CN110533667B - Lung tumor CT image 3D segmentation method based on image pyramid fusion - Google Patents

Lung tumor CT image 3D segmentation method based on image pyramid fusion Download PDF

Info

Publication number
CN110533667B
CN110533667B CN201910688043.6A CN201910688043A CN110533667B CN 110533667 B CN110533667 B CN 110533667B CN 201910688043 A CN201910688043 A CN 201910688043A CN 110533667 B CN110533667 B CN 110533667B
Authority
CN
China
Prior art keywords
image
lung tumor
segmentation
voxel
pyramid
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910688043.6A
Other languages
Chinese (zh)
Other versions
CN110533667A (en
Inventor
王蓉芳
杨靖
陈佳伟
郝红侠
缑水平
刘红英
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xidian University
Original Assignee
Xidian University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xidian University filed Critical Xidian University
Priority to CN201910688043.6A priority Critical patent/CN110533667B/en
Publication of CN110533667A publication Critical patent/CN110533667A/en
Application granted granted Critical
Publication of CN110533667B publication Critical patent/CN110533667B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10081—Computed x-ray tomography [CT]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20016—Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30061—Lung
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30096—Tumor; Lesion

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Apparatus For Radiation Diagnosis (AREA)
  • Image Processing (AREA)

Abstract

The invention provides a lung tumor CT image 3D segmentation method based on image pyramid fusion, which is used for solving the technical problems of low segmentation precision and low robustness of a lung tumor CT image in the prior art. The method comprises the following implementation steps: extracting a 3D lung tumor CT image A; normalizing the 3D lung tumor CT image A to obtain a 3D lung tumor CT image B with normalized pixel values; constructing an image pyramid S of the 3D lung tumor CT image B; for each image S in the image pyramid S i Carrying out voxel pre-segmentation to obtain a pre-segmentation region set F of the image pyramid S; for each element F in the set F of pre-divided regions i All the voxel pre-segmentation areas are combined to form an image pyramid segmentation result set L; and fusing the image pyramid segmentation result L to obtain a final segmentation result Z. The invention improves the lung tumor CT image segmentation precision, has robust segmentation result and can be used for lung tumor CT image segmentation in the field of computer vision.

