confounds.py

# emacs: -*- mode: python; py-indent-offset: 4; indent-tabs-mode: nil -*-
# vi: set ft=python sts=4 ts=4 sw=4 et:
"""
Calculate BOLD confounds
^^^^^^^^^^^^^^^^^^^^^^^^

.. autofunction:: init_bold_confs_wf
.. autofunction:: init_ica_aroma_wf

"""
from os import getenv

from nipype.algorithms import confounds as nac
from nipype.interfaces import utility as niu, fsl
from nipype.pipeline import engine as pe
from templateflow.api import get as get_template

from ...config import DEFAULT_MEMORY_MIN_GB
from ...interfaces import (
    GatherConfounds, ICAConfounds, FMRISummary, DerivativesDataSink
)


def init_bold_confs_wf(
    mem_gb,
    metadata,
    regressors_all_comps,
    regressors_dvars_th,
    regressors_fd_th,
    freesurfer=False,
    name="bold_confs_wf",
):
    """
    Build a workflow to generate and write out confounding signals.

    This workflow calculates confounds for a BOLD series, and aggregates them
    into a :abbr:`TSV (tab-separated value)` file, for use as nuisance
    regressors in a :abbr:`GLM (general linear model)`.
    The following confounds are calculated, with column headings in parentheses:

    #. Region-wise average signal (``csf``, ``white_matter``, ``global_signal``)
    #. DVARS - original and standardized variants (``dvars``, ``std_dvars``)
    #. Framewise displacement, based on head-motion parameters
       (``framewise_displacement``)
    #. Temporal CompCor (``t_comp_cor_XX``)
    #. Anatomical CompCor (``a_comp_cor_XX``)
    #. Cosine basis set for high-pass filtering w/ 0.008 Hz cut-off
       (``cosine_XX``)
    #. Non-steady-state volumes (``non_steady_state_XX``)
    #. Estimated head-motion parameters, in mm and rad
       (``trans_x``, ``trans_y``, ``trans_z``, ``rot_x``, ``rot_y``, ``rot_z``)


    Prior to estimating aCompCor and tCompCor, non-steady-state volumes are
    censored and high-pass filtered using a :abbr:`DCT (discrete cosine
    transform)` basis.
    The cosine basis, as well as one regressor per censored volume, are included
    for convenience.

    Workflow Graph
        .. workflow::
            :graph2use: orig
            :simple_form: yes

            from fmriprep.workflows.bold.confounds import init_bold_confs_wf
            wf = init_bold_confs_wf(
                mem_gb=1,
                metadata={},
                regressors_all_comps=False,
                regressors_dvars_th=1.5,
                regressors_fd_th=0.5,
            )

    Parameters
    ----------
    mem_gb : :obj:`float`
        Size of BOLD file in GB - please note that this size
        should be calculated after resamplings that may extend
        the FoV
    metadata : :obj:`dict`
        BIDS metadata for BOLD file
    name : :obj:`str`
        Name of workflow (default: ``bold_confs_wf``)
    regressors_all_comps : :obj:`bool`
        Indicates whether CompCor decompositions should return all
        components instead of the minimal number of components necessary
        to explain 50 percent of the variance in the decomposition mask.
    regressors_dvars_th : :obj:`float`
        Criterion for flagging DVARS outliers
    regressors_fd_th : :obj:`float`
        Criterion for flagging framewise displacement outliers

    Inputs
    ------
    bold
        BOLD image, after the prescribed corrections (STC, HMC and SDC)
        when available.
    bold_mask
        BOLD series mask
    movpar_file
        SPM-formatted motion parameters file
    rmsd_file
        Framewise displacement as measured by ``fsl_motion_outliers``.
    skip_vols
        number of non steady state volumes
    t1w_mask
        Mask of the skull-stripped template image
    t1w_tpms
        List of tissue probability maps in T1w space
    t1_bold_xform
        Affine matrix that maps the T1w space into alignment with
        the native BOLD space

    Outputs
    -------
    confounds_file
        TSV of all aggregated confounds
    rois_report
        Reportlet visualizing white-matter/CSF mask used for aCompCor,
        the ROI for tCompCor and the BOLD brain mask.
    confounds_metadata
        Confounds metadata dictionary.

