Source code for dipy.reconst.dki

#!/usr/bin/python
"""
Classes and functions for fitting the diffusion kurtosis model.
"""

import warnings

import numpy as np
import scipy.optimize as opt

from dipy.core.geometry import cart2sphere, perpendicular_directions, sphere2cart
from dipy.core.gradients import check_multi_b
from dipy.core.ndindex import ndindex
from dipy.core.optimize import PositiveDefiniteLeastSquares
import dipy.core.sphere as dps
from dipy.data import get_fnames, get_sphere, load_sdp_constraints
from dipy.reconst.base import ReconstModel
from dipy.reconst.dti import (
    MIN_POSITIVE_SIGNAL,
    TensorFit,
    decompose_tensor,
    from_lower_triangular,
    iterative_fit_tensor,
    lower_triangular,
    mean_diffusivity,
    nlls_fit_tensor,
    radial_diffusivity,
    restore_fit_tensor,
    robust_fit_tensor_nlls,
    robust_fit_tensor_wls,
)
from dipy.reconst.multi_voxel import multi_voxel_fit
from dipy.reconst.recspeed import local_maxima
from dipy.reconst.utils import dki_design_matrix as design_matrix
from dipy.reconst.vec_val_sum import vec_val_vect
from dipy.reconst.weights_method import (
    weights_method_wls_m_est,
)
from dipy.testing.decorators import warning_for_keywords


@warning_for_keywords()
def _positive_evals(L1, L2, L3, *, er=2e-7):
    """Helper function that identifies which voxels in an array have all
    eigenvalues significantly larger than zero

    Parameters
    ----------
    L1 : ndarray
        First independent variable of the integral.
    L2 : ndarray
        Second independent variable of the integral.
    L3 : ndarray
        Third independent variable of the integral.
    er : float, optional
        A eigenvalues is classified as larger than zero if it is larger than er

    Returns
    -------
    ind : boolean (n,)
        Array that marks the voxels that have all eigenvalues are larger than
        zero.

    """
    ind = np.logical_and(L1 > er, np.logical_and(L2 > er, L3 > er))

    return ind


[docs] @warning_for_keywords() def carlson_rf(x, y, z, *, errtol=3e-4): r"""Compute the Carlson's incomplete elliptic integral of the first kind. Carlson's incomplete elliptic integral of the first kind is defined as :footcite:p:`Carlson1995`: .. math:: R_F = \frac{1}{2} \int_{0}^{\infty} \left [(t+x)(t+y)(t+z) \right ] ^{-\frac{1}{2}}dt Parameters ---------- x : ndarray First independent variable of the integral. y : ndarray Second independent variable of the integral. z : ndarray Third independent variable of the integral. errtol : float Error tolerance. Integral is computed with relative error less in magnitude than the defined value Returns ------- RF : ndarray Value of the incomplete first order elliptic integral Notes ----- x, y, and z have to be nonnegative and at most one of them is zero. References ---------- .. footbibliography:: """ xn = x.copy() yn = y.copy() zn = z.copy() An = (xn + yn + zn) / 3.0 Q = (3.0 * errtol) ** (-1 / 6.0) * np.max( np.abs([An - xn, An - yn, An - zn]), axis=0 ) # Convergence has to be done voxel by voxel index = ndindex(x.shape) for v in index: n = 0 # Convergence condition while 4.0 ** (-n) * Q[v] > abs(An[v]): xnroot = np.sqrt(xn[v]) ynroot = np.sqrt(yn[v]) znroot = np.sqrt(zn[v]) lamda = xnroot * (ynroot + znroot) + ynroot * znroot n = n + 1 xn[v] = (xn[v] + lamda) * 0.250 yn[v] = (yn[v] + lamda) * 0.250 zn[v] = (zn[v] + lamda) * 0.250 An[v] = (An[v] + lamda) * 0.250 # post convergence calculation X = 1.0 - xn / An Y = 1.0 - yn / An Z = -X - Y E2 = X * Y - Z * Z E3 = X * Y * Z RF = An ** (-1 / 2.0) * ( 1 - E2 / 10.0 + E3 / 14.0 + (E2**2) / 24.0 - 3 / 44.0 * E2 * E3 ) return RF
[docs] @warning_for_keywords() def carlson_rd(x, y, z, *, errtol=1e-4): r"""Compute the Carlson's incomplete elliptic integral of the second kind. Carlson's incomplete elliptic integral of the second kind is defined as :footcite:p:`Carlson1995`: .. math:: R_D = \frac{3}{2} \int_{0}^{\infty} (t+x)^{-\frac{1}{2}} (t+y)^{-\frac{1}{2}}(t+z) ^{-\frac{3}{2}} Parameters ---------- x : ndarray First independent variable of the integral. y : ndarray Second independent variable of the integral. z : ndarray Third independent variable of the integral. errtol : float Error tolerance. Integral is computed with relative error less in magnitude than the defined value Returns ------- RD : ndarray Value of the incomplete second order elliptic integral Notes ----- x, y, and z have to be nonnegative and at most x or y is zero. References ---------- .. footbibliography:: """ xn = x.copy() yn = y.copy() zn = z.copy() A0 = (xn + yn + 3.0 * zn) / 5.0 An = A0.copy() Q = (errtol / 4.0) ** (-1 / 6.0) * np.max( np.abs([An - xn, An - yn, An - zn]), axis=0 ) sum_term = np.zeros(x.shape, dtype=x.dtype) n = np.zeros(x.shape) # Convergence has to be done voxel by voxel index = ndindex(x.shape) for v in index: # Convergence condition while 4.0 ** (-n[v]) * Q[v] > abs(An[v]): xnroot = np.sqrt(xn[v]) ynroot = np.sqrt(yn[v]) znroot = np.sqrt(zn[v]) lamda = xnroot * (ynroot + znroot) + ynroot * znroot sum_term[v] = sum_term[v] + 4.0 ** (-n[v]) / (znroot * (zn[v] + lamda)) n[v] = n[v] + 1 xn[v] = (xn[v] + lamda) * 0.250 yn[v] = (yn[v] + lamda) * 0.250 zn[v] = (zn[v] + lamda) * 0.250 An[v] = (An[v] + lamda) * 0.250 # post convergence calculation X = (A0 - x) / (4.0**n * An) Y = (A0 - y) / (4.0**n * An) Z = -(X + Y) / 3.0 E2 = X * Y - 6.0 * Z * Z E3 = (3.0 * X * Y - 8.0 * Z * Z) * Z E4 = 3.0 * (X * Y - Z * Z) * Z**2.0 E5 = X * Y * Z**3.0 RD = ( 4 ** (-n) * An ** (-3 / 2.0) * ( 1 - 3 / 14.0 * E2 + 1 / 6.0 * E3 + 9 / 88.0 * (E2**2) - 3 / 22.0 * E4 - 9 / 52.0 * E2 * E3 + 3 / 26.0 * E5 ) + 3 * sum_term ) return RD
