Skip to content

Commit 1981549

Browse files
bclenetbclenet
committed
Preprocessing inspired from Chen et al
1 parent 4caaf30 commit 1981549

File tree

1 file changed

+152
-9
lines changed

1 file changed

+152
-9
lines changed

‎narps_open/pipelines/team_O21U.py‎

Lines changed: 152 additions & 9 deletions
Original file line numberDiff line numberDiff line change
@@ -7,15 +7,16 @@
77
from itertools import product
88

99
from nipype import Workflow, Node, MapNode
10-
from nipype.interfaces.utility import IdentityInterface, Function, Split
10+
from nipype.algorithms.modelgen import SpecifyModel
11+
from nipype.interfaces.utility import IdentityInterface, Function, Split, Merge
1112
from nipype.interfaces.io import SelectFiles, DataSink
12-
from nipype.interfaces.fsl import (
13-
IsotropicSmooth, Level1Design, FEATModel,
14-
L2Model, Merge, FLAMEO, FILMGLS, MultipleRegressDesign,
13+
from nipype.interfaces.fsl.maths import ImageMaths, ImageStats, MultiImageMaths
14+
from nipype.interfaces.fsl.preprocess import SUSAN
15+
from nipype.interfaces.fsl.model import (
16+
Level1Design, FEATModel, L2Model, FLAMEO, FILMGLS, MultipleRegressDesign,
1517
FSLCommand, Cluster
1618
)
17-
from nipype.algorithms.modelgen import SpecifyModel
18-
from nipype.interfaces.fsl.maths import MultiImageMaths
19+
from nipype.interfaces.fsl.utils import Merge as FSLMerge
1920

2021
from narps_open.utils.configuration import Configuration
2122
from narps_open.pipelines import Pipeline
@@ -41,8 +42,150 @@ def __init__(self):
4142
]
4243

