磁共振影像分析之: 基于FSL的VBM分析(2)

统计分析

生成好study-specific模板, 下来要做统计分析组间差异了. 如果用fsl提供的vbm方案, 这一步对应的就是fslvbm_3_proc这个命令了.

fslvbm_3_proc

这一步和实验设计和研究问题本身很有关系. 这里讲讲fslvbm_3_proc命令的内部过程吧.

这个文件保存在 $FSLDIR/bin中, 可以用vim查看vim $FSLDIR/bin/fslvbm_3_proc. shell脚本的主体如下(省略帮助和说明文字)

#!/bin/sh
export LC_ALL=C

echo [`date`] [`hostname`] [`uname -a`] [`pwd`] [$0 $@] >> .fslvbmlog

mkdir -p stats
cd struc

echo "Now running the preprocessing steps and the pre-analyses"

/bin/rm -f fslvbm3a
for g in `$FSLDIR/bin/imglob *_struc.*` ; do
  echo $g
  echo "${FSLDIR}/bin/fsl_reg ${g}_GM template_GM ${g}_GM_to_template_GM -fnirt \"--config=GM_2_MNI152GM_2mm.cnf --jout=${g}_JAC_nl\"; \
        $FSLDIR/bin/fslmaths ${g}_GM_to_template_GM -mul ${g}_JAC_nl ${g}_GM_to_template_GM_mod -odt float" >> fslvbm3a
done
chmod a+x fslvbm3a
fslvbm3a_id=`${FSLDIR}/bin/fsl_sub -T 40 -N fslvbm3a -t ./fslvbm3a`
echo Doing registrations: ID=$fslvbm3a_id

cd ../stats

cat < fslvbm3b
#!/bin/sh

\$FSLDIR/bin/imcp ../struc/template_GM template_GM

\$FSLDIR/bin/fslmerge -t GM_merg     \`\${FSLDIR}/bin/imglob ../struc/*_GM_to_template_GM.*\`
\$FSLDIR/bin/fslmerge -t GM_mod_merg \`\${FSLDIR}/bin/imglob ../struc/*_GM_to_template_GM_mod.*\`

\$FSLDIR/bin/fslmaths GM_merg -Tmean -thr 0.01 -bin GM_mask -odt char

/bin/cp ../design.* .

for i in GM_mod_merg ; do
  for j in 2 3 4 ; do
    \$FSLDIR/bin/fslmaths \$i -s \$j \${i}_s\${j} 
    \$FSLDIR/bin/randomise -i \${i}_s\${j} -o \${i}_s\${j} -m GM_mask -d design.mat -t design.con -V
  done
done

stage_preproc2

chmod a+x fslvbm3b

fslvbm3b_id=`${FSLDIR}/bin/fsl_sub -T 15 -N fslvbm3b -j $fslvbm3a_id ./fslvbm3b`

echo Doing subject concatenation and initial randomise: ID=$fslvbm3b_id

echo "Once this has finished, run randomise with 5000 permutations on the 'best' smoothed 4D GM_mod_merg. We recommend using the -T (TFCE) option. For example:"
echo "randomise -i GM_mod_merg_s3 -o GM_mod_merg_s3 -m GM_mask -d design.mat -t design.con -n 5000 -T -V"

这个shell脚本中, 采用cat命令配合EOF符号生成了另两个脚本: fslvbm3a 和fslvbm3b . 值得一提的是, 这里采用fsl_sub命令调用并行计算进行计算加速. 但是这个命令是基于Sun grid cluster的, 所以对于普通的台式机, 或者工作站, 建议将这一步去掉, 直接运行fslvbm3a或者fslvbm3b.

我们再从头看看这个脚本. 首先来说说生成脚本fslvbm3a的部分.

fslvbm3a

这一步是对统计分析前的准备工作.

使用fsl_reg将所有灰质图像和study-specific的模板配准, 生成${g}_GM_to_template_GM, 之后乘以Jac_nl, 生成${g}_GM_to_template_GM_mod ${g}是指被试样本的名称.

fslvbm3b

1) 进入states 文件夹, 拷贝template_GM到该文件夹.

2) 用fslmerge将(../struc中的所有_GM_to_template_GM)图像沿着”时间”轴拼接concatenate到一起, 生成GM_merge
同时, 将(../struc中所有_GM_to_template_GM_mod)图像沿着”时间”轴concatenate到一起, 生成GM_mod_merg.

当然, 这里的”时间”轴, 只是一个序号而已. 并不没有时间属性. 以下是fslmerge命令的用法:

Usage: fslmerge <-x/y/z/t/a/tr> 
  [tr value in seconds]
     -t : concatenate images in time
     -x : concatenate images in the x direction
     -y : concatenate images in the y direction
     -z : concatenate images in the z direction
     -a : auto-choose: single slices -> volume, volumes -> 4D (time series)
     -tr : concatenate images in time and set the output image tr to the final option value

3) 对GM_merg 所有图像中信号强度是否0.01做binary(非0即1)的voxel沿着”时间”轴做平均, 设置输出的结果为字符char.
fslmaths GM_merg -Tmean -thr 0.01 -bin GM_mask -odt char

4) 将design.*文件全部拷贝到./stat中. 之后对GM_mod_merg中的每个图片,
对每个图片做高斯平滑, 分别选择平滑核宽度为2mm 3mm 和4mm. 然后用randomise命令做非参数统计. 将刚才进行过高斯平滑的图像作为输入, 输出文件名称为${g}_s${j} ${g}为被试文件名称, ${j}是高斯核宽度. GM_mask 是前一步生成的mask文件, -d后面是设计文件design.mat, -t后面紧跟着的是t检验设置文件,design.con 最后的-V是指在t检验时使用方差平滑.

fslmaths \$i -s \$j \${i}_s\${j} 
\$FSLDIR/bin/randomise -i \${i}_s\${j} -o \${i}_s\${j} -m GM_mask -d design.mat -t design.con -V

附录:

1) fslmaths的基本调用格式如下:

fslmaths [-dt ] [operations and inputs]

[-odt ]

fslmaths 

Usage: fslmaths [-dt ]  [operations and inputs] 
 [-odt ]

Datatype information:
 -dt sets the datatype used internally for calculations (default float for all except double images)
 -odt sets the output datatype ( default is float )
 Possible datatypes are: char short int float double input
 "input" will set the datatype to that of the original image

Binary operations:
  (some inputs can be either an image or a number)
 -add   : add following input to current image
 -sub   : subtract following input from current image
 -mul   : multiply current image by following input
 -div   : divide current image by following input
 -rem   : modulus remainder - divide current image by following input and take remainder
 -mas   : use (following image>0) to mask current image
 -thr   : use following number to threshold current image (zero anything below the number)
 -thrp  : use following percentage (0-100) of ROBUST RANGE to threshold current image (zero anything below the number)
 -thrP  : use following percentage (0-100) of ROBUST RANGE of non-zero voxels and threshold below
 -uthr  : use following number to upper-threshold current image (zero anything above the number)
 -uthrp : use following percentage (0-100) of ROBUST RANGE to upper-threshold current image (zero anything above the number)
 -uthrP : use following percentage (0-100) of ROBUST RANGE of non-zero voxels and threshold above
 -max   : take maximum of following input and current image
 -min   : take minimum of following input and current image
 -seed  : seed random number generator with following number
 -restart : replace the current image with input for future processing operations
 -save : save the current working image to the input filename

Basic unary operations:
 -exp   : exponential
 -log   : natural logarithm
 -sin   : sine function
 -cos   : cosine function
 -tan   : tangent function
 -asin  : arc sine function
 -acos  : arc cosine function
 -atan  : arc tangent function
 -sqr   : square
 -sqrt  : square root
 -recip : reciprocal (1/current image)
 -abs   : absolute value
 -bin   : use (current image>0) to binarise
 -binv  : binarise and invert (binarisation and logical inversion)
 -fillh : fill holes in a binary mask (holes are internal - i.e. do not touch the edge of the FOV)
 -fillh26 : fill holes using 26 connectivity
 -index : replace each nonzero voxel with a unique (subject to wrapping) index number
 -grid <value>  : add a 3D grid of intensity <value> with grid spacing 
 -edge  : edge strength
 -tfce   : enhance with TFCE, e.g. -tfce 2 0.5 6 (maybe change 6 to 26 for skeletons)
 -tfceS       : show support area for voxel (X,Y,Z)
 -nan   : replace NaNs (improper numbers) with 0
 -nanm  : make NaN (improper number) mask with 1 for NaN voxels, 0 otherwise
 -rand  : add uniform noise (range 0:1)
 -randn : add Gaussian noise (mean=0 sigma=1)
 -inm  :  (-i i ip.c) intensity normalisation (per 3D volume mean)
 -ing  :  (-I i ip.c) intensity normalisation, global 4D mean)
 -range : set the output calmin/max to full data range

Matrix operations:
 -tensor_decomp : convert a 4D (6-timepoint )tensor image into L1,2,3,FA,MD,MO,V1,2,3 (remaining image in pipeline is FA)

Kernel operations (set BEFORE filtering operation if desired):
 -kernel 3D : 3x3x3 box centered on target voxel (set as default kernel)
 -kernel 2D : 3x3x1 box centered on target voxel
 -kernel box         : all voxels in a cube of width  mm centered on target voxel
 -kernel boxv        : all voxels in a cube of width  voxels centered on target voxel, CAUTION: size should be an odd number
 -kernel boxv3    : all voxels in a cuboid of dimensions X x Y x Z centered on target voxel, CAUTION: size should be an odd number
 -kernel gauss      : gaussian kernel (sigma in mm, not voxels)
 -kernel sphere      : all voxels in a sphere of radius  mm centered on target voxel
 -kernel file    : use external file as kernel

Spatial Filtering operations: N.B. all options apart from -s use the default kernel or that _previously_ specified by -kernel
 -dilM    : Mean Dilation of non-zero voxels
 -dilD    : Modal Dilation of non-zero voxels
 -dilF    : Maximum filtering of all voxels
 -dilall  : Apply -dilM repeatedly until the entire FOV is covered
 -ero     : Erode by zeroing non-zero voxels when zero voxels found in kernel
 -eroF    : Minimum filtering of all voxels
 -fmedian : Median Filtering 
 -fmean   : Mean filtering, kernel weighted (conventionally used with gauss kernel)
 -fmeanu  : Mean filtering, kernel weighted, un-normalised (gives edge effects)
 -s  : create a gauss kernel of sigma mm and perform mean filtering
 -subsamp2  : downsamples image by a factor of 2 (keeping new voxels centred on old)
 -subsamp2offc  : downsamples image by a factor of 2 (non-centred)

Dimensionality reduction operations:
  (the "T" can be replaced by X, Y or Z to collapse across a different dimension)
 -Tmean   : mean across time
 -Tstd    : standard deviation across time
 -Tmax    : max across time
 -Tmaxn   : time index of max across time
 -Tmin    : min across time
 -Tmedian : median across time
 -Tperc  : nth percentile (0-100) of FULL RANGE across time
 -Tar1    : temporal AR(1) coefficient (use -odt float and probably demean first)

Basic statistical operations:
 -pval    : Nonparametric uncorrected P-value, assuming timepoints are the permutations; first timepoint is actual (unpermuted) stats image
 -pval0   : Same as -pval, but treat zeros as missing data
 -cpval   : Same as -pval, but gives FWE corrected P-values
 -ztop    : Convert Z-stat to (uncorrected) P
 -ptoz    : Convert (uncorrected) P to Z
 -rank    : Convert data to ranks (over T dim)
 -ranknorm: Transform to Normal dist via ranks

Multi-argument operations:
 -roi         : zero outside roi (using voxel coordinates). Inputting -1 for a size will set it to the full image extent for that dimension.
 -bptf    : (-t in ip.c) Bandpass temporal filtering; nonlinear highpass and Gaussian linear lowpass (with sigmas in volumes, not seconds); set either sigma<0 to skip that filter
 -roc   [4Dnoiseonly]  : take (normally binary) truth and test current image in ROC analysis against truth.  is usually 0.05 and is limit of Area-under-ROC measure FP axis.  is a text file of the ROC curve (triplets of values: FP TP threshold). If the truth image contains negative voxels these get excluded from all calculations. If  is positive then the [4Dnoiseonly] option needs to be set, and the FP rate is determined from this noise-only data, and is set to be the fraction of timepoints where any FP (anywhere) is seen, as found in the noise-only 4d-dataset. This is then controlling the FWE rate. If  is negative the FP rate is calculated from the zero-value parts of the  image, this time averaging voxelwise FP rate over all timepoints. In both cases the TP rate is the average fraction of truth=positive voxels correctly found.

Combining 4D and 3D images:
 If you apply a Binary operation (one that takes the current image and a new image together), when one is 3D and the other is 4D,
 the 3D image is cloned temporally to match the temporal dimensions of the 4D image.

e.g. fslmaths inputVolume -add inputVolume2 output_volume
     fslmaths inputVolume -add 2.5 output_volume
     fslmaths inputVolume -add 2.5 -mul inputVolume2 output_volume

     fslmaths 4D_inputVolume -Tmean -mul -1 -add 4D_inputVolume demeaned_4D_inputVolume

2) randomise 非参数统计命令

randomise的详细使用帮助: https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/Randomise

randomise



Part of FSL (ID: 5.0.10)
randomise v2.9

Usage: 
randomise -i  -o 
 -d  -t  [options]

Compulsory arguments (You MUST set one or more of):
    -i   4D input image
    -o    output file-rootname

Optional arguments (You may optionally specify one or more of):
    -D      demean data temporally before model fitting ( demean model as well if required )
    -1      perform 1-sample group-mean test instead of generic permutation test
    -m    mask image
    -d  design matrix file
    -t  t contrasts file
    -f  f contrasts file
    -e  exchangeability block labels file
    --effective_design     alternative design for determining valid permutations
    -q      print out how many unique permutations would be generated and exit
    -Q      print out information required for parallel mode and exit
    -n  number of permutations (default 5000, set to 0 for exhaustive)
    -x      output voxelwise corrected p-value images
    --fonly     calculate f-statistics only
    -T      carry out Threshold-Free Cluster Enhancement
    --T2        carry out Threshold-Free Cluster Enhancement with 2D optimisation (e.g. for TBSS data); H=2, E=1, C=26
    -c  carry out cluster-based thresholding
    -C  carry out cluster-mass-based thresholding
    -F  carry out f cluster thresholding
    -S  carry out f cluster-mass thresholding
    -v     use variance smoothing for t-stats (std is in mm)
    -h,--help   display this message
    --quiet     switch off diagnostic messages
    --twopass   carry out cluster normalisation thresholding
    -R      output raw ( unpermuted ) statistic images
    --uncorrp   output uncorrected p-value images
    -P      output permutation vector text file
    -N      output null distribution text files
    --norcmask  don't remove constant voxels from mask
    --seed    specific integer seed for random number generator
    --tfce_H <H>    TFCE height parameter (default=2)
    --tfce_D <H>    TFCE delta parameter overide
    --tfce_E <E>    TFCE extent parameter (default=0.5)
    --tfce_C <C>    TFCE connectivity (6 or 26; default=6)
    --vxl       list of numbers indicating voxelwise EVs position in the design matrix (list order corresponds to files in vxf option). caution BETA option.
    --vxf       list of 4D images containing voxelwise EVs (list order corresponds to numbers in vxl option). caution BETA option.
    --permuteBlocks permute exchangeability blocks. Caution BETA option
    --glm_output    output glm information for t-statistics ( unpermuted case only )
    --film  output stats to simulate the output of film