    """
    from niworkflows.engine.workflows import LiterateWorkflow as Workflow
    from niworkflows.interfaces.confounds import ExpandModel, SpikeRegressors
    from niworkflows.interfaces.fixes import FixHeaderApplyTransforms as ApplyTransforms
    from niworkflows.interfaces.images import SignalExtraction
    from niworkflows.interfaces.masks import ROIsPlot
    from niworkflows.interfaces.nibabel import ApplyMask, Binarize
    from niworkflows.interfaces.patches import (
        RobustACompCor as ACompCor,
        RobustTCompCor as TCompCor,
    )
    from niworkflows.interfaces.plotting import (
        CompCorVariancePlot, ConfoundsCorrelationPlot
    )
    from niworkflows.interfaces.utils import (
        AddTSVHeader, TSV2JSON, DictMerge
    )
    from ...interfaces.confounds import aCompCorMasks

    gm_desc = (
        "dilating a GM mask extracted from the FreeSurfer's *aseg* segmentation" if freesurfer
        else "thresholding the corresponding partial volume map at 0.05"
    )

    workflow = Workflow(name=name)
    workflow.__desc__ = f"""\
Several confounding time-series were calculated based on the
*preprocessed BOLD*: framewise displacement (FD), DVARS and
three region-wise global signals.
FD was computed using two formulations following Power (absolute sum of
relative motions, @power_fd_dvars) and Jenkinson (relative root mean square
displacement between affines, @mcflirt).
FD and DVARS are calculated for each functional run, both using their
implementations in *Nipype* [following the definitions by @power_fd_dvars].
The three global signals are extracted within the CSF, the WM, and
the whole-brain masks.
Additionally, a set of physiological regressors were extracted to
allow for component-based noise correction [*CompCor*, @compcor].
Principal components are estimated after high-pass filtering the
*preprocessed BOLD* time-series (using a discrete cosine filter with
128s cut-off) for the two *CompCor* variants: temporal (tCompCor)
and anatomical (aCompCor).
tCompCor components are then calculated from the top 2% variable
voxels within the brain mask.
For aCompCor, three probabilistic masks (CSF, WM and combined CSF+WM)
are generated in anatomical space.
The implementation differs from that of Behzadi et al. in that instead
of eroding the masks by 2 pixels on BOLD space, the aCompCor masks are
subtracted a mask of pixels that likely contain a volume fraction of GM.
This mask is obtained by {gm_desc}, and it ensures components are not extracted
from voxels containing a minimal fraction of GM.
Finally, these masks are resampled into BOLD space and binarized by
thresholding at 0.99 (as in the original implementation).
Components are also calculated separately within the WM and CSF masks.
For each CompCor decomposition, the *k* components with the largest singular
values are retained, such that the retained components' time series are
sufficient to explain 50 percent of variance across the nuisance mask (CSF,
WM, combined, or temporal). The remaining components are dropped from
consideration.
The head-motion estimates calculated in the correction step were also
placed within the corresponding confounds file.
The confound time series derived from head motion estimates and global
signals were expanded with the inclusion of temporal derivatives and
quadratic terms for each [@confounds_satterthwaite_2013].
Frames that exceeded a threshold of {regressors_fd_th} mm FD or
{regressors_dvars_th} standardised DVARS were annotated as motion outliers.
"""
    inputnode = pe.Node(niu.IdentityInterface(
        fields=['bold', 'bold_mask', 'movpar_file', 'rmsd_file',
                'skip_vols', 't1w_mask', 't1w_tpms', 't1_bold_xform']),
        name='inputnode')
    outputnode = pe.Node(niu.IdentityInterface(
        fields=['confounds_file', 'confounds_metadata', 'acompcor_masks', 'tcompcor_mask']),
        name='outputnode')

    # DVARS
    dvars = pe.Node(nac.ComputeDVARS(save_nstd=True, save_std=True, remove_zerovariance=True),
                    name="dvars", mem_gb=mem_gb)

    # Frame displacement
    fdisp = pe.Node(nac.FramewiseDisplacement(parameter_source="SPM"),
                    name="fdisp", mem_gb=mem_gb)