Description

Lung tumor CT image 3D segmentation method based on image pyramid fusion
Technical Field
The invention relates to the technical field of medical image processing, in particular to lung tumor image segmentation, and specifically relates to a lung tumor CT image 3D segmentation method based on image pyramid fusion, which is used for lung CT image processing.
Background
The use of medical images to assist disease diagnosis is an important tool in clinical practice, and the quantitative analysis results given by computer-aided diagnosis systems can help doctors to make better diagnosis decisions. Tumor segmentation has a wide range of applications, including measuring treatment response, radiation treatment planning, and facilitating the extraction of robust features for high-throughput radiology for subsequent tumor prediction or assessment of treatment efficacy, particularly for radiology analysis. Structural features of a complete tumor are often required to be extracted in the radiologic analysis, and the segmentation precision of the tumor can seriously affect the final analysis result.
In particular, the use of CT images of the lung to observe structural and functional features of the lung is an important diagnostic method for lung diseases in clinical practice today. Despite improvements in surgery, radiation therapy and chemotherapy, respectively, over the last three decades, lung tumors are the most common cause of cancer death worldwide, with an overall 5-year survival rate of about 13%. At present, with the further and intensive pathological research on lung tumors, a new generation of targeted therapy has begun to show clinical prospects in lung tumor therapy. In patients with lung tumors that employ systemic targeted therapy, the relative benefit of the therapy is typically determined by measuring changes in the size of the tumor lesion, usually using volumetric measurements, which require the imaging physician to segment the precise contours of the different treatment stages of the lung tumor for analysis of the effects of the therapeutic intervention. However, delineating tumors requires a significant amount of manpower and material resources, and statistics indicate that it takes approximately 40 minutes for a skilled imaging physician to accurately delineate all lung tumors of a single patient.
Currently, the field of segmentation of lung tumor images generally utilizes the gray scale information, gradient information and texture information of pixels of the images to perform segmentation. For example, the patent application with the application publication number CN109767421A entitled "a method for delineating a semi-automatic segmentation method for radiotherapy target area of lung tumor with regional growth". A method is disclosed for detecting the minimum value of the image gradient in a region by calculating the CT image gradient of the lung parenchyma, scanning the neighborhood of each pixel in the gradient image, which is defined as the local minimum gradient pixel; predefining an empty stack to store local minimum gradient pixels, comparing each new local minimum gradient pixel size to all existing element gradient pixel sizes in the stack, the new minimum value being pushed onto the stack and used to label the source pixel, the value in the stack having the greatest similarity to the new local minimum value being returned to label the original source gradient pixel; if a gradient pixel is not labeled, its neighborhood will be searched for a local minimum, and the process is repeated until all gradient pixels in the image are segmented; and finally, calculating the focus central point, extracting the lesion contour of the central slice and the adjacent slices thereof, calculating the distance between the boundary point on the adjacent slices and the focus central point, and detecting a more accurate lesion boundary by taking the obtained average distance between the boundary point on the adjacent slices and the focus central point as a reference. The method has the defects that the 2D gradient information of local image pixels is used for calculating the initial segmentation result of the image, and although the subsequent steps are optimized on the basis of the initial segmentation result, the final image segmentation precision is still low; at the same time, the segmentation result is not robust enough.
Disclosure of Invention
The invention aims to provide a lung tumor CT image 3D segmentation method based on image pyramid fusion aiming at the defects of the prior art and aims to improve the precision and robustness of the segmentation result of the lung tumor CT image.
