title          :: Alignment (part 3/4): EPI-anatomical
presentation   :: afni14_alignment.pdf
scripts        :: 
date           :: 2020-03
speaker        :: Taylor
outline        :: 
EPI-anatomical (mainly T1w) alignment, within the same subject
This uses linear affine alignment (same subject, assume not huge distortion)
The main tool in AFNI for this is align_epi_anat.py (wrapper of 3dAllineate)
    + default cost function (lpc, lpc+ZZ) good for dsets with opposite contrasts
    + can combine EPI-anat transform with other ones (deoblique, etc.)
    + slice timing correction can also be concatenated
    + can check left-right consistency between anat and EPI (hard to see flips)
Defaults work well for most standard cases
There are a number of options for other/problem cases, such as:
    + dsets are not "near" each other to start
    + poor tissue contrast or different contrasts (i.e., applied contrast agent)
    + non-uniformity of brightness in dset
    + clinical cases- stroke/lesion/tumors present
    + non-human cases- monkey, rat, etc.
    + non-(f)MRI cases- CT, PET, DTI, etc.
Make sure you visually check your alignment outputs!
    + also, check matching of internal features (ventricles, gyri, etc.)
    + several programs to help with this, generate edge-y overlay images
GUI has several features to look at alignment
    + change olay opacity
    + toggle ulay/olay
    + slider windows between dsets
    + blend dsets
    + toggle ulay with "edge-ified" version of it
    + use "-AddEdge" option in align_epi_anat.py to get edgey-images 
Check left-right consistency between EPI and anatomical dsets
    + use "-check_flip" option in align_epi_anat.py
    + quantitative check based on cost function between original and 'flipped' ver
    + AND visual check for both cases
    + we have found public repos with consistently flipped dsets, unfortunately
      - data analysis is hard and full of details (DICOMs can be tricky...)
    + let the researcher beware- use this simple tool to check your own data!