    # Generate aCompCor probseg maps
    acc_masks = pe.Node(aCompCorMasks(is_aseg=freesurfer), name="acc_masks")

    # Resample probseg maps in BOLD space via T1w-to-BOLD transform
    acc_msk_tfm = pe.MapNode(ApplyTransforms(
        interpolation='Gaussian', float=False), iterfield=["input_image"],
        name='acc_msk_tfm', mem_gb=0.1)
    acc_msk_brain = pe.MapNode(ApplyMask(), name="acc_msk_brain",
                               iterfield=["in_file"])
    acc_msk_bin = pe.MapNode(Binarize(thresh_low=0.99), name='acc_msk_bin',
                             iterfield=["in_file"])
    acompcor = pe.Node(
        ACompCor(components_file='acompcor.tsv', header_prefix='a_comp_cor_', pre_filter='cosine',
                 save_pre_filter=True, save_metadata=True, mask_names=['CSF', 'WM', 'combined'],
                 merge_method='none', failure_mode='NaN'),
        name="acompcor", mem_gb=mem_gb)

    tcompcor = pe.Node(
        TCompCor(components_file='tcompcor.tsv', header_prefix='t_comp_cor_', pre_filter='cosine',
                 save_pre_filter=True, save_metadata=True, percentile_threshold=.02,
                 failure_mode='NaN'),
        name="tcompcor", mem_gb=mem_gb)

    # Set number of components
    if regressors_all_comps:
        acompcor.inputs.num_components = 'all'
        tcompcor.inputs.num_components = 'all'
    else:
        acompcor.inputs.variance_threshold = 0.5
        tcompcor.inputs.variance_threshold = 0.5

    # Set TR if present
    if 'RepetitionTime' in metadata:
        tcompcor.inputs.repetition_time = metadata['RepetitionTime']
        acompcor.inputs.repetition_time = metadata['RepetitionTime']

    # Global and segment regressors
    signals_class_labels = [
        "global_signal", "csf", "white_matter", "csf_wm", "tcompcor",
    ]
    merge_rois = pe.Node(niu.Merge(3, ravel_inputs=True), name='merge_rois',
                         run_without_submitting=True)
    signals = pe.Node(SignalExtraction(class_labels=signals_class_labels),
                      name="signals", mem_gb=mem_gb)

    # Arrange confounds
    add_dvars_header = pe.Node(
        AddTSVHeader(columns=["dvars"]),
        name="add_dvars_header", mem_gb=0.01, run_without_submitting=True)
    add_std_dvars_header = pe.Node(
        AddTSVHeader(columns=["std_dvars"]),
        name="add_std_dvars_header", mem_gb=0.01, run_without_submitting=True)
    add_motion_headers = pe.Node(
        AddTSVHeader(columns=["trans_x", "trans_y", "trans_z", "rot_x", "rot_y", "rot_z"]),
        name="add_motion_headers", mem_gb=0.01, run_without_submitting=True)
    add_rmsd_header = pe.Node(
        AddTSVHeader(columns=["rmsd"]),
        name="add_rmsd_header", mem_gb=0.01, run_without_submitting=True)
    concat = pe.Node(GatherConfounds(), name="concat", mem_gb=0.01, run_without_submitting=True)

    # CompCor metadata
    tcc_metadata_fmt = pe.Node(
        TSV2JSON(index_column='component', drop_columns=['mask'], output=None,
                 additional_metadata={'Method': 'tCompCor'}, enforce_case=True),
        name='tcc_metadata_fmt')
    acc_metadata_fmt = pe.Node(
        TSV2JSON(index_column='component', output=None,
                 additional_metadata={'Method': 'aCompCor'}, enforce_case=True),
        name='acc_metadata_fmt')
    mrg_conf_metadata = pe.Node(niu.Merge(3), name='merge_confound_metadata',
                                run_without_submitting=True)
    mrg_conf_metadata.inputs.in3 = {label: {'Method': 'Mean'}
                                    for label in signals_class_labels}
    mrg_conf_metadata2 = pe.Node(DictMerge(), name='merge_confound_metadata2',
                                 run_without_submitting=True)

    # Expand model to include derivatives and quadratics
    model_expand = pe.Node(ExpandModel(
        model_formula='(dd1(rps + wm + csf + gsr))^^2 + others'),
        name='model_expansion')

    # Add spike regressors
    spike_regress = pe.Node(SpikeRegressors(
        fd_thresh=regressors_fd_th,
        dvars_thresh=regressors_dvars_th),
        name='spike_regressors')

    # Generate reportlet (ROIs)
    mrg_compcor = pe.Node(niu.Merge(2, ravel_inputs=True),
                          name='mrg_compcor', run_without_submitting=True)
    rois_plot = pe.Node(ROIsPlot(colors=['b', 'magenta'], generate_report=True),
                        name='rois_plot', mem_gb=mem_gb)

    ds_report_bold_rois = pe.Node(
        DerivativesDataSink(desc='rois', datatype="figures", dismiss_entities=("echo",)),
        name='ds_report_bold_rois', run_without_submitting=True,
        mem_gb=DEFAULT_MEMORY_MIN_GB)

