usage: __main__.py [-h] [--mask MASK] [--tolerance tol] [--skip_b0_check]
[--min_k MIN_K] [--max_k MAX_K] [--smooth SMOOTH]
[--not_all] [--ak file] [--mk file] [--rk file]
[--msk file] [--dki_fa file] [--dki_md file]
[--dki_ad file] [--dki_rd file] [--dki_residual file]
[--msd file] [-v [{DEBUG,INFO,WARNING}]] [-f]
in_dwi in_bval in_bvec
Script to compute the Diffusion Kurtosis Imaging (DKI) and Mean Signal DKI
(MSDKI) metrics. DKI is a multi-shell diffusion model. The input DWI needs
to be multi-shell, i.e. multi-bvalued.
Since the diffusion kurtosis model involves the estimation of a large number
of parameters and since the non-Gaussian components of the diffusion signal
are more sensitive to artefacts, you should really denoise your DWI volume
before using this DKI script (e.g. scil_denoising_nlmeans.py). Moreover, to
remove biases due to fiber dispersion, fiber crossings and other mesoscopic
properties of the underlying tissue, MSDKI does a powder-average of DWI for all
directions, thus removing the orientational dependencies and creating an
alternative mean kurtosis map.
DKI is also known to be vulnerable to artefacted voxels induced by the
low radial diffusivities of aligned white matter (CC, CST voxels). Since it is
very hard to capture non-Gaussian information due to the low decays in radial
direction, its kurtosis estimates have very low robustness.
Noisy kurtosis estimates tend to be negative and its absolute values can have
order of magnitudes higher than the typical kurtosis values. Consequently,
these negative kurtosis values will heavily propagate to the mean and radial
kurtosis metrics. This is well-reported in [Rafael Henriques MSc thesis 2012,
chapter 3]. Two ways to overcome this issue: i) compute the kurtosis values
from powder-averaged MSDKI, and ii) perform 3D Gaussian smoothing. On
powder-averaged signal decays, you don't have this low diffusivity issue and
your kurtosis estimates have much higher precision (additionally they are
independent to the fODF).
By default, will output all available metrics, using default names. Specific
names can be specified using the metrics flags that are listed in the "Metrics
files flags" section. If --not_all is set, only the metrics specified
explicitly by the flags will be output.
This script directly comes from the DIPY example gallery and references
therein.
[1] examples_built/reconst_dki/#example-reconst-dki
[2] examples_built/reconst_msdki/#example-reconst-msdki
Formerly: scil_compute_kurtosis_metrics.py
positional arguments:
in_dwi Path of the input multi-shell DWI dataset.
in_bval Path of the b-value file, in FSL format.
in_bvec Path of the b-vector file, in FSL format.
options:
-h, --help show this help message and exit
--mask MASK Path to a binary mask.
Only data inside the mask will be used for computations and reconstruction.
[Default: None]
--tolerance tol The tolerated gap between the b-values to extract and the current b-value.
[Default: 20]
* Note. We would expect to find at least one b-value in the
range [0, tolerance]. To skip this check, use --skip_b0_check.
--skip_b0_check By default, we supervise that at least one b0 exists in your data
(i.e. b-values below the default --tolerance). Use this option to
allow continuing even if the minimum b-value is suspiciously high.
Use with care, and only if you understand your data.
--min_k MIN_K Minimum kurtosis value in the output maps
(ak, mk, rk). In theory, -3/7 is the min kurtosis
limit for regions that consist of water confined
to spherical pores (see DIPY example and
documentation) [Default: 0.0].
--max_k MAX_K Maximum kurtosis value in the output maps
(ak, mk, rk). In theory, 10 is the max kurtosis
limit for regions that consist of water confined
to spherical pores (see DIPY example and
documentation) [Default: 3.0].
--smooth SMOOTH Smooth input DWI with a 3D Gaussian filter with
full-width-half-max (fwhm). Kurtosis fitting is
sensitive and outliers occur easily. According to
tests on HCP, CB_Brain, Penthera3T, this smoothing
is thus turned ON by default with fwhm=2.5.
[Default: 2.5].
--not_all If set, will only save the metrics explicitly
specified using the other metrics flags.
[Default: not set].
-v [{DEBUG,INFO,WARNING}]
Produces verbose output depending on the provided level.
Default level is warning, default when using -v is info.
-f Force overwriting of the output files.
Metrics files flags:
--ak file Output filename for the axial kurtosis.
--mk file Output filename for the mean kurtosis.
--rk file Output filename for the radial kurtosis.
--msk file Output filename for the mean signal kurtosis.
--dki_fa file Output filename for the fractional anisotropy from DKI.
--dki_md file Output filename for the mean diffusivity from DKI.
--dki_ad file Output filename for the axial diffusivity from DKI.
--dki_rd file Output filename for the radial diffusivity from DKI.
Quality control files flags:
--dki_residual file Output filename for the map of the residual of the tensor fit.
Note. In previous versions, the resulting map was normalized.
It is not anymore.
--msd file Output filename for the mean signal diffusion (powder-average).