In order to achieve the purpose, the technical scheme adopted by the invention comprises the following steps:
(1) Extracting a 3D lung tumor CT image A:
marking J pixels of a single lung tumor to be segmented in a CT image containing the lung tumor to obtain a pixel set M, M = { (x) 1 ,y 1 ,z 1 ),(x 2 ,y 2 ,z 2 ),...,(x j ,y j ,z j ),...,(x J ,y J ,z J ) And calculating the central coordinate c = (x) of the geometric area formed by M M ,y M ,z M ) Then taking c as the center and v as the side length to extract a 3D lung tumor CT image A, wherein x is j ,y j ,z j Respectively representing an x coordinate, a y coordinate and a z coordinate of the jth pixel, wherein J is more than or equal to 2, m/8 is more than or equal to v is more than or equal to m/2, and m is the minimum side length of the CT image containing the lung tumor;
(2) Normalization of 3D lung tumor CT image a:
normalizing the gray value of each pixel in the 3D lung tumor CT image A to obtain a 3D lung tumor CT image B with normalized pixel values;
(3) Constructing an image pyramid S of the 3D lung tumor CT image B:
let the image pyramid S to be constructed comprise I images, S = { S = } 1 ,s 2 ,...,s i-1 ,s i ,...,s I In which s is i Represents the ith image:
Figure BDA0002146987500000021
wherein upsample () represents a 2 times upsampling function, I ≧ 3;
(4) For each image S in the image pyramid S i Carrying out voxel pre-segmentation:
for each image S in the image pyramid S i Performing hyper-voxel pre-segmentation to obtain a pre-segmentation region set F of the I images, wherein F = { F = { (F) 1 ,F 2 ,...,F i ,...F I In which F i Representing the ith set of hyper-voxel pre-segmented regions,
Figure BDA0002146987500000031
Figure BDA0002146987500000032
represents the kth hyper-voxel area, and K is more than or equal to 2;
(5) For each F in the set F of pre-divided regions i Carrying out region merging;
for each element F in the pre-divided region set F i All the hyper-voxel pre-segmentation areas are merged to obtain a merging result l i I.e. the image segmentation result of the ith layer in the image pyramid, all the merged results constitute an image pyramid segmentation result set L, L = { L 1 ,...,l i ,...,l I }；
(6) Fusing an image pyramid segmentation result L;
fusing the image pyramid segmentation result L to obtain a final segmentation result Z of the original 3D lung tumor CT image A:
Figure BDA0002146987500000033
wherein, the downlink sample (d) i ) Is shown to d i And (5) downsampling, wherein the size of the downsampled image is consistent with that of the original 3D lung tumor CT image A.
Compared with the prior art, the invention has the following advantages:
(1) The segmentation result is robust, the image is segmented in different scale ranges by the image pyramid fusion-based method, and finally the segmentation results in different scale ranges are fused. The method avoids the problem that when the image is segmented in a single scale range, the segmentation result is easily influenced by tissues around the tumor, the segmentation result comprises a large number of non-lung regions such as blood vessels and thoracic tissues, the robust segmentation result can still be obtained when the tumor with a complex background is processed and segmented, and meanwhile, the method has good adaptability to tumors with different sizes.
(2) The segmentation result has high precision, the scheme adopts a method for pre-segmenting the voxels, generates a 3D voxel pre-segmentation region, and performs region combination on the basis; compared with the method that the initial segmentation result is generated by utilizing the 2D gradient information of local image pixels, effective 3D contour boundary information is extracted in the pre-segmentation stage, and meanwhile, the 3D gray level accumulation-based feature vector is used as similarity measurement in the region fusion stage, so that the statistical significance is higher; the segmentation result of this scheme is more accurate and substantially contains no region other than the tumor.
Drawings
FIG. 1 is a flow chart of an implementation of the present invention;
FIG. 2 is a CT image of a tumor containing lung;
FIG. 3 is a CT image A of a 3D lung tumor;
fig. 4 is a diagram of the final segmentation result.
Detailed Description
The invention is described in further detail below with reference to the figures and the specific embodiments.
Referring to fig. 1, a lung tumor CT image 3D segmentation method based on image pyramid fusion comprises the following steps
(1) Extracting a 3D lung tumor CT image A:
as shown in FIG. 2, the itk-snap software is used to open a CT image containing a lung tumor, and the mouse is used to mark and store J pixels of a single lung tumor to be segmented, so as to obtain a pixel set M, M = { (x) 1 ,y 1 ,z 1 ),(x 2 ,y 2 ,z 2 ),...,(x j ,y j ,z j ),...,(x J ,y J ,z J ) And calculating the central coordinate c = (x) of the geometric area formed by M M ,y M ,z M ) Then taking c as the center and v as the side length to extract a 3D lung tumor CT image A, wherein x j ,y j ,z j Respectively representing the x coordinate, the y coordinate and the z coordinate of the jth pixel, J is more than or equal to 2, M/8 is more than or equal to v is less than or equal to M/2, M is the minimum side length of the CT image containing the lung tumor, and the central coordinate c = (x is more than or equal to 8) of the geometric region formed by M (x is less than or equal to M is less than or equal to 2) M ,y M ,z M ) The calculation formula is as follows:
Figure BDA0002146987500000041
as a rule of thumb, a single tumor will not generally occupy the entire lung cavity, so this is chosen here
Figure BDA0002146987500000042
(2) Normalization of 3D lung tumor CT image a:
normalizing the gray value of each pixel in the 3D lung tumor CT image A to obtain a 3D lung tumor CT image B with normalized pixel values, wherein the normalization formula is as follows:
Figure BDA0002146987500000043
min () represents a minimum function, max () represents a maximum function, and because different CT images have different radioactivity levels, different imaging devices and different environmental noises during imaging, the CT images often have different gray value ranges and environmental noises, and the influence of the gray value ranges and the image noises of the CT images on the final segmentation result can be reduced after the images are normalized;
(3) Constructing an image pyramid S of the 3D lung tumor CT image B:
let the image pyramid S to be constructed comprise I images, S = { S = } 1 ,s 2 ,...,s i-1 ,s i ,...,s I In which s is i Represents the ith image:
Figure BDA0002146987500000051
the upsample () represents a 2-time upsampling function, I is more than or equal to 3, the method sequentially performs upsampling, the result of the upsampling of the previous time is used as the input of the upsampling of the next time, the upsampling of different scales on the original image is avoided, and the image sawtooth effect caused by the upsampling is reduced, so that the original image can still contain good detail information after being subjected to the upsampling for many times;
(4) For each image S in the image pyramid S i Performing hyper-voxel pre-segmentation:
each image S in the image pyramid S is segmented using a 3D SLICO (zero simple iterative Cluster) hyper-voxel segmentation method i Performing hyper-voxel pre-segmentation, SLICO is short for a zero-parameter simple linear iterative clustering algorithm, the SLICO algorithm takes space and color distance between pixel points into consideration, and segments the image into a hyper-voxel region comprising a plurality of pixel pointsThe robustness of the whole algorithm to tumors of different sizes can be improved by the aid of the pixel regions, and finally, a pre-segmentation region set F, F = { F) of the I images is obtained 1 ,F 2 ,...,F i ,...F I In which F i Representing the ith set of hyper-voxel pre-segmented regions,
Figure BDA0002146987500000052
Figure BDA0002146987500000053
represents the kth hyper-voxel area, and K is more than or equal to 2;
(5) For each F in the set F of pre-divided regions i Carrying out region merging;
(5a) F is to be i All the hyper-voxel pre-segmented regions containing lung tumor markers are combined into a set,
Figure BDA0002146987500000054
(5b) Find each one of them
Figure BDA0002146987500000055
Neighboring sets of hyper-voxel pre-segmented regions
Figure BDA0002146987500000056
Figure BDA0002146987500000057
Judgment of
Figure BDA0002146987500000058
And
Figure BDA0002146987500000059
when the two regions are adjacent to each other, the rule is that whether more than 20 8 adjacent pixel pairs exist between the two regions or not, and the strict adjacent rule can ensure that the two regions are strictly adjacent to each other certainly on 3D, so that the phenomenon that burrs exist in a final segmentation result is avoided;
(5c) Find out
Figure BDA00021469875000000510
Each of which
Figure BDA00021469875000000511
Adjacent area of
Figure BDA00021469875000000512
And calculate
Figure BDA00021469875000000513
And
Figure BDA00021469875000000514
similarity between them
Figure BDA00021469875000000515
Judgment of
Figure BDA00021469875000000516
If true, then
Figure BDA00021469875000000517
