orig = genfromtxt(orig, delimiter=' ', dtype=None, skip_header=0)
new = 'func_roi_ts.txt'
savetxt(new, orig, delimiter=' ')
new_file = abspath(new)
return (new_file)
converthex = Node(name='converthex',
interface=Function(input_names=['orig'],
output_names=['new_file'],
function=converthex))
# model ROI connectivity
glm = Node(GLM(out_file='betas.nii', out_cope='cope.nii'),
name='glm',
iterfield='design')
# In[4]:
sbc1_workflow = Workflow(name='sbc1_workflow')
sbc1_workflow.connect([
(infosource, selectfiles, [('subject_id', 'subject_id')]),
(selectfiles, ROI_timeseries, [('orig_func', 'in_file')]),
(infosource, ROI_timeseries, [('ROIs', 'mask')]),
(ROI_timeseries, converthex, [('out_file', 'orig')]),
(converthex, glm, [('new_file', 'design')]),
(selectfiles, glm, [('orig_func', 'in_file')]),
(converthex, datasink, [('new_file', 'roi_ts')]),
(glm, datasink, [('out_cope', 'glm_seed_copes')]),
def rs_preprocess(in_file, fwhm, work_dir, output_dir):
from nipype.workflows.fmri.fsl.preprocess import create_susan_smooth
from nipype.interfaces.fsl.model import GLM
from nipype.interfaces import fsl as fsl
# define nodes and workflows
rs_preproc_workflow = pe.Workflow(name="rs_preproc_workflow")
rs_preproc_workflow.base_dir = work_dir
inputnode = pe.Node(interface=util.IdentityInterface(fields=['func', 'fwhm']), name='inputspec')
inputnode.inputs.func = in_file
inputnode.inputs.fwhm = fwhm
#make a brain mask
immask = pe.Node(interface=fsl.ImageMaths(op_string = '-abs -bin -Tmin'), name='immask')
rs_preproc_workflow.connect(inputnode, 'func', immask, 'in_file')
#calculate mean image from smoothed data, for adding back after GSR
immean = pe.Node(interface=fsl.ImageMaths(op_string = '-Tmean'), name='immean')
rs_preproc_workflow.connect(inputnode, 'func', immean, 'in_file')
#get time-series for GSR
meants = pe.Node(interface=fsl.utils.ImageMeants(), name='meants')
rs_preproc_workflow.connect(inputnode, 'func', meants, 'in_file')
rs_preproc_workflow.connect(immask, 'out_file', meants, 'mask')
#removing global signal
glm = pe.Node(interface=GLM(), name='glm')
glm.inputs.out_res_name = op.join(work_dir, 'res4d.nii.gz')
rs_preproc_workflow.connect(inputnode, 'func', glm, 'in_file')
rs_preproc_workflow.connect(immask, 'out_file', glm, 'mask')
rs_preproc_workflow.connect(meants, 'out_file', glm, 'design')
#add mean back to GSR'ed image
maths = pe.Node(interface=fsl.maths.BinaryMaths(operation = 'add'), name='maths')
rs_preproc_workflow.connect(glm, 'out_res', maths, 'in_file')
rs_preproc_workflow.connect(immean, 'out_file', maths, 'operand_file')
#do the smoothing
smooth = create_susan_smooth()
rs_preproc_workflow.connect(maths, 'out_file', smooth, 'inputnode.in_files')
rs_preproc_workflow.connect(inputnode, 'fwhm', smooth, 'inputnode.fwhm')
rs_preproc_workflow.connect(immask, 'out_file', smooth, 'inputnode.mask_file')
datasink = pe.Node(nio.DataSink(), name='sinker')
datasink.inputs.base_directory = work_dir
rs_preproc_workflow.connect(maths, 'out_file', datasink, 'gsr')
rs_preproc_workflow.connect(immask, 'out_file', datasink, 'mask')
rs_preproc_workflow.connect(smooth.get_node('smooth'), ('smoothed_file', pickfirst), datasink, 'gsr_smooth')
rs_preproc_workflow.run()
#copy data to directory
gsr_fn = glob(op.join(work_dir, 'gsr', '*.nii.gz'))[0]
mask_fn = glob(op.join(work_dir, 'mask', '*.nii.gz'))[0]
gsr_smooth_fn = glob(op.join(work_dir, 'gsr_smooth', '*', '*.nii.gz'))[0]
gsr_fn2 = op.join(output_dir, '{0}.nii.gz'.format(op.basename(in_file).split('.')[0]))
mask_fn2 = op.join(output_dir, '{0}_mask.nii.gz'.format(op.basename(in_file).split('.')[0]))
gsr_smooth_fn2 = op.join(output_dir, '{0}_smooth.nii.gz'.format(op.basename(in_file).split('.')[0]))
shutil.copyfile(gsr_fn, gsr_fn2)
shutil.copyfile(mask_fn, mask_fn2)
shutil.copyfile(gsr_smooth_fn, gsr_smooth_fn2)
shutil.rmtree(work_dir)
noise_file = 'noise_matrix.txt'
savetxt(noise_file, noise_matrix, delimiter='\t')
noise_filepath = path.abspath(noise_file)
return (noise_filepath)
noise_mat = Node(name='noise_mat',
interface=Function(input_names=[
'vols_to_censor', 'motion_params', 'comp_noise'
],
output_names=['noise_filepath'],
function=create_noise_matrix))
denoise = Node(GLM(out_res_name='denoised_residuals.nii',
out_data_name='denoised_func.nii'),
name='denoise')
# band pass filtering- all rates are in Hz (1/TR or samples/second)
def bandpass_filter(in_file, lowpass, highpass, TR):
import numpy as np
import nibabel as nb
from os import path
from nipype.interfaces.afni.preprocess import Bandpass
from nipype import config, logging
config.enable_debug_mode()
logging.update_logging(config)
out_file = 'func_filtered.nii'
bp = Bandpass()
from os.path import abspath
orig = genfromtxt(orig, delimiter=' ', dtype=None, skip_header=0)
new = 'func_roi_ts.txt'
savetxt(new, orig, delimiter=' ')
new_file = abspath(new)
return(new_file)
converthex = Node(name='converthex',
interface=Function(input_names=['orig'],
output_names=['new_file'],
function=converthex))
# model ROI connectivity
glm = Node(GLM(out_file='betas.nii',out_cope='cope.nii'), name='glm', iterfield='design')
# In[4]:
sbc1_workflow2 = Workflow(name='sbc1_workflow2')
sbc1_workflow2.connect([(infosource,selectfiles,[('subject_id','subject_id')]),
(selectfiles,ROI_timeseries,[('orig_func','in_file')]),
(infosource,ROI_timeseries,[('ROIs','mask')]),
(ROI_timeseries,converthex,[('out_file','orig')]),
(converthex,glm,[('new_file','design')]),
(selectfiles,glm,[('orig_func','in_file')]),
(converthex, datasink, [('new_file','roi_ts')]),
(glm,datasink,[('out_cope','glm_seed_copes')]),
(glm,datasink,[('out_file','glm_betas')])
