CN106898000B - Automatic segmentation method for magnetic resonance imaging brain gray matter nuclei - Google Patents

Automatic segmentation method for magnetic resonance imaging brain gray matter nuclei Download PDF

Info

Publication number
CN106898000B
CN106898000B CN201710080824.8A CN201710080824A CN106898000B CN 106898000 B CN106898000 B CN 106898000B CN 201710080824 A CN201710080824 A CN 201710080824A CN 106898000 B CN106898000 B CN 106898000B
Authority
CN
China
Prior art keywords
image
segmented
segmentation
nuclei
magnetic resonance
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
CN201710080824.8A
Other languages
Chinese (zh)
Other versions
CN106898000A (en
Inventor
郭天
杨光
李建奇
薄斌仕
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.)
East China Normal University
Original Assignee
East China Normal 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 East China Normal University filed Critical East China Normal University
Priority to CN201710080824.8A priority Critical patent/CN106898000B/en
Publication of CN106898000A publication Critical patent/CN106898000A/en
Application granted granted Critical
Publication of CN106898000B publication Critical patent/CN106898000B/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
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10072—Tomographic images
    • G06T2207/10088—Magnetic resonance imaging [MRI]
    • 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
    • G06T2207/30016—Brain

Landscapes

  • Magnetic Resonance Imaging Apparatus (AREA)

Abstract

The invention discloses an automatic segmentation method of a magnetic resonance imaging brain gray matter nuclear group, which can automatically obtain seed points positioned in each brain gray matter nuclear group by calculation by utilizing a standard reference image and a matched atlas; preprocessing an image to be segmented by utilizing the incoherence among the seed points, eliminating fuzzy areas among adjacent nuclei, and automatically obtaining the segmentation contour of the target nuclei by combining a level set segmentation method. The invention can more effectively deal with the variability of the nuclear group morphology in different images, and the segmentation result is more accurate. The invention can accurately segment specific nuclei in the brain, such as substantia nigra, red nucleus and the like, and can help the diagnosis and pathological research of diseases such as Parkinson's disease and the like to a great extent.

