the accuracy and automation of 3D bone reconstruction, providing a more precise and intuitive view of bone structure, and assisting in the formulation of clinical diagnosis and treatment plans.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119048537A_ABST
    Figure CN119048537A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of medical image processing and computer-aided diagnosis, in particular to a plain-scan CT skeleton segmentation and three-dimensional reconstruction method based on KNN multi-dimensional data. The method comprises the following steps: 1) preprocessing a CT image by using an image processing technology to highlight the ROI of a skeleton; 2) automatically identifying and segmenting a skeleton structure in the CT image by applying a K-nearest neighbor (KNN) algorithm; 3) performing enhancement processing on the segmented image, and constructing a three-dimensional form of a skeleton from different directions through a multi-view reconstruction technology; and 4) the three-dimensional reconstruction system performs three-dimensional reconstruction on the segmented and labeled image data, and performs optimization display on the generated model. According to the invention, through accurate three-dimensional modeling of the skeleton, a doctor can more clearly master the mutual positions of human anatomical structures, so that the safety and the treatment effect of an operation are improved, and the communication between the doctor and a patient is promoted.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of medical imaging technology, and in particular to an innovative method for plain scan CT bone segmentation and three-dimensional reconstruction based on KNN multi-dimensional data. Background Art

[0002] The present invention belongs to the field of medical image processing technology, and in particular refers to a multi-dimensional data processing technology based on the K nearest neighbor (KNN) algorithm, which is used for bone segmentation and three-dimensional reconstruction in plain scan CT images. With the advancement of medical technology, traditional two-dimensional imaging technology has gradually been unable to meet the high standards of clinical accurate diagnosis. Therefore, three-dimensional reconstruction technology has emerged, which can provide a more detailed and intuitive view of the human skeletal structure. The present invention receives image data obtained by CT scanning, and uses efficient image processing technology and machine learning algorithms, especially the KNN algorithm, to achieve automatic segmentation of bones under multi-dimensional data and accurate three-dimensional model construction. The innovative application of this technology not only improves the automation level of medical image processing, but also significantly enhances its accuracy, providing strong technical support for the formulation of clinical diagnosis and treatment plans. Summary of the invention

[0003] The present invention aims to overcome the shortcomings of existing medical imaging technology in three-dimensional reconstruction of bones, and proposes a multi-dimensional data processing method based on the K nearest neighbor (KNN) algorithm, which is used to realize accurate segmentation and three-dimensional reconstruction of bones from plain scan CT images. The method uses image processing and machine learning technology to automatically perform bone structure recognition and modeling by analyzing CT image data, thereby providing doctors with an intuitive and accurate view of the bone spatial structure. This method not only helps doctors to accurately locate and evaluate diseases, but also helps to formulate more accurate surgical plans, while improving patients' understanding of their own conditions. The implementation of the present invention has demonstrated its broad prospects and practical application value in clinical applications, and has brought innovation and progress to the field of medical image processing.

[0004] The present invention adopts the following technical scheme for the plain scan CT bone segmentation and three-dimensional reconstruction method based on KNN multi-dimensional data:

[0005] A method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data, comprising the following steps:

[0006] 1) Use image processing technology to preprocess CT images to highlight the bone ROI;

[0007] 2) Apply the K-nearest neighbor (KNN) algorithm to automatically identify and segment the bone structure in CT images;

[0008] 3) Enhance the segmented image;

[0009] 4) The 3D reconstruction system reconstructs the segmented and annotated image data in 3D and optimizes the display of the generated model.

[0010] Preferably, step 1) performs Gaussian smoothing filtering on the CT plain scan image as part of image preprocessing, namely:

[0011]

[0012] Preferably, in step 1), the contrast enhancement process is performed on the CT plain scan image as part of the image preprocessing to ensure

[0013] High contrast of the image, that is:

[0014] I'=αI+β (2);

[0015] Preferably, the image preprocessing in step 1) includes a histogram equalization operation, and the equalization operation is performed according to the histogram of the CT plain scan data, that is:

[0016]

[0017] Preferably, the image preprocessing in step 1) includes a Canny edge check operation, and a Canny edge check operation is performed on the CT plain scan data after the histogram is equalized, that is:

[0018] Edge=Canny(I,σ,T1,T2)(4);

[0019] Preferably, the image preprocessing in step 1) includes a sharpening enhancement operation, which is performed on the CT plain scan data after Canny edge detection, namely:

[0020] I sharpened (x,y)=I(x,y)+λL(x,y)(5);

[0021] Preferably, the operation of automatically identifying and segmenting the bone structure in the CT image in step 2) extracts multi-dimensional data information of the CT image, namely:

[0022] Feature=[x,y,z,I(x,y,z)](6);

[0023] Preferably, the operation of automatically identifying and segmenting the bone structure in the CT image in step 2) uses a KNN algorithm to classify the feature vectors, namely:

[0024] C(Feature)=KNN(Feature,K,Training Data)(7);

[0025] Preferably, the operation of automatically identifying and segmenting the bone structure in the CT image in step 2) automatically segments the bone structure according to the classification result of the KNN algorithm, that is:

[0026] S bone ={(x,y,z)|C(Feature)=bone}(8);

[0027] Preferably, in step 3), the segmented image is enhanced by using a morphological closing operation to optimize the segmentation effect of the skeleton, and the formula is expressed as follows:

[0028] D(I,B)(x)=max{I(y):y∈neighborhood(x,B)}(9);

[0029] E(I,B)(x)=min{I(y):y∈neighborhood(x,B)}(10);

[0030] Substitute (9)(10) into (11)

[0031] closing(I)=D(E(I,B))(11);

[0032] Preferably, in step 4), the segmented and annotated image data is 3D reconstructed. First, a volume rendering method is used to stack a series of segmented 2D slices to form a 3D volume data. In the process, linear interpolation is used to generate pixel values ​​between slices to ensure the smoothness of the 3D reconstruction. Linear interpolation formula:

[0033]

[0034] Preferably, in step 4), the segmented and annotated image data is reconstructed in three dimensions, and an isosurface is extracted from the three-dimensional volume data using the MarchingCubes algorithm to form a three-dimensional model. The isosurface is generated by interpolating all points in the three-dimensional data that exceed a certain threshold, that is:

[0035] V(p i )≥V iso or V(p i )≤V iso (13);

[0036] Preferably, in step 4), the segmented and annotated image data is 3D reconstructed, and the unnecessary black part (background) is removed by setting a threshold, and the pixels of the target organ are retained, that is:

[0037]

[0038] Preferably, in step 4), the segmented and annotated image data is three-dimensionally reconstructed, and the density value in the volume data is mapped to color and transparency through linear color mapping for visualization of the three-dimensional model, that is:

[0039] C=C min +d norm ×(C max -C min ) (15);

[0040] α=α min +d norm ×(α max -α min ) (16);

[0041] The beneficial effects of the present invention are:

[0042] 1. Improve the accuracy and reliability of 3D reconstruction of bones: The present invention significantly improves the accuracy of 3D reconstruction of bones by adopting a multi-dimensional data processing method based on the K nearest neighbor (KNN) algorithm. When processing large amounts of data, the KNN algorithm can effectively identify and distinguish bones from other tissues, thereby achieving more accurate bone segmentation. This high-precision segmentation is the basis for subsequent 3D reconstruction work, ensuring the accuracy and reliability of the reconstructed model. In addition, the present invention further optimizes the 3D morphology of bones through image enhancement processing and multi-view reconstruction technology, making the final 3D model richer in details, providing a more accurate reference for clinical diagnosis.

[0043] 2. Automated and efficient image processing flow: The present invention realizes a fully automated process from image preprocessing to three-dimensional reconstruction, greatly improving processing efficiency. In the preprocessing stage, the region of interest (ROI) of the skeleton is highlighted through image processing technology, laying a good foundation for subsequent skeleton segmentation and modeling. The introduction of the KNN algorithm makes the automatic recognition and segmentation process of the skeleton structure more efficient, reducing the need for manual intervention. In addition, the present invention also realizes the effective integration of image information from different perspectives by dynamically adjusting the weight distribution in the image fusion process, further improving the automation and efficiency of the processing flow.

[0044] 3. Enhance the auxiliary role of clinical diagnosis and treatment planning: The three-dimensional bone model provided by the present invention provides powerful assistance for doctors to locate, evaluate and formulate surgical plans for diseases. Compared with traditional two-dimensional imaging technology, three-dimensional reconstruction technology can provide more intuitive and detailed bone structure information, which helps doctors to more accurately understand the spatial relationship and degree of lesions. In addition, the present invention can also optimize the display of the three-dimensional model according to clinical needs, so that doctors can more clearly observe key anatomical structures and lesion characteristics, thereby formulating more accurate and personalized treatment plans.