    # Generate reportlet (CompCor)
    mrg_cc_metadata = pe.Node(niu.Merge(2), name='merge_compcor_metadata',
                              run_without_submitting=True)
    compcor_plot = pe.Node(
        CompCorVariancePlot(variance_thresholds=(0.5, 0.7, 0.9),
                            metadata_sources=['tCompCor', 'aCompCor']),
        name='compcor_plot')
    ds_report_compcor = pe.Node(
        DerivativesDataSink(desc='compcorvar', datatype="figures", dismiss_entities=("echo",)),
        name='ds_report_compcor', run_without_submitting=True,
        mem_gb=DEFAULT_MEMORY_MIN_GB)

    # Generate reportlet (Confound correlation)
    conf_corr_plot = pe.Node(
        ConfoundsCorrelationPlot(reference_column='global_signal', max_dim=20),
        name='conf_corr_plot')
    ds_report_conf_corr = pe.Node(
        DerivativesDataSink(desc='confoundcorr', datatype="figures", dismiss_entities=("echo",)),
        name='ds_report_conf_corr', run_without_submitting=True,
        mem_gb=DEFAULT_MEMORY_MIN_GB)

    def _last(inlist):
        return inlist[-1]

    def _select_cols(table):
        import pandas as pd
        return [
            col for col in pd.read_table(table, nrows=2).columns
            if not col.startswith(("a_comp_cor_", "t_comp_cor_", "std_dvars"))
        ]

    workflow.connect([
        # connect inputnode to each non-anatomical confound node
        (inputnode, dvars, [('bold', 'in_file'),
                            ('bold_mask', 'in_mask')]),
        (inputnode, fdisp, [('movpar_file', 'in_file')]),

        # aCompCor
        (inputnode, acompcor, [("bold", "realigned_file"),
                               ("skip_vols", "ignore_initial_volumes")]),
        (inputnode, acc_masks, [("t1w_tpms", "in_vfs"),
                                (("bold", _get_zooms), "bold_zooms")]),
        (inputnode, acc_msk_tfm, [("t1_bold_xform", "transforms"),
                                  ("bold_mask", "reference_image")]),
        (inputnode, acc_msk_brain, [("bold_mask", "in_mask")]),
        (acc_masks, acc_msk_tfm, [("out_masks", "input_image")]),
        (acc_msk_tfm, acc_msk_brain, [("output_image", "in_file")]),
        (acc_msk_brain, acc_msk_bin, [("out_file", "in_file")]),
        (acc_msk_bin, acompcor, [("out_file", "mask_files")]),

        # tCompCor
        (inputnode, tcompcor, [("bold", "realigned_file"),
                               ("skip_vols", "ignore_initial_volumes"),
                               ("bold_mask", "mask_files")]),
        # Global signals extraction (constrained by anatomy)
        (inputnode, signals, [('bold', 'in_file')]),
        (inputnode, merge_rois, [('bold_mask', 'in1')]),
        (acc_msk_bin, merge_rois, [('out_file', 'in2')]),
        (tcompcor, merge_rois, [('high_variance_masks', 'in3')]),
        (merge_rois, signals, [('out', 'label_files')]),

        # Collate computed confounds together
        (inputnode, add_motion_headers, [('movpar_file', 'in_file')]),
        (inputnode, add_rmsd_header, [('rmsd_file', 'in_file')]),
        (dvars, add_dvars_header, [('out_nstd', 'in_file')]),
        (dvars, add_std_dvars_header, [('out_std', 'in_file')]),
        (signals, concat, [('out_file', 'signals')]),
        (fdisp, concat, [('out_file', 'fd')]),
        (tcompcor, concat, [('components_file', 'tcompcor'),
                            ('pre_filter_file', 'cos_basis')]),
        (acompcor, concat, [('components_file', 'acompcor')]),
        (add_motion_headers, concat, [('out_file', 'motion')]),
        (add_rmsd_header, concat, [('out_file', 'rmsd')]),
        (add_dvars_header, concat, [('out_file', 'dvars')]),
        (add_std_dvars_header, concat, [('out_file', 'std_dvars')]),

