View_img slide order / masking is weird

Hello again.

When I run this (where D is a freshly-loaded EPI.nii, and array_data is a numpy array computed on D and its friends):

array_img = nl.image.new_img_like(D, array_data)

resamp_ai = nl.image.resample_to_img(array_img,anatpath+anatname,interpolation='linear')

fig = plt.figure(figsize=(20,5.5))
cc = (2,-53,10)

I get a nice image (cannot upload because i am new, see reply below) where everything is there where it should be (seems to me) - i.e. the stat map is drawn onto the T1, and everything looks right.

But then when I try:

cc = (2,-53,10)
view = plotting.view_img(resamp_ai,anatpath+anatname,threshold=0.25)


I get:

At first glance seems ok, but in fact the volumes are all masked incorrectly (you can see black ventricle holes where there shouldn’t be, round clipping around the edges), and the coordinates are not the same.

So, view_img thinks the images are oriented differently and in a different position than they actually are.

I suspect there is something wrong in one of the nifti objects and only view_img is picking it up, but… what could it be?

thanks again

here’s the reasonable-looking result from plot_stat_map

Is there any possibility to share your image, so that we can reproduce the behavior ?
Also it may be an avatar of `view_img` always masks `bg_img` with MNI mask · Issue #3120 · nilearn/nilearn · GitHub ?
Bertrand Thirion

1 Like

It does sound similar to that issue… I’m starting over, re-preprocessing everything anew, we’ll see if it gets better.