[0045] 4. Promote the innovation and development of medical image processing technology: The present invention introduces a multi-dimensional data processing method based on the KNN algorithm in the field of medical image processing, providing a new technical path for skeletal three-dimensional reconstruction. This method not only improves the accuracy of bone segmentation and three-dimensional reconstruction, but also significantly improves the processing efficiency through automated and efficient image processing processes. In addition, the present invention further enriches the details and expressiveness of the three-dimensional model through multi-perspective reconstruction technology and model optimization display. These innovations not only improve the technical content of the present invention, but also provide new ideas and directions for the development of medical image processing technology, and have important scientific research value and application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0047] Figure 1 It is a flow chart of a method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data of the present invention;

[0048] Figure 2 It is the plain scan CT human skeleton segmentation map of the present invention;

[0049] Figure 3 A schematic diagram of the skeleton modeling of the present invention;

[0050] Figure 4 This is a schematic diagram of modeling after the skeleton and CT image are registered in the present invention; DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0052] An embodiment of a method for plain scan CT bone segmentation and three-dimensional reconstruction based on KNN multi-dimensional data of the present invention is as follows: Figure 1 As shown, including:

[0053] See also Figure 1 The method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data includes the following steps:

[0054] 1) Use image processing technology to preprocess CT images to highlight the bone ROI;

[0055] 2) Apply the K-nearest neighbor (KNN) algorithm to automatically identify and segment the bone structure in CT images;

[0056] 3) Enhance the segmented image;

[0057] 4) The 3D reconstruction system reconstructs the segmented and annotated image data in 3D and optimizes the display of the generated model.

[0058] Furthermore, step 1) performs Gaussian smoothing filtering on the CT plain scan image, which is a part of image preprocessing, namely:

[0059]

[0060] Where (x, y) represents the distance from the center point, and σ is the standard deviation, which controls the degree of smoothing.

[0061] Furthermore, step 1) performs contrast enhancement on the CT plain scan image, which is the next part of the image preprocessing after the Gaussian smoothing filter to ensure high contrast of the image, namely:

[0062] I'=αI+β (2);

[0063] Where I represents the original image, I' is the enhanced image, α is the contrast control parameter, and β is the brightness control parameter.

[0064] Furthermore, step 1) performs histogram equalization on the CT plain scan image, which is the next part of the image preprocessing after contrast enhancement, and its purpose is to make the brightness distribution of the image more uniform, that is:

[0065]

[0066] Where L is the number of gray levels, T is the cumulative distribution function, g -1 It is the inverse of the equalized cumulative distribution function.

[0067] Furthermore, step 1) performs Canny edge detection on the CT plain scan image, which is the next part of the image preprocessing after histogram equalization, and is used for the blurred edge of the image, namely:

[0068] Edge=Canny(I,σ,T1,T2)(4);

[0069] Where I is the input image, σ is the standard deviation of the Gaussian filter, and T1 and T2 are double thresholds for edge connections.

[0070] Furthermore, step 1) performs sharpening enhancement on the CT plain scan image, which is the next part of the processing after Canny edge detection in image preprocessing, and is used to enhance the details and edges in the image, namely:

[0071] I sharpened (x,y)=I(x,y)+λL(x,y)(5);

[0072] Among them I sharpened (x, y) is the gray value of the sharpened image at (x, y), λ is a positive sharpening strength parameter, I(x, y) represents the gray value of the image at the coordinate (x, y), and L(x, y) represents the Laplace response at the coordinate (x, y).

[0073] Furthermore, step 2) extracts multi-dimensional data information of the CT image from the CT plain scan image, so as to facilitate the subsequent automatic recognition and segmentation of the bone structure in the CT image, namely:

[0074] Feature=[x,y,z,I(x,y,z)](6);

[0075] Among them, Feature is the extracted multi-dimensional information, (x, y, z) is the three-dimensional coordinate information, and I(x, y, z) is the grayscale value corresponding to the (x, y, z) three-dimensional coordinates.

[0076] Further, in step 2), the multi-dimensional data information of the CT image extracted from the CT plain scan image is classified using the KNN algorithm, that is:

[0077] C(Feature)=KNN(Feature,K,Training Data)(7);

[0078] Among them, C (Feature) is the classification result, K is the number of neighbors, and Training Data is the training data set.

