[CCPPETMR/SIRF] examples/Python/MR/Gadgetron/fully_sampled_recon_single_chain.py (#167)

[David Atkinson]

1. Incorrectly displays modulus data when labelled as imaginary. Problem is with 'abs' in:
   show_3D_array(abs(image_array[1::2,:,:])

2. Appears to reconstruct 3D data, but fails in show_3D_array:
   sirfuser@ubuntu-xenial:~/CCP-Elba$ python /home/sirfuser/devel/buildVM/sources/SIRF/examples/Python/MR/Gadgetron/fully_sampled_recon_single_chain.py -f m3d.h5 -o m3d_out.h5 -p .
   Input stream has terminated
   Traceback (most recent call last):
   File "/home/sirfuser/devel/buildVM/sources/SIRF/examples/Python/MR/Gadgetron/fully_sampled_recon_single_chain.py", line 118, in
   main()
   File "/home/sirfuser/devel/buildVM/sources/SIRF/examples/Python/MR/Gadgetron/fully_sampled_recon_single_chain.py", line 99, in main
   show = False)
   File "/home/sirfuser/devel/install/python/pUtilities.py", line 138, in show_3D_array
   rows = int(round(math.sqrt(n*nx/ny)))
   ZeroDivisionError: integer division or modulo by zero

3. When applied to our dataset 'simulated_MR_2D_cartesian.h5', it labels output images as two slices, but I believe they are 2 repetitions of one slice.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub, or mute the thread.

[evgueni-ovtchinnikov]

1. corrected
2. need your data to debug
3. slice number means its number in the whole set of slices wherever they come from (different z, different repetition etc.) - do not see how to label slices otherwise, too many dimensions:(

[David Atkinson]

1. corrected - Thanks
2. need your data to debug - run siemens_to_ismrmrd on this 3D dataset to generate an input h5 file: https://www.dropbox.com/s/tz7q02fziskq9u7/meas_MID00103_FID57244_test.dat?dl=0
3. slice number means its number in the whole set of slices wherever they come from (different z, different repetition etc.) - do not see how to label slices otherwise, too many dimensions:( . ‘slice’ implies spatial position, suggest subplot title is something like ‘3rd dimension index’ or ‘ind3’.

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub<https://github.com/CCPPETMR/SIRF/issues/167#issuecomment-388024019>, or mute the thread<https://github.com/notifications/unsubscribe-auth/AF2bn4rFndNK9S-OuYzyOBuXAltJSQQCks5txB-3gaJpZM4T0BHK>.

[evgueni-ovtchinnikov]

2. I do not have siemens_to_ismrmrd - can you not just let me have m3d.h5?
   By the way, what makes you think the reconstruction succeeded? Can you print image_data.number() and image_data.norm()? image_array.shape is clearly wrong as show_3D_array found that ny = image_array.shape[1] = 0.

3. So you suggest to have 'slice with 3rd dimension index 1' or 'slice with ind3 = 1' instead of 'slice 1'? The former is too long and the latter rather cryptic, are they not?
   @ckolbPTB: any ideas?
   @KrisThielemans: I suggest we have a discussion on this on a Friday tcon, and for the time being shift it to milestone 1.2.

[Kris Thielemans]

2. for the data, you have siemens_to_ismrmrd on the VM. would that be ok?

3. let's definitely discuss, but I suppose that the data issue needs to be fixed before the course. can just as well adjust the plot then?

[evgueni-ovtchinnikov]

2. the error is due to failure to read the saved image - have no idea why. for now commented out the reading.
3. the point is - how to adjust. i considered David's suggestions and pointed out at their limitations. it will not take more than 5 min to implement any of the two.

[evgueni-ovtchinnikov]

found a bug in image read from file, now reads ok

however, gadgetrton sometimes crashes with the message:

terminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injectorGadgetron::bad_alloc >'
what(): x

in fact, it was this crash that caused zero ny and hence division by zero in show_3D_array

crash might be due to insufficient memory on VM - the raw data size is 1.3G

an annoying thing is that the crash happens on the server and the client does not know about it

[Kris Thielemans]

how much memory does your VM have? (check in Machine Settings in the VB interface). If you have spare, try to increase it (shutdown VM, change memory, start it).

[evgueni-ovtchinnikov]

on laptop i use VM with 4096MB, which leaves me with about 3GB free memory

tried 8192MB VM on desktop, ran 6 times - no crashes, so must be the VM memory issue

btw, implemented empty-chain connection check, now have correct error traceback (to Reconstructor.process) and more useful message, @DANAJK please try again on m3d.h5

[evgueni-ovtchinnikov]

Closed #167.