CN106909947B - Mean Shift algorithm-based CT image metal artifact elimination method and system - Google Patents
Mean Shift algorithm-based CT image metal artifact elimination method and system Download PDFInfo
- Publication number
- CN106909947B CN106909947B CN201710123972.3A CN201710123972A CN106909947B CN 106909947 B CN106909947 B CN 106909947B CN 201710123972 A CN201710123972 A CN 201710123972A CN 106909947 B CN106909947 B CN 106909947B
- Authority
- CN
- China
- Prior art keywords
- image
- metal
- mean shift
- algorithm
- projection data
- 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
Links
- 239000002184 metal Substances 0.000 title claims abstract description 121
- 229910052751 metal Inorganic materials 0.000 title claims abstract description 121
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 74
- 238000000034 method Methods 0.000 title claims abstract description 54
- 230000008030 elimination Effects 0.000 title description 2
- 238000003379 elimination reaction Methods 0.000 title description 2
- 238000009499 grossing Methods 0.000 claims abstract description 40
- 230000011218 segmentation Effects 0.000 claims abstract description 36
- 238000007781 pre-processing Methods 0.000 claims abstract description 14
- 238000004364 calculation method Methods 0.000 claims abstract description 9
- 238000012545 processing Methods 0.000 claims description 18
- 230000003044 adaptive effect Effects 0.000 claims description 9
- 238000001914 filtration Methods 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 4
- 230000008520 organization Effects 0.000 claims description 4
- 238000010606 normalization Methods 0.000 claims description 3
- 230000008439 repair process Effects 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 29
- 238000002591 computed tomography Methods 0.000 description 112
- 210000001519 tissue Anatomy 0.000 description 14
- 238000010586 diagram Methods 0.000 description 13
- 210000000988 bone and bone Anatomy 0.000 description 5
- 230000008569 process Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 150000002739 metals Chemical class 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 230000000747 cardiac effect Effects 0.000 description 1
- 238000003759 clinical diagnosis Methods 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 239000007943 implant Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000005855 radiation Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/003—Reconstruction from projections, e.g. tomography
- G06T11/008—Specific post-processing after tomographic reconstruction, e.g. voxelisation, metal artifact correction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2210/00—Indexing scheme for image generation or computer graphics
- G06T2210/41—Medical
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a method and a system for eliminating metal artifacts of CT images based on Mean Shift algorithm. Removing noise and smoothing slight artifacts by preprocessing a CT image, segmenting a metal region in the CT image, performing linear interpolation on projection data of the metal region to generate an interpolation image, restoring tissue information to obtain a prior image, and replacing the original projection data with the projection data of the prior image to obtain a corrected CT image. Compared with the prior art, the method and the system for eliminating the metal artifact of the CT image based on the Mean Shift algorithm can eliminate the mild metal artifact in the CT image, have good smoothing effect and do not lose detailed information; the metal area segmentation effect is good, and the calculation efficiency is higher; the artifact removing effect is obvious, and the human tissue structure information in the original CT image is also protected.
Description
Technical Field
The invention relates to the technical field of reducing metal artifacts in Computed Tomography (CT), in particular to a method and a system for eliminating metal artifacts in CT images based on a Mean Shift algorithm.
Background
With the development of medical technology, CT images have been widely used in the field of clinical medicine, and the level of medical diagnosis has been greatly improved. However, in the CT technique, the X-ray beam is used to scan the human body surface and then receive the image, and for some special patients with metal implants in their bodies, such as metal dental fillings, artificial joints, cardiac pacemakers, etc., due to the hardening, scattering, noise and partial volume effect of the radiation, there is a high possibility that CT images with metal artifacts will be obtained. The metal artifacts cause the quality of the CT image to be poor, the tissue structure is difficult to judge, and misdiagnosis may result. Therefore, Metal Artifact Reduction (MAR) of CT images is of great importance to improve the accuracy of clinical diagnosis. The MAR technique is a popular research direction in the field of medical image processing, and many scholars make contributions to this, and the involved image processing techniques can be roughly divided into three major categories, namely a projection repairing method, an iterative method and a hybrid method of the two methods. The projection repairing method mainly generates new projection data by methods of interpolation, prior images and the like to replace the projection data of corroded metal areas, thereby eliminating the metal artifacts of CT images. The iterative method continuously iteratively corrects the image according to a set optimization criterion by utilizing an algebraic or statistical method, can obtain a better result when the noise is larger, but has relatively higher algorithm complexity. The hybrid method is an algorithm which combines the two methods, makes up for the deficiencies of each other, obviously improves the image quality and has higher calculation efficiency. The projection repairing method is a characteristic processing method different from other image processing methods of the MAR technology, many researchers propose own projection repairing algorithms, but ideal results are not obtained yet, and the direction still has an exploration space.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a CT image metal artifact removing system based on Mean Shift algorithm, which has the characteristics of obvious artifact removing effect and capability of protecting human tissue structure.