[0079] Furthermore, in step 2), the multi-dimensional data information of the CT image extracted from the CT plain scan image is automatically segmented into the bone structure according to the classification result of the KNN algorithm, that is:

[0080] S bone ={(x,y,z)|C(Feature)=bone}(8);

[0081] Where S bone is the bone pixel point obtained by segmentation, (x, y, z) is the three-dimensional coordinate information, and C (Feature) is the result of classifying the feature vector using the KNN algorithm.

[0082] Furthermore, in step 3), the segmented image is enhanced by using a morphological closing operation to optimize the segmentation effect of the skeleton, and the formula is expressed as follows:

[0083] D(I,B)(x)=max{I(y):y∈neighborhood(x,B)}(9);

[0084] Where I is the input 3D image, B is the structural element, x is a pixel position in the image, neighborhood(x,B) represents the neighborhood defined by the structural element B with x as the center, and D(I,B)(x) is the pixel value of the image at position x after the dilation operation, which is the maximum value of all pixel values ​​of the input image B in the defined neighborhood.

[0085] E(I,B)(x)=min{I(y):y∈neighborhood(x,B)}(10);

[0086] Where I is the input 3D image, B is the structural element, x is a pixel position in the image, neighborhood(x,B) represents the neighborhood defined by the structural element B with x as the center, and E(I,B)(x) is the pixel value of the image at position x after the erosion operation, which is the minimum value of all pixel values ​​of the input image B in the defined neighborhood.

[0087] Substitute (9)(10) into (11)

[0088] closing(I)=D(E(I,B))(11);

[0089] Here, the closing operation first applies the erosion operation E to the image I, and then applies the dilation operation D to the eroded result, using the same structure element B.

[0090] Furthermore, in step 4), for the reconstruction of the skeleton model, the volume rendering method is first used to stack a series of segmented 2D slices to form a 3D volume data. In the process, linear interpolation is used to generate pixel values ​​between slices to ensure the smoothness of the 3D reconstruction. Linear interpolation formula:

[0091]

[0092] Where f(x) is the interpolated value at position x, x0 and x1 are known points, and f(x0) and f(x1) are the corresponding pixel values.

[0093] Furthermore, in step 4), the Marching Cubes algorithm is used to extract isosurfaces from the three-dimensional volume data to form a three-dimensional model. The isosurface is generated by interpolating all points in the three-dimensional data that exceed a certain threshold, that is:

[0094] V(p i )≥V iso or V(p i )≤V iso (13);

[0095] Where V(p i ) is point p i The voxel value, V iso is the isosurface threshold.

[0096] Furthermore, in step 4), the unnecessary black part (background) is removed by setting a threshold, and the pixels of the target organ are retained, that is:

[0097]

[0098] Among them, I(x, y, z) is the pixel value of the original volume data, I'(x, y, z) is the pixel value after threshold processing, and T is the threshold. Further, in step 4), the density value in the volume data is mapped to color and transparency through linear color mapping for visualization of the three-dimensional model.

[0099] Color mapping formula:

[0100]

[0101] Among them, C(v) is the color corresponding to the voxel value v, C min and C max are the colors corresponding to the minimum and maximum voxel values.

[0102] Transparency mapping formula:

[0103]

[0104] Among them, α(v) is the color corresponding to the voxel value v, α min and α max are the colors corresponding to the minimum and maximum voxel values.

[0105] After executing the above process on the plain scan image, the following Figure 2-4 effect.

[0106] Advantages of the present invention:

[0107] 1) Accurate bone segmentation and three-dimensional reconstruction: The multi-dimensional data processing method based on the K nearest neighbor (KNN) algorithm adopted by the present invention significantly improves the accuracy of bone segmentation. The KNN algorithm can effectively identify the difference between bones and other tissues by analyzing CT image data, and realize accurate bone segmentation. This accurate segmentation is the basis for subsequent three-dimensional reconstruction, ensuring the accuracy of the reconstructed model. In addition, the present invention further optimizes the three-dimensional morphology of the skeleton through image enhancement processing and multi-view reconstruction technology, so that the final three-dimensional model is richer in details, providing a more accurate reference for clinical diagnosis.