def _F1m(a, b, c): r""" Helper function that computes function $F_1$ which is required to compute the analytical solution of the Mean kurtosis Parameters ---------- a : ndarray Array containing the values of parameter $\lambda_1$ of function $F_1$ b : ndarray Array containing the values of parameter $\lambda_2$ of function $F_1$ c : ndarray Array containing the values of parameter $\lambda_3$ of function $F_1$ Returns ------- F1 : ndarray Value of the function $F_1$ for all elements of the arrays a, b, and c Notes ----- Function $F_1$ is defined as :footcite:p:`Tabesh2011`: .. math:: F_1(\lambda_1,\lambda_2,\lambda_3)= \frac{(\lambda_1+\lambda_2+\lambda_3)^2} {18(\lambda_1-\lambda_2)(\lambda_1-\lambda_3)} [\frac{\sqrt{\lambda_2\lambda_3}}{\lambda_1} R_F(\frac{\lambda_1}{\lambda_2},\frac{\lambda_1}{\lambda_3},1)+\\ \frac{3\lambda_1^2-\lambda_1\lambda_2-\lambda_2\lambda_3- \lambda_1\lambda_3} {3\lambda_1 \sqrt{\lambda_2 \lambda_3}} R_D(\frac{\lambda_1}{\lambda_2},\frac{\lambda_1}{\lambda_3},1)-1 ] References ---------- .. footbibliography:: """ # Eigenvalues are considered equal if they are not 2.5% different to each # other. This value is adjusted according to the analysis reported in: # https://gsoc2015dipydki.blogspot.com/2015/08/rnh-post-13-start-wrapping-up-test.html er = 2.5e-2 # Initialize F1 F1 = np.zeros(a.shape) # Only computes F1 in voxels that have all eigenvalues larger than zero cond0 = _positive_evals(a, b, c) # Apply formula for non problematic plausible cases, i.e. a!=b and a!=c cond1 = np.logical_and( cond0, np.logical_and(abs(a - b) >= a * er, abs(a - c) >= a * er) ) if np.sum(cond1) != 0: L1 = a[cond1] L2 = b[cond1] L3 = c[cond1] RFm = carlson_rf(L1 / L2, L1 / L3, np.ones(len(L1))) RDm = carlson_rd(L1 / L2, L1 / L3, np.ones(len(L1))) F1[cond1] = ( ((L1 + L2 + L3) ** 2) / (18 * (L1 - L2) * (L1 - L3)) * ( np.sqrt(L2 * L3) / L1 * RFm + (3 * L1**2 - L1 * L2 - L1 * L3 - L2 * L3) / (3 * L1 * np.sqrt(L2 * L3)) * RDm - 1 ) ) # Resolve possible singularity a==b cond2 = np.logical_and( cond0, np.logical_and(abs(a - b) < a * er, abs(a - c) > a * er) ) if np.sum(cond2) != 0: L1 = (a[cond2] + b[cond2]) / 2.0 L3 = c[cond2] F1[cond2] = _F2m(L3, L1, L1) / 2.0 # Resolve possible singularity a==c cond3 = np.logical_and( cond0, np.logical_and(abs(a - c) < a * er, abs(a - b) > a * er) ) if np.sum(cond3) != 0: L1 = (a[cond3] + c[cond3]) / 2.0 L2 = b[cond3] F1[cond3] = _F2m(L2, L1, L1) / 2 # Resolve possible singularity a==b and a==c cond4 = np.logical_and( cond0, np.logical_and(abs(a - c) < a * er, abs(a - b) < a * er) ) if np.sum(cond4) != 0: F1[cond4] = 1 / 5.0 return F1 def _F2m(a, b, c): r""" Helper function that computes function $F_2$ which is required to compute the analytical solution of the Mean kurtosis Parameters ---------- a : ndarray Array containing the values of parameter $\lambda_1$ of function $F_2$ b : ndarray Array containing the values of parameter $\lambda_2$ of function $F_2$ c : ndarray Array containing the values of parameter $\lambda_3$ of function $F_2$ Returns ------- F2 : ndarray Value of the function $F_2$ for all elements of the arrays a, b, and c Notes ----- Function $F_2$ is defined as :footcite:p:`Tabesh2011`: .. math:: F_2(\lambda_1,\lambda_2,\lambda_3)= \frac{(\lambda_1+\lambda_2+\lambda_3)^2} {3(\lambda_2-\lambda_3)^2} [\frac{\lambda_2+\lambda_3}{\sqrt{\lambda_2\lambda_3}} R_F(\frac{\lambda_1}{\lambda_2},\frac{\lambda_1}{\lambda_3},1)+\\ \frac{2\lambda_1-\lambda_2-\lambda_3}{3\sqrt{\lambda_2 \lambda_3}} R_D(\frac{\lambda_1}{\lambda_2},\frac{\lambda_1}{\lambda_3},1)-2] References ---------- .. footbibliography:: """ # Eigenvalues are considered equal if they are not 2.5% different to each # other. This value is adjusted according to the analysis reported in: # https://gsoc2015dipydki.blogspot.com/2015/08/rnh-post-13-start-wrapping-up-test.html er = 2.5e-2 # Initialize F2 F2 = np.zeros(a.shape) # Only computes F2 in voxels that have all eigenvalues larger than zero cond0 = _positive_evals(a, b, c) # Apply formula for non problematic plausible cases, i.e. b!=c cond1 = np.logical_and(cond0, (abs(b - c) > b * er)) if np.sum(cond1) != 0: L1 = a[cond1] L2 = b[cond1] L3 = c[cond1] RF = carlson_rf(L1 / L2, L1 / L3, np.ones(len(L1))) RD = carlson_rd(L1 / L2, L1 / L3, np.ones(len(L1))) F2[cond1] = (((L1 + L2 + L3) ** 2) / (3.0 * (L2 - L3) ** 2)) * ( ((L2 + L3) / (np.sqrt(L2 * L3))) * RF + ((2.0 * L1 - L2 - L3) / (3.0 * np.sqrt(L2 * L3))) * RD - 2.0 ) # Resolve possible singularity b==c cond2 = np.logical_and( cond0, np.logical_and(abs(b - c) < b * er, abs(a - b) > b * er) ) if np.sum(cond2) != 0: L1 = a[cond2] L3 = (c[cond2] + b[cond2]) / 2.0 # Compute alfa :footcite:p:`Tabesh2011` x = 1.0 - (L1 / L3) alpha = np.zeros(len(L1)) for i in range(len(x)): if x[i] > 0: alpha[i] = 1.0 / np.sqrt(x[i]) * np.arctanh(np.sqrt(x[i])) else: alpha[i] = 1.0 / np.sqrt(-x[i]) * np.arctan(np.sqrt(-x[i])) F2[cond2] = ( 6.0 * ((L1 + 2.0 * L3) ** 2) / (144.0 * L3**2 * (L1 - L3) ** 2) * (L3 * (L1 + 2.0 * L3) + L1 * (L1 - 4.0 * L3) * alpha) ) # Resolve possible singularity a==b and a==c cond3 = np.logical_and( cond0, np.logical_and(abs(b - c) < b * er, abs(a - b) < b * er) ) if np.sum(cond3) != 0: F2[cond3] = 6 / 15.0 return F2
