Efficient ROI analysis

[tashrifbillah]

Consider a problem where TBSS has been run without ROI analysis. Now the user wants to know tract statistics over a random ROI defining image e.g. ENIGMA, FMRIB, JHU label map.

In the current design, we require the user to provide a space-defining image of the skeleton. The template used in the above TBSS is registered to the space of the skeleton. Finally, antsApplyTransforms performs two transforms to warp the FAs from:

subject-->template-->labelmap sapce

Since antsApplyTransforms accepts multiple -t flag, interpolation is done only once which is a good thing. Regardless, instead of doing the two transformations, a more efficient approach can be to warp the label map to template space. Thus, we get away with only one warping of all FAs:

subject-->template

And perform just one additional warping of the label map.

[tashrifbillah]

A similar quality analysis is done at https://github.com/pnlbwh/pnlNipype/blob/master/scripts/DWIqc/dwi_quality.py