And performing step (5 d), otherwise not merging, and performing step (5 d), wherein p () represents a similarity calculation function,
Figure BDA0002146987500000061
the similarity function is calculated in the following specific manner:
Figure BDA0002146987500000062
wherein
Figure BDA0002146987500000063
And
Figure BDA0002146987500000064
respectively represent regions
Figure BDA0002146987500000065
And
Figure BDA0002146987500000066
gray scale cumulative feature vector, region
Figure BDA0002146987500000067
And
Figure BDA0002146987500000068
the calculation formulas of the gray scale cumulative feature vectors are as follows:
p＝[p 1 ,...,p q ,....,p Q ]
wherein the content of the first and second substances,
Figure BDA0002146987500000069
q is the gray scale of the pixel, n q Is the number of pixels in the region having a gray scale q, MN represents s i The number of pixels in (1) is used as a feature vector of a final region for calculating the similarity between regions, the gray scale accumulated feature vector utilizes the number of pixels on each gray scale as a component of the feature vector, so that the influence caused by environmental noise, imaging equipment and CT dose level can be reduced to the maximum extent, and the final gray scale accumulated feature vector uses the total number of pixels in the region for normalization, so that the advantages of: the aim of the method is to find the region pair of the most similar regions between two regions (the regions have the same gray distribution), but the pre-divided regions are inevitably different in size, if the two regions are larger, the similarity calculated based on the gray cumulative feature vector before normalization is larger, which is contrary to the original purpose of calculating the similarity, so that the method of normalizing by using the number of pixels contained in the regions can lead us to find the regions really having the same gray distribution, and the interference of the factor of the size of the regions is eliminated;
(5d) If each one
Figure BDA00021469875000000610
Can not find out the satisfaction
Figure BDA00021469875000000611
Adjacent area of
Figure BDA00021469875000000612
Are combined, then O i I.e. the final merged result l i Otherwise, executing step (5 b);
the calculation of two adjacent regions is carried out in the merging process, the calculation of the two adjacent regions finds a superpixel pre-segmentation region of the regions which are the most similar to each other, if one of the two regions already contains a tumor mark, the other pixel also contains the tumor mark, which is the basis for the success of the whole technical scheme, in the process, the region information containing the accurate contour is always used, and then the regions are merged, so that the final segmentation result is more accurate and robust;
(6) Fusing an image pyramid segmentation result L;
fusing the image pyramid segmentation result L to obtain a final segmentation result Z of the original 3D lung tumor CT image A:
Figure BDA0002146987500000071
wherein, the downlink sample (d) i ) Is shown to d i Down-sampling, wherein the size of the down-sampled image is consistent with that of an original 3D lung tumor CT image A, segmentation results in different scale ranges are obtained by combining areas on an image pyramid, then the segmentation results in the different scale ranges are down-sampled to the same size as that of an original tumor image block, and finally the results in the different scale ranges are fused in a mean value fusion mode to serve as a final segmentation resultAll the contour boundaries are more accurate according to the pyramid fusion result of the obtained image;
the technical effects of the present invention will be further described with reference to simulation experiments.
1. Simulation conditions are as follows: the invention was performed on WINDOWS 10 systems using Matlab R2016a platform.
2. And (5) simulating content and result analysis.
Simulation 1:
the CT image containing the lung tumor adopted in the embodiment of the invention is shown as 2, a crescent-shaped dark region on the right half part of the image is lung parenchyma, a brighter irregular connected region in the lung parenchyma is a tumor to be segmented, the CT image of the lung tumor extracted based on the CT image is shown as figure 3, a z section of a 3D image at the upper left corner in figure 3 represents a y section of the 3D image at the upper right corner, an x section of the 3D image at the lower right corner, a final segmentation result image is shown as 4, and a final 3D segmentation result is represented at the lower left corner in figure 4.

