[MINC-users] minctracc/masking bug?

[Andrew Janke]

Hi Alex,

> - take 100 subjects, already in linear template (stx) space
> - register these with an nlfit*-like process, up to 4mm grid step
> size, to a symmetric template+mask
> - flip the subject image and its mask about x=0, repeat the registration
> - flip the resampled image and the grid from the registration again about x=0
> - calculate the magnitude of the difference between the 'normal' and
> '2xflipped' grid (per subject)
>
> - average the 'normal' and '2xflipped' grid files across subjects
> - calculate the magnitude of the difference between the average
> 'normal' and average '2xflipped' grid files
>
> In the end, what I am left with is the asymmetry in the deformation
> field(s) due solely to the registration process; if minctracc were
> unbiased, these difference images would be 0. This image:
>
> https://dl.dropboxusercontent.com/u/5709165/def_diff_mag.jpeg
>
> shows the average of the non-linearly registered images over these 100
> subjects; and the magnitude of the difference of the average
> deformation estimated 'normally' and '2xflipped'. Note that the in
> these population averages, the maximum deformation magnitude is 2.86;
> in individual subjects, the max magnitude of the deformation
> difference is typically in the 10-20 range. In other words, the
> asymmetry in the estimated deformation field purely caused by the
> methodology, can be as high as 20mm.

Something seems off. The previous tests I ran along this line didn't
find such large differences. What I found for for human sized heads
were typically differences of 0.1mm and less. The differences also
didn't show any similarity to underlying structures like what you
have.

I had a go at finding this analysis just then but drew a blank, what I
do remember is that I (like Claude) wasn't using masks. Is there any
chance than you can put the code up somewhere for what you are doing
in a more distilled test case?

My suggestion would be to keep this very simple, drop the masks and
flip both the source and target image to avoid issues of grid
sampling. minctracc goes through some interesting logic to ensure that
the grid aligns properly with the data.  Is it possible that your twin
grid images (one flipped) aren't aligned which can cause a resampling
error?

One trick I have often employed here is to seed minctracc with an
initial transform which includes an identity grid transform with the
exact sampling I want. There is a perl script here to do this:

    http://packages.bic.mni.mcgill.ca/scripts/gennlxfm


a