[docs] @warning_for_keywords() def directional_diffusion(dt, V, *, min_diffusivity=0): r"""Compute apparent diffusion coefficient (adc). Calculate the apparent diffusion coefficient (adc) in each direction of a sphere for a single voxel :footcite:t:`NetoHenriques2015`. Parameters ---------- dt : array (6,) elements of the diffusion tensor of the voxel. V : array (g, 3) g directions of a Sphere in Cartesian coordinates min_diffusivity : float, optional Because negative eigenvalues are not physical and small eigenvalues cause quite a lot of noise in diffusion-based metrics, diffusivity values smaller than `min_diffusivity` are replaced with `min_diffusivity`. Default = 0 Returns ------- adc : ndarray (g,) Apparent diffusion coefficient (ADC) in all g directions of a sphere for a single voxel. References ---------- .. footbibliography:: """ adc = ( V[:, 0] * V[:, 0] * dt[0] + 2 * V[:, 0] * V[:, 1] * dt[1] + V[:, 1] * V[:, 1] * dt[2] + 2 * V[:, 0] * V[:, 2] * dt[3] + 2 * V[:, 1] * V[:, 2] * dt[4] + V[:, 2] * V[:, 2] * dt[5] ) if min_diffusivity is not None: adc = adc.clip(min=min_diffusivity) return adc
[docs] def directional_diffusion_variance(kt, V): r"""Calculate the apparent diffusion variance (adv) in each direction of a sphere for a single voxel See :footcite:p:`Jensen2005`, :footcite:p:`NetoHenriques2015`, and :footcite:p:`NetoHenriques2021a` for further details about the method. Parameters ---------- kt : array (15,) elements of the kurtosis tensor of the voxel. V : array (g, 3) g directions of a Sphere in Cartesian coordinates. Returns ------- adv : ndarray (g,) Apparent diffusion variance (adv) in all g directions of a sphere for a single voxel. References ---------- .. footbibliography:: """ adv = ( V[:, 0] * V[:, 0] * V[:, 0] * V[:, 0] * kt[0] + V[:, 1] * V[:, 1] * V[:, 1] * V[:, 1] * kt[1] + V[:, 2] * V[:, 2] * V[:, 2] * V[:, 2] * kt[2] + 4 * V[:, 0] * V[:, 0] * V[:, 0] * V[:, 1] * kt[3] + 4 * V[:, 0] * V[:, 0] * V[:, 0] * V[:, 2] * kt[4] + 4 * V[:, 0] * V[:, 1] * V[:, 1] * V[:, 1] * kt[5] + 4 * V[:, 1] * V[:, 1] * V[:, 1] * V[:, 2] * kt[6] + 4 * V[:, 0] * V[:, 2] * V[:, 2] * V[:, 2] * kt[7] + 4 * V[:, 1] * V[:, 2] * V[:, 2] * V[:, 2] * kt[8] + 6 * V[:, 0] * V[:, 0] * V[:, 1] * V[:, 1] * kt[9] + 6 * V[:, 0] * V[:, 0] * V[:, 2] * V[:, 2] * kt[10] + 6 * V[:, 1] * V[:, 1] * V[:, 2] * V[:, 2] * kt[11] + 12 * V[:, 0] * V[:, 0] * V[:, 1] * V[:, 2] * kt[12] + 12 * V[:, 0] * V[:, 1] * V[:, 1] * V[:, 2] * kt[13] + 12 * V[:, 0] * V[:, 1] * V[:, 2] * V[:, 2] * kt[14] ) return adv
[docs] @warning_for_keywords() def directional_kurtosis( dt, md, kt, V, *, min_diffusivity=0, min_kurtosis=-3 / 7, adc=None, adv=None ): r"""Calculate the apparent kurtosis coefficient (akc) in each direction of a sphere for a single voxel. See :footcite:p:`NetoHenriques2015` and :footcite:p:`NetoHenriques2021a` for further details about the method. Parameters ---------- dt : array (6,) elements of the diffusion tensor of the voxel. md : float mean diffusivity of the voxel kt : array (15,) elements of the kurtosis tensor of the voxel. V : array (g, 3) g directions of a Sphere in Cartesian coordinates min_diffusivity : float, optional Because negative eigenvalues are not physical and small eigenvalues cause quite a lot of noise in diffusion-based metrics, diffusivity values smaller than `min_diffusivity` are replaced with `min_diffusivity`. Default = 0 min_kurtosis : float, optional Because high-amplitude negative values of kurtosis are not physically and biologicaly pluasible, and these cause artefacts in kurtosis-based measures, directional kurtosis values smaller than `min_kurtosis` are replaced with `min_kurtosis`. (theoretical kurtosis limit for regions that consist of water confined to spherical pores :footcite:p:`Jensen2005`). adc : ndarray(g,), optional Apparent diffusion coefficient (ADC) in all g directions of a sphere for a single voxel. adv : ndarray(g,), optional Apparent diffusion variance (advc) in all g directions of a sphere for a single voxel. Returns ------- akc : ndarray (g,) Apparent kurtosis coefficient (AKC) in all g directions of a sphere for a single voxel. References ---------- .. footbibliography:: """ if adc is None: adc = directional_diffusion(dt, V, min_diffusivity=min_diffusivity) if adv is None: adv = directional_diffusion_variance(kt, V) akc = adv * (md / adc) ** 2 if min_kurtosis is not None: akc = akc.clip(min=min_kurtosis) return akc