Claims (5)

1. A lung tumor CT image 3D segmentation method based on image pyramid fusion is characterized by comprising the following steps:
(1) Extracting a 3D lung tumor CT image A:
marking J pixels of a single lung tumor to be segmented in a CT image containing the lung tumor to obtain a pixel set M, M = { (x) 1 ,y 1 ,z 1 ),(x 2 ,y 2 ,z 2 ),...,(x j ,y j ,z j ),...,(x J ,y J ,z J ) And calculating the central coordinate c = (x) of the geometric area formed by M M ,y M ,z M ) Then taking c as the center and v as the side length to extract a 3D lung tumor CT image A, wherein x is j ,y j ,z j Respectively representing an x coordinate, a y coordinate and a z coordinate of the jth pixel, wherein J is more than or equal to 2, m/8 is more than or equal to v is more than or equal to m/2, and m is the minimum side length of the CT image containing the lung tumor;
(2) Normalization of 3D lung tumor CT image a:
normalizing the gray value of each pixel in the 3D lung tumor CT image A to obtain a 3D lung tumor CT image B with normalized pixel values;
(3) Constructing an image pyramid S of the 3D lung tumor CT image B:
let the image pyramid S to be constructed comprise I images, S = { S = { (S) } 1 ,s 2 ,...,s i-1 ,s i ,...,s I In which s i Represents the ith image:
Figure FDA0002146987490000011
wherein upsample () represents a 2 times upsampling function, I ≧ 3;
(4) For each image S in the image pyramid S i Carrying out voxel pre-segmentation:
for each image S in the image pyramid S i Performing hyper-voxel pre-segmentation to obtain a pre-segmentation region set F of the I images, wherein F = { F = { (F) 1 ,F 2 ,...,F i ,...F I In which F i Representing the ith set of hyper-voxel pre-segmented regions,
Figure FDA0002146987490000012
Figure FDA0002146987490000013
represents the kth hyper-voxel area, and K is more than or equal to 2;
(5) For each F in the set F of pre-divided regions i Carrying out region merging;
for each element F in the set F of pre-divided regions i All the voxel pre-segmentation areas are merged to obtain a merged result l i I.e. the image segmentation result of the ith layer in the image pyramid, and all the merged results form an image pyramid segmentation result set L, L = { L = 1 ,...,l i ,...,l I }；
(6) Fusing an image pyramid segmentation result L;
fusing the image pyramid segmentation result L to obtain a final segmentation result Z of the original 3D lung tumor CT image A:
Figure FDA0002146987490000021
wherein, the download sample (d) i ) Represents a pair d i And (5) downsampling, wherein the size of the downsampled image is consistent with that of the original 3D lung tumor CT image A.
2. The image pyramid fusion-based 3D segmentation method for CT image of tumor in lung according to claim 1, wherein the calculation of the center coordinates c = (x) of the geometric region consisting of M in step (1) M ,y M ,z M ) The calculation formula is as follows:
Figure FDA0002146987490000022
wherein x is M ,y M ,z M An x-coordinate, a y-coordinate and a z-coordinate respectively representing the center coordinate c.
3. The image pyramid fusion-based 3D segmentation method for CT images of lung tumors as claimed in claim 1, wherein the gray-level value of each pixel in the 3D CT image of lung tumors A is normalized in step (2) to obtain a normalized pixel-value image B, and the normalization formula is as follows:
Figure FDA0002146987490000023
min () represents the minimum function and max () represents the maximum function.
4. The image pyramid fusion-based lung tumor CT image 3D segmentation method of claim 1, whichCharacterized in that step (5) is performed for each element F i All the hyper-voxel pre-segmentation areas in the image are combined to obtain a combined result l i The method comprises the following implementation steps:
(5a) F is to be i All the hyper-voxel pre-segmented regions containing lung tumor markers are combined into a set,
Figure FDA0002146987490000031
(5b) Find each one of them
Figure FDA0002146987490000032
Neighboring set of hyper-voxel pre-segmented regions
Figure FDA0002146987490000033
(5c) Find out
Figure FDA0002146987490000034
Each of which
Figure FDA0002146987490000035
Adjacent area of
Figure FDA0002146987490000036
And calculate
Figure FDA0002146987490000037
And with
Figure FDA0002146987490000038
Similarity between them
Figure FDA0002146987490000039
Judgment of
Figure FDA00021469874900000310
If true, then
Figure FDA00021469874900000311
And performing step (5 d), otherwise not merging, and performing step (5 d), wherein p () represents a similarity calculation function,
Figure FDA00021469874900000312
(5d) If each one
Figure FDA00021469874900000313
Can not find out the satisfaction
Figure FDA00021469874900000314
Adjacent area of (2)
Figure FDA00021469874900000315
Are combined, then O i I.e. the final merged result l i Otherwise, step (5 b) is performed.
5. The image pyramid fusion-based lung tumor CT image 3D segmentation method as claimed in claim 4, wherein the calculation in step (5 c)
Figure FDA00021469874900000316
And with
Figure FDA00021469874900000317
Similarity between them
Figure FDA00021469874900000318
The calculation formula is as follows:
Figure FDA00021469874900000319
wherein
Figure FDA00021469874900000320
And
Figure FDA00021469874900000321
respectively represent regions
Figure FDA00021469874900000322
And
Figure FDA00021469874900000323
gray scale cumulative feature vector, region
Figure FDA00021469874900000324
And
Figure FDA00021469874900000325
the calculation formulas of the gray scale accumulation feature vectors are as follows:
p＝[p 1 ,...,p q ,....,p Q ]
wherein the content of the first and second substances,
Figure FDA00021469874900000326
q is the gray scale of the pixel, n q Is the number of pixels in the region having a gray scale q, MN represents s i The number of pixels in (1).
CN201910688043.6A 2019-07-29 2019-07-29 Lung tumor CT image 3D segmentation method based on image pyramid fusion Active CN110533667B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910688043.6A CN110533667B (en) 2019-07-29 2019-07-29 Lung tumor CT image 3D segmentation method based on image pyramid fusion