The invention provides a method for eliminating metal artifacts of CT images based on Mean Shift algorithm, which comprises the following steps:
s1, preprocessing the CT image through a self-adaptive Mean Shift smoothing algorithm, and performing smoothing filtering on the CT image containing the metal artifacts to eliminate noise in the CT image and smooth mild artifacts;
s2, obtaining a metal region in the CT image by analyzing the characteristic space and the clustering method of the CT image by adopting a simplified Mean Shift segmentation algorithm;
s3, projecting the metal area and the pre-processed CT image to obtain projection data, and obtaining corrected projection data by adopting a linear interpolation algorithm for the projection data at the coordinate x
Wherein β is the projection angle, { pβ,qβIs the interpolation space;
s4, carrying out back projection reconstruction according to the projection data after linear interpolation to obtain a CT image v without metal areas after artifact correctionli＝{vli(i)|i∈IliWhere I is a pixel point in the CT image, IliRepresenting the image after linear interpolation processing;
s5, replacing the metal area in the original CT image by the image obtained by linear interpolation to obtain the CT image v without metal area and with good organization informationnom＝{vnom(i)|i∈InometalGet the difference image vdiff＝{vnom(i)-vli(i)|i∈IdiffV, the final prior image vpriorCan be expressed as: v. ofprior(i)＝vli(i)+vdiff(i) W (I), wherein InometalRepresenting CT images containing no metal, IdiffDifference image representing original CT image and linear interpolation image, and weight coefficient
V in pair formuladiff(i) Normalization processing is carried out, wherein the sigma value interval is (0, 1);
s6, carrying out front projection on the prior image to obtain corrected projection data, replacing the projection data of a metal area in the original CT image, and completing projection repair;
and S7, carrying out back projection reconstruction on the obtained projection data, adding the metal regions obtained by the segmentation in the step S2, and generating a final corrected image.
Preferably, the step of smoothing by the adaptive Mean Shift smoothing algorithm includes:
(1) forming a characteristic space vector x by the position and the pixel value of each pixel point in the CT image;
(2) starting from the first pixel point i at the upper left of the CT image, the formula is utilized
Calculating an offset mean vector M (x), wherein w (x) is the weight of the sample point, G (x) is a Gaussian kernel function, and h is the bandwidth;
(3) setting a stop condition epsilon, if M (x) > epsilon, assigning the value of M (x) + x to x, iteratively executing the step (2) until M (x) ≦ epsilon or reaching the upper limit of iteration times to end the iteration, and starting the calculation of the next pixel until the whole CT image is traversed.
Preferably, the step of determining the bandwidth h includes:
(1) given an initial bandwidth h0Setting step length delta h and threshold delta;
(2) if the difference | p-p between the CT values of the point within the bandwidth and the center point0If | < delta, then the two points are judged to be similar, and the similar point n is counted0And a total sampling point n;
(3) if n is0And h + delta h is larger than or equal to 75% of x n, and the step (2) is executed iteratively until n is satisfied0<65% n, or the upper limit of the number of iterations is reached and the iteration is ended.
Preferably, the upper limit of the number of iterations is five.
Preferably, in step S2, x is calculated for n sample points in d-dimensional space i1.. n, the basic form of the offset mean vector at point x is:
wherein ShIs a high dimensional sphere of radius h, Sh(x)＝{y:(y-x)T(y-x)≤h2K is the number of points falling into the high dimensional sphere of the n sample points, then Mh(x) The offset of the sample point falling into the high-dimensional sphere area relative to the x point is averaged。
Preferably, the Mean Shift segmentation algorithm is implemented by the following steps:
(1) forming a characteristic space vector x by the position and the pixel value of each pixel point, and setting a bandwidth h;
(2) randomly selecting an initial clustering center x, and calculating the distance from each pixel point to the clustering center if (x)i-x)T(xi-x)≤h2Is marked by xiAnd voting the cluster;
(3) using the above formula
Calculating Mh(x) If the distance between x' and x is smaller than h/2, merging the two clusters, otherwise, generating a new cluster, and iteratively executing the step (2) until all points are marked;
(4) and each pixel point belongs to the cluster with the largest ticket number, so that all the pixel points of each cluster are obtained, and the cluster with the highest pixel value is a metal area.