[docs] @warning_for_keywords() def apparent_kurtosis_coef( dki_params, sphere, *, min_diffusivity=0, min_kurtosis=-3.0 / 7 ): r"""Calculate the apparent kurtosis coefficient (AKC) in each direction of a sphere. See :footcite:p:`NetoHenriques2015` and :footcite:p:`NetoHenriques2021a` for further details about the method. Parameters ---------- dki_params : ndarray (x, y, z, 27) or (n, 27) All parameters estimated from the diffusion kurtosis model. Parameters are ordered as follows: 1) Three diffusion tensor's eigenvalues 2) Three lines of the eigenvector matrix each containing the first, second and third coordinates of the eigenvectors respectively 3) Fifteen elements of the kurtosis tensor sphere : a Sphere class instance The AKC will be calculated for each of the vertices in the sphere min_diffusivity : float, optional Because negative eigenvalues are not physical and small eigenvalues cause quite a lot of noise in diffusion-based metrics, diffusivity values smaller than `min_diffusivity` are replaced with `min_diffusivity`. min_kurtosis : float, optional Because high-amplitude negative values of kurtosis are not physically and biologicaly pluasible, and these cause artefacts in kurtosis-based measures, directional kurtosis values smaller than `min_kurtosis` are replaced with `min_kurtosis`. Default = -3./7 (theoretical kurtosis limit for regions that consist of water confined to spherical pores :footcite:p:`Jensen2005`). Returns ------- akc : ndarray (x, y, z, g) or (n, g) Apparent kurtosis coefficient (AKC) for all g directions of a sphere. Notes ----- For each sphere direction with coordinates $(n_{1}, n_{2}, n_{3})$, the calculation of AKC is done using formula :footcite:p:`NetoHenriques2015`: .. math:: AKC(n)=\frac{MD^{2}}{ADC(n)^{2}}\sum_{i=1}^{3}\sum_{j=1}^{3} \sum_{k=1}^{3}\sum_{l=1}^{3}n_{i}n_{j}n_{k}n_{l}W_{ijkl} where $W_{ijkl}$ are the elements of the kurtosis tensor, MD the mean diffusivity and ADC the apparent diffusion coefficient computed as: .. math:: ADC(n)=\sum_{i=1}^{3}\sum_{j=1}^{3}n_{i}n_{j}D_{ij} where $D_{ij}$ are the elements of the diffusion tensor. References ---------- .. footbibliography:: """ # Flat parameters outshape = dki_params.shape[:-1] dki_params = dki_params.reshape((-1, dki_params.shape[-1])) # Split data evals, evecs, kt = split_dki_param(dki_params) # Initialize AKC matrix V = sphere.vertices akc = np.zeros((len(kt), len(V))) # select relevant voxels to process rel_i = _positive_evals(evals[..., 0], evals[..., 1], evals[..., 2]) kt = kt[rel_i] evecs = evecs[rel_i] evals = evals[rel_i] akci = akc[rel_i] # Compute MD and DT md = mean_diffusivity(evals) dt = lower_triangular(vec_val_vect(evecs, evals)) # loop over all relevant voxels for vox in range(len(kt)): akci[vox] = directional_kurtosis( dt[vox], md[vox], kt[vox], V, min_diffusivity=min_diffusivity, min_kurtosis=min_kurtosis, ) # reshape data according to input data akc[rel_i] = akci return akc.reshape((outshape + (len(V),)))
[docs] @warning_for_keywords() def mean_kurtosis( dki_params, *, min_kurtosis=-3.0 / 7, max_kurtosis=3, analytical=True ): r""" Compute mean kurtosis (MK) from the kurtosis tensor. See :footcite:p:`Tabesh2011` and :footcite:p:`NetoHenriques2021a` for further details about the method. Parameters ---------- dki_params : ndarray (x, y, z, 27) or (n, 27) All parameters estimated from the diffusion kurtosis model. Parameters are ordered as follows: 1) Three diffusion tensor's eigenvalues 2) Three lines of the eigenvector matrix each containing the first, second and third coordinates of the eigenvector 3) Fifteen elements of the kurtosis tensor min_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, mean kurtosis values that are smaller than `min_kurtosis` are replaced with `min_kurtosis`. Default = -3./7 (theoretical kurtosis limit for regions that consist of water confined to spherical pores :footcite:p:`Jensen2005`). max_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, mean kurtosis values that are larger than `max_kurtosis` are replaced with `max_kurtosis`. analytical : bool, optional If True, MK is calculated using its analytical solution, otherwise an exact numerical estimator is used (see Notes). Returns ------- mk : array Calculated MK. Notes ----- The MK is defined as the average of directional kurtosis coefficients across all spatial directions, which can be formulated by the following surface integral :footcite:p:`Tabesh2011`, :footcite:p:`NetoHenriques2021a`: .. math:: MK \equiv \frac{1}{4\pi} \int d\Omega_\mathbf{n} K(\mathbf{n}) This integral can be numerically solved by averaging directional kurtosis values sampled for directions of a spherical t-design :footcite:p:`Hardin1996`. Alternatively, MK can be solved from the analytical solution derived by :footcite:p:`Tabesh2011`, :footcite:p:`NetoHenriques2021a`. This solution is given by: .. math:: MK=F_1(\lambda_1,\lambda_2,\lambda_3)\hat{W}_{1111}+ F_1(\lambda_2,\lambda_1,\lambda_3)\hat{W}_{2222}+ F_1(\lambda_3,\lambda_2,\lambda_1)\hat{W}_{3333}+ \\ F_2(\lambda_1,\lambda_2,\lambda_3)\hat{W}_{2233}+ F_2(\lambda_2,\lambda_1,\lambda_3)\hat{W}_{1133}+ F_2(\lambda_3,\lambda_2,\lambda_1)\hat{W}_{1122} where $\hat{W}_{ijkl}$ are the components of the $W$ tensor in the coordinates system defined by the eigenvectors of the diffusion tensor $\mathbf{D}$ and .. math:: F_1(\lambda_1,\lambda_2,\lambda_3)= \frac{(\lambda_1+\lambda_2+\lambda_3)^2} {18(\lambda_1-\lambda_2)(\lambda_1-\lambda_3)} [\frac{\sqrt{\lambda_2\lambda_3}}{\lambda_1} R_F(\frac{\lambda_1}{\lambda_2},\frac{\lambda_1}{\lambda_3},1)+\\ \frac{3\lambda_1^2-\lambda_1\lambda_2-\lambda_2\lambda_3- \lambda_1\lambda_3} {3\lambda_1 \sqrt{\lambda_2 \lambda_3}} R_D(\frac{\lambda_1}{\lambda_2},\frac{\lambda_1}{\lambda_3},1)-1 ] F_2(\lambda_1,\lambda_2,\lambda_3)= \frac{(\lambda_1+\lambda_2+\lambda_3)^2} {3(\lambda_2-\lambda_3)^2} [\frac{\lambda_2+\lambda_3}{\sqrt{\lambda_2\lambda_3}} R_F(\frac{\lambda_1}{\lambda_2},\frac{\lambda_1}{\lambda_3},1)+\\ \frac{2\lambda_1-\lambda_2-\lambda_3}{3\sqrt{\lambda_2 \lambda_3}} R_D(\frac{\lambda_1}{\lambda_2},\frac{\lambda_1}{\lambda_3},1)-2] where $R_f$ and $R_d$ are the Carlson's elliptic integrals. References ---------- .. footbibliography:: """ # Flat parameters. For numpy versions more recent than 1.6.0, this step # isn't required outshape = dki_params.shape[:-1] dki_params = dki_params.reshape((-1, dki_params.shape[-1])) if analytical: # Split the model parameters to three variable containing the evals, # evecs, and kurtosis elements evals, evecs, kt = split_dki_param(dki_params) # Rotate the kurtosis tensor from the standard Cartesian coordinate # system to another coordinate system in which the 3 orthonormal # eigenvectors of DT are the base coordinate Wxxxx = Wrotate_element(kt, 0, 0, 0, 0, evecs) Wyyyy = Wrotate_element(kt, 1, 1, 1, 1, evecs) Wzzzz = Wrotate_element(kt, 2, 2, 2, 2, evecs) Wxxyy = Wrotate_element(kt, 0, 0, 1, 1, evecs) Wxxzz = Wrotate_element(kt, 0, 0, 2, 2, evecs) Wyyzz = Wrotate_element(kt, 1, 1, 2, 2, evecs) # Compute MK MK = ( _F1m(evals[..., 0], evals[..., 1], evals[..., 2]) * Wxxxx + _F1m(evals[..., 1], evals[..., 0], evals[..., 2]) * Wyyyy + _F1m(evals[..., 2], evals[..., 1], evals[..., 0]) * Wzzzz + _F2m(evals[..., 0], evals[..., 1], evals[..., 2]) * Wyyzz + _F2m(evals[..., 1], evals[..., 0], evals[..., 2]) * Wxxzz + _F2m(evals[..., 2], evals[..., 1], evals[..., 0]) * Wxxyy ) else: # Numerical Solution using t-design of 45 directions V = np.loadtxt(get_fnames(name="t-design")) sph = dps.Sphere(xyz=V) KV = apparent_kurtosis_coef(dki_params, sph, min_kurtosis=min_kurtosis) MK = np.mean(KV, axis=-1) if min_kurtosis is not None: MK = MK.clip(min=min_kurtosis) if max_kurtosis is not None: MK = MK.clip(max=max_kurtosis) return MK.reshape(outshape)
def _G1m(a, b, c): r"""Helper function that computes function $G_1$ which is required to compute the analytical solution of the Radial kurtosis Parameters ---------- a : ndarray Array containing the values of parameter $\lambda_1$ of function $G_1$ b : ndarray Array containing the values of parameter $\lambda_2$ of function $G_1$ c : ndarray Array containing the values of parameter $\lambda_3$ of function $G_1$ Returns ------- G1 : ndarray Value of the function $G_1$ for all elements of the arrays a, b, and c Notes ----- Function $G_1$ is defined as :footcite:p:`Tabesh2011`: .. math:: G_1(\lambda_1,\lambda_2,\lambda_3)= \frac{(\lambda_1+\lambda_2+\lambda_3)^2}{18\lambda_2(\lambda_2- \lambda_3)} \left (2\lambda_2 + \frac{\lambda_3^2-3\lambda_2\lambda_3}{\sqrt{\lambda_2\lambda_3}} \right) References ---------- .. footbibliography:: """ # Float error used to compare two floats, abs(l1 - l2) < er for l1 = l2 # Error is defined as five orders of magnitude larger than system's epslon er = np.finfo(a.ravel()[0]).eps * 1e5 # Initialize G1 G1 = np.zeros(a.shape) # Only computes G1 in voxels that have all eigenvalues larger than zero cond0 = _positive_evals(a, b, c) # Apply formula for non problematic plausible cases, i.e. b!=c cond1 = np.logical_and(cond0, (abs(b - c) > er)) if np.sum(cond1) != 0: L1 = a[cond1] L2 = b[cond1] L3 = c[cond1] G1[cond1] = ( (L1 + L2 + L3) ** 2 / (18 * L2 * (L2 - L3) ** 2) * (2.0 * L2 + (L3**2 - 3 * L2 * L3) / np.sqrt(L2 * L3)) ) # Resolve possible singularity b==c cond2 = np.logical_and(cond0, abs(b - c) < er) if np.sum(cond2) != 0: L1 = a[cond2] L2 = b[cond2] G1[cond2] = (L1 + 2.0 * L2) ** 2 / (24.0 * L2**2) return G1 def _G2m(a, b, c): r"""Helper function that computes function $G_2$ which is required to compute the analytical solution of the Radial kurtosis Parameters ---------- a : ndarray Array containing the values of parameter $\lambda_1$ of function $G_2$ b : ndarray Array containing the values of parameter $\lambda_2$ of function $G_2$ c : ndarray (n,) Array containing the values of parameter $\lambda_3$ of function $G_2$ Returns ------- G2 : ndarray Value of the function $G_2$ for all elements of the arrays a, b, and c Notes ----- Function $G_2$ is defined as :footcite:p:`Tabesh2011`: .. math:: G_2(\lambda_1,\lambda_2,\lambda_3)= \frac{(\lambda_1+\lambda_2+\lambda_3)^2}{(\lambda_2-\lambda_3)^2} \left ( \frac{\lambda_2+\lambda_3}{\sqrt{\lambda_2\lambda_3}}-2\right ) References ---------- .. footbibliography:: """ # Float error used to compare two floats, abs(l1 - l2) < er for l1 = l2 # Error is defined as five order of magnitude larger than system's epsilon er = np.finfo(a.ravel()[0]).eps * 1e5 # Initialize G2 G2 = np.zeros(a.shape) # Only computes G2 in voxels that have all eigenvalues larger than zero cond0 = _positive_evals(a, b, c) # Apply formula for non problematic plausible cases, i.e. b!=c cond1 = np.logical_and(cond0, (abs(b - c) > er)) if np.sum(cond1) != 0: L1 = a[cond1] L2 = b[cond1] L3 = c[cond1] G2[cond1] = ( (L1 + L2 + L3) ** 2 / (3 * (L2 - L3) ** 2) * ((L2 + L3) / np.sqrt(L2 * L3) - 2) ) # Resolve possible singularity b==c cond2 = np.logical_and(cond0, abs(b - c) < er) if np.sum(cond2) != 0: L1 = a[cond2] L2 = b[cond2] G2[cond2] = (L1 + 2.0 * L2) ** 2 / (12.0 * L2**2) return G2
