Merge branch 'main' into config

Author: bclenet

.github/workflows/code_quality.yml

@@ -32,7 +32,7 @@ jobs:
       with:
         python-version: 3.9
 
-    - uses: actions/cache@v3
+    - uses: actions/cache@v4
       with:
         path: ~/.cache/pip
         key: ${{ runner.os }}-pip-${{ hashFiles('setup.py') }}

.github/workflows/pipeline_status.yml

@@ -37,7 +37,7 @@ jobs:
       with:
         python-version: 3.9
 
-    - uses: actions/cache@v3
+    - uses: actions/cache@v4
       with:
         path: ~/.cache/pip
         key: ${{ runner.os }}-pip-${{ hashFiles('setup.py') }}

docs/core.md

@@ -124,6 +124,17 @@ This module contains a set of functions dedicated to computations on images.
 # Get dimensions of voxels along x, y, and z in mm (returns e.g.: [1.0, 1.0, 1.0]).
 get_voxel_dimensions('/path/to/the/image.nii.gz')
 ```
+
+ * `get_image_timepoint` : extracts the 3D volume in a given time point range from a 4D Nifti image
+
+```python
+# Create a nifti 3D file containing time point 5 of the input image (returns a path to the generated 3D image)
+get_image_timepoints('/path/to/the/image.nii.gz', 5, 5):
+
+# Create a nifti 3D file containing time points 5 to 10 of the input image (returns a path to the generated 3D image)
+get_image_timepoints('/path/to/the/image.nii.gz', 5, 10):
+```
+
 ## narps_open.core.interfaces
 
 This module contains a set of interface creators inheriting form the `narps_open.core.interfaces.InterfaceCreator` abstract class.

narps_open/core/image.py

@@ -23,3 +23,32 @@ def get_voxel_dimensions(image: str) -> list:
         float(voxel_dimensions[1]),
         float(voxel_dimensions[2])
         ]
+
+def get_image_timepoints(in_file: str, start_time_point: int, end_time_point: int) -> str:
+    """
+    Extract the 3D volume at a given time point from a 4D Nifti image.
+    Create a Nifti file containing the 3D volume.
+    Return the filename of the created 3D output file.
+
+    Arguments:
+        in_file: str, string that represent an absolute path to a Nifti 4D image.
+        start_time_point: int, zero-based index of the first volume to extract from the 4D image.
+        end_time_point: int, zero-based index of the last volume to extract from the 4D image.
+
+    Returns:
+        str, path to the created file
+    """
+    # These imports must stay inside the function, as required by Nipype
+    from os import extsep
+    from os.path import abspath, basename
+    import nibabel as nib
+
+    # The output filename is based on the base filename of the input file.
+    # Note that we use abspath to write the file in the base_directory of a nipype Node.
+    out_file = abspath(
+        basename(in_file).split(extsep)[0]+f'_timepoint-{start_time_point}-{end_time_point}.nii')
+
+    # Perform timepoints extraction
+    nib.save(nib.load(in_file).slicer[..., start_time_point:end_time_point+1], out_file)
+
+    return out_file

narps_open/data/results/team_2T6S.py

@@ -7,8 +7,7 @@
 
 from shutil import copyfile
 
-from nibabel import load
-from nilearn.image import math_img
+from nibabel import load, save, Nifti1Image
 
 from narps_open.data.results import ResultsCollection
 
@@ -44,5 +43,4 @@ def rectify(self):
 
             # Change the signs of values inside the input image
             input_image = load(input_file)
-            output_image = math_img('img*-1', img = input_image)
-            output_image.to_filename(rectified_file)
+            save(Nifti1Image(input_image.get_fdata()*-1, input_image.affine), rectified_file)

narps_open/pipelines/team_0H5E.py

@@ -17,16 +17,14 @@
    EstimateModel, EstimateContrast, Threshold
    )
 from nipype.interfaces.spm.base import Info as SPMInfo
-from nipype.interfaces.fsl import (
-    ExtractROI
-    )
 
 from narps_open.pipelines import Pipeline
 from narps_open.data.task import TaskInformation
 from narps_open.data.participants import get_group
 from narps_open.core.common import (