[0108] 2) Automated and efficient image processing flow: The present invention realizes a fully automated flow from image preprocessing to three-dimensional reconstruction, greatly improving processing efficiency. In the preprocessing stage, the region of interest (ROI) of the skeleton is highlighted through image processing technology, laying a good foundation for subsequent skeleton segmentation and modeling. The introduction of the KNN algorithm makes the automatic recognition and segmentation process of the skeleton structure more efficient, reducing the need for manual intervention. In addition, the present invention also realizes the effective integration of image information from different perspectives by dynamically adjusting the weight distribution in the image fusion process, further improving the automation and efficiency of the processing flow.

[0109] 3) Enhance the auxiliary role of clinical diagnosis and treatment planning: The three-dimensional bone model provided by the present invention provides powerful assistance for doctors to locate, evaluate and formulate surgical plans for diseases. Compared with traditional two-dimensional imaging technology, three-dimensional reconstruction technology can provide more intuitive and detailed bone structure information, which helps doctors to more accurately understand the spatial relationship and degree of lesions. In addition, the present invention can also optimize the display of the three-dimensional model according to clinical needs, so that doctors can more clearly observe the key anatomical structures and lesion characteristics, so as to formulate more accurate and personalized treatment plans. Improve model quality: In the process of building a three-dimensional model, the present invention uses insert technology and optimization algorithms to enhance the quality and smoothness of the model. This processing reduces the noise and holes on the surface of the model, ensures the continuity and accuracy of the three-dimensional model, and is conducive to subsequent medical analysis and surgical planning.

[0110] The description of the present invention is only a preferred embodiment and does not limit its protection scope. Any modification, replacement or improvement following the principle of the present invention is protected. The exact protection scope of the present invention is defined by the claims.

Claims

1. A method for bone segmentation and 3D reconstruction of plain scan CT based on KNN multi-dimensional data. Features: The following steps are involved: 1) Use image processing technology to preprocess CT images to highlight the bone ROI; 2) Apply the K-nearest neighbor (KNN) algorithm to automatically identify and segment the bone structure in CT images; 3) Enhance the segmented image and construct the three-dimensional shape of the bone from different directions through multi-view reconstruction technology; 4) The 3D reconstruction system reconstructs the segmented and annotated image data in 3D and optimizes the display of the generated model.

2. According to the method of claim 1, a method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data, Features: In step 1), Gaussian smoothing filtering is performed on the CT plain scan image as part of image preprocessing. The formula is as follows: Where (x, y) represents the distance from the center point, and σ is the standard deviation, which controls the degree of smoothing.

3. According to the method of claim 1, a method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data, Features: In step 1), the contrast enhancement process is performed on the CT plain scan image, which is a part of the image preprocessing to ensure the high contrast of the image. The formula is as follows: I'=αI+β(2); Where I represents the original image, I' is the enhanced image, α is the contrast control parameter, and β is the brightness control parameter.

4. The method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data according to claim 1, Features: The image preprocessing in step 1) includes a histogram equalization operation, which is performed based on the histogram of the CT scan data. The formula is as follows: Where L is the number of gray levels, T is the cumulative distribution function, g -1 It is the inverse of the equalized cumulative distribution function.

5. The method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data according to claim 1, Features: The image preprocessing in step 1) includes a Canny edge check operation, which is performed on the CT plain scan data after histogram equalization. The formula is as follows: Edge=Canny(I,σ,T 1 ,T 2 )(4); Where I is the input image, σ is the standard deviation of the Gaussian filter, and T 1 and T 2 is a double threshold for edge connections.

6. The method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data according to claim 1, Features: The image preprocessing in step 1) includes a sharpening enhancement operation, which is performed on the CT plain scan data after Canny edge detection. The formula is as follows: I sharpened (x,y)=I(x,y)+λL(x,y)(5); Among them I sharpened (x, y) is the gray value of the sharpened image at (x, y), λ is a positive sharpening strength parameter, I(x, y) represents the gray value of the image at the coordinate (x, y), and L(x, y) represents the Laplace response at the coordinate (x, y).

7. The method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data according to claim 1, Features: The operation of automatically identifying and segmenting the bone structure in the CT image in step 2) extracts the multi-dimensional data information of the CT image. The formula is as follows: Feature=[x,y,z,I(x,y,z)](6); Among them, Feature is the extracted multi-dimensional information, (x, y, z) is the three-dimensional coordinate information, and I(x, y, z) is the grayscale value corresponding to the (x, y, z) three-dimensional coordinates.

8. The method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data according to claim 1, Features: In step 2), the operation of automatically identifying and segmenting the bone structure in the CT image uses the KNN algorithm to classify the feature vectors. The formula is as follows: C(Feature)=KNN(Feature,K,Training Data)(7); Among them, C (Feature) is the classification result, K is the number of neighbors, and Training Data is the training data set.

9. The method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data according to claim 1, Features: In step 2), the operation of automatically identifying and segmenting the bone structure in the CT image is to automatically segment the bone structure according to the classification result of the KNN algorithm. The formula is as follows: S bone ={(x,y,z)|C(Feature)=bone}(8); Where S bone is the bone pixel point obtained by segmentation, (x, y, z) is the three-dimensional coordinate information, and C (Feature) is the result of classifying the feature vector using the KNN algorithm.

10. The method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data according to claim 1, Features: In step 3), the segmented image is enhanced by using morphological closing operation to optimize the bone segmentation effect. The formula is as follows: D(I,B)(x)=max{I(y):y∈neighborhood(x,B)}(9); Where I is the input 3D image, B is the structural element, x is a pixel position in the image, neighborhood(x,B) represents the neighborhood defined by the structural element B with x as the center, and D(I,B)(x) is the pixel value of the image at position x after the dilation operation, which is the maximum value of all pixel values ​​of the input image B in the defined neighborhood. E(I,B)(x)=min{I(y):y∈neighborhood(x,B)}(10); Where I is the input 3D image, B is the structural element, x is a pixel position in the image, neighborhood(x,B) represents the neighborhood defined by the structural element B with x as the center, and E(I,B)(x) is the pixel value of the image at position x after the erosion operation, which is the minimum value of all pixel values ​​of the input image B in the defined neighborhood. Substitute (9)(10) into (11) closing(I)=D(E(I,B))(11); Here, the closing operation first applies the erosion operation E to the image I, and then applies the dilation operation D to the eroded result, using the same structure element B.

11. The method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data according to claim 1, Features: In step 4), the segmented and annotated image data is reconstructed in three dimensions. First, the volume rendering method is used to stack a series of segmented 2D slices to form a three-dimensional volume data. In the process, linear interpolation is used to generate pixel values ​​between slices to ensure the smoothness of the three-dimensional reconstruction. Linear interpolation formula: where f(x) is the interpolated value at position x, 0 and x 1 is a known point, f(x 0 ) and f(x 1 ) is the corresponding pixel value.

12. The method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data according to claim 1, Features: In step 4), the segmented and annotated image data is reconstructed in three dimensions, and the MarchingCubes algorithm is used to extract isosurfaces from the three-dimensional volume data to form a three-dimensional model. The isosurface is generated by interpolating all points in the three-dimensional data that exceed a certain threshold. The specific formula is as follows: V(p i )≥V iso or V(p i )≤V iso (13); Where V(p i ) is point p i The voxel value, V iso is the isosurface threshold.

13. The method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data according to claim 1, Features: In step 4), the segmented and annotated image data is reconstructed in three dimensions. By setting a threshold, the unnecessary black part (background) is removed and the pixels of the target organ are retained. The specific formula is as follows: Among them, I(x, y, z) is the pixel value of the original volume data, I'(x, y, z) is the pixel value after threshold processing, and T is the threshold. Further, in step 4), the density value in the volume data is mapped to color and transparency through linear color mapping for visualization of the three-dimensional model.

14. The method for bone segmentation and three-dimensional reconstruction of plain scan CT based on KNN multi-dimensional data according to claim 1, Features: In step 4), the segmented and annotated image data is reconstructed in three dimensions by linear color mapping. Color Mapping) maps the density value in the volume data to color and transparency for visualization of the 3D model. The specific formula is as follows: C=C min +d norm ×(C max -C min ) (15); Among them, C(v) is the color corresponding to the voxel value v, C min and C max are the colors corresponding to the minimum and maximum voxel values. α=α min +d norm ×(α max -α min ) (16); Among them, α(v) is the color corresponding to the voxel value v, α min and α max are the colors corresponding to the minimum and maximum voxel values.

Citation Information

Cited By

  • Bone image data model construction method based on CT (Computed Tomography) image

    CN121074239A