vbm8使用手册 中文
http://wenku.baidu.com/view/ea101c30eefdc8d376ee323f.html
vbm5处理流程
http://wenku.baidu.com/view/cd9b5ac20c22590102029d98.html?re=view
vbm8使用手册 英文
静息态论坛的一些讨论:
http://restfmri.net/forum/node/1997全脑体积的影响作为协变量去除掉,但是应该使用原始空间的c1还是使用wc1或者mwc1作为协变量呢?
http://restfmri.net/forum/node/1345 How
to get a mean grey matter map as a mask?
mwc1***.img的是灰质体积wc1***.img的是灰质密度mwc2***.img的是白质体积wc2***.img的是白质密度
c3那个是csf
http://restfmri.net/forum/node/1784 你说的各种segmentation的办法,现在有直接分割,即vbm和spm里面的segmentation,我记得这种segmentation叫unified
segmentation,
http://restfmri.net/forum/node/2014 wc1*(wc2*),mwc1*(mwc2*)及smwc1*(smwc2*)用于后续计算灰质(白质)密度和体积。其余的都可以删掉。
http://restfmri.net/forum/node/1686 我想加入灰质萎缩做协变量,那么是加入wc1*图像还是mwc1图像?mwc1图像是灰质密度还是灰质体积?谢谢指导
dparsf-A 版本里也有vbm的处理模板。
全脑体积计算软件:volume one和IBASPM-----vbm可以计算
==================================================================
VBM-学习汇总
==================================================================
非常感谢Li老师的耐心指导VBM的学习。我以前误认为:在功能像与结构像进行配准,结构像分割的那部的数据即可用于VBM分析。Dpabi-dparsfA有对VBM的处理功能,非常简单;FSL-VBM;SPM-VBM如下将详细介绍:
prepare your T1-weighted images in
the right format
fslvbm_1_bet -
carry out brain extraction on all T1 images
fslvbm_2_template -
create the study-specific symmetric grey matter
template 生成模板这步:需要将不同组人数不同的弄成等组人数,来产生模板。即:A组15人,B组20人。那生成模板时,B组要选取15人,A组继续使用15人,即等组人数来生成模板。----这也是与SPM-VBM不同的地方:即SPM-VBM使用该程序里默认模板,不需要生成组模板再进行后续分析。
fslvbm_3_proc -
register all the grey matter images to the template, modulate and
smooth them with different kernel sizes and finally runs an initial
GLM analysis for qualitative evaluation
randomise - carry
out voxelwise GLM analysis using permutation testing
SPM-VBM的汇总(基于VBM8-manual的汇总)
大致流程:“preprocessing---> quality
check ---> smoothing---> statistical analysis”
其中,preprocessing主要基于VBM下Estimate and write文件
quality check主要基于VBM下check data quality文件
Smoothing和statistical analysis主要基于SPM8的工作目录进行。
preprocessing主要基于VBM下Estimate and write文件可以直接用默认文件:在manual中,也详细介绍了非默认的方法。
1.
Estimation Options-对于非儿童被试,都选择默认参数,对于儿童被试需要制作自己的模板customized
Tissue Probability
Maps。关于儿童这块制作,manual后面有详细介绍。
2.
Extended Options-可以选择high-dimensional DARTEL
normalization,而非默认的low—dimensional: SPM default. 即采用DARTEL的方法;对于非脑组织(即头皮之类的)的去除,可以选择light
cleanup或thorough cleanup(彻底的清除)。这对一些脑萎缩的头部处理非常有用。之后,有两种去噪的方法,选择默认的Denoising
(multi-threaded).MRF也是选择默认数值0.15即可,但如填0就表示两种SANLM&MRF两种滤波都不选择操作。
3.
Writing Options 是分别对GM/WM和CSF进行操作。注意最好选择默认的“Modulated
normalized-non linear only”.
该步骤可根据个体脑大小来对比the
absolute amount of tissue. 这步骤很关键,必须选择这个。
但如果不选择“Modulated normalized-non linear
only”的话,“Affine+non-linear”&“global normalization”(这步是后续的统计模型中操作,具体看manual介绍)的组合操作就与“Modulated
normalized-non linear only”
操作相似!或者选择“Affine+non-linear”与numeric
brain volume为covariate。但后面这两组方法,都不及“Modulated
normalized-non linear only”。
“Native space”指在原始脑数据中产生tissue
class images in spatial。该方法可以估计global tissue volumes全脑体积(即GM+WM+CSF),但由于其失去全脑voxel-wise,故在VBM处理中不适合做。那不做此步VBM中GM/WM和CSF可以选择它“Native
space”(No),那在“Estimate
and write”里每一名被试生成的“*_seg8.txt”文件里就含有每一名被试的GM、WM、CSF的值,将其相加即为全脑体积,即操作步骤为VBM-data
presentation-calculate raw volume for GM/WM/CSF。-----不过,这步骤并不影响结果,可以选择不做,那不做此步VBM中GM/WM和CSF可以选择它“Native
space”(No)。
“Normalized” 在脑模板中产生tissue class images in spatial。这对于VBM分析非常有用。所以VBM中GM/WM和CSF可以选择它“Normalized”(No)。-----不过,这步骤并不影响结果,可以选择不做。
DARTEL export选择affine的方法。
“bias
corrected image volume”用于移除MRI inhomogeneities和噪音,并将其写进normalized
或
native 空间。这步骤对于quality control和产生所有组的normalized T1 images非常有用,用于display.
“ partial volume effect
(PVE) label image volume”写进”normalized”或“native space“,该步骤作为quality control和将来用image重构surface非常有用.
“Jacobian
determinant”将每个voxel写进normalized space 这步骤在FSL里面也有。选择“Jacobian
determinant”(YES)。
“deformation
fields”----inverse + forward 原始空间和标准空间相互转换,用于后续研究使用。
之后运行,质量监控这步就是选择“wm*”文件查看,对于不合格的数据。就得重新操作,这种操作类似dparsfa里面的reorient,将结构像的调整至原始点。
注意:最后标准化和smooth后会产生一个1.5*1.5*1.5的数据,以进行统计分析时,
a.可将每个被试用于统计分析的数据进行reslice 2*2*2 (结构像一般都是重新采样成这体素大小的,我尝试过采样成1*1*1的结果,统计分析后结果矩阵太大,都无法用rest软件查看结果;而采样成3*3*3的结果时,有时结果跑到脑外去了。),之后,再用一个“与重采样相同的dimension,相同的voxel,相同的origin的灰质mask或全脑mask(即dimension91*109*91;voxel2*2*2;orgin:46,64,37)”进行统计分析;
b.或者将smooth后会产生一个1.5*1.5*1.5的数据,以Estimation
Options中的Tissue Probability
Maps(TPM,它共有6层,第一层为灰质,第二层为白质...可以用spm里面reg查看,点击图像后,再点击1:6)的第一层灰质为mask进行统计分析,之后再将结果文件重新采样成2*2*2的结果进行查看。
Note:Estimation
Options中的Tissue Probability
Maps(TPM,它共有6层,第一层为灰质,第二层为白质...可以用spm里面reg查看,点击图像后,再点击1:6),我们尝试将TPM的第一层提取出来,之后重新采样了,发现:重新采样后的图像的origin为46,55,46与smooth后会产生一个1.5*1.5*1.5的数据再重采样的orgin46,64,37是不同的。所以不能将TPM第一层提取出来再采样再进行统计分析!
==================================================================
CAT-学习汇总
==================================================================
_______ ___
_______
| ____/ / _ \ \_
_/
| |___ / /_\ \
| | Computational Anatomy Toolbox
|____/ /_/ \_\ |_|
CAT12 -
http://www.neuro.uni-jena.de
CAT default file:
D:\Program
Files\MATLAB\R2012a\toolbox\spm12\toolbox\cat12\cat_defaults.m
------------------------------------------------------------------------
Running job #1
------------------------------------------------------------------------
Running 'CAT12: Segmentation'
WARNING: Please note that no additional modules in the batch
can be run
except CAT12 segmentation. Any
dependencies will be broken for
subsequent modules if you
split the job into separate processes.
Running job 1:
.\1_t1_mpr_sag_iso_mww.img,1
.\2_t1_mpr_sag_iso_mww.img,1
.\3_t1_mpr_sag_iso_mww.img,1
.\4_t1_mpr_sag_iso_mww.img,1
.\5_t1_mpr_sag_iso_mww.img,1
.\6_t1_mpr_sag_iso_mww.img,1
.\7_t1_mpr_sag_iso_mww.img,1
.\8_t1_mpr_sag_iso_mww.img,1
.\9_t1_mpr_sag_iso_mww.img,1
.\10_t1_mpr_sag_iso_mww.img,1
WARNING: No background processes possible because your SPM
installation is located in
a folder that contains spaces.
Please set the number of processes in the
GUI
to '0'. In order to split your
job into different processes,
please do not use any spaces
in folder names!
------------------------------------------------------------------------
CAT12 r1092: .\1_t1_mpr_sag_iso_mww.img,1
------------------------------------------------------------------------
SANLM denoising (NCstr=Inf): 32s
APP: Rough bias correction:
Initialize 4s
Estimate background 5s
Initial correction 6s
Refine background 4s
Final correction 5s
Background correction 1s
Final scaling 2s
28s
Coarse affine registration: 4s
Affine registration 8s
SPM preprocessing 1: 105s
SPM preprocessing 2: 80s
Global intensity correction: 20s
SANLM noise correction: 3s
Local adaptive segmentation (LASstr=0.50):
Prepare maps 4s
Prepare partitions 2s
Prepare segments 15s
Estimate local tissue thresholds
41s
SANLM noise correction for LAS
4s
66s
ROI segmentation (partitioning):
Atlas -> subject space
8s
Major structures 4s
Ventricle detection 4s
NO WMH detection (CSF ~7%)
0s
Closing of deep structures
1s
Side alignment 4s
Final corrections 2s
24s
Skull-stripping using graph-cut (gcutstr=0.50):
WM initialisation 4s
GM region growing 1s
GM-CSF region growing 4s
CSF region growing 4s
Ventricle filling 3s
15s
Amap using initial SPM12 segmentations (MRF filter strength
0.06): 21s
AMAP peaks: [CSF,GM,WM]
= [0.42±0.07,0.71±0.10,0.99±0.02]
Final cleanup (gcutstr=0.50):
Level 1 cleanup (ROI estimation)
4s
Level 1 cleanup (brain masking)
2s
Level 2 cleanup (CSF correction)
2s
Level 3 cleanup (CSF/WM PVE)
1s
9s
Internal WMH correction for spatial normalization
(WMHCstr=0.50): 4s
Dartel registration with 1.50 mm:
167s
Write result maps: 55s
ROI estimation:
Data mapping to normalized space
20s
ROI estimation of 'hammers' atlas
4s
ROI estimation of 'neuromorphometrics'
atlas 6s
ROI estimation of 'lpba40' atlas
2s
33s
Quality check: 12s
Print 'Graphics' figure to:
I:\VBM\T1Img-CAT\1_t1_mpr_sag_iso_mww.pdf
------------------------------------------------------------------------
CAT preprocessing takes 10 minute(s) and 35 second(s).
Image Quality Rating (IQR): 85.44% (B)
Segmentations are saved in I:\VBM\T1Img-CAT\1\mri
Reports are saved in I:\VBM\T1Img-CAT\1\report
Labels are saved in I:\VBM\T1Img-CAT\1\label
------------------------------------------------------------------------
------------------------------------------------------------------------
CAT12 r1092: .\2_t1_mpr_sag_iso_mww.img,1
------------------------------------------------------------------------
SANLM denoising (NCstr=Inf): 30s
APP: Rough bias correction:
Initialize 4s
Estimate background 3s
Initial correction 6s
Refine background 4s
Final correction 5s
Background correction 1s
Final scaling 3s
25s
Coarse affine registration: 8s
Affine registration 7s
SPM preprocessing 1: 130s
SPM preprocessing 2: 82s
Global intensity correction: 19s
SANLM noise correction: 3s
Local adaptive segmentation (LASstr=0.50):
Prepare maps 4s
Prepare partitions 2s
Prepare segments 15s
Estimate local tissue thresholds
41s
SANLM noise correction for LAS
4s
66s
ROI segmentation (partitioning):
Atlas -> subject space
8s
Major structures 4s
Ventricle detection 9s
NO WMH detection (CSF ~9%)
0s
Closing of deep structures
1s
Side alignment 4s
Final corrections 2s
29s
Skull-stripping using graph-cut (gcutstr=0.50):
WM initialisation 4s
GM region growing >>
========================= 上述成两组处理呈现的内容
=========================如下是一起处理后,matlab呈现的内容==================
------------------------------------------------------------------------
Running job #2
------------------------------------------------------------------------
Running 'CAT12: Segmentation'
------------------------------------------------------------------------
CAT12 r1092: .\1_002_t1_mpr_sag_iso_mww.img,1
------------------------------------------------------------------------
SANLM denoising (NCstr=Inf): 31s
APP: Rough bias correction:
Initialize 4s
Estimate background 3s
Initial correction 6s
Refine background 4s
Final correction 5s
Background correction 1s
Final scaling 2s
24s
Coarse affine registration: 4s
Affine registration 7s
SPM preprocessing 1: 99s
SPM preprocessing 2: 82s
Global intensity correction: 19s
SANLM noise correction: 4s
Local adaptive segmentation (LASstr=0.50):
Prepare maps 4s
Prepare partitions 2s
Prepare segments 17s
Estimate local tissue thresholds
42s
SANLM noise correction for LAS
4s
70s
ROI segmentation (partitioning):
Atlas -> subject space
9s
Major structures 5s
Ventricle detection 6s
NO WMH detection (CSF ~8%)
0s
Closing of deep structures
1s
Side alignment 5s
Final corrections 2s
28s
Skull-stripping using graph-cut (gcutstr=0.50):
WM initialisation 4s
GM region growing 2s
GM-CSF region growing 4s
CSF region growing 4s
Ventricle filling 3s
17s
Amap using initial SPM12 segmentations (MRF filter strength
0.05): 23s
AMAP peaks: [CSF,GM,WM]
= [0.42±0.08,0.72±0.10,0.99±0.02]
Final cleanup (gcutstr=0.50):
Level 1 cleanup (ROI estimation)
4s
Level 1 cleanup (brain masking)
2s
Level 2 cleanup (CSF correction)
2s
Level 3 cleanup (CSF/WM PVE)
2s
11s
Internal WMH correction for spatial normalization
(WMHCstr=0.50): 3s
Dartel registration with 1.50 mm:
162s
Write result maps: 55s
ROI estimation:
Data mapping to normalized space
20s
ROI estimation of 'hammers' atlas
4s
ROI estimation of 'neuromorphometrics'
atlas 6s
ROI estimation of 'lpba40' atlas
2s
32s
Quality check: 12s
Print 'Graphics' figure to:
I:\VBM\T1Img-CAT\1_002_t1_mpr_sag_iso_mww.pdf
------------------------------------------------------------------------
CAT preprocessing takes 10 minute(s) and 45 second(s).
Image Quality Rating (IQR): 86.00% (B)
Segmentations are saved in I:\VBM\T1Img-CAT\1\mri
Reports are saved in I:\VBM\T1Img-CAT\1\report
Labels are saved in I:\VBM\T1Img-CAT\1\label
------------------------------------------------------------------------