     remove_parent_directory, list_intersection, elements_in_string, clean_list
     )
+from narps_open.core.image import get_image_timepoints
 from narps_open.utils.configuration import Configuration
 
 class PipelineTeam0H5E(Pipeline):
@@ -93,10 +91,13 @@ def get_preprocessing(self):
 
         # EXTRACTROI - remove first image of func
         # > Removal of "dummy" scans (deleting the first four volumes from each run)
-        remove_first_image = Node(ExtractROI(), name = 'remove_first_image')
-        remove_first_image.inputs.t_min = 4
-        remove_first_image.inputs.t_size = 449 # number of time points - 4
-        remove_first_image.inputs.output_type='NIFTI'
+        remove_first_image = Node(Function(
+            function = get_image_timepoints,
+            input_names = ['in_file', 'start_time_point', 'end_time_point'],
+            output_names = ['roi_file']
+            ), name = 'remove_first_image')
+        remove_first_image.inputs.start_time_point = 4
+        remove_first_image.inputs.end_time_point = 453 # last image
         preprocessing.connect(gunzip_func, 'out_file', remove_first_image, 'in_file')
 
         # REALIGN - motion correction

narps_open/pipelines/team_98BT.py

@@ -16,7 +16,6 @@
     DARTELNorm2MNI, FieldMap, Threshold
     )
 from nipype.interfaces.spm.base import Info as SPMInfo
-from nipype.interfaces.fsl import ExtractROI
 from nipype.algorithms.modelgen import SpecifySPMModel
 from niflow.nipype1.workflows.fmri.spm import create_DARTEL_template
 
@@ -26,6 +25,7 @@
 from narps_open.core.common import (
     remove_parent_directory, list_intersection, elements_in_string, clean_list
     )
+from narps_open.core.image import get_image_timepoints
 from narps_open.utils.configuration import Configuration
 
 class PipelineTeam98BT(Pipeline):
@@ -264,10 +264,13 @@ def get_preprocessing_sub_workflow(self):
         preprocessing.connect(slice_timing, 'timecorrected_files', motion_correction, 'in_files')
 
         # Intrasubject coregistration
-        extract_first = Node(ExtractROI(), name = 'extract_first')
-        extract_first.inputs.t_min = 1
-        extract_first.inputs.t_size = 1
-        extract_first.inputs.output_type = 'NIFTI'
+        extract_first = Node(Function(
+            function = get_image_timepoints,
+            input_names = ['in_file', 'start_time_point', 'end_time_point'],
+            output_names = ['roi_file']
+            ), name = 'extract_first')
+        extract_first.inputs.start_time_point = 1
+        extract_first.inputs.end_time_point = 1
         preprocessing.connect(
             motion_correction, 'realigned_unwarped_files', extract_first, 'in_file')
 
@@ -421,9 +424,9 @@ def get_parameters_file(
         from os import makedirs
         from os.path import join
         from nibabel import load
+        from nibabel.processing import resample_from_to
         from numpy import mean
-        from pandas import read_table
-        from nilearn.image import iter_img, resample_to_img
+        from pandas import read_csv
 
         # Ignore all future warnings
         from warnings import simplefilter
@@ -432,22 +435,23 @@ def get_parameters_file(
         simplefilter(action = 'ignore', category = RuntimeWarning)
 
         # Load wc2 file and create a mask out of it
-        wc2 = load(wc2_file)
-        wc2_mask = wc2.get_fdata() > 0.6
-        wc2_mask = wc2_mask.astype(int)
+        wm_class_image = load(wc2_file)
+        wm_mask_data = wm_class_image.get_fdata() > 0.6
+        wm_mask_data = wm_mask_data.astype(int)
 
         # Compute the mean signal in white matter, for each slice of the functional data
         mean_wm = []
-        for current_slice in iter_img(load(func_file)):
-            slice_data = resample_to_img(
-                current_slice, wc2, interpolation = 'nearest', clip = True).get_fdata()
+        func_image = load(func_file)
+        for index in range(func_image.header.get_data_shape()[3]):
