lesion_align¶
Overview¶
Script to align a subject structural data with large lesion lesions
to a template and invert the warps to compute the segmentation in the
subject's original, native space.
This program uses basic AFNI commands to compute affine and nonlinear
alignments. The program works by first aligning centers of the subject
to that of the template. Affine and nonlinear alignment follow. The
inverse warp is computed to bring the template and atlas segmentation
into the center-shifted grid. Skullstripping is provided by masking
with the template. Finally, the grids are adjusted back to the
original center. Mirrored brains with "repaired" lesions are also
computed.
Usage Example¶
lesion_align -input Subj2+orig \
-base MNI152_T1_2009c+tlrc \
-atlas MNI_Glasser_HCP_v1.0.nii.gz \
-outdir lesion_align -goodside right
Note only the input dset and template_dset are required. If no
segmentation is given, then only the alignment steps are performed.
Options¶
-input dset :required input dataset to align to template
-goodside left/right/both : specify good side of brain
-base base_dataset :required template. Can be in a standard AFNI
location or fully specified path.
Note, if the template has no skull, then a
masked (skullstripped) version of the input
is produced in the output
-atlas atlas_dataset :atlas can also be in a standard AFNI location
or fully specified
-outdir outputdir :create new directory and do all processing there.
Default is template_align
-template_prefix templatename
:select name for template and segmentation for
output naming. Uses template space of template
if available in template header
-seg_followers segdset1 segdset2 ...
:warp related datasets back to native space
-cost costfunction :cost function for affine transformation.
Default is lpa. Choose nmi, lpa+ZZ, cru for
noisy or difficult datasets. See 3dAllineate
help for more information.
-lesion_mask ldset :provide lesion mask as input dataset.
Used to determine bad and good sides
-center_split :split input dataset on left-right center for affine
alignment keeping either the left or right side for
the computation. Nonlinear alignment uses the full
dataset
-maxlev nn :maximum level for nonlinear warping. Determines
neighborhood size that is searched. See 3dQwarp
help for information on maxlev. Default is 11.
Use smaller values for testing
-no_unifize :turn off unifizing for mirror/heal output
-keep_temp :keep temporary files including awpy directory and
other intermediate datasets
-ok_to_exist :reuse and do not overwrite existing datasets.
This option is used for faster restarts or with
limited alignment options
QC images¶
The following quality control (QC) images are automatically generated
during processing, to help with speedy checking of processing. In
each case, there are three sets of montages (one for sag, cor and axi
views) and a copy of the colorbar used (same prefix as file name,
*.jpg). Additionally, there is also a *.txt file of ranges of values
related to the ulay and olay, which might be useful for QC or
figure-generation.
Inside the output directory is a subdirectory called QC/ that contains
the following semi-cryptically named files:
qc_00_e_temp+wrpd_inp.*
[ulay] edges of the template (in template space)
[olay] warped input dset
qc_01_e_wrpd_temp+orig_inp.*
[ulay] edges of the template (warped to orig space)
[olay] original input dset
qc_02_orig_inp+mask.*
[ulay] original input dset (in orig space)
[olay] estimated mask, showing skullstripping
qc_03_ee_orig_inp+wrpd_atlas.*
[ulay] 'edge enhanced' original input dset (in orig space)
[olay] warped atlas dset
References¶
Please cite:
Maallo, AMS, et al. Large-scale resculpting of cortical circuits in
children after surgical resection. Sci Rep 10, 21589 (2020).
https://doi.org/10.1038/s41598-020-78394-z
For questions about this program, please ask on AFNI message board or
email glend at mail.nih.gov
Comments¶
lesion_align provides multiple outputs to assist in registering your anatomicals and associated MRI data to the template: Subject scans registered to the template + mydset_shft.nii.gz - dataset center aligned to the template center + mydset_shft_at.nii.gz - dataset affine aligned to the template and on the template grid + mydset_nl.nii.gz - dataset nonlinearly warped to the template Registration datasets for alignment to template + mydset_shft_at.aff12.1D - affine transformation to the template + mydset_shft_qw_WARP.nii.gz - warp deformations to the template from nonlinear alignment only Registration datasets for Template alignment to Subject + mydset_shft_inv.aff12.1D - inverse of mydset_shft_at.aff12.1D Atlas aligned to Subject (Optional - only if atlas provided) + seg_in_mydset.nii.gz - atlas segmentation aligned to native scan Template aligned_to Subject + template_in_mydset.nii.gz** - template (e.g. TT_N27,MNI152_2009c) aligned to native scan Mirrored/Healed brains - see descriptions below of output datasets Text description of output + lesion_outs.txt - list of important datasets and transforms in output directory ***-NOTE: lesion_align is part of the AFNI software package *** Here all occurrences of mydset in the output file names would be replaced with the name of your dataset. To help with other kinds of processing, the program also computes "healed" brains in native and subject spaces using these methods. 1. template-filled subject - template values fill in values of subject 2. template-half subject - template values fill in half of the subject 3. mirrored transformed subject - subject is masked, left-right flipped, and transformed to native space. Interpolated in both directions will make this a little blurrier 4. mirrored - filled subject. Blends original subject with the mirrored subject