Hi Everyone,
Im getting a simple trait error regarding a .bvec file, as follows :
traits.trait_errors.TraitError: The 'bvecs' trait of a XFibres5InputSpec instance must be a pathlike object or string representing an existing file, but a value of '/path/to/dti_raw_rot.bvecs' <class 'str'> was specified.
The file path itself exists. I am using bedpostx_parallel. How can i provide a path like object to Xfibres instead of a string?
Hi @Ziad_Motagaly and welcome to neurostars!
This error indicates that the program can not find the file at that path. Some more details of your code and your computing environment would help us debug your issue.
Best,
Steven
Hi @Steven ,
Thank you !
The computing environment is a SLURM-based HPC cluster. ( sbatch, srun… ) . I am using Nipype bedpostx_parallel (this is where xfibres is called from)
I am currently running it linearly on CPU (and eventually run it on GPU as well).
#!/bin/env python
import os
import sys
import subprocess
import configparser
from nipype import DataGrabber, DataSink, IdentityInterface, Node, Workflow, MapNode, JoinNode, Merge
from niflow.nipype1.workflows.dmri.fsl.dti import bedpostx_parallel
from nipype.interfaces import utility
import nipype.interfaces.fsl as fsl
import nipype.interfaces.ants as ants
from nipype.interfaces.freesurfer import ReconAll, MRIsConvert, MRIConvert
from nipype.interfaces.utility import Function
def create_diffusion_prep_pipeline(name='dMRI_preprocessing', bet_frac=0.34):
input_subject = Node(
IdentityInterface(
fields=['dwi', 'bval', 'bvec'],
),
name='input_subject'
)
def bvec_flip(bvecs_in, flip):
from os.path import join, basename
from os import getcwd
import numpy as np
print(bvecs_in)
bvecs = np.loadtxt(bvecs_in).T * flip
output_file = str(join(getcwd(), basename(bvecs_in)))
np.savetxt(output_file, bvecs)
return output_file
if __name__ == '__main__':
dmri_preprocess_workflow = create_diffusion_prep_pipeline(
'dmri_preprocess')
config = configparser.ConfigParser()
config.read(sys.argv[1])
PATH='/path/to/my/script'
subject_list = config['DEFAULT']['id_list'].split(" ")
visits = list(config['DEFAULT']['visits'])
subjects_dir = config['DEFAULT']['subjects_dir']
sessions = list(config['DEFAULT']['sessions'])
infosource = Node(IdentityInterface(fields=['subject_id', 'visit','session']),
name='subjects')
infosource.iterables = [('subject_id', subject_list), ('visit', visits), ('session', sessions)]
subject_id_visit = Node(
interface=Function(
input_names=['subject_id', 'visit'], output_names=['composite_id'],
function=lambda subject_id, visit: '{}_{}'.format(subject_id, visit)
),
name='subject_id_visit'
)
data_source = Node(DataGrabber(infields=['subject_id', 'visit','session'],
outfields=['dwi', 'bval', 'bvec', 'T1']),
name='data_grabber')
data_source.inputs.sort_filelist = True
data_source.inputs.base_directory = config['DEFAULT']['base_directory']
data_source.inputs.template = ''
data_source.inputs.field_template = {
'T1': '%s/visit%s/session%s/anat/T1w.nii',
'dwi': '%s/visit%s/session%s/dwi/dwi_raw.nii.gz',
'bval': '%s/visit%s/session%s/dwi/dti_raw.bvals',
'bvec': '%s/visit%s/session%s/dwi/dti_raw.bvecs',
#'bval': '%s/visit%s/session%s/dwi/dwepi.150.bvals',
#'bvec': '%s/visit%s/session%s/dwi/dwepi.150.grads'
}
data_source.inputs.template_args = {
template: [['subject_id', 'visit','session']]
for template in data_source.inputs.field_template.keys()
}
flip_bvectors_node = Node(
interface=Function(
input_names=['bvecs_in', 'flip'], output_names=['bvecs_out'],
function=bvec_flip
),
name='flip_bvecs',
)
flip_bvectors_node.inputs.flip = (-1, 1, 1)
recon_all = Node(interface=ReconAll(), name='recon_all')
recon_all.inputs.directive = 'all'
recon_all.inputs.subjects_dir = subjects_dir
recon_all.inputs.openmp = 20
recon_all.inputs.mprage = True
recon_all.inputs.parallel = True
recon_all.interface.num_threads = 20
recon_all.inputs.flags = "-no-isrunning"
bedpostx_parallel_wf_params = dict(n_fibres=3,
fudge=1,
burn_in=1000,
n_jumps=1250,
sample_every=25)
bedpostx_parallel_wf = bedpostx_parallel(
'nipype_bedpostx_parallel', params=bedpostx_parallel_wf_params)
mri_convert = Node(interface=MRIConvert(), name='mri_convert')
mri_convert.inputs.out_type = 'nii'
mri_convert.inputs.subjects_dir = subjects_dir
workflow = Workflow('diffusion_workflow_new_mgt', base_dir=PATH)
workflow.connect([
(infosource, data_source, [('subject_id', 'subject_id'),
('visit', 'visit'),
('session','session')]),
(infosource, subject_id_visit, [
('subject_id', 'subject_id'),
('visit', 'visit')
]),
(data_source, flip_bvectors_node, [('bvec', 'bvecs_in')]),
(data_source, dmri_preprocess_workflow,
[
('dwi', 'input_subject.dwi'),
('bval', 'input_subject.bval'),
]),
(data_source, recon_all, [('T1', 'T1_files')]),
(subject_id_visit, recon_all, [('composite_id', 'subject_id')]),
(flip_bvectors_node, dmri_preprocess_workflow, [('bvecs_out', 'input_subject.bvec')]),
(recon_all, mri_convert, [('brain', 'in_file')]),
(mri_convert, dmri_preprocess_workflow, [
('out_file', 'input_template.T1'),
('out_file', 'input_template.T2')
]),
(
dmri_preprocess_workflow,
bedpostx_parallel_wf,
[('output.bval', 'inputnode.bvals'),
('output.bvec_rotated', 'inputnode.bvecs'),
('output.dwi_rigid_registered', 'inputnode.dwi'),
('output.mask', 'inputnode.mask')],
),
])
workflow.write_graph(format='pdf', simple_form=False)
if config['DEFAULT'].get('server', '').lower() == 'margaret':
workflow.run(plugin='Linear')
#workflow.run(plugin='MultiProc', plugin_args={'n_procs':220, 'memory_gb': 320})
The main two nodes that are involved seem to be “dmri_preprocess_workflow” and “bedpostx_parallel_wf”
diffusion_preprocessing file is as follows :
#!/bin/env python
import inspect
import os
from nipype import IdentityInterface, Node, Workflow
import nipype.interfaces.fsl as fsl
import nipype.interfaces.ants as ants
from niflow.nipype1.workflows.dmri.fsl.epi import create_eddy_correct_pipeline
from nipype.interfaces import utility
from nipype.interfaces.utility.wrappers import Function
def convert_affine_itk_2_ras(input_affine):
import subprocess
import os, os.path
output_file = os.path.join(
os.getcwd(),
f'{os.path.basename(input_affine)}.ras'
)
subprocess.check_output(
f'c3d_affine_tool '
f'-itk {input_affine} '
f'-o {output_file} -info-full ',
shell=True
).decode('utf8')
return output_file
ConvertAffine2RAS = Function(
input_names=['input_affine'], output_names=['affine_ras'],
function=convert_affine_itk_2_ras
)
def rotate_gradients_(input_affine, gradient_file):
import os
import os.path
import numpy as np
from scipy.linalg import polar
affine = np.loadtxt(input_affine)
u, p = polar(affine[:3, :3], side='right')
gradients = np.loadtxt(gradient_file)
new_gradients = np.linalg.solve(u, gradients.T).T
name, ext = os.path.splitext(os.path.basename(gradient_file))
output_name = os.path.join(
os.getcwd(),
f'{name}_rot{ext}'
)
np.savetxt(output_name, new_gradients)
return output_name
RotateGradientsAffine = Function(
input_names=['input_affine', 'gradient_file'],
output_names=['rotated_gradients'],
function=rotate_gradients_
)
def create_diffusion_prep_pipeline(name='dMRI_preprocessing', bet_frac=0.34):
input_subject = Node(
IdentityInterface(
fields=['dwi', 'bval', 'bvec'],
),
name='input_subject'
)
input_template = Node(
IdentityInterface(
fields=['T1', 'T2'],
),
name='input_template'
)
output = Node(
IdentityInterface(
fields=[
'dwi_rigid_registered', 'bval', 'bvec_rotated', 'mask', 'rigid_dwi_2_template'
]
),
name='output'
)
fslroi = Node(interface=fsl.ExtractROI(), name='fslroi')
fslroi.inputs.t_min = 0
fslroi.inputs.t_size = 1
bet = Node(interface=fsl.BET(), name='bet')
bet.inputs.mask = True
bet.inputs.frac = bet_frac
eddycorrect = create_eddy_correct_pipeline('eddycorrect')
eddycorrect.inputs.inputnode.ref_num = 0
rigid_registration = Node(
interface=ants.RegistrationSynQuick(),
name='affine_reg'
)
rigid_registration.inputs.num_threads = 8
rigid_registration.inputs.transform_type = 'a'
conv_affine = Node(
interface=ConvertAffine2RAS,
name='convert_affine_itk_2_ras'
)
rotate_gradients = Node(
interface=RotateGradientsAffine,
name='rotate_gradients'
)
transforms_to_list = Node(
interface=utility.Merge(1),
name='transforms_to_list'
)
apply_registration = Node(
interface=ants.ApplyTransforms(),
name='apply_registration'
)
apply_registration.inputs.dimension = 3
apply_registration.inputs.input_image_type = 3
apply_registration.inputs.interpolation = 'NearestNeighbor'
apply_registration_mask = Node(
interface=ants.ApplyTransforms(),
name='apply_registration_mask'
)
apply_registration_mask.inputs.dimension = 3
apply_registration_mask.inputs.input_image_type = 3
apply_registration_mask.inputs.interpolation = 'NearestNeighbor'
workflow = Workflow(
name=name,
)
workflow.connect([
(input_subject, fslroi, [('dwi', 'in_file')]),
(fslroi, bet, [('roi_file', 'in_file')]),
(input_subject, eddycorrect, [('dwi', 'inputnode.in_file')]),
(fslroi, rigid_registration, [('roi_file', 'moving_image')]),
(input_template, rigid_registration, [('T2', 'fixed_image')]),
(rigid_registration, transforms_to_list, [('out_matrix', 'in1')]),
(rigid_registration, conv_affine, [('out_matrix', 'input_affine')]),
(input_subject, rotate_gradients, [('bvec', 'gradient_file')]),
(conv_affine, rotate_gradients, [('affine_ras', 'input_affine')]),
(transforms_to_list, apply_registration, [('out', 'transforms')]),
(eddycorrect, apply_registration, [('outputnode.eddy_corrected', 'input_image')]),
(input_template, apply_registration, [('T2', 'reference_image')]),
(transforms_to_list, apply_registration_mask, [('out', 'transforms')]),
(bet, apply_registration_mask, [('mask_file', 'input_image')]),
(input_template, apply_registration_mask, [('T2', 'reference_image')]),
(conv_affine, output, [('affine_ras', 'rigid_dwi_2_template')]),
(apply_registration, output, [('output_image', 'dwi_rigid_registered')]),
(rotate_gradients, output, [('rotated_gradients', 'bvec_rotated')]),
(input_subject, output, [('bval', 'bval')]),
(apply_registration_mask, output, [('output_image', 'mask')]),
])
return workflow
MANAGED BY INCF