Description

Automatic segmentation method for magnetic resonance imaging brain gray matter nuclei
Technical Field
The invention relates to the technical field of magnetic resonance imaging, in particular to a method for segmenting a brain gray matter nucleus in magnetic resonance imaging.
Background
The magnetic resonance imaging technology is widely applied to the field of medical diagnosis at present, and as an important diagnosis and treatment means, the magnetic resonance imaging technology has the greatest characteristic of providing rich image contrast information. With the continuous improvement of the spatial resolution of magnetic resonance imaging, the magnetic resonance imaging technology can show the small gray matter tissues of the living body in the deep region of the human brain, such as red nucleus, black stroma, dentate nucleus and other brain nuclei in the basal ganglia, which cannot be observed before. A number of studies show that the nuclei are greatly related to certain neurodegenerative diseases (such as Parkinson's disease and Wilson's disease), and especially that substantia nigra is important in relation to Parkinson's disease. Therefore, accurate segmentation of these specific nuclei can provide important help for the study of effective surgical treatments (such as deep brain stimulation) and the pathological mechanisms of related diseases. However, conventional magnetic resonance T1 weighted images do not clearly show these fine nuclear tissues. Quantitative magnetic susceptibility mapping is a new and important advance in magnetic resonance imaging technology in recent years, which can provide new image contrast information based on the calculation of spatial magnetic susceptibility. Meanwhile, the quantitative susceptibility imaging mode is more sensitive to quantitatively measuring the change of the histopathological structure, so that the method is more meaningful for segmenting a specific nuclear mass in a quantitative susceptibility map.
However, due to these various forms of organization, the current common segmentation method based on atlas cannot achieve accurate and effective segmentation, and the result is greatly different from that of the golden standard segmentation map. The level set segmentation method is a classical and effective automatic segmentation method, which can cope with variable tissue morphology, but because the space positions of the tiny nuclei in the deep brain region are close, the classical level set method cannot effectively distinguish the adjacent nuclei such as substantia nigra and red nucleus. This affects the results of the diagnosis and study.
Disclosure of Invention
The invention aims to provide an automatic segmentation method of a magnetic resonance imaging brain gray matter nucleus, which overcomes the defects in the prior segmentation technology and provides an image preprocessing means combining the discontinuity of seed points and a segmentation method of a level set.
The specific technical scheme for realizing the purpose of the invention is as follows:
an automatic segmentation method of magnetic resonance imaging brain gray matter nuclei is characterized in that: the method comprises the following specific steps:
step 1: registering an image to be segmented with a standard reference image to obtain a conversion matrix for registering the image;
step 2: according to the conversion matrix, transforming an original image set matched with the standard reference image to obtain a new image set matched with the image to be segmented;
and step 3: linearly mapping the obtained new image set to an image to be segmented to obtain an interested area of the gray matter nuclear group;
and 4, step 4: obtaining the positions of different nuclear group seed points by calculating the gravity centers in the regions of interest;
and 5: preprocessing an image to be segmented by utilizing the incoherence among seed points in different nuclear group tissues in the magnetic resonance image;
step 6: adjusting parameters, segmenting the preprocessed image by using a level set segmentation model, and drawing a segmented contour map; wherein:
the step 5 specifically includes:
i) setting the iteration number N;
ii) searching a shortest spatial path connected between the two seed points;
iii) drawing an intensity profile of the corresponding gray scale image in the space path;
iv) judging whether each pixel point passed by the space path belongs to a certain kernel group by using a threshold-based method; removing pixel points which do not belong to any nuclear group from the image, namely setting the gray value as 0;
v) repeating the steps ii), iii) and iv) N times, and ending the iteration.
In the iteration process, the shortest spatial path searched each time cannot pass through the pixel points eliminated in the last iteration, and the spatial path connected between the two sub-points changes along with the iteration times.
The level set segmentation model is represented by the following formula:
Figure BDA0001225835340000031
in the formula, omega1And Ω2Represent the areas inside and outside the contour C, respectively; lambda [ alpha ]1、λ2And v represent positive weighting factors, respectively. x and y respectively represent the positions of pixel points in the image; f. of1(x) And f2(x) Is two functions that approximate the gray values of the inner and outer regions of the contour C; kσIs a gaussian function with a scaling factor of σ; and solving the process of minimizing the energy function epsilon, namely obtaining the contour of the nuclear mass to be segmented. The invention solves the minimum value and the outline of an energy function epsilon by using a split Brazilian iteration methodC may evolve to the boundaries of the tissue. In the iterative calculation, a level set function φ (x) is referenced to represent the contour C.
The invention utilizes the incoherence among various sub-points to preprocess the image and combines the level set segmentation method to segment the specific nuclei in the human brain. Compared with the classical level set segmentation result, the method can effectively distinguish the nuclear groups with close spatial positions. Compared with the current common method based on atlas segmentation, the method can more effectively cope with the variability of the nuclear cluster morphology in different images, and the segmentation result is more accurate. The invention can accurately segment specific nuclei (such as substantia nigra, red nucleus and the like) in the brain, and can help the diagnosis and pathological research of diseases such as Parkinson's disease and the like to a great extent.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a flowchart illustrating the image preprocessing steps of the present invention;
FIG. 3 is a detailed diagram of the image preprocessing step according to the embodiment of the present invention;
FIG. 4 is a graph comparing the segmentation results of red nucleus and black stroma with the prior segmentation method in the embodiment of the present invention;
FIG. 5 is a graph comparing the segmentation accuracy of the red nucleus using different segmentation methods according to the present invention;
FIG. 6 is a graph comparing the segmentation accuracy of different segmentation methods applied to the substantia nigra according to the present invention.
Detailed Description