[docs] @warning_for_keywords() def radial_kurtosis( dki_params, *, min_kurtosis=-3.0 / 7, max_kurtosis=10, analytical=True ): r"""Compute radial kurtosis (RK) of a diffusion kurtosis tensor. See :footcite:p:`Tabesh2011` and :footcite:p:`NetoHenriques2021a` for further details about the method. Parameters ---------- dki_params : ndarray (x, y, z, 27) or (n, 27) All parameters estimated from the diffusion kurtosis model. Parameters are ordered as follows: 1) Three diffusion tensor's eigenvalues 2) Three lines of the eigenvector matrix each containing the first, second and third coordinates of the eigenvector 3) Fifteen elements of the kurtosis tensor min_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, radial kurtosis values that are smaller than `min_kurtosis` are replaced with `min_kurtosis`. Default = -3./7 (theoretical kurtosis limit for regions that consist of water confined to spherical pores :footcite:p:`Jensen2005`). max_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, radial kurtosis values that are larger than `max_kurtosis` are replaced with `max_kurtosis`. analytical : bool, optional If True, RK is calculated using its analytical solution, otherwise an exact numerical estimator is used (see Notes). Default is set to True. Returns ------- rk : array Calculated RK. Notes ----- RK is defined as the average of the directional kurtosis perpendicular to the fiber's main direction e1 :footcite:p:`Tabesh2011`, :footcite:p:`NetoHenriques2021a`: .. math:: RK \equiv \frac{1}{2\pi} \int d\Omega _\mathbf{\theta} K(\mathbf{\theta}) \delta (\mathbf{\theta}\cdot \mathbf{e}_1) This equation can be numerically computed by averaging apparent directional kurtosis samples for directions perpendicular to e1 :footcite:p:`NetoHenriques2021a`. Otherwise, RK can be calculated from its analytical solution :footcite:p:`Tabesh2011`: .. math:: K_{\bot} = G_1(\lambda_1,\lambda_2,\lambda_3)\hat{W}_{2222} + G_1(\lambda_1,\lambda_3,\lambda_2)\hat{W}_{3333} + G_2(\lambda_1,\lambda_2,\lambda_3)\hat{W}_{2233} where: .. math:: G_1(\lambda_1,\lambda_2,\lambda_3)= \frac{(\lambda_1+\lambda_2+\lambda_3)^2}{18\lambda_2(\lambda_2- \lambda_3)} \left (2\lambda_2 + \frac{\lambda_3^2-3\lambda_2\lambda_3}{\sqrt{\lambda_2\lambda_3}} \right) and .. math:: G_2(\lambda_1,\lambda_2,\lambda_3)= \frac{(\lambda_1+\lambda_2+\lambda_3)^2}{(\lambda_2-\lambda_3)^2} \left ( \frac{\lambda_2+\lambda_3}{\sqrt{\lambda_2\lambda_3}}-2\right ) References ---------- .. footbibliography:: """ outshape = dki_params.shape[:-1] dki_params = dki_params.reshape((-1, dki_params.shape[-1])) # Split the model parameters to three variable containing the evals, # evecs, and kurtosis elements evals, evecs, kt = split_dki_param(dki_params) if analytical: # Rotate the kurtosis tensor from the standard Cartesian coordinate # system to another coordinate system in which the 3 orthonormal # eigenvectors of DT are the base coordinate Wyyyy = Wrotate_element(kt, 1, 1, 1, 1, evecs) Wzzzz = Wrotate_element(kt, 2, 2, 2, 2, evecs) Wyyzz = Wrotate_element(kt, 1, 1, 2, 2, evecs) # Compute RK RK = ( _G1m(evals[..., 0], evals[..., 1], evals[..., 2]) * Wyyyy + _G1m(evals[..., 0], evals[..., 2], evals[..., 1]) * Wzzzz + _G2m(evals[..., 0], evals[..., 1], evals[..., 2]) * Wyyzz ) else: # Numerical Solution using 10 perpendicular directions samples npa = 10 # Initialize RK RK = np.zeros(kt.shape[:-1]) # select relevant voxels to process rel_i = _positive_evals(evals[..., 0], evals[..., 1], evals[..., 2]) dki_params = dki_params[rel_i] evecs = evecs[rel_i] RKi = RK[rel_i] # loop over all voxels KV = np.zeros((dki_params.shape[0], npa)) for vox in range(len(dki_params)): V = perpendicular_directions(np.array(evecs[vox, :, 0]), num=npa, half=True) sph = dps.Sphere(xyz=V) KV[vox, :] = apparent_kurtosis_coef( dki_params[vox], sph, min_kurtosis=min_kurtosis ) RKi = np.mean(KV, axis=-1) RK[rel_i] = RKi if min_kurtosis is not None: RK = RK.clip(min=min_kurtosis) if max_kurtosis is not None: RK = RK.clip(max=max_kurtosis) return RK.reshape(outshape)
[docs] @warning_for_keywords() def axial_kurtosis( dki_params, *, min_kurtosis=-3.0 / 7, max_kurtosis=10, analytical=True ): r"""Compute axial kurtosis (AK) from the kurtosis tensor. See :footcite:p:`Tabesh2011` and :footcite:`NetoHenriques2021a` for further details about the method. Parameters ---------- dki_params : ndarray (x, y, z, 27) or (n, 27) All parameters estimated from the diffusion kurtosis model. Parameters are ordered as follows: 1) Three diffusion tensor's eigenvalues 2) Three lines of the eigenvector matrix each containing the first, second and third coordinates of the eigenvector 3) Fifteen elements of the kurtosis tensor min_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, axial kurtosis values that are smaller than `min_kurtosis` are replaced with `min_kurtosis`. Default = -3./7 (theoretical kurtosis limit for regions that consist of water confined to spherical pores :footcite:p:`Jensen2005`). max_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, axial kurtosis values that are larger than `max_kurtosis` are replaced with `max_kurtosis`. analytical : bool, optional If True, AK is calculated from rotated diffusion kurtosis tensor, otherwise it will be computed from the apparent diffusion kurtosis values along the principal axis of the diffusion tensor (see notes). Returns ------- ak : array Calculated AK. Notes ----- AK is defined as the directional kurtosis parallel to the fiber's main direction e1 :footcite:p:`Tabesh2011`, :footcite:`NetoHenriques2021a`. You can compute AK using to approaches: 1) AK is calculated from rotated diffusion kurtosis tensor :footcite:p:`Tabesh2011`, i.e.: .. math:: AK = \hat{W}_{1111} \frac{(\lambda_{1}+\lambda_{2}+\lambda_{3})^2}{(9 \lambda_{1}^2)} 2) AK can be sampled from the principal axis of the diffusion tensor :footcite:`NetoHenriques2021a`: .. math:: AK = K(\mathbf{e}_1) Although both approaches leads to an exact calculation of AK, the first approach will be referred to as the analytical method while the second approach will be referred to as the numerical method based on their analogy to the estimation strategies for MK and RK. References ---------- .. footbibliography:: """ # Flat parameters outshape = dki_params.shape[:-1] dki_params = dki_params.reshape((-1, dki_params.shape[-1])) # Split data evals, evecs, kt = split_dki_param(dki_params) # Initialize AK AK = np.zeros(kt.shape[:-1]) # select relevant voxels to process rel_i = _positive_evals(evals[..., 0], evals[..., 1], evals[..., 2]) kt = kt[rel_i] evecs = evecs[rel_i] evals = evals[rel_i] AKi = AK[rel_i] # Compute mean diffusivity md = mean_diffusivity(evals) if analytical: # Rotate the kurtosis tensor from the standard Cartesian coordinate # system to another coordinate system in which the 3 orthonormal # eigenvectors of DT are the base coordinate Wxxxx = Wrotate_element(kt, 0, 0, 0, 0, evecs) AKi = Wxxxx * (md**2) / (evals[..., 0] ** 2) else: # Compute apparent directional kurtosis along evecs[0] dt = lower_triangular(vec_val_vect(evecs, evals)) for vox in range(len(kt)): AKi[vox] = directional_kurtosis( dt[vox], md[vox], kt[vox], np.array([evecs[vox, :, 0]]) ).item() # reshape data according to input data AK[rel_i] = AKi if min_kurtosis is not None: AK = AK.clip(min=min_kurtosis) if max_kurtosis is not None: AK = AK.clip(max=max_kurtosis) return AK.reshape(outshape)
def _kt_maximum_converge(ang, dt, md, kt): """Helper function that computes the inverse of the directional kurtosis of a voxel along a given direction in polar coordinates Parameters ---------- ang : array (2,) array containing the two polar angles dt : array (6,) elements of the diffusion tensor of the voxel. md : float mean diffusivity of the voxel kt : array (15,) elements of the kurtosis tensor of the voxel. Returns ------- neg_kt : float The inverse value of the apparent kurtosis for the given direction. Notes ----- This function is used to refine the kurtosis maximum estimate See Also -------- dipy.reconst.dki.kurtosis_maximum """ n = np.array([sphere2cart(1, ang[0], ang[1])]) return -1.0 * directional_kurtosis(dt, md, kt, n) @warning_for_keywords() def _voxel_kurtosis_maximum(dt, md, kt, sphere, *, gtol=1e-2): """Compute the maximum value of a single voxel kurtosis tensor Parameters ---------- dt : array (6,) elements of the diffusion tensor of the voxel. md : float mean diffusivity of the voxel kt : array (15,) elements of the kurtosis tensor of the voxel. sphere : Sphere class instance, optional The sphere providing sample directions for the initial search of the maximum value of kurtosis. gtol : float, optional This input is to refine kurtosis maximum under the precision of the directions sampled on the sphere class instance. The gradient of the convergence procedure must be less than gtol before successful termination. If gtol is None, fiber direction is directly taken from the initial sampled directions of the given sphere object Returns ------- max_value : float kurtosis tensor maximum value max_dir : array (3,) Cartesian coordinates of the direction of the maximal kurtosis value """ # Estimation of maximum kurtosis candidates akc = directional_kurtosis(dt, md, kt, sphere.vertices) max_val, ind = local_maxima(akc, sphere.edges) n = len(max_val) # case that none maximum was find (spherical or null kurtosis tensors) if n == 0: return np.mean(akc), np.zeros(3) max_dir = sphere.vertices[ind] # Select the maximum from the candidates max_value = max(max_val) max_direction = max_dir[np.argmax(max_val.argmax)] # refine maximum direction if gtol is not None: for p in range(n): r, theta, phi = cart2sphere(max_dir[p, 0], max_dir[p, 1], max_dir[p, 2]) ang = np.array([theta, phi]) ang[:] = opt.fmin_bfgs( _kt_maximum_converge, ang, args=(dt, md, kt), disp=False, retall=False, gtol=gtol, ) k_dir = np.array([sphere2cart(1.0, ang[0], ang[1])]) k_val = directional_kurtosis(dt, md, kt, k_dir) if k_val > max_value: max_value = k_val max_direction = k_dir return max_value.item(), max_direction
