notes about neuroimaging using ants

1. piperdream

• series names
  • T1: T1 MPRAGE
  • DTI: DTI.nii.gz
  • RSBOLD: BOLD rsbold.nii.gz
• dicom2series <sorted_MRI_dir> 0 0 <raw_subject_exam_dicom_dir> # 0: don't empty field; 0: don't rename files
  • dicom2series MRI/S1 0 0 raw/S1
  • cd raw; foreach s in *; dicom2series ../MRI/$s 0 0 $s; end
• dicom2nii "none" <subject_list_file> <protocol_list_file> <data_dir> <output_dir> [subdir_name]
  • dicom2nii "none" subject.txt protocol.txt ./MRI ./input MRI

2. NeuroBattery # for multimodalities

diff --git a/scripts/run_test.sh b/scripts/run_test.sh
index 494fae8..60dc08a 100755
--- a/scripts/run_test.sh
+++ b/scripts/run_test.sh
@@ -49,5 +49,5 @@ fi

 ./process_t1.sh
 ./process_modalities.sh
-./warp_labels.h PEDS012 20131101
+./warp_labels.sh PEDS012 20131101

• battery_all.sh <input_subj_dir> <output_subj_dir>
  • set template dir (ANTS_TDIR) in battery\all.sh first
  • install parallel first
  • will scan input_subj_dir for each subject each timepoint, and do
    • T1:
      • antsCorticalThickness.sh
        • KellyKapowski -> _CorticalThickness.nii.gz (DiReCt)
        • To measure mean cortical thickness, measure mean value of the ROI
          • thick.test = as.array(antsImageRead(thick.test, 3))
      • mask.test = as.array(antsImageRead(mask.label.image.file, 3))
      • mask.label = ifelse( mask.test == label.id, 1, 0)
      • mean( thick.test[mask.label>0] ) # this is the mean thickness of the specific label
    • DTI: nii2dt.sh and antsNeuroimagingBattery for DTI
    • warp_label:
      • antsApplyTransforms, applied to an input image, transforms it according to a reference image and a transform (or a set of transforms): Transform atlas label (PTBP_T1_AAL.nii.gz) to subject space (${SUBJECTID}_BrainExtractionMask.nii.gz) using transforms (${SUBJECTID}_TemplateToSubject1GenericAffine.mat + ${SUBJECTID}_TemplateToSubject0Warp.nii.gz) -> ${SUBJECTID}_aal.nii.gz
      • ImageMath, ImageMath 3 ${SUBJECTID}_aal.nii.gz m ${SUBJECTID}_aal.nii.gz ${SUBJECTID}_BrainExtractionMask.nii.gz: applying brain mask

3. LabelGeometryMeasures imageDimension labelImage [intensityImage=none] [csvFile]

• LabelGeometryMeasures 3 PEDS012_20131101_BrainSegmentation.nii.gz

4. JointLabelFusion

• antsJointLabelFusion.sh: used for multi-atlas segmentation with joint label fusion

5. Atlas

• Mindboggle Data
• Mindboggle Label

6. Registration

• antsRegistrationSyN.sh -d 3 -o <ID> -f <fixed_img.nii.gz> -m <moving_img.nii.gz> -t bo -s 26
• antsRegistrationSyN.sh uses CC metric (slower) and antsRegistrationSyNQuick.sh uses MI

7. rsfMRI

• melodic --report -v -i BOLD_rsbold.nii.gz -m BOLD_brainmask.nii.gz
• rsbold = ANTsR::antsBOLDNetworkAnalysis()
  • ANTsR::heatmap(rsbold$mynetwork$corrmat)

8. TBSS

• FSL
  • tbss_1_preproc
    • Erode FA_i slightly; zero the end slices
  • tbss_2_reg
    • Nonlinear registration to a target (template FMRIB58_FA or specific subject)
  • tbss_3_postreg
    • apply transform to target space
  • tbss_4_prestats
    • threshold mean FA skeleton image (0.2)
  • design_ttest2 design 7 11
  • Summary: Do stats in (FA -> template) space
• camino
  • dt.nii.gz: contains the 8 components normally produced by dtfit
• mrtrix 3
• ANTS
  • Local circularity in VBA
    • FA_i -> T1_i normalization
    • The idea is, to maintain independence between the features used in normalization and in hypothesis testing
    • The features that drive the minimization of a similarity metric should be independent of the features that will be used in hypothesis testing.
    • eg. one should avoid combining the SSD and Demons metrics for normalization with Student's t test for assessing image-derived differences.
    • Local circularity will increase Type 1 errors (false positive rate)
  • To prevent Local circularity in VBA
    • Build T1_n template using T1_i, get Transform_i->n
    • Normalize FA_i to T1_i space, and to T1_n template space
    • Do stats at warped FA_i in T1_n template space
    • Summary: Do stats in (FA_i -> T1_i -> T1_n) space
  • mapping:
    • for faImage in faImages*nii.gz do antsApplyTransforms .... -i faImage -o faImageWarped -t faImageWarp.nii.gz -t faImage0GenericAffine.mat -r faTemplate done
  • skel.sh: AverageImages 3 averageFA.nii.gz 0 *fa.nii.gz then call skel.sh on averageFA.nii.gz

ThresholdImage 3 id002_TransformationMNI.nii.gz wm.nii.gz 0.25 Inf SmoothImage 3 wm.nii.gz 1 wms.nii.gz ThresholdImage 3 wms.nii.gz wm.nii.gz 0.25 Inf skel.sh wm.nii.gz wm_skel 1 snap -g wm.nii.gz -s wm_skel_topo_skel.nii.gz + bigLMstats

• ImageMath

  • ImageMath 3 $ref m $anat $mask: ref used for MOCO options: b0, meanDWI, medianDWI

  • ResampleImageBySpacing 3 $ref $ref 2.0 2.0 2.0

  • dt_template.nii.gz: dt in template space

  • ImageMath 3 fa_template.nii.gz TensorFA dt_template.nii.gz

  • ImageMath 3 md_template.nii.gz TensorMeanDiffusion dt_template.nii.gz

  • ImageMath 3 ad_template.nii.gz TensorAxialDiffusion dt_template.nii.gz

  • ImageMath 3 rd_template.nii.gz TensorRadialDiffusion dt_template.nii.gz

  • ImageMath 3 l1_template.nii.gz TensorEigenvalue dt_template.nii.gz 2 = ad_template.nii.gz

  • ImageMath 3 l2_template.nii.gz TensorEigenvalue dt_template.nii.gz 1

  • ImageMath 3 l3_template.nii.gz TensorEigenvalue dt_template.nii.gz 0

  • ImageMath 3 fa_template.nii.gz m fa_template.nii.gz $mask_template

9. Building Template

• antsMultivariateTemplateConstruction.sh to build template image. (about 2+ days for 100+ subjects with 10 cores)
• antsMultivariateTemplateConstruction.sh uses ANTS; antsMultivariateTemplateConstruction2.sh uses antsRegistration
• antsJointLabelFusion.sh to jointly fuse the labels into template space.
  • template image T1_n from T1_i (produced by antsMultivariateTemplateConstruction.sh)
  • labels in each T1_i

10. VBM pipeline

• antsMultivariateTemplateConstruction2.sh: Build study specific template
• antsCorticalThickness.sh: T1/MM normalization
• Use warped images for VBM

11. Jacobian

• the log jacobian transforms jacobian values > 1 to > 0 and values < 1 to < 0
• geodesic shooting (Ashburner & Friston Neuroimage 2011)

looks like you need two images to estimate this ... let me accentuate estimate.
image 1 is the jacobian of the transformation:
CreateJacobianDeterminantImage
​image 2 is the difference between the transformed image and the template
you can compute this (in ANTsR) via:
template = antsImageRead("template.nii.gz")
jimg = createJacobianDeterminantImage( template, fileNameOfWarp )
dimg = template - antsImageRead("warpedImage.nii.gz")
​scalarmom = dimg * jimg
i might have a sign error but this should get you close to what you want.

12. To create a probability mask

• SmoothImage 3 $mask 1.0 $probabilityMask 1