+            resampled_func = resample_from_to(
+                func_image.slicer[..., index], wm_class_image, order = 0, mode = 'nearest')
+
             # Append mean value of masked data
-            mean_wm.append(mean(slice_data * wc2_mask))
+            mean_wm.append(mean(resampled_func.get_fdata() * wm_mask_data))
 
         # Create new parameters file
-        data_frame = read_table(parameters_file, sep = '  ', header = None)
+        data_frame = read_csv(parameters_file, sep = '\t', dtype = str)
         data_frame['Mean_WM'] = mean_wm
-
         new_parameters_file = join(working_dir, 'parameters_files',
             f'parameters_file_sub-{subject_id}_run-{run_id}.tsv')
 

narps_open/pipelines/team_UK24.py

@@ -68,23 +68,21 @@ def get_average_values(
         """
         from os import makedirs
         from os.path import join, abspath
-
-        from nilearn.image import load_img, math_img, index_img, get_data
+        from nibabel import load
+        from numpy import mean
 
         # Load input images
-        mask_img = load_img(mask)
-        in_file_img = load_img(in_file)
+        mask_image = load(mask)
+        in_file_image = load(in_file)
+
+        # Create mask
+        mask_data = mask_image.get_fdata() > 0.0
+        mask_data = mask_data.astype(int)
 
         # Loop through time points
         average_values = []
-        for frame_index in range(in_file_img.shape[3]):
-            average_values.append(
-                get_data(math_img(
-                    'np.mean(image * (mask > 0.0))',
-                    image = index_img(in_file_img, frame_index),
-                    mask = mask_img
-                ))
-            )
+        for index in range(in_file_image.shape[3]):
+            average_values.append(mean(in_file_image.slicer[..., index].get_fdata() * mask_data))
 
         # Write confounds to a file
         out_file_name = abspath(f'sub-{subject_id}_run-{run_id}_' + out_file_suffix)

narps_open/pipelines/team_V55J.py

@@ -14,7 +14,6 @@
     Coregister, Smooth, OneSampleTTestDesign, EstimateModel, EstimateContrast,
     Level1Design, TwoSampleTTestDesign, RealignUnwarp,
     Normalize12, NewSegment, FieldMap, Threshold)
-from nipype.interfaces.fsl import ExtractROI
 from nipype.algorithms.modelgen import SpecifySPMModel
 from nipype.interfaces.spm.base import Info as SPMInfo
 
@@ -24,6 +23,7 @@
 from narps_open.core.common import (
     remove_parent_directory, list_intersection, elements_in_string, clean_list
     )
+from narps_open.core.image import get_image_timepoints
 from narps_open.utils.configuration import Configuration
 
 class PipelineTeamV55J(Pipeline):
@@ -100,10 +100,13 @@ def get_preprocessing(self):
 
         # EXTRACTROI - get the image 10 in func file
         # "For each run, we selected image 10 to distortion correct and asked to match the VDM file"
-        extract_tenth_image = Node(ExtractROI(), name = 'extract_tenth_image')
-        extract_tenth_image.inputs.t_min = 10
-        extract_tenth_image.inputs.t_size = 1
-        extract_tenth_image.inputs.output_type='NIFTI'
+        extract_tenth_image = Node(Function(
+            function = get_image_timepoints,
+            input_names = ['in_file', 'start_time_point', 'end_time_point'],
+            output_names = ['roi_file']
+            ), name = 'extract_tenth_image')
+        extract_tenth_image.inputs.start_time_point = 9 # 0-based 10th image
+        extract_tenth_image.inputs.end_time_point = 9
         preprocessing.connect(gunzip_func, 'out_file', extract_tenth_image, 'in_file')
 
         # FIELDMAP - Calculate VDM routine of the FieldMap tool in SPM12

setup.py

@@ -17,7 +17,6 @@
     'importlib_resources>=5.10.2,<5.11',
     'tomli>=2.0.1,<2.1',
     'networkx>=2.0,<3.0', # a workaround to nipype's bug (issue 3530)
-    'nilearn>=0.10.0,<0.11',
     'nipype>=1.8.6,<1.9',
     'pandas>=1.5.2,<1.6',
     'niflow-nipype1-workflows>=0.0.5,<0.1.0'