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910688043.6A CN110533667B (en) 2019-07-29 2019-07-29 Lung tumor CT image 3D segmentation method based on image pyramid fusion

Publications (2)

Publication Number Publication Date
CN110533667A CN110533667A (en) 2019-12-03
CN110533667B true CN110533667B (en) 2022-12-02

Family

ID=68661952

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910688043.6A Active CN110533667B (en) 2019-07-29 2019-07-29 Lung tumor CT image 3D segmentation method based on image pyramid fusion

Country Status (1)

Country Link
CN (1) CN110533667B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114022491B (en) * 2021-10-27 2022-05-10 安徽医科大学 Small data set esophageal cancer target area image automatic delineation method based on improved spatial pyramid model
CN115294122B (en) * 2022-10-08 2023-04-07 江苏诺阳家居科技有限公司 Lung tumor feature extraction method and system based on CT image

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2013091186A1 (en) * 2011-12-21 2013-06-27 中国科学院自动化研究所 Multi-parametric 3d magnetic resonance image brain tumor segmentation method
CN107230206A (en) * 2017-06-02 2017-10-03 太原理工大学 A kind of 3D Lung neoplasm dividing methods of the super voxel sequence lung images based on multi-modal data

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10304192B2 (en) * 2017-07-11 2019-05-28 Sony Corporation Fast, progressive approach to supervoxel-based spatial temporal video segmentation

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2013091186A1 (en) * 2011-12-21 2013-06-27 中国科学院自动化研究所 Multi-parametric 3d magnetic resonance image brain tumor segmentation method
CN107230206A (en) * 2017-06-02 2017-10-03 太原理工大学 A kind of 3D Lung neoplasm dividing methods of the super voxel sequence lung images based on multi-modal data

Also Published As

Publication number Publication date
CN110533667A (en) 2019-12-03

Similar Documents

Publication Publication Date Title
Wu et al. Automatic liver segmentation on volumetric CT images using supervoxel‐based graph cuts
EP2693951B1 (en) Image analysis for specific objects
US8577129B2 (en) System and method for three-dimensional image rendering and analysis
US7876938B2 (en) System and method for whole body landmark detection, segmentation and change quantification in digital images
US7760941B2 (en) Method and apparatus of segmenting an object in a data set and of determination of the volume of segmented object
US8045770B2 (en) System and method for three-dimensional image rendering and analysis
Zacharaki et al. Non-diffeomorphic registration of brain tumor images by simulating tissue loss and tumor growth
US8073210B2 (en) Methods of smoothing segmented regions and related devices
US20050207630A1 (en) Lung nodule detection and classification
CN109753997B (en) Automatic accurate robust segmentation method for liver tumor in CT image
Marias et al. A registration framework for the comparison of mammogram sequences
US9092867B2 (en) Methods for segmenting images and detecting specific structures
Bardera et al. Semi-automated method for brain hematoma and edema quantification using computed tomography
Foruzan et al. A knowledge-based technique for liver segmentation in CT data
US7480401B2 (en) Method for local surface smoothing with application to chest wall nodule segmentation in lung CT data
US8229189B2 (en) Visual enhancement of interval changes using temporal subtraction, convolving, and non-rigid transformation field mapping
CN110533667B (en) Lung tumor CT image 3D segmentation method based on image pyramid fusion
Hasan A hybrid approach of using particle swarm optimization and volumetric active contour without edge for segmenting brain tumors in MRI scan
Garg et al. Spinal cord MRI segmentation techniques and algorithms: A survey
Lee et al. Hybrid airway segmentation using multi-scale tubular structure filters and texture analysis on 3D chest CT scans
Qiao et al. Edge extraction method for medical images based on improved local binary pattern combined with edge-aware filtering
CN116309551B (en) Method, device, electronic equipment and readable medium for determining focus sampling area
US8224046B2 (en) Visual enhancement of interval changes using rigid and non-rigid transformations
CN116051553B (en) Method and device for marking inside three-dimensional medical model
Kurugol et al. Centerline extraction with principal curve tracing to improve 3D level set esophagus segmentation in CT images

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant