preprocess.py

# -*- coding: utf-8 -*-

import os.path as op
import nibabel as nb
import numpy as np

from looseversion import LooseVersion
from ... import logging
from ..base import traits, TraitedSpec, File, isdefined
from .base import (
    HAVE_DIPY,
    dipy_version,
    dipy_to_nipype_interface,
    get_dipy_workflows,
    DipyBaseInterface,
)

IFLOGGER = logging.getLogger("nipype.interface")

if HAVE_DIPY and LooseVersion(dipy_version()) >= LooseVersion("0.15"):
    from dipy.workflows import denoise, mask

    l_wkflw = get_dipy_workflows(denoise) + get_dipy_workflows(mask)
    for name, obj in l_wkflw:
        new_name = name.replace("Flow", "")
        globals()[new_name] = dipy_to_nipype_interface(new_name, obj)
    del l_wkflw

else:
    IFLOGGER.info(
        "We advise you to upgrade DIPY version. This upgrade will"
        " open access to more function"
    )


class ResampleInputSpec(TraitedSpec):
    in_file = File(
        exists=True, mandatory=True, desc="The input 4D diffusion-weighted image file"
    )
    vox_size = traits.Tuple(
        traits.Float,
        traits.Float,
        traits.Float,
        desc=(
            "specify the new voxel zooms. If no vox_size"
            " is set, then isotropic regridding will "
            "be performed, with spacing equal to the "
            "smallest current zoom."
        ),
    )
    interp = traits.Int(
        1,
        mandatory=True,
        usedefault=True,
        desc=("order of the interpolator (0 = nearest, 1 = linear, etc."),
    )


class ResampleOutputSpec(TraitedSpec):
    out_file = File(exists=True)


class Resample(DipyBaseInterface):
    """
    An interface to reslicing diffusion datasets.
    See
    http://nipy.org/dipy/examples_built/reslice_datasets.html#example-reslice-datasets.

    Example
    -------

    >>> import nipype.interfaces.dipy as dipy
    >>> reslice = dipy.Resample()
    >>> reslice.inputs.in_file = 'diffusion.nii'
    >>> reslice.run() # doctest: +SKIP
    """

    input_spec = ResampleInputSpec
    output_spec = ResampleOutputSpec

    def _run_interface(self, runtime):
        order = self.inputs.interp
        vox_size = None

        if isdefined(self.inputs.vox_size):
            vox_size = self.inputs.vox_size

        out_file = op.abspath(self._gen_outfilename())
        resample_proxy(
            self.inputs.in_file, order=order, new_zooms=vox_size, out_file=out_file
        )

        IFLOGGER.info("Resliced image saved as %s", out_file)
        return runtime

    def _list_outputs(self):
        outputs = self._outputs().get()
        outputs["out_file"] = op.abspath(self._gen_outfilename())
        return outputs

    def _gen_outfilename(self):
        fname, fext = op.splitext(op.basename(self.inputs.in_file))
        if fext == ".gz":
            fname, fext2 = op.splitext(fname)
            fext = fext2 + fext
        return op.abspath("%s_reslice%s" % (fname, fext))


class DenoiseInputSpec(TraitedSpec):
    in_file = File(
        exists=True, mandatory=True, desc="The input 4D diffusion-weighted image file"
    )
    in_mask = File(exists=True, desc="brain mask")
    noise_model = traits.Enum(
        "rician",
        "gaussian",
        mandatory=True,
        usedefault=True,
        desc=("noise distribution model"),
    )
    signal_mask = File(
        desc=("mask in which the mean signal " "will be computed"), exists=True
    )
    noise_mask = File(
        desc=("mask in which the standard deviation of noise " "will be computed"),
        exists=True,
    )
    patch_radius = traits.Int(1, usedefault=True, desc="patch radius")
    block_radius = traits.Int(5, usedefault=True, desc="block_radius")
    snr = traits.Float(desc="manually set an SNR")


class DenoiseOutputSpec(TraitedSpec):
    out_file = File(exists=True)


class Denoise(DipyBaseInterface):
    """
    An interface to denoising diffusion datasets [Coupe2008]_.
    See
    http://nipy.org/dipy/examples_built/denoise_nlmeans.html#example-denoise-nlmeans.

    .. [Coupe2008] Coupe P et al., `An Optimized Blockwise Non Local Means
      Denoising Filter for 3D Magnetic Resonance Images
      <https://doi.org/10.1109%2FTMI.2007.906087>`_,
      IEEE Transactions on Medical Imaging, 27(4):425-441, 2008.


    Example
    -------

    >>> import nipype.interfaces.dipy as dipy
    >>> denoise = dipy.Denoise()
    >>> denoise.inputs.in_file = 'diffusion.nii'
    >>> denoise.run() # doctest: +SKIP
    """

    input_spec = DenoiseInputSpec
    output_spec = DenoiseOutputSpec

    def _run_interface(self, runtime):
        out_file = op.abspath(self._gen_outfilename())

        settings = dict(mask=None, rician=(self.inputs.noise_model == "rician"))

        if isdefined(self.inputs.in_mask):
            settings["mask"] = np.asanyarray(nb.load(self.inputs.in_mask).dataobj)

        if isdefined(self.inputs.patch_radius):
            settings["patch_radius"] = self.inputs.patch_radius

        if isdefined(self.inputs.block_radius):
            settings["block_radius"] = self.inputs.block_radius

        snr = None
        if isdefined(self.inputs.snr):
            snr = self.inputs.snr

        signal_mask = None
        if isdefined(self.inputs.signal_mask):
            signal_mask = np.asanyarray(nb.load(self.inputs.signal_mask).dataobj)
        noise_mask = None
        if isdefined(self.inputs.noise_mask):
            noise_mask = np.asanyarray(nb.load(self.inputs.noise_mask).dataobj)

        _, s = nlmeans_proxy(
            self.inputs.in_file,
            settings,
            snr=snr,
            smask=signal_mask,
            nmask=noise_mask,
            out_file=out_file,
        )
        IFLOGGER.info("Denoised image saved as %s, estimated SNR=%s", out_file, str(s))
        return runtime

    def _list_outputs(self):
        outputs = self._outputs().get()
        outputs["out_file"] = op.abspath(self._gen_outfilename())
        return outputs

    def _gen_outfilename(self):
        fname, fext = op.splitext(op.basename(self.inputs.in_file))
        if fext == ".gz":
            fname, fext2 = op.splitext(fname)
            fext = fext2 + fext
        return op.abspath("%s_denoise%s" % (fname, fext))


def resample_proxy(in_file, order=3, new_zooms=None, out_file=None):
    """
    Performs regridding of an image to set isotropic voxel sizes using dipy.
    """
    from dipy.align.reslice import reslice

    if out_file is None:
        fname, fext = op.splitext(op.basename(in_file))
        if fext == ".gz":
            fname, fext2 = op.splitext(fname)
            fext = fext2 + fext
        out_file = op.abspath("./%s_reslice%s" % (fname, fext))

    img = nb.load(in_file)
    hdr = img.header.copy()
    data = img.get_fdata(dtype=np.float32)
    affine = img.affine
    im_zooms = hdr.get_zooms()[:3]

    if new_zooms is None:
        minzoom = np.array(im_zooms).min()
        new_zooms = tuple(np.ones((3,)) * minzoom)

    if np.all(im_zooms == new_zooms):
        return in_file

    data2, affine2 = reslice(data, affine, im_zooms, new_zooms, order=order)
    tmp_zooms = np.array(hdr.get_zooms())
    tmp_zooms[:3] = new_zooms[0]
    hdr.set_zooms(tuple(tmp_zooms))
    hdr.set_data_shape(data2.shape)
    hdr.set_xyzt_units("mm")
    nb.Nifti1Image(data2.astype(hdr.get_data_dtype()), affine2, hdr).to_filename(
        out_file
    )
    return out_file, new_zooms


def nlmeans_proxy(in_file, settings, snr=None, smask=None, nmask=None, out_file=None):
    """
    Uses non-local means to denoise 4D datasets
    """
    from dipy.denoise.nlmeans import nlmeans
    from scipy.ndimage.morphology import binary_erosion
    from scipy import ndimage

    if out_file is None:
        fname, fext = op.splitext(op.basename(in_file))
        if fext == ".gz":
            fname, fext2 = op.splitext(fname)
            fext = fext2 + fext
        out_file = op.abspath("./%s_denoise%s" % (fname, fext))

    img = nb.load(in_file)
    hdr = img.header
    data = img.get_fdata()
    aff = img.affine

    if data.ndim < 4:
        data = data[..., np.newaxis]

    data = np.nan_to_num(data)

    if data.max() < 1.0e-4:
        raise RuntimeError("There is no signal in the image")

    df = 1.0
    if data.max() < 1000.0:
        df = 1000.0 / data.max()
        data *= df

    b0 = data[..., 0]

    if smask is None:
        smask = np.zeros_like(b0)
        smask[b0 > np.percentile(b0, 85.0)] = 1

    smask = binary_erosion(smask.astype(np.uint8), iterations=2).astype(np.uint8)

    if nmask is None:
        nmask = np.ones_like(b0, dtype=np.uint8)
        bmask = settings["mask"]
        if bmask is None:
            bmask = np.zeros_like(b0)
            bmask[b0 > np.percentile(b0[b0 > 0], 10)] = 1
            label_im, nb_labels = ndimage.label(bmask)
            sizes = ndimage.sum(bmask, label_im, range(nb_labels + 1))
            maxidx = np.argmax(sizes)
            bmask = np.zeros_like(b0, dtype=np.uint8)
            bmask[label_im == maxidx] = 1
        nmask[bmask > 0] = 0
    else:
        nmask = np.squeeze(nmask)
        nmask[nmask > 0.0] = 1
        nmask[nmask < 1] = 0
        nmask = nmask.astype(bool)

    nmask = binary_erosion(nmask, iterations=1).astype(np.uint8)

    den = np.zeros_like(data)

    est_snr = True
    if snr is not None:
        snr = [snr] * data.shape[-1]
        est_snr = False
    else:
        snr = []

    for i in range(data.shape[-1]):
        d = data[..., i]
        if est_snr:
            s = np.mean(d[smask > 0])
            n = np.std(d[nmask > 0])
            snr.append(s / n)

        den[..., i] = nlmeans(d, snr[i], **settings)

    den = np.squeeze(den)
    den /= df

    nb.Nifti1Image(den.astype(hdr.get_data_dtype()), aff, hdr).to_filename(out_file)
    return out_file, snr