4344
def get_preprocessing(self):
44-
""" No preprocessing has been done by team O21U """
45-
return None
45+
""" Create the preprocessing workflow """
46+
47+
# Initiate the workflow
48+
preprocessing = Workflow(
49+
base_dir = self.directories.working_dir,
50+
name = 'preprocessing'
51+
)
52+
53+
# IdentityInterface Node - Iterate on subject and runs
54+
information_source = Node(IdentityInterface(
55+
fields = ['subject_id', 'run_id']),
56+
name = 'information_source')
57+
information_source.iterables = [
58+
('subject_id', self.subject_list),
59+
('run_id', self.run_list)
60+
]
61+
62+
# SelectFiles - Get necessary files
63+
templates = {
64+
'func' : join('derivatives', 'fmriprep', 'sub-{subject_id}', 'func',
65+
'sub-{subject_id}_task-MGT_run-{run_id}_bold_space-MNI152NLin2009cAsym_preproc.nii.gz'),
66+
'mask' : join('derivatives', 'fmriprep', 'sub-{subject_id}', 'func',
67+
'sub-{subject_id}_task-MGT_run-{run_id}_bold_space-MNI152NLin2009cAsym_brainmask.nii.gz')
68+
}
69+
select_files = Node(SelectFiles(templates), name = 'select_files')
70+
select_files.inputs.base_directory = self.directories.dataset_dir
71+
preprocessing.connect(information_source, 'subject_id', select_files, 'subject_id')
72+
preprocessing.connect(information_source, 'run_id', select_files, 'run_id')
73+
74+
# ImageMaths - Convert func to float representation
75+
func_to_float = Node(ImageMaths(), name = 'func_to_float')
76+
func_to_float.inputs.out_data_type = 'float'
77+
func_to_float.inputs.op_string = ''
78+
func_to_float.inputs.suffix = '_dtype'
79+
preprocessing.connect(select_files, 'func', func_to_float, 'in_file')
80+
81+
# ImageMaths - Mask the functional image
82+
mask_func = Node(ImageMaths(), name = 'mask_func')
83+
mask_func.inputs.suffix = '_thresh'
84+
mask_func.inputs.op_string = '-mas'
85+
preprocessing.connect(func_to_float, 'out_file', mask_func, 'in_file')
86+
preprocessing.connect(select_files, 'mask', mask_func, 'in_file2')
87+
88+
# ImageMaths - Compute the mean image of each time point
89+
mean_func = Node(ImageMaths(), name = 'mean_func')
90+
mean_func.inputs.suffix = '_mean'
91+
mean_func.inputs.op_string = '-Tmean'
92+
preprocessing.connect(mask_func, 'out_file', mean_func, 'in_file')
93+
94+
# ImageStats - Compute the median value of each time point
95+
# (only because it's needed by SUSAN)
96+
median_value = Node(ImageStats(), name = 'median_value')
97+
median_value.op_string='-k %s -p 50'
98+
preprocessing.connect(select_files, 'func', median_value, 'in_file')
99+
preprocessing.connect(select_files, 'mask', median_value, 'mask_file')
100+
101+
# Merge - Merge the median values with the mean functional images into a coupled list
102+
merge_median = Node(Merge(2, axis='hstack'), name = 'merge_median')
103+
preprocessing.connect(mean_func, 'out_file', merge_median, 'in1')
104+
preprocessing.connect(median_value, 'out_stat', merge_median, 'in2')
105+
106+
# SUSAN - Smoothing funk
107+
smooth_func = Node(SUSAN(), name = 'smooth_func')
108+
smooth_func.inputs.fwhm = 4.9996179300001655 # see Chen et. al. 2022
109+
110+
# Define a function to get the brightness threshold for SUSAN
111+
get_brightness_threshold = lambda median : 0.75 * median
112+
113+
# Define a function to get the usans for SUSAN
114+
get_usans = lambda value : [tuple([val[0], 0.75 * val[1]])]
115+
116+
preprocessing.connect(mask_func, 'out_file', smooth_func, 'in_file')
117+
preprocessing.connect(
118+
median_value, ('out_stat', get_brightness_threshold),
119+
smooth_func, 'brightness_threshold')
120+
preprocessing.connect(merge_median, ('out', getusans), smooth_func, 'usans')
121+
122+
# TODO : Mask the smoothed data ?
123+
"""
124+
maskfunc3 = pe.MapNode(
125+
interface=fsl.ImageMaths(op_string='-mas'),
126+
name='maskfunc3',
127+
iterfield=['in_file','in_file2'])
128+
wf.connect(smooth, 'smoothed_file', maskfunc3, 'in_file')
129+
wf.connect(dilatemask, 'out_file', maskfunc3, 'in_file2')
130+
"""
131+
132+
# Define a function to get the scaling factor for intensity normalization
133+
get_intensity_normalization_scale = lambda median : '-mul %.10f' % (10000. / median)
134+
135+
# ImageMaths - Scale each time point so that its median value is 10000
136+
normalize_intensity = Node(ImageMaths(), name = 'normalize_intensity')
137+
normalize_intensity.inputs.suffix = '_intnorm'
138+
preprocessing.connect(smooth_func, 'out_file', normalize_intensity, 'in_file')
139+
preprocessing.connect(
140+
median_value, ('out_stat', get_intensity_normalization_scale),
141+
normalize_intensity, 'op_string')
142+
143+
# TODO : temporal filtering ???
144+
145+
# DataSink Node - store the wanted results in the wanted repository
146+
data_sink = Node(DataSink(), name = 'data_sink')
147+
data_sink.inputs.base_directory = self.directories.output_dir
148+
preprocessing.connect(
149+
normalize_intensity, 'out_file', data_sink, 'preprocessing.@normalized_file')
150+
151+
# Remove large files, if requested
152+
if Configuration()['pipelines']['remove_unused_data']:
153+
154+
# Merge Node - Merge func file names to be removed after datasink node is performed
155+
merge_removable_files = Node(Merge(4), name = 'merge_removable_files')
156+
merge_removable_files.inputs.ravel_inputs = True
157+
preprocessing.connect(func_to_float, 'out_file', merge_removable_files, 'in1')
158+
preprocessing.connect(mask_func, 'out_file', merge_removable_files, 'in2')
159+
preprocessing.connect(mean_func, 'out_file', merge_removable_files, 'in3')
160+
preprocessing.connect(smooth_func, 'out_file', merge_removable_files, 'in4')
161+
162+
# Function Nodes remove_files - Remove sizeable func files once they aren't needed
163+
remove_dirs = MapNode(Function(
164+
function = remove_parent_directory,
165+
input_names = ['_', 'file_name'],
166+
output_names = []
167+
), name = 'remove_dirs', iterfield = 'file_name')
168+
preprocessing.connect(merge_removable_files, 'out', remove_dirs, 'file_name')
169+
preprocessing.connect(data_sink, 'out_file', remove_dirs, '_')
170+
171+
return preprocessing
172+
173+
def get_preprocessing_outputs(self):
174+
""" Return the names of the files the preprocessing is supposed to generate. """
175+
176+
# Outputs from preprocessing (intensity normalized files)
177+
parameters = {
178+
'subject_id': self.subject_list,
179+
'run_id': self.run_list,
180+
}
181+
parameter_sets = product(*parameters.values())
182+
template = join(
183+
self.directories.output_dir, 'preprocessing',
184+
'_subject_id_{subject_id}', '_run_id_{run_id}',
185+
'wc2sub-{subject_id}_T1w.nii')
186+
187+
return [template.format(**dict(zip(parameters.keys(), parameter_values)))\
188+
for parameter_values in parameter_sets]
46189

47190
def get_subject_information(event_file):
48191
"""
@@ -121,7 +264,7 @@ def get_run_level_analysis(self):
121264
Returns:
122265
- run_level : nipype.WorkFlow
123266
"""
124-
# Create run level analysis workflow and connect its nodes
267+
# Create run level analysis workflow
125268
run_level = Workflow(
126269
base_dir = self.directories.working_dir,
127270
name = 'run_level_analysis'

0 commit comments

Comments
 (0)