[docs] @warning_for_keywords() def kurtosis_maximum(dki_params, *, sphere="repulsion100", gtol=1e-2, mask=None): """Compute kurtosis maximum value. See :footcite:`NetoHenriques2021a` for further details about the method. Parameters ---------- dki_params : ndarray (x, y, z, 27) or (n, 27) All parameters estimated from the diffusion kurtosis model. Parameters are ordered as follows: 1) Three diffusion tensor's eingenvalues 2) Three lines of the eigenvector matrix each containing the first, second and third coordinates of the eigenvector 3) Fifteen elements of the kurtosis tensor sphere : Sphere class instance, optional The sphere providing sample directions for the initial search of the maximal value of kurtosis. gtol : float, optional This input is to refine kurtosis maximum under the precision of the directions sampled on the sphere class instance. The gradient of the convergence procedure must be less than gtol before successful termination. If gtol is None, fiber direction is directly taken from the initial sampled directions of the given sphere object mask : ndarray A boolean array used to mark the coordinates in the data that should be analyzed that has the shape dki_params.shape[:-1] Returns ------- max_value : float kurtosis tensor maximum value max_dir : array (3,) Cartesian coordinates of the direction of the maximal kurtosis value References ---------- .. footbibliography:: """ shape = dki_params.shape[:-1] # load gradient directions if not isinstance(sphere, dps.Sphere): sphere = get_sphere(name="repulsion100") # select voxels where to find fiber directions if mask is None: mask = np.ones(shape, dtype="bool") else: if mask.shape != shape: raise ValueError("Mask is not the same shape as dki_params.") evals, evecs, kt = split_dki_param(dki_params) # select non-zero voxels pos_evals = _positive_evals(evals[..., 0], evals[..., 1], evals[..., 2]) mask = np.logical_and(mask, pos_evals) kt_max = np.zeros(mask.shape) dt = lower_triangular(vec_val_vect(evecs, evals)) md = mean_diffusivity(evals) for idx in ndindex(shape): if not mask[idx]: continue kt_max[idx], da = _voxel_kurtosis_maximum( dt[idx], md[idx], kt[idx], sphere, gtol=gtol ) return kt_max
[docs] @warning_for_keywords() def mean_kurtosis_tensor(dki_params, *, min_kurtosis=-3.0 / 7, max_kurtosis=10): r"""Compute mean of the kurtosis tensor (MKT). See :footcite:p:`Hansen2013` for further details about the method. Parameters ---------- dki_params : ndarray (x, y, z, 27) or (n, 27) All parameters estimated from the diffusion kurtosis model. Parameters are ordered as follows: 1) Three diffusion tensor's eigenvalues 2) Three lines of the eigenvector matrix each containing the first, second and third coordinates of the eigenvector 3) Fifteen elements of the kurtosis tensor min_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, mean kurtosis values that are smaller than `min_kurtosis` are replaced with `min_kurtosis`. Default = -3./7 (theoretical kurtosis limit for regions that consist of water confined to spherical pores :footcite:p:`Jensen2005`). max_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, mean kurtosis values that are larger than `max_kurtosis` are replaced with `max_kurtosis`. Returns ------- mkt : array Calculated mean kurtosis tensor. Notes ----- The MKT is defined as :footcite:p:`Hansen2013`: .. math:: MKT \equiv \frac{1}{4\pi} \int d \Omega_{\mathnbf{n}} n_i n_j n_k n_l W_{ijkl} which can be directly computed from the trace of the kurtosis tensor: .. math:: MKT = \frac{1}{5} Tr(\mathbf{W}) = \frac{1}{5} (W_{1111} + W_{2222} + W_{3333} + 2W_{1122} + 2W_{1133} + 2W_{2233}) References ---------- .. footbibliography:: """ MKT = ( 1 / 5 * ( dki_params[..., 12] + dki_params[..., 13] + dki_params[..., 14] + 2 * dki_params[..., 21] + 2 * dki_params[..., 22] + 2 * dki_params[..., 23] ) ) if min_kurtosis is not None: MKT = MKT.clip(min=min_kurtosis) if max_kurtosis is not None: MKT = MKT.clip(max=max_kurtosis) return MKT
[docs] def radial_tensor_kurtosis(dki_params, *, min_kurtosis=-3.0 / 7, max_kurtosis=10): r"""Compute the rescaled radial tensor kurtosis (RTK). See :footcite:p:`Hansen2013` for further details about the method. Parameters ---------- dki_params : ndarray (x, y, z, 27) or (n, 27) All parameters estimated from the diffusion kurtosis model. Parameters are ordered as follows: 1) Three diffusion tensor's eigenvalues 2) Three lines of the eigenvector matrix each containing the first, second and third coordinates of the eigenvector 3) Fifteen elements of the kurtosis tensor min_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, radial kurtosis values that are smaller than `min_kurtosis` are replaced with `min_kurtosis`. max_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, radial kurtosis values that are larger than `max_kurtosis` are replaced with `max_kurtosis`. Returns ------- rtk : array Calculated rescaled radial tensor kurtosis (RTK). Notes ----- Rescaled radial tensor kurtosis (RTK) is defined as :footcite:p:`Hansen2013`: .. math:: RKT = \frac{3}{8} \frac{MD^2}{RD^2} (W_{2222} + W_{3333} + 2*W_{2233}) where W is the kurtosis tensor rotated to a coordinate system in which the 3 orthonormal eigenvectors of DT are the base coordinate, MD is the mean diffusivity, and RD is the radial diffusivity. References ---------- .. footbibliography:: """ outshape = dki_params.shape[:-1] dki_params = dki_params.reshape((-1, dki_params.shape[-1])) # Split the model parameters to three variable containing the evals, # evecs, and kