Closed paskino closed 6 months ago
@paskino im_ms
needs to be NiftiImageData
, so after im_ms = recon.get_output()
add
im_ms = pReg.NiftiImageData(im_ms)
Notes after discussion with @evgueni-ovtchinnikov.
Root-cause is #1044. The input data in the example has 2 repetitions. NiftyResampler::norm
constructs a NiftiImageData
from the MRImageData
. The NiftiImageData
has effectively only 1 repetition, but all the image-data gets copied into it, therefore causing buffer overflow and segfault. More notes added in #1044.
Just what I was afraid of - repetition is not the only extra dimension:
MR data have many different dimensions, obviously 3 spatial ones, but the additional ones (at least defined in ISMRMRD) are:
contrast phase repetition set
and I have just remembered David mentioning 11 dimensions.
Makes the situation still more complicated...
With https://github.com/SyneRBI/SIRF/pull/1186
norm
on theNiftyResampler
was introduced. While running the MCIR reconstruction I call it and python crashes.A minimal example to reproduce it is
I am running this on SIRF from the https://github.com/SyneRBI/SIRF/tree/ignore-acq branch