        # Confounds metadata
        (tcompcor, tcc_metadata_fmt, [('metadata_file', 'in_file')]),
        (acompcor, acc_metadata_fmt, [('metadata_file', 'in_file')]),
        (tcc_metadata_fmt, mrg_conf_metadata, [('output', 'in1')]),
        (acc_metadata_fmt, mrg_conf_metadata, [('output', 'in2')]),
        (mrg_conf_metadata, mrg_conf_metadata2, [('out', 'in_dicts')]),

        # Expand the model with derivatives, quadratics, and spikes
        (concat, model_expand, [('confounds_file', 'confounds_file')]),
        (model_expand, spike_regress, [('confounds_file', 'confounds_file')]),

        # Set outputs
        (spike_regress, outputnode, [('confounds_file', 'confounds_file')]),
        (mrg_conf_metadata2, outputnode, [('out_dict', 'confounds_metadata')]),
        (tcompcor, outputnode, [("high_variance_masks", "tcompcor_mask")]),
        (acc_msk_bin, outputnode, [("out_file", "acompcor_masks")]),
        (inputnode, rois_plot, [('bold', 'in_file'),
                                ('bold_mask', 'in_mask')]),
        (tcompcor, mrg_compcor, [('high_variance_masks', 'in1')]),
        (acc_msk_bin, mrg_compcor, [(('out_file', _last), 'in2')]),
        (mrg_compcor, rois_plot, [('out', 'in_rois')]),
        (rois_plot, ds_report_bold_rois, [('out_report', 'in_file')]),
        (tcompcor, mrg_cc_metadata, [('metadata_file', 'in1')]),
        (acompcor, mrg_cc_metadata, [('metadata_file', 'in2')]),
        (mrg_cc_metadata, compcor_plot, [('out', 'metadata_files')]),
        (compcor_plot, ds_report_compcor, [('out_file', 'in_file')]),
        (concat, conf_corr_plot, [('confounds_file', 'confounds_file'),
                                  (('confounds_file', _select_cols), 'columns')]),
        (conf_corr_plot, ds_report_conf_corr, [('out_file', 'in_file')]),
    ])

    return workflow


def init_carpetplot_wf(mem_gb, metadata, cifti_output, name="bold_carpet_wf"):
    """
    Build a workflow to generate *carpet* plots.

    Resamples the MNI parcellation (ad-hoc parcellation derived from the
    Harvard-Oxford template and others).

    Parameters
    ----------
    mem_gb : :obj:`float`
        Size of BOLD file in GB - please note that this size
        should be calculated after resamplings that may extend
        the FoV
    metadata : :obj:`dict`
        BIDS metadata for BOLD file
    name : :obj:`str`
        Name of workflow (default: ``bold_carpet_wf``)

    Inputs
    ------
    bold
        BOLD image, after the prescribed corrections (STC, HMC and SDC)
        when available.
    bold_mask
        BOLD series mask
    confounds_file
        TSV of all aggregated confounds
    t1_bold_xform
        Affine matrix that maps the T1w space into alignment with
        the native BOLD space
    std2anat_xfm
        ANTs-compatible affine-and-warp transform file
    cifti_bold
        BOLD image in CIFTI format, to be used in place of volumetric BOLD

    Outputs
    -------
    out_carpetplot
        Path of the generated SVG file

    """
    from niworkflows.engine.workflows import LiterateWorkflow as Workflow
    from niworkflows.interfaces.fixes import FixHeaderApplyTransforms as ApplyTransforms

    inputnode = pe.Node(niu.IdentityInterface(
        fields=['bold', 'bold_mask', 'confounds_file',
                't1_bold_xform', 'std2anat_xfm', 'cifti_bold']),
        name='inputnode')

    outputnode = pe.Node(niu.IdentityInterface(
        fields=['out_carpetplot']), name='outputnode')

    # List transforms
    mrg_xfms = pe.Node(niu.Merge(2), name='mrg_xfms')

    # Warp segmentation into EPI space
    resample_parc = pe.Node(ApplyTransforms(
        dimension=3,
        input_image=str(get_template(
            'MNI152NLin2009cAsym', resolution=1, desc='carpet',
            suffix='dseg', extension=['.nii', '.nii.gz'])),
        interpolation='MultiLabel'),
        name='resample_parc')

    # Carpetplot and confounds plot
    conf_plot = pe.Node(FMRISummary(
        tr=metadata['RepetitionTime'],
        confounds_list=[
            ('global_signal', None, 'GS'),
            ('csf', None, 'GSCSF'),
            ('white_matter', None, 'GSWM'),
            ('std_dvars', None, 'DVARS'),
            ('framewise_displacement', 'mm', 'FD')]),
        name='conf_plot', mem_gb=mem_gb)
    ds_report_bold_conf = pe.Node(
        DerivativesDataSink(desc='carpetplot', datatype="figures", extension="svg",
                            dismiss_entities=("echo",)),
        name='ds_report_bold_conf', run_without_submitting=True,
        mem_gb=DEFAULT_MEMORY_MIN_GB)

    workflow = Workflow(name=name)
    # no need for segmentations if using CIFTI
    if cifti_output:
        workflow.connect(inputnode, 'cifti_bold', conf_plot, 'in_func')
    else:
        workflow.connect([
            (inputnode, mrg_xfms, [('t1_bold_xform', 'in1'),
                                   ('std2anat_xfm', 'in2')]),
            (inputnode, resample_parc, [('bold_mask', 'reference_image')]),
            (mrg_xfms, resample_parc, [('out', 'transforms')]),
            # Carpetplot
            (inputnode, conf_plot, [
                ('bold', 'in_func'),
                ('bold_mask', 'in_mask')]),
            (resample_parc, conf_plot, [('output_image', 'in_segm')])
        ])

    workflow.connect([
        (inputnode, conf_plot, [('confounds_file', 'confounds_file')]),
        (conf_plot, ds_report_bold_conf, [('out_file', 'in_file')]),
        (conf_plot, outputnode, [('out_file', 'out_carpetplot')]),
    ])
    return workflow


def init_ica_aroma_wf(
    mem_gb,
    metadata,
    omp_nthreads,
    aroma_melodic_dim=-200,
    err_on_aroma_warn=False,
    name='ica_aroma_wf',
    susan_fwhm=6.0,
):
    """
    Build a workflow that runs `ICA-AROMA`_.

    This workflow wraps `ICA-AROMA`_ to identify and remove motion-related
    independent components from a BOLD time series.

    The following steps are performed:

    #. Remove non-steady state volumes from the bold series.
    #. Smooth data using FSL `susan`, with a kernel width FWHM=6.0mm.
    #. Run FSL `melodic` outside of ICA-AROMA to generate the report
    #. Run ICA-AROMA
    #. Aggregate identified motion components (aggressive) to TSV
    #. Return ``classified_motion_ICs`` and ``melodic_mix`` for user to complete
       non-aggressive denoising in T1w space
    #. Calculate ICA-AROMA-identified noise components
       (columns named ``AROMAAggrCompXX``)

    Additionally, non-aggressive denoising is performed on the BOLD series
    resampled into MNI space.

    There is a current discussion on whether other confounds should be extracted
    before or after denoising `here
    <http://nbviewer.jupyter.org/github/nipreps/fmriprep-notebooks/blob/922e436429b879271fa13e76767a6e73443e74d9/issue-817_aroma_confounds.ipynb>`__.

    .. _ICA-AROMA: https://github.com/maartenmennes/ICA-AROMA

    Workflow Graph
        .. workflow::
            :graph2use: orig
            :simple_form: yes

            from fmriprep.workflows.bold.confounds import init_ica_aroma_wf
            wf = init_ica_aroma_wf(
                mem_gb=3,
                metadata={'RepetitionTime': 1.0},
                omp_nthreads=1)

    Parameters
    ----------
    metadata : :obj:`dict`
        BIDS metadata for BOLD file
    mem_gb : :obj:`float`
        Size of BOLD file in GB
    omp_nthreads : :obj:`int`
        Maximum number of threads an individual process may use
    name : :obj:`str`
        Name of workflow (default: ``bold_tpl_trans_wf``)
    susan_fwhm : :obj:`float`
        Kernel width (FWHM in mm) for the smoothing step with
        FSL ``susan`` (default: 6.0mm)
    err_on_aroma_warn : :obj:`bool`
        Do not fail on ICA-AROMA errors
    aroma_melodic_dim : :obj:`int`
        Set the dimensionality of the MELODIC ICA decomposition.
        Negative numbers set a maximum on automatic dimensionality estimation.
        Positive numbers set an exact number of components to extract.
        (default: -200, i.e., estimate <=200 components)