The invention also provides a system for eliminating the metal artifact of the CT image based on the Mean Shift algorithm, which comprises the following steps:
the preprocessing module is used for preprocessing the CT image through a self-adaptive Mean Shift smoothing algorithm, performing smooth filtering on the CT image containing the metal artifacts, eliminating noise in the CT image and smoothing mild artifacts;
the metal segmentation module is used for obtaining a metal region in the CT image by analyzing the characteristic space and the clustering method of the CT image by adopting a simplified Mean Shift segmentation algorithm; and
and the projection repairing module is used for performing linear interpolation on the projection data of the metal area to generate an interpolation image, repairing the tissue information to obtain a prior image, and replacing the original projection data with the projection data of the prior image to obtain a corrected image.
Compared with the prior art, the method and the system for eliminating the metal artifacts of the CT image based on the Mean Shift algorithm can remove the light metal artifacts in the CT image by preprocessing the CT image, and in addition, the size of a smooth window is selected in a self-adaptive manner, so that the smoothing effect is good, and the detail information is not lost; by adopting a simplified Mean Shift segmentation algorithm for the CT image, the metal region segmentation effect is good, and the calculation efficiency is higher; the artifact removing effect is obvious by repairing the tissue information of the interpolation image, and the human tissue structure information in the original CT image is also protected.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts, wherein:
FIG. 1 is a flow chart of a CT image metal artifact removing method based on Mean Shift algorithm of the present invention;
FIG. 2(a) is an effect diagram before smoothing CT images by using the method for eliminating metal artifacts in CT images based on Mean Shift algorithm of the present invention;
FIG. 2(b) is an effect diagram of a CT image smoothed by the method for eliminating metal artifacts of a CT image based on Mean Shift algorithm according to the present invention;
FIG. 3(a) is an original CT image without metal segmentation;
FIG. 3(b) is a diagram of a metal region segmented by a threshold value for the original CT image shown in FIG. 3 (a);
FIG. 3(c) is a diagram of a metal region segmented by a region growing method for the original CT image shown in FIG. 3 (a);
FIG. 3(d) is a diagram of a metal region segmented from the original CT image shown in FIG. 3(a) by using Mean Shift algorithm;
FIG. 4(a) is a raw CT image map containing metal artifacts;
FIG. 4(b) is a CT image effect graph of preprocessing the original CT image graph containing metal artifacts shown in FIG. 4 (a);
FIG. 4(c) is a CT image effect diagram of metal segmentation performed on the original CT image diagram containing metal artifacts shown in FIG. 4 (a);
FIG. 4(d) is a CT image effect graph of linear interpolation of the original CT image graph containing metal artifacts shown in FIG. 4 (a);
FIG. 4(e) is a prior image effect graph after processing the original CT image graph containing metal artifacts shown in FIG. 4 (a);
FIG. 4(f) is a graph of the effect of the corrected image obtained by processing the original CT image containing metal artifacts shown in FIG. 4 (a);
FIG. 5(a) is an original CT image which is not processed by the Mean Shift algorithm-based CT image metal artifact removal method of the present invention;
FIG. 5(b) is a diagram illustrating the effect of conventional linear interpolation processing on the original CT image shown in FIG. 5 (a);
FIG. 5(c) is an effect diagram of the original CT image shown in FIG. 5(a) being processed by the method for removing metal artifacts in CT images based on Mean Shift algorithm according to the present invention;
FIG. 6 is a frame diagram of a system for eliminating metal artifacts in CT images based on the Mean Shift algorithm according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1, the present invention provides a method for removing metal artifacts in a CT image based on a Mean Shift algorithm, which includes the following steps:
s1, preprocessing the CT image through a self-adaptive Mean Shift smoothing algorithm, and performing smoothing filtering on the CT image containing the metal artifacts to eliminate noise in the CT image and smooth mild artifacts;
s2, obtaining a metal region in the CT image by analyzing the characteristic space and the clustering method of the CT image by adopting a simplified Mean Shift segmentation algorithm;
s3, projecting the metal area and the pre-processed CT image to obtain projection data, for the coordinate xProjection data, and linear interpolation algorithm is adopted to obtain corrected projection data
Wherein β is the projection angle, { pβ,qβIs the interpolation space;
s4, carrying out back projection reconstruction according to the corrected projection data to obtain a corrected CT image v without metal areasli＝{vli(i)|i∈IliWhere I is a pixel point in the CT image, IliIs an image after linear interpolation processing;
s5, replacing the metal area in the original CT image by the image obtained by linear interpolation to obtain the CT image v without metal area and with good organization informationnom＝{vnom(i)|i∈InometalGet the difference image vdiff＝{vnom(i)-vli(i)|i∈IdiffV, the final prior image vpriorCan be expressed as: v. ofprior(i)＝vli(i)+vdiff(i) W (I), wherein InometalRepresenting CT images containing no metal, IdiffA difference image representing the linearly interpolated image and the original CT image, and weight coefficients
V in pair formuladiff(i) Normalization processing is carried out, wherein the sigma value interval is (0, 1);
s6, carrying out front projection on the prior image to obtain corrected projection data, replacing the projection data of a metal area in the original CT image, and completing projection repair;
and S7, carrying out back projection reconstruction on the obtained projection data, adding the metal regions obtained by the segmentation in the step S2, and generating a final corrected image.
Step S1 is used to perform image preprocessing on the CT image, and an adaptive Mean Shift smoothing algorithm is applied, so that artifacts with a low degree can be removed, the occurrence of overcorrection can be avoided, and the tissue structure of the original CT image cannot be damaged.
The basic principle of the self-adaptive Mean Shift smoothing algorithm is to calculate a Shift Mean vector according to an initial point, move the initial point to the Shift Mean, recalculate the Shift Mean vector, and continuously iterate the process until an ending condition is met, so that the self-adaptive Mean Shift smoothing algorithm is an estimation algorithm based on a probability density gradient function of a feature space.
Based on the principle, the concept of kernel function and sample point weight is introduced, the kernel function enables the distances between the samples and the central point to be different, the offset of the kernel function contributes to the offset mean vector to be different, and the weight reflects the difference of the importance of each sample point. In summary, the offset mean vector can be described as:
w (x) is the weight of a sample point, and the closer to x, the higher the weight of the sample point; g (x) is a kernel function, two commonly used are a unit kernel function and a Gaussian kernel function, and the system adopts the Gaussian kernel function; h is the bandwidth, the larger h, the flatter the kernel function, and the smoother the image.
For n sample points in d-dimensional space, the density estimation function is:
wherein the kernel function K (x) ck,dk(||x||2)，k(||x||2) A profile function called kernel, h is the bandwidth, and w (x) is the weight of the sample point.
The compound can be obtained by the formula,
let k (x) have a negative derivative function g (x), and its corresponding kernel function g (x) ═ cg,dg(||x||2) Then, there are:
in view of the above, it can be seen that,
description of Mh,G(x) Always pointing in the direction in which the probability density increases the fastest, is a normalized probability density gradient function. The Mean Shift smoothing algorithm has edge-preserving property, does not destroy tissue structure information, and can be used for smoothing metal artifacts in CT images.
The main steps of the adaptive Mean Shift smoothing algorithm for smoothing comprise:
(1) forming a characteristic space vector x by the position and the pixel value of each pixel point in the CT image;
(2) starting from the first pixel point i at the upper left of the CT image, the formula is utilized
Calculating an offset mean vector M (x), wherein w (x) is the weight of the sample point, G (x) is a Gaussian kernel function, and h is the bandwidth;
(3) setting a stop condition epsilon, if M (x) > epsilon, assigning the value of M (x) + x to x, iteratively executing the step (2) until M (x) ≦ epsilon or reaching the upper limit of iteration times to end the iteration, and starting the calculation of the next pixel until the whole CT image is traversed.
The bandwidth h is an important parameter in smoothing CT images with the adaptive Mean Shift smoothing algorithm. If h is too large, the CT image may be too smooth and lose detail, and if h is too small, the smoothing effect may not be achieved. When a fixed bandwidth is manually selected, the above situation is easily caused, and the processing result of the final image is affected. Therefore, the invention adaptively selects the bandwidth h, and the step of determining the bandwidth h comprises:
(1) given an initial bandwidth h0Setting step length delta h and threshold delta;
(2) if the difference | p-p between the CT values of the point within the bandwidth and the center point0If | < delta, then the two points are judged to be similar, and the similar point n is counted0And a total sampling point n;
(3) if n is0And h + delta h, and iteratively executing the step (2) until n is satisfied0<65% n, or the upper limit of the number of iterations is reached and the iteration is ended. The number of iterationsThe upper limit of the number is five.
The self-adaptive bandwidth selection enables the bandwidth to be larger in the area with less detail information, and the calculation efficiency is improved, and the bandwidth to be smaller in the area with more dense detail and edge information, and the detail information is not lost, so that a better smoothing effect is achieved.
Referring to fig. 2(a) and (b), fig. 2(a) is an effect diagram before smoothing the CT image by using the Mean Shift algorithm-based CT image metal artifact removing method of the present invention, and fig. 2(a) is an effect diagram after smoothing the CT image by using the Mean Shift algorithm-based CT image metal artifact removing method of the present invention. Therefore, for the artifacts with light degrees of No. 1 and No. 2, the Mean Shift smoothing algorithm can achieve a relatively ideal smoothing effect without damaging tissue structures such as bones. The severe artifact of the 3 rd bit has a certain effect but is not obvious enough, and further processing is needed.
The step S2 is to perform metal region segmentation on the CT image, where the segmentation of the metal region is a key factor for determining the final corrected image quality, and although the conventional threshold segmentation method is simple and efficient, it is easy to misclassify a bone tissue with a high CT value into a metal, and an accurate segmentation result cannot be obtained. The region growing method solves the problem of misclassification to a certain extent, the segmentation result is more accurate than that of a threshold value method, but based on the principle that the seed points expand to the outer region according to the similarity, the segmented metal edge is not clearly expanded, and the segmentation result is larger than the correct metal region. The system adopts a simplified MeanShift algorithm to perform metal segmentation, can sharpen the metal boundary, solves the problem of bone misclassification far away from the metal, and obtains a more satisfactory segmentation result.
Therefore, the invention adopts a simplified Mean Shift segmentation algorithm and achieves the purpose of segmentation by analyzing the characteristic space of the image and a clustering method. For n sample points x in d-dimensional space i1.. n, the basic form of the offset mean vector at point x is:
wherein ShIs a high dimension of radius hBall, Sh(x)＝{y:(y-x)T(y-x)≤h2K is the number of points falling into the high dimensional sphere of the n sample points, then Mh(x) The offset of the sample points falling within the high-dimensional sphere region with respect to x is then averaged.
Based on the principle, the Mean Shift segmentation algorithm can be simplified, and the implementation steps comprise:
(1) forming a characteristic space vector x by the position and the pixel value of each pixel point, and setting a bandwidth h;
(2) randomly selecting an initial clustering center x, and calculating the distance from each pixel point to the clustering center if (x)i-x)T(xi-x)≤h2Is marked by xiAnd voting the cluster;
(3) using formulas
Calculating Mh(x) And obtaining a new cluster center x ', merging the two clusters if the distance between x' and x is less than h/2, and otherwise, generating a new cluster. Iteratively performing step (2) until all points are marked;
(4) and each pixel point belongs to the cluster with the largest ticket number, so that all the pixel points of each cluster are obtained, and the cluster with the highest pixel value is a metal area.
As shown in fig. 3(a), the segmentation results of the metal object obtained by processing the same CT image as shown in fig. 3(a) by the threshold method, the region growing method and the Mean Shift algorithm are shown in fig. 3(b) to (d), the threshold segmentation shown in fig. (b) and the region growing method shown in fig. (c) both have bone misclassification, and the edge is not clear enough and the detail is poor, while the Mean Shift algorithm shown in fig. (d) can more accurately segment the metal object from the background.
The steps S3 to S7 are used for projection patch to obtain the final corrected image. Since the metal projection value in the original projection data is much higher than the metal projection value of other human tissues, artifacts are easily generated when a CT image is reconstructed, and the image quality is affected. There are many interpolation methods used in the metal artifact correction of CT images, such as linear interpolation, nonlinear interpolation, cubic spline interpolation, etc. Through comparative analysis, the system adopts linear interpolation to patch projection data, the linear interpolation can effectively eliminate most metal artifacts, and the system is simple, efficient and easy to implement.
Referring to the effect diagrams of each stage shown in fig. 4(a) - (f), it can be seen that the image is smoothed by using the adaptive Mean Shift smoothing algorithm, the metal region is segmented by the simplified Mean Shift segmentation algorithm, the artifact is repaired by using the linear interpolation technique, the prior image is generated after the difference image of the original CT image without the metal object and the interpolated image is processed, and the algorithm flow for removing the metal artifact in the CT image of the final corrected image is obtained from the prior image.
The metal artifact eliminating algorithm for the CT image provided by the system is adopted to process a series of CT images with metal artifacts, wherein the CT images comprise light artifacts and heavy artifacts, three groups of data such as single metal objects and a plurality of metal objects in the images, and the experimental results and the processing results of the traditional linear interpolation algorithm are shown in figures 5(a) - (c).
When the artifacts are light, the traditional linear interpolation may cause the result of over-correction, and the preprocessed image obtained by the smoothing algorithm provided by the system can remove the metal artifacts and protect the original organization information in the image. When the image only contains one metal object, the two image processing algorithms can effectively remove the artifact, but compared with a linear interpolation method, the algorithm provided by the system can avoid the generation of the secondary artifact and has a better processing effect. When the image contains a plurality of metals and bone tissues are arranged around the metals, the tissue structure is possibly damaged by linear interpolation, but the algorithm provided by the system removes artifacts and considers the protection of human tissue information in the original image, so that the method has a better processing effect.
Referring to fig. 6, the present invention further provides a system 100 for eliminating metal artifacts in CT images based on Mean Shift algorithm, including:
the preprocessing module 1 is used for preprocessing the CT image by a self-adaptive Mean Shift smoothing algorithm, and performing smooth filtering on the CT image containing metal artifacts to eliminate noise in the CT image and smooth mild artifacts;
the metal segmentation module 2 is used for obtaining a metal region in the CT image by analyzing the characteristic space and the clustering method of the CT image by adopting a simplified Mean Shift segmentation algorithm; and
and the projection repairing module 3 is used for performing linear interpolation on the projection data of the metal area to generate an interpolation image, repairing the tissue information to obtain a prior image, and replacing the original projection data with the projection data of the prior image to obtain a corrected image.
Compared with the prior art, the method and the system for eliminating the metal artifacts of the CT image based on the Mean Shift algorithm can remove the light metal artifacts in the CT image by preprocessing the CT image, and in addition, the size of a smooth window is selected in a self-adaptive manner, so that the smoothing effect is good, and the detail information is not lost; by adopting a simplified Mean Shift segmentation algorithm for the CT image, the metal region segmentation effect is good, and the calculation efficiency is higher; the artifact removing effect is obvious by repairing the tissue information of the interpolation image, and the human tissue structure information in the original CT image is also protected.
The above description is only an embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (5)
1. A CT image metal artifact eliminating method based on Mean Shift algorithm is characterized by comprising the following steps:
s1, preprocessing the CT image through an adaptive Mean Shift smoothing algorithm, and performing smoothing filtering on the CT image containing metal artifacts to eliminate noise in the CT image and smooth mild artifacts, wherein the step of performing smoothing through the adaptive Mean Shift smoothing algorithm comprises the following steps:
(1) forming a characteristic space vector x by the position and the pixel value of each pixel point in the CT image;
(2) starting from the first pixel point i at the upper left of the CT image, the formula is utilized
Calculating an offset mean vector M (x), wherein w (x) is the weight of the sample point, G (x) is a Gaussian kernel function, and h is the bandwidth;
(3) setting a stop condition epsilon, if M (x) > epsilon, assigning the value of M (x) + x to x, iteratively executing the step (2) until M (x) ≦ epsilon or the iteration time upper limit is reached to end the iteration, and starting the calculation of the next pixel until the whole CT image is traversed;
s2, obtaining a metal region in the CT image by analyzing the characteristic space and the clustering method of the CT image by adopting a simplified Mean Shift segmentation algorithm, wherein the simplified Mean Shift segmentation algorithm comprises the following implementation steps:
(1) forming a characteristic space vector x by the position and the pixel value of each pixel point, and setting a bandwidth h;
(2) randomly selecting an initial clustering center x, and calculating the distance from each pixel point to the clustering center if (x)i-x)T(xi-x)≤h2Is marked by xiAnd voting the cluster;
(3) using the above formula
Calculating Mh(x) If the distance between x' and x is smaller than h/2, merging the two clusters, otherwise, generating a new cluster, and iteratively executing the step (2) until all points are marked;
(4) each pixel point belongs to the cluster with the largest ticket number, so that all the pixel points of each cluster are obtained, and the cluster with the highest pixel value is a metal area;
s3, projecting the metal area and the pre-processed CT image to obtain projection data, and obtaining corrected projection data by adopting a linear interpolation algorithm for the projection data at the coordinate x
Wherein β is the projection angle, { pβ,qβIs the interpolation space;
s4, carrying out back projection reconstruction on the projection data after linear interpolation to obtain a CT image v without a metal region after artifact correctionli＝{vli(i)|i∈IliWhere I is a pixel point in the CT image, IliRepresenting the image after linear interpolation processing;
s5, replacing the metal area in the original CT image by the image obtained by linear interpolation to obtain the CT image v without metal area and with good organization informationnom＝{vnom(i)|i∈InometalGet the difference image vdiff＝{vnom(i)-vli(i)|i∈IdiffV, the final prior image vpriorCan be expressed as: v. ofprior(i)＝vli(i)+vdiff(i) W (I), wherein InometalRepresenting CT images containing no metal, IdiffDifference image representing original CT image and linear interpolation image, and weight coefficient
V in pair formuladiff(i) Normalization processing is carried out, wherein the sigma value interval is (0, 1);
s6, carrying out front projection on the prior image to obtain corrected projection data, replacing the projection data of a metal area in the original CT image, and completing projection repair;
and S7, carrying out back projection reconstruction on the obtained projection data, adding the metal regions obtained by the segmentation in the step S2, and generating a final corrected image.
2. The Mean Shift algorithm-based CT image metal artifact removal method according to claim 1, wherein the step of determining the bandwidth h comprises:
(1) given an initial bandwidth h0Setting step length delta h and threshold delta;
(2) if the difference | p-p between the CT values of the point within the bandwidth and the center point0If | is less than or equal to δ, thenJudging that the two points are similar, counting the similar points n0And a total sampling point n;
(3) if n is0And h + delta h is larger than or equal to 75% of x n, and the step (2) is executed iteratively until n is satisfied0<65% n, or the upper limit of the number of iterations is reached and the iteration is ended.
3. The Mean Shift algorithm-based CT image metal artifact removal method according to claim 2, wherein the upper limit of the number of iterations is five.
4. The Mean Shift algorithm-based CT image metal artifact removal method as claimed in claim 1, wherein in step S2, x is applied to n sample points in d-dimensional spacei1.. n, the basic form of the offset mean vector at point x is:
wherein ShIs a high dimensional sphere of radius h, Sh(x)＝{y:(y-x)T(y-x)≤h2K is the number of points falling into the high dimensional sphere of the n sample points, then Mh(x) The offset of the sample points falling within the high-dimensional sphere region from the x point is then averaged.
5. A system for eliminating metal artifacts in CT images based on Mean Shift algorithm is characterized by comprising:
the preprocessing module is used for preprocessing the CT image through an adaptive Mean Shift smoothing algorithm, smoothing and filtering the CT image containing metal artifacts, eliminating noise in the CT image and smoothing mild artifacts, wherein the smoothing step of the adaptive Mean Shift smoothing algorithm comprises the following steps:
(1) forming a characteristic space vector x by the position and the pixel value of each pixel point in the CT image;
(2) starting from the first pixel point i at the upper left of the CT image, the formula is utilized
Calculating an offset mean vector M (x), wherein w (x) is the weight of the sample point, G (x) is a Gaussian kernel function, and h is the bandwidth;
(3) setting a stop condition epsilon, if M (x) > epsilon, assigning the value of M (x) + x to x, iteratively executing the step (2) until M (x) ≦ epsilon or the iteration time upper limit is reached to end the iteration, and starting the calculation of the next pixel until the whole CT image is traversed;
the metal segmentation module is used for obtaining a metal region in the CT image by analyzing the characteristic space and the clustering method of the CT image by adopting a simplified Mean Shift segmentation algorithm, wherein the simplified Mean Shift segmentation algorithm comprises the following implementation steps of:
(1) forming a characteristic space vector x by the position and the pixel value of each pixel point, and setting a bandwidth h;
(2) randomly selecting an initial clustering center x, and calculating the distance from each pixel point to the clustering center if (x)i-x)T(xi-x)≤h2Is marked by xiAnd voting the cluster;
(3) using the above formula
Calculating Mh(x) If the distance between x' and x is smaller than h/2, merging the two clusters, otherwise, generating a new cluster, and iteratively executing the step (2) until all points are marked;
(4) each pixel point belongs to the cluster with the largest ticket number, so that all the pixel points of each cluster are obtained, and the cluster with the highest pixel value is a metal area; and
and the projection repairing module is used for performing linear interpolation on the projection data of the metal area to generate an interpolation image, repairing the tissue information to obtain a prior image, and replacing the original projection data with the projection data of the prior image to obtain a corrected image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710123972.3A CN106909947B (en) | 2017-03-03 | 2017-03-03 | Mean Shift algorithm-based CT image metal artifact elimination method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710123972.3A CN106909947B (en) | 2017-03-03 | 2017-03-03 | Mean Shift algorithm-based CT image metal artifact elimination method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106909947A CN106909947A (en) | 2017-06-30 |
| CN106909947B true CN106909947B (en) | 2020-06-26 |
Family
ID=59186370
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710123972.3A Active CN106909947B (en) | 2017-03-03 | 2017-03-03 | Mean Shift algorithm-based CT image metal artifact elimination method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106909947B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108074223B (en) * | 2017-12-28 | 2021-09-28 | 中国矿业大学(北京) | Automatic extraction method of crack network in coal rock sequence CT (computed tomography) image |
| CN111652953B (en) * | 2020-06-29 | 2022-12-30 | 中国电子科技集团公司第十四研究所 | Security image artifact removing method utilizing region search and pixel value suppression |
| CN112652031A (en) * | 2020-12-30 | 2021-04-13 | 上海联影智能医疗科技有限公司 | Artifact correction analysis method, electronic device and storage medium |
| CN113256529B (en) * | 2021-06-09 | 2021-10-15 | 腾讯科技（深圳）有限公司 | Image processing method, image processing device, computer equipment and storage medium |
| CN113520441B (en) * | 2021-08-03 | 2022-11-29 | 浙江大学 | Tissue imaging method and system for eliminating CT high-impedance artifact interference |
| CN114241070B (en) * | 2021-12-01 | 2022-09-16 | 北京长木谷医疗科技有限公司 | Method and device for removing metal artifacts from CT image and training model |
| CN116309923A (en) * | 2023-05-24 | 2023-06-23 | 吉林大学 | CT metal artifact eliminating method and system based on graph neural network |
| CN117808718B (en) * | 2024-02-29 | 2024-05-24 | 江西科技学院 | Method and system for improving medical image data quality based on Internet |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101777177A (en) * | 2009-12-29 | 2010-07-14 | 上海维宏电子科技有限公司 | Attenuation filter-based metal artifact removing mixed reconstruction method for CT images |
| CN102567958B (en) * | 2010-12-31 | 2014-06-25 | 飞利浦(中国)投资有限公司 | Image postprocessing method for removing metal artifact from computed tomography (CT) image |
| CN103106675A (en) * | 2013-02-05 | 2013-05-15 | 南方医科大学 | X-ray computed tomography (CT) metal artifact processing method |
| CN103279929B (en) * | 2013-05-25 | 2015-11-18 | 北京工业大学 | A kind of prediction of the CT image metallic traces based on integral cosine and artifact minimizing technology |
| CN104992409B (en) * | 2014-09-30 | 2018-01-12 | 中国科学院苏州生物医学工程技术研究所 | A kind of metal artifacts reduction method of CT images |
| US9679219B2 (en) * | 2015-08-12 | 2017-06-13 | International Business Machines Corporation | Image feature classification |
| CN105741244B (en) * | 2016-01-28 | 2019-09-10 | 山东鲁能智能技术有限公司 | The method of shade and halation is removed under a kind of interior crusing robot dim light |
| CN106296615B (en) * | 2016-08-16 | 2017-09-29 | 广州华端科技有限公司 | The method and system of metal artifacts is corrected in CT images |
-
2017
- 2017-03-03 CN CN201710123972.3A patent/CN106909947B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN106909947A (en) | 2017-06-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106909947B (en) | Mean Shift algorithm-based CT image metal artifact elimination method and system | |
| CN110998602B (en) | Classification and 3D modeling of 3D dentofacial structures using deep learning methods | |
| CN108053417B (en) | lung segmentation device of 3D U-Net network based on mixed rough segmentation characteristics | |
| Erdogan et al. | Ordered subsets algorithms for transmission tomography | |
| CN110827216A (en) | Multi-generator generation countermeasure network learning method for image denoising | |
| Naranjo et al. | Metal artifact reduction in dental CT images using polar mathematical morphology | |
| CN110728729B (en) | Attention mechanism-based unsupervised CT projection domain data recovery method | |
| CN110503614B (en) | Magnetic resonance image denoising method based on sparse dictionary learning | |
| CN111815521B (en) | Cone beam CT metal artifact correction algorithm based on priori image | |
| CN111402261B (en) | Improved skull segmentation method based on Hessian matrix enhancement | |
| CN112017131B (en) | CT image metal artifact removing method and device and computer readable storage medium | |
| CN112634388B (en) | Optimization method of CT iterative reconstruction cost function, CT image reconstruction method, system and CT | |
| CN113421226B (en) | CT-DR multi-modal esophageal image registration method and system based on mutual information | |
| CN117474823B (en) | CT data processing system for pediatric infectious inflammation detection assistance | |
| CN113516586A (en) | Low-dose CT image super-resolution denoising method and device | |
| Arakeri et al. | An effective and efficient approach to 3D reconstruction and quantification of brain tumor on magnetic resonance images | |
| CN108038840B (en) | Image processing method and device, image processing equipment and storage medium | |
| CN111223156A (en) | Metal artifact eliminating method for dental cone beam CT system | |
| CN110599530B (en) | MVCT image texture enhancement method based on double regular constraints | |
| CN111260583A (en) | Multi-discriminant-based multi-analysis network missing CT projection data estimation method | |
| CN116894783A (en) | Metal artifact removal method for countermeasure generation network model based on time-varying constraint | |
| CN113706643A (en) | Homomorphic adaptation learning-based head CT metal artifact correction method | |
| Roy et al. | MDL-IWS: multi-view deep learning with iterative watershed for pulmonary fissure segmentation | |
| CN109767396B (en) | Oral cavity CBCT image denoising method based on image dynamic segmentation | |
| CN113706687A (en) | Nose environment modeling method and device for path planning |
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 |