The present invention will be described in further detail with reference to the following specific examples and the accompanying drawings. The procedures, conditions, experimental protocol methods and the like for carrying out the present invention are general knowledge and common general knowledge in the art except for the contents specifically mentioned below, and the present invention is not particularly limited.
On the basis of a quantitative magnetic susceptibility map with very good contrast ratio of brain deep nuclei, the method adopts an image preprocessing mode aiming at the incoherence of seed points and combines a level set method to segment the specific gray matter nuclei in the brain deep.
The invention is described in detail below with reference to the accompanying drawings and examples.
Examples
Segmenting the deep nuclei, red nuclei and substantia nigra of the brain
The acquired magnetic resonance image data is acquired by a multi-echo gradient echo sequence, the data is from a 3T magnetic resonance imaging equipment system (Siemens MAGNETOM Trio a Tim 3T), and the number of adopted echoes is 8.
The craniocerebral cross-sectional position magnetic susceptibility diagram is reconstructed by the steps of phase fitting, phase unwrapping, background field removing, Morphology-based Dipole Inversion algorithm (MEDI) and the like of complex data acquired by a gradient echo sequence.
Referring to fig. 1, which is a flow chart of the present invention, after obtaining a quantitative magnetic susceptibility map by calculation, it is first necessary to register the calculated images of the layers with a standard template image. By using the same registration mode, the segmentation atlas corresponding to the standard template space can be deformed, and the atlas corresponding to the image space to be segmented is obtained. In this embodiment, the above steps register the images using the software "Diffeomap" provided by Issel ane l.lim. And secondly, obtaining interested areas by using the atlas through a graph searching method, wherein each interested area respectively represents the approximate position of different nuclei. In this embodiment, the seed points are selected based on the gravity center positions of the regions of interest obtained through calculation, and the gravity center positions are obtained by calculating the geometric center of the regions according to the gray values of the pixel points in the regions as weight factors.
Since the region of interest can approximately find the location of the nuclei, the various sub-points found in the above manner lie within the red nucleus or the substantia nigra. The level set contour cannot distinguish two adjacent nuclei, but because the seed points in the two adjacent nuclei are known to be unconnected by segmentation, the image to be segmented can be preprocessed by using the incoherence as a priori knowledge. The flow chart of the preprocessing step is shown in fig. 2, and is described in detail as follows:
1) in the process of each iteration, the A-STAR algorithm is used for finding the shortest spatial path connecting the two seed points. It should be noted that the shortest path searched in each iteration changes continuously, because the subsequent processing will remove some pixel points (points in black circles in the left image of fig. 3) passed in the previous path, and the shortest path between two sub-points searched again in each iteration will bypass the pixel positions removed before.
2) The intensity profile of the corresponding gray scale image in this path is plotted (as shown in the right image of fig. 3).
3) Due to the influence of uneven image gray level distribution, a simple threshold value method cannot judge. And judging whether each pixel point in the path belongs to black or red nucleus or neither red nucleus nor black by adopting a mode similar to full width at half maximum. And eliminating the pixel points which do not belong to the red nucleus and the black texture from the image (the gray value is set as 0).
4) Repeating the steps 1) to 3); after enough image pixel points which do not belong to the black matter and the red nucleus are removed, the fuzzy area between the red nucleus and the black matter disappears.
And solving the model by adopting an RSF model proposed by Chan and Vese and introducing a level set function to the preprocessed image so as to obtain a final segmentation contour. The concrete expression of the model and solution is as follows:
Figure BDA0001225835340000051
here H∈Is a smooth step function of the temperature of the sample,
Figure BDA0001225835340000061
Figure BDA0001225835340000062
in the formula, λ1、λ2And v represent positive weighting factors, respectively. x and y represent the positions of pixel points in the image, respectively. f. of1(x) And f2(x) Is two approximations to the gray values of the inner and outer regions of the contour CA function. KσIs a gaussian function whose scaling coefficient is σ. The level set function phi (x) is used for representing the contour, and the invention solves the energy function epsilon (phi, f) by using a split Brazilian iteration method1,f2) The contour will gradually evolve to the edge of the tissue along with the iteration to complete the segmentation.
The embodiment is brain data of a healthy normal volunteer, the data is from a Siemens 3.0T magnetic resonance imaging system, a 12-channel head coil is adopted for data acquisition, an 8-echo gradient echo sequence is adopted for scanning, and the parameters are as follows: TR 60ms, Δ TE 6.8ms, flip angle 15 °, field of view (FOV) 240x180mm2Image resolution 384 x 288, pixel size 0.625 x 0.625mm2And a total of 96 layers, the layer thickness being 2 mm. The results of the image segmentation are compared as shown in fig. 4 (the contour represents the segmentation result of the black texture and the red nucleus): (a) is the result of an atlas-based segmentation method; (b) segmenting a result by a classical level set method; (c) the segmentation result of the invention; (d) and (5) dividing the gold standard. The segmentation accuracy was evaluated by an expert manually delineating the deep nuclei ROI on the quantitative susceptibility map using the software ITK-SNAP 3.2.
The statistical results for the quantitative evaluation of the segmentation are shown in fig. 5 and 6. The Dice coefficient is used as a quantitative index of the segmentation accuracy, and is obtained by the following formula:
Figure BDA0001225835340000063
the similarity Dice coefficient is the ratio of the number of pixels of the correct segmentation result to the whole segmentation region (including all regions of manual segmentation and automatic segmentation), is sensitive to the difference of the sizes and the positions of the two regions, and has a value range of [0, 1], and 1 represents complete consistency.
As can be seen from the segmentation results of FIG. 4, FIG. 5 and FIG. 6, the accuracy of the nuclear group segmentation obtained by the present invention is superior to that based on the atlas registration method and the classical level set segmentation method.
The protection of the present invention is not limited to the above embodiments. Variations and advantages that may occur to those skilled in the art may be incorporated into the invention without departing from the spirit and scope of the inventive concept, and the scope of the appended claims is intended to be protected.

Claims (2)

1. An automatic segmentation method of magnetic resonance imaging brain gray matter nuclei is characterized by comprising the following specific steps:
step 1: registering an image to be segmented with a standard reference image to obtain a conversion matrix for registering the image;
step 2: according to the conversion matrix, transforming an original image set matched with the standard reference image to obtain a new image set matched with the image to be segmented;
and step 3: linearly mapping the obtained new image set to an image to be segmented to obtain an interested area of the gray matter nuclear group;
and 4, step 4: obtaining the positions of different nuclear group seed points by calculating the gravity centers in the regions of interest;
and 5: preprocessing an image to be segmented by utilizing the incoherence among seed points in different nuclear group tissues in the magnetic resonance image;
step 6: adjusting parameters, segmenting the preprocessed image by using a level set segmentation model, and drawing a segmented contour map; wherein:
the step 5 specifically includes:
i) setting the iteration number N;
ii) searching a shortest spatial path connected between the two seed points;
iii) drawing an intensity profile of the corresponding gray scale image in the space path;
iv) judging whether each pixel point passed by the space path belongs to a certain kernel group by using a threshold-based method; removing pixel points which do not belong to any nuclear group from the image, namely setting the gray value as 0;
v) repeating the steps ii), iii) and iv) N times, and ending the iteration;
in the iteration process, the shortest spatial path searched each time cannot pass through the pixel points eliminated in the last iteration, and the spatial path connected between the two sub-points changes along with the iteration times.
2. The automatic segmentation method of claim 1, wherein the level set segmentation model is represented by the following equation:
Figure FDA0002140354300000011
in the formula, omega"i-1, 2 represent the regions inside and outside the contour C, respectively; lambda [ alpha ]1、λ2And v represent positive weighting factors, respectively; x and y respectively represent the positions of pixel points in the image; f. of1(x) And f2(x) Is two functions that approximate the gray values of the inner and outer regions of the contour C; k is a radical ofσIs a gaussian function with a scaling factor of σ; and solving the process of minimizing the energy function epsilon, namely obtaining the contour of the nuclear mass to be segmented.
CN201710080824.8A 2017-02-15 2017-02-15 Automatic segmentation method for magnetic resonance imaging brain gray matter nuclei Active CN106898000B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710080824.8A CN106898000B (en) 2017-02-15 2017-02-15 Automatic segmentation method for magnetic resonance imaging brain gray matter nuclei

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710080824.8A CN106898000B (en) 2017-02-15 2017-02-15 Automatic segmentation method for magnetic resonance imaging brain gray matter nuclei

Publications (2)

Publication Number Publication Date
CN106898000A CN106898000A (en) 2017-06-27
CN106898000B true CN106898000B (en) 2020-02-18

Family

ID=59199023

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710080824.8A Active CN106898000B (en) 2017-02-15 2017-02-15 Automatic segmentation method for magnetic resonance imaging brain gray matter nuclei

Country Status (1)

Country Link
CN (1) CN106898000B (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108323389B (en) * 2018-01-18 2019-11-08 华南农业大学 The rice transplanting rice shoot spacing in the rows of rice transplanter and the detection method and device of cave rice shoot number
CN109242849A (en) * 2018-09-26 2019-01-18 上海联影智能医疗科技有限公司 Medical image processing method, device, system and storage medium
KR102097738B1 (en) * 2020-03-03 2020-04-06 주식회사 휴런 PARKINSON'S DISEASE DIAGNOSIS APPARATUS BASED ON AI(Artificial Intelligence) AND METHOD
CN111358585A (en) * 2020-03-18 2020-07-03 山东大学 CT image-based porous implant manufacturing method and system
CN111583189B (en) * 2020-04-16 2024-02-27 武汉联影智融医疗科技有限公司 Brain nucleus positioning method, device, storage medium and computer equipment

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2011061511A (en) * 2009-09-10 2011-03-24 Dainippon Printing Co Ltd Fish-eye monitoring system
CN102663700A (en) * 2012-04-13 2012-09-12 福州大学 Segmentation method for adhering grain binary image
CN104899849A (en) * 2014-01-21 2015-09-09 武汉联影医疗科技有限公司 Multi-target interactive image segmentation method and device

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2011061511A (en) * 2009-09-10 2011-03-24 Dainippon Printing Co Ltd Fish-eye monitoring system
CN102663700A (en) * 2012-04-13 2012-09-12 福州大学 Segmentation method for adhering grain binary image
CN104899849A (en) * 2014-01-21 2015-09-09 武汉联影医疗科技有限公司 Multi-target interactive image segmentation method and device

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
复杂环境中鸡蛋识别及粘连分离方法研究;劳凤丹 等;《农业机械学报》;20130430;第44卷(第4期);第213-227页 *

Also Published As

Publication number Publication date
CN106898000A (en) 2017-06-27

Similar Documents

Publication Publication Date Title
CN106898000B (en) Automatic segmentation method for magnetic resonance imaging brain gray matter nuclei
Tomas-Fernandez et al. A model of population and subject (MOPS) intensities with application to multiple sclerosis lesion segmentation
Eskildsen et al. BEaST: brain extraction based on nonlocal segmentation technique
Le Troter et al. Volume measurements of individual muscles in human quadriceps femoris using atlas-based segmentation approaches
EP3133993B1 (en) Lean tissue volume quantification
US20150015253A1 (en) Automated image analysis for magnetic resonance imaging
Stieltjes et al. Diffusion tensor imaging in primary brain tumors: reproducible quantitative analysis of corpus callosum infiltration and contralateral involvement using a probabilistic mixture model
WO2018176985A1 (en) Quantitative calculation method for level of brain atrophy based on automatic segmentation
Peng et al. Development of a human brain diffusion tensor template
Auger et al. Semi-automated left ventricular segmentation based on a guide point model approach for 3D cine DENSE cardiovascular magnetic resonance
Shyu et al. Measuring complexity of fetal cortical surface from MR images using 3-D modified box-counting method
Alsenan et al. A Deep Learning Model based on MobileNetV3 and UNet for Spinal Cord Gray Matter Segmentation
Grivalsky et al. Segmentation of gliomas in magnetic resonance images using recurrent neural networks
CN111445553B (en) Depth learning-based intra-voxel incoherent motion imaging acceleration method and device
Barta et al. A stochastic model for studying the laminar structure of cortex from MRI
Pai et al. Evaluation of fiber bundles across subjects through brain mapping and registration of diffusion tensor data
CN117177704A (en) System and method for template-based automatic detection of anatomical structures
CN103514607A (en) Dynamic contrast enhancement magnetic resonance image detection method
Eskildsen et al. Quantitative comparison of two cortical surface extraction methods using MRI phantoms
Noorizadeh et al. Multi-atlas based neonatal brain extraction using atlas library clustering and local label fusion
Park et al. Deep learning-based brain metastatic detection and treatment response assessment system on 3D MRI
Bourgeat et al. Improved cortical thickness measurement from MR images using partial volume estimation
Lemieux Automatic, accurate, and reproducible segmentation of the brain and cerebro-spinal fluid in T1-weighted volume MRI scans and its application to serial cerebral and intracranial volumetry
Jørgensen et al. Unsupervised assessment of subcutaneous and visceral fat by MRI
KR20160053228A (en) Field map meansuring method for fat quantification of high field magnetic resonance imaging

Legal Events

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