    Inputs
    ------
    itk_bold_to_t1
        Affine transform from ``ref_bold_brain`` to T1 space (ITK format)
    anat2std_xfm
        ANTs-compatible affine-and-warp transform file
    name_source
        BOLD series NIfTI file
        Used to recover original information lost during processing
    skip_vols
        number of non steady state volumes
    bold_split
        Individual 3D BOLD volumes, not motion corrected
    bold_mask
        BOLD series mask in template space
    hmc_xforms
        List of affine transforms aligning each volume to ``ref_image`` in ITK format
    movpar_file
        SPM-formatted motion parameters file

    Outputs
    -------
    aroma_confounds
        TSV of confounds identified as noise by ICA-AROMA
    aroma_noise_ics
        CSV of noise components identified by ICA-AROMA
    melodic_mix
        FSL MELODIC mixing matrix
    nonaggr_denoised_file
        BOLD series with non-aggressive ICA-AROMA denoising applied

    """
    from niworkflows.engine.workflows import LiterateWorkflow as Workflow
    from niworkflows.interfaces.segmentation import ICA_AROMARPT
    from niworkflows.interfaces.utility import KeySelect
    from niworkflows.interfaces.utils import TSV2JSON

    workflow = Workflow(name=name)
    workflow.__postdesc__ = """\
Automatic removal of motion artifacts using independent component analysis
[ICA-AROMA, @aroma] was performed on the *preprocessed BOLD on MNI space*
time-series after removal of non-steady state volumes and spatial smoothing
with an isotropic, Gaussian kernel of 6mm FWHM (full-width half-maximum).
Corresponding "non-aggresively" denoised runs were produced after such
smoothing.
Additionally, the "aggressive" noise-regressors were collected and placed
in the corresponding confounds file.
"""

    inputnode = pe.Node(niu.IdentityInterface(
        fields=[
            'bold_std',
            'bold_mask_std',
            'movpar_file',
            'name_source',
            'skip_vols',
            'spatial_reference',
        ]), name='inputnode')

    outputnode = pe.Node(niu.IdentityInterface(
        fields=['aroma_confounds', 'aroma_noise_ics', 'melodic_mix',
                'nonaggr_denoised_file', 'aroma_metadata']), name='outputnode')

    # extract out to BOLD base
    select_std = pe.Node(KeySelect(fields=['bold_mask_std', 'bold_std']),
                         name='select_std', run_without_submitting=True)
    select_std.inputs.key = 'MNI152NLin6Asym_res-2'

    rm_non_steady_state = pe.Node(niu.Function(function=_remove_volumes,
                                               output_names=['bold_cut']),
                                  name='rm_nonsteady')

    calc_median_val = pe.Node(fsl.ImageStats(op_string='-k %s -p 50'), name='calc_median_val')
    calc_bold_mean = pe.Node(fsl.MeanImage(), name='calc_bold_mean')

    def _getusans_func(image, thresh):
        return [tuple([image, thresh])]
    getusans = pe.Node(niu.Function(function=_getusans_func, output_names=['usans']),
                       name='getusans', mem_gb=0.01)

    smooth = pe.Node(fsl.SUSAN(fwhm=susan_fwhm), name='smooth')

    # melodic node
    melodic = pe.Node(fsl.MELODIC(
        no_bet=True, tr_sec=float(metadata['RepetitionTime']), mm_thresh=0.5, out_stats=True,
        dim=aroma_melodic_dim), name="melodic")

    # ica_aroma node
    ica_aroma = pe.Node(ICA_AROMARPT(
        denoise_type='nonaggr', generate_report=True, TR=metadata['RepetitionTime'],
        args='-np'), name='ica_aroma')

    add_non_steady_state = pe.Node(niu.Function(function=_add_volumes,
                                                output_names=['bold_add']),
                                   name='add_nonsteady')

    # extract the confound ICs from the results
    ica_aroma_confound_extraction = pe.Node(ICAConfounds(err_on_aroma_warn=err_on_aroma_warn),
                                            name='ica_aroma_confound_extraction')

    ica_aroma_metadata_fmt = pe.Node(
        TSV2JSON(index_column='IC', output=None, enforce_case=True,
                 additional_metadata={'Method': {
                                      'Name': 'ICA-AROMA',
                                      'Version': getenv('AROMA_VERSION', 'n/a')}}),
        name='ica_aroma_metadata_fmt')

    ds_report_ica_aroma = pe.Node(
        DerivativesDataSink(desc='aroma', datatype="figures", dismiss_entities=("echo",)),
        name='ds_report_ica_aroma', run_without_submitting=True,
        mem_gb=DEFAULT_MEMORY_MIN_GB)

    def _getbtthresh(medianval):
        return 0.75 * medianval

    # connect the nodes
    workflow.connect([
        (inputnode, select_std, [('spatial_reference', 'keys'),
                                 ('bold_std', 'bold_std'),
                                 ('bold_mask_std', 'bold_mask_std')]),
        (inputnode, ica_aroma, [('movpar_file', 'motion_parameters')]),
        (inputnode, rm_non_steady_state, [
            ('skip_vols', 'skip_vols')]),
        (select_std, rm_non_steady_state, [
            ('bold_std', 'bold_file')]),
        (select_std, calc_median_val, [
            ('bold_mask_std', 'mask_file')]),
        (rm_non_steady_state, calc_median_val, [
            ('bold_cut', 'in_file')]),
        (rm_non_steady_state, calc_bold_mean, [
            ('bold_cut', 'in_file')]),
        (calc_bold_mean, getusans, [('out_file', 'image')]),
        (calc_median_val, getusans, [('out_stat', 'thresh')]),
        # Connect input nodes to complete smoothing
        (rm_non_steady_state, smooth, [
            ('bold_cut', 'in_file')]),
        (getusans, smooth, [('usans', 'usans')]),
        (calc_median_val, smooth, [(('out_stat', _getbtthresh), 'brightness_threshold')]),
        # connect smooth to melodic
        (smooth, melodic, [('smoothed_file', 'in_files')]),
        (select_std, melodic, [
            ('bold_mask_std', 'mask')]),
        # connect nodes to ICA-AROMA
        (smooth, ica_aroma, [('smoothed_file', 'in_file')]),
        (select_std, ica_aroma, [
            ('bold_mask_std', 'report_mask'),
            ('bold_mask_std', 'mask')]),
        (melodic, ica_aroma, [('out_dir', 'melodic_dir')]),
        # generate tsvs from ICA-AROMA
        (ica_aroma, ica_aroma_confound_extraction, [('out_dir', 'in_directory')]),
        (inputnode, ica_aroma_confound_extraction, [
            ('skip_vols', 'skip_vols')]),
        (ica_aroma_confound_extraction, ica_aroma_metadata_fmt, [
            ('aroma_metadata', 'in_file')]),
        # output for processing and reporting
        (ica_aroma_confound_extraction, outputnode, [('aroma_confounds', 'aroma_confounds'),
                                                     ('aroma_noise_ics', 'aroma_noise_ics'),
                                                     ('melodic_mix', 'melodic_mix')]),
        (ica_aroma_metadata_fmt, outputnode, [('output', 'aroma_metadata')]),
        (ica_aroma, add_non_steady_state, [
            ('nonaggr_denoised_file', 'bold_cut_file')]),
        (select_std, add_non_steady_state, [
            ('bold_std', 'bold_file')]),
        (inputnode, add_non_steady_state, [
            ('skip_vols', 'skip_vols')]),
        (add_non_steady_state, outputnode, [('bold_add', 'nonaggr_denoised_file')]),
        (ica_aroma, ds_report_ica_aroma, [('out_report', 'in_file')]),
    ])

    return workflow


def _remove_volumes(bold_file, skip_vols):
    """Remove skip_vols from bold_file."""
    import nibabel as nb
    from nipype.utils.filemanip import fname_presuffix

    if skip_vols == 0:
        return bold_file

    out = fname_presuffix(bold_file, suffix='_cut')
    bold_img = nb.load(bold_file)
    bold_img.__class__(bold_img.dataobj[..., skip_vols:],
                       bold_img.affine, bold_img.header).to_filename(out)
    return out


def _add_volumes(bold_file, bold_cut_file, skip_vols):
    """Prepend skip_vols from bold_file onto bold_cut_file."""
    import nibabel as nb
    import numpy as np
    from nipype.utils.filemanip import fname_presuffix

    if skip_vols == 0:
        return bold_cut_file

    bold_img = nb.load(bold_file)
    bold_cut_img = nb.load(bold_cut_file)

    bold_data = np.concatenate((bold_img.dataobj[..., :skip_vols],
                                bold_cut_img.dataobj), axis=3)

    out = fname_presuffix(bold_cut_file, suffix='_addnonsteady')
    bold_img.__class__(bold_data, bold_img.affine, bold_img.header).to_filename(out)
    return out


def _get_zooms(in_file):
    import nibabel as nb
    return tuple(nb.load(in_file).header.get_zooms()[:3])