配准ANTs工具的官方介绍文档_ants配准,ants配准资源-CSDN文库

需积分: 50 162 浏览量 2022-05-26 22:24:43 上传评论 1 收藏 6.18MB PDF 举报

资源详情

资源评论

资源推荐

Advanced Normalization Tools (ANTS)

Release 2.x

Brian B. Avants

, Nick Tustison

and Hans Johnson

July 10, 2014

University of Pennsylvania

University of Virginia

University of Iowa

Abstract

We provide examples and highlights of Advanced Normalization Tools (ANTs), versions 2.x, that ad-

dress practical problems in real data.

Contents

1 Introduction 2

1.1 Structure of this document and its examples . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.2 Example 1: Quick SyN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

1.3 The antsRegistration executable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

Initializing antsRegistration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.4 The antsApplyTransforms executable . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.5 Using antsApplyTransforms or antsApplyTransformsToPoints . . . . . . . . . . . . . 5

1.6 The antsMotionCorr executable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

1.7 I/O data formats in ANTs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Image volumes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Afﬁne transformation ﬁle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Deformation ﬁeld ﬁle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Labeled point sets—Currently not supported by antsRegistration, only by old ANTS . . . . . 6

1.8 The ImageMath executable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.9 ANTs/Scripts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

2 Image registration with ANTs 8

2.1 World coordinates: Use your header! . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.2 ANTs transformation models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Afﬁne and rigid registration—FIXME—this section and ﬁgures . . . . . . . . . . . . . . . 10

Deformable registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2.3 ANTs similarity terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.4 Choosing a metric . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

2.5 Notes on basic brain mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2.6 Normalization across different modalities: E.g. DTI . . . . . . . . . . . . . . . . . . . . . . 20

2.7 Multivariate normalization with ANTs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

2.8 Notes on large deformation mapping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.9 Optimal template construction with ANTs . . . . . . . . . . . . . . . . . . . . . . . . . . . 21

2.10 2D to 3D registration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2.11 More ANTs examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

3 Image segmentation and labeling 25

3.1 Basic segmentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.2 Prior and template-based image segmentation . . . . . . . . . . . . . . . . . . . . . . . . . 25

3.3 Cortical thickness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26

4 Data visualization with ANTs 28

4.1 Creating faux-colormapped images with ConvertScalarImageToRGB . . . . . . . . . . . . 28

4.2 Figure production and large-scale data inspection using CreateTiledMosaic . . . . . . . . 28

4.3 Volumetric visualizations with antsSurf . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

5 ANTs-based studies 32

5.1 Brain mapping in the presence of lesions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

5.2 Statistical mapping with ANTs: Morphometry, function, jacobian, thickness . . . . . . . . . 34

5.3 Statistics with ANTs and R: ANTsR . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

6 Dependencies and Related Software 36

6.1 How to build ANTs/ANTsR for users who are new to scientiﬁc computing . . . . . . . . . . 36

6.2 Dependencies and Compilation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

6.3 Pipelining with ANTs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

7 Annotated Bibliography (Old) 38

1 Introduction

This update to ANTs documentation was initiated April 29, 2014. This document does not cover all of ANTs

functionality — but summarizes the most frequently used components. Originally, the ANTs framework

provided open-source functionality for deformable image registration with small or large deformations, as

shown in ﬁgure 1. Independent evaluation of the 0.0 version of ANTs software, applied to “control” data,

placed the toolkit as a top performer amongst 14 methods [26]. Developer evaluation showed stronger differ-

ences with other methodology in neurodegenerative neuroimaging data, where large deformation is required

[10]. ANTs has since grown to include N4 bias correction [37], additional evaluation of multiple modalities

and organ systems [36, 38, 28], univariate or multivariate image segmentation [19, 40], tighter integra-

tion with the Insight ToolKit [35, 17], a well-evaluated cortical thickness pipeline [41] and, more recently,

visualization tools and integration with R[42]. ANTs serves as both a core library for further algorithm

development and also as a command-line application-oriented toolkit. ANTs also has a permissive software

license that allows it to be employed freely by industry [31]. ANTs enables diffeomorphic normalization

with a variety of transformation models, optimal template construction, multiple types of diffeomorphisms,

multivariate similarity metrics, diffusion tensor processing and warping, image segmentation with and with-

1.1 Structure of this document and its examples 3

Figure 1: The original goal of ANTs was to develop public, open source large deformation image registration. This is a

classic example showing the progress of deforming a half C to a full C along a geodesic diffeomorphism. The deforming

grid accompanies each deformed image. See http://stnava.github.io/C/ for example data and code.

out priors and measurement of cortical thickness from probabilistic segmentations. The normalization tools,

alone, provide a near limitless range of functionality and allow the user to develop customized objective

functions. Objective functions in ANTs are of the form:

Deformation Cost + Data Terms,

and the command line reﬂects this balance of two terms. As mentioned above, the data

term may combine multiple different measures of similarity that are optimized in parallel, for

instance, image similarity and landmark terms. This document seeks to provide a practical

overview of basic functionality and some of the common use cases that users seek. Addi-

tional information is available online – see http://stnava.github.io/ANTs/. For compilation de-

tails, see: https://brianavants.wordpress.com/2012/04/13/updated-ants-compile-instructions-april-12-2012/

or section 6.2. The most important core C++-based ANTs programs are: antsRegistration,

antsApplyTransforms, Atropos for segmentation, N4BiasFieldCorrection for inhomogeneity cor-

rection, KellyKapowski for estimating thickness, ImageMath image processing utilities and, ﬁ-

nally, sccan for sparse dimensionality reduction. There are many other programs which are listed

in https://github.com/stnava/ANTs/tree/master/Examples. The scripts in ANTs wrap the core pro-

grams and are in https://github.com/stnava/ANTs/tree/master/Scripts. Perhaps the most important are

antsCorticalThickness.sh as a wrapper of several sub-programs that support our cortical thick-

ness pipeline [41], antsMultivariateTemplateConstruction2.sh for template construction and

antsRegistrationSyN.sh as a basic interface to a few commonly used registration approaches.

1.1 Structure of this document and its examples

This document is generated with L

X and is version controlled at https://github.com/stnava/ANTsDoc.

The examples, here, have data and example scripts stored at dedicated github repositories, to which we will

refer. Other (simpler) examples will use data and scripts that are stored in the ANTsDoc git repository.

These scripts are (loosely speaking) tested and should serve as reproducible examples for the reader to try.

All data and code is available via ANTs-related repositories. The compile script builds both latex and tests

the example scripts.

1.2 Example 1: Quick SyN

If you want a decent, fast registration you might run something like this:

1 antsRegistrationSyNQuick.sh -d 2 -f r16slice.nii.gz -m r64slice.nii.gz -o $op

This document is a work in progress. Please check for updates with each release.

1.3 The antsRegistration executable 4

The variable $op represents the output preﬁx for the ﬁlename. This will be the case in many of the follow-

ing examples. The -d option denotes image dimensionality, -f option denotes the “ﬁxed” image and the -m

option denotes the “moving” image. The moving image will be deformed to match the ﬁxed image. The in-

verse of these maps deform the ﬁxed to the moving image. Output is determined by -o where the output will

be named with preﬁx=outputEx1.sh (the word output concatenated with the name of the Example 1 script)

and include a preﬁx0GenericAfﬁne.mat (the low-dimensional afﬁne transform which may be inverted nu-

merically), the preﬁx1Warp.nii.gz (the diffeomorphic transformation pulling the afﬁne transformed moving

image toward the ﬁxed image), and the preﬁx1InverseWarp.nii.gz (the inverse diffeomorphic transforma-

tion pulling the ﬁxed image toward the afﬁne transformed moving image). The caveats for this “canned”

approach include: 1. registration performance can always be improved by using prior knowledge [44]; 2.

there are many assumptions about the data embedded in the above call and they may not be appropriate

for whatever problem is at hand; 3. you must have some facility with the command line to run a shell

script. It is, in general, better to understand a little bit about image registration rather than running methods

blindly. To aid readers in this, we have two options: 1. github issues https://github.com/stnava/ANTs/issues;

2. sourceforge discussion or help sites http://sourceforge.net/p/advants/discussion/. Feel free to use either

to ask clarifying questions. We note that many issues have been discussed previously and you might ﬁnd

answers by searching the archives.

1.3 The antsRegistration executable

The antsRegistration program itself is the central program encapsulating normalization/registra-

tion functionality. Its main output is an afﬁne transform ﬁle and a deformation ﬁeld, potentially

with inverse. Options allow the user to navigate the similarity and transformations that are avail-

able. antsRegistration allows multiple similarity and optimization criteria as options. The program

is wrapped in antsRegistrationSyN.sh for normalization with “out of the box” parameters and in

antsMultivariateTemplateConstruction2.sh for computationally distributed optimal (multivariate)

template construction.

Initializing antsRegistration

You can use the -r option in antsRegistration to initialize a registration with an ITK .mat format transfor-

mation matrix, with a deformable mapping or with a center of mass alignment. See the scripts for examples.

The output transformation will include the initial transformation. FIXME need to check this.

1.4 The antsApplyTransforms executable

The antsApplyTransforms program applies ANTs mappings to images including scalars, tensors, time-

series and vector images. It also composes transforms together and is able to compute inverses of low-

dimensional (afﬁne, rigid) maps. antsApplyTransformsToPoints similarly works on point sets (see

http://stnava.github.io/chicken/ for details). One may apply an arbitrarily long series of transformations

through these programs. Thus, they enable one to compose a series of afﬁne and deformable mappings

and/or their inverses. One may therefore avoid repeated interpolations of a single image. Several different

interpolation options are available and multiple image types may be transformed including: tensors, vectors,

timeseries and d-dimensional scalar images where d = 2,3,4.

1.5 Using antsApplyTransforms or antsApplyTransformsToPoints 5

1.5 Using antsApplyTransforms or antsApplyTransformsToPoints

For example: to apply the transform to the moving image and the inverse to the ﬁxed image:

1 antsRegistrationSyNQuick.sh -d 2 -f B.nii.gz -m A.nii.gz -o RegA2B

2 antsApplyTransforms -d 2 -i A.nii.gz -o ADeformed.nii.gz -r B.nii.gz -t RegA2B1Warp.

nii.gz -t RegA2B0GenericAffine.mat

3 antsApplyTransforms -d 2 -i B.nii.gz -o BDeformed.nii.gz -r A.nii.gz -t [

RegA2B0GenericAffine.mat,1] -t RegA2B1InverseWarp.nii.gz

The numbers of the transformations (here, 0 and 1 because there is only a deformation and afﬁne mapping)

relate to the order in which the transforms are computed, during optimization, and also the order in which

they should be applied. There is reasonably complete discussion of this framework in [17].

The usage of antsApplyTransformsToPoints is nearly identical. However, it is critical to recognize that

transforms that are one-to-one and onto in image space may not be in point space. Therefore, points are

transformed by the inverses of the transformation that is applied to images. This is discussed in several

places in the image registration literature but we ﬁrst discussed this in [16].

1.6 The antsMotionCorr executable

Performs motion correction of time-series data. Control parameters are similar to antsRegistration.

See the example http://stnava.github.io/fMRIANTs/. This example also shows how to run basic CompCor

on fmri data. Our minimal fMRI pipeline involves running antsMotionCorr and CompCorr to factor out

nuisance variables. More complex approaches require ANTsR.

1.7 I/O data formats in ANTs

ANTs supports 2D, 3D and, in some cases, 4D images. Since ANTs is implemented in concert with the

Insight ToolKit (ITK) http://www.itk.org/, it is able to read and write the popular data formats supported

through ITK. ANTs also uses the ITK view of world-coordinates which can be confusing, at times, when

one is mixing software. There is much discussion of this on the web and in ANTs discussion sites/issues.

Whatever is relevant for ITK world coordinates is relevant to ANTs as these two software systems agree, in

their entirety, about what the voxel to physical space coordinate mapping should be. There are four basic

types of data for ANTs.

Image volumes

ANTs supports 2D, 3D, 4D images, including

• Nifti (.nii, .nii.gz)

• Analyze (.hdr + .img / .img.gz)

• MetaImage (.mha)

• Other formats through itk::ImageFileWriter / itk::ImageFileWriter such as jpg, tiff, etc. See ITK

documentation.

剩余40页未读，继续阅读

评论收藏

内容反馈

Joker007

粉丝: 8752
资源: 19

配准ANTs工具的官方介绍文档

评论0

最新资源

配准ANTs工具的官方介绍文档

评论0

ants工具箱.zip

Advanced Normalization Tools ( ANTs ):多维图像配准，分割和统计-开源

ANTS-GUIDE

医学配准软件和库的安装介绍.pdf

ants 工具箱

ANTs:先进的标准化工具（ANTs）

ants代码性能检查工具

ANTsDoc:ANT 文档

Algorithms for Image Registration: Advanced Normalization Tools (ANTS)

图像配准工具包

ArcGIS影像自动配准工具

Matlab实现多种图像配准

elastix-5.0.0_配准_医学图像配准程序_图像配准_

图像配准，imagematching

异源图像配准文档

Google下载影像批量自动配准

基于ITK的图像配准实现.zip_ITK_ITK配准_itk图像配准_图像配准_基于ITK的配准

ArcGIS 影像图配准

simpleITK医学图像配准

RANSAC算法在图像配准中的应用文档

基于ICP+sift算法的点云特征提取与配准matlab仿真,含仿真操作录像+word说明文档

手动配准基于Matlab

SAC-IA粗配准+ICP精配准

ICP_MATLAB_Implementation-master_点云_粗配准_icp_点云配准_粗配准matlab_源码

图像配准-医学图像配准技术

InSAR图像配准程序

医学图像配准方法

YOLOv8-deepsort 实现智能车辆目标检测+车辆跟踪+车辆计数

YOLOv8网络结构图，自制visio文件，yolov8.vsds，需要的自取，在原有的基础上直接改就行了

最新资源

Algorithms for Image Registration:Advanced Normalization Tools (ANTS)