Open mszinte opened 3 years ago
Hey Martin, I haven't implemented the Cifti stuff. Perhaps @ecasimiro can take a look at this?
Hey Martin, could you try adding something to the source code for me? Starting on line 272 of utils.py (below if cfg['cifti'] ...:)
runs = len(ddict['funcs'])
data_splitted = [data[:,x:x + runs] for x in range(0, data.shape[1], runs)]
full_data = dict()
for idx, run_data in enumerate(data_splitted):
subc_data = cfg['subc_original']
pos = cfg['pos']
subc_data[pos] = run_data.T
full_data[idx] = subc_data
data_to_save = np.concatenate([v for k, v in full_data.items()], 3)
np.save(f_out + '_subc.npy', data_to_save)
Ok it was still incorrect, i'm working on a solution now, i will share it later... it might take some time as after fixing this error i received some other ones. Thanks.
I was expecting more errors to pop up. If you need any help let me know, but then I'll need to look for a similar cifti file with multiple runs to test.
Hi @ecasimiro you have written:
but after loading, 'tr' is in seconds, so it should be *=1000, isn't it ? Althought after doing that there is a warning showing up :
/home/mszinte/anaconda3/envs/idp/lib/python3.6/site-packages/nilearn/glm/first_level/design_matrix.py:108: UserWarning: High-pass filter will span all accessible frequencies and saturate the design matrix. You may want to reduce the high_pass value.The provided value is 0.01 Hz
'The provided value is {0} Hz'.format(high_pass))
What do you think shoudl be the correct value ?
Hey @mszinte, I think you're right that the cifti one does not need the /= 1000 and should just be the TR. I added that line because I was following the load_gifti as an example (that part was already there) to set up my code. The load_gifti one I think returns TRs as msec, so does need it.
Hi,
I have an error when running latest pybest with cifti files. Here is the command I use :
the code goes through the first steps but next reach a broadcasting error :
indeed the TR indice is incorrect as it is multiplied by the number of run (208 x 2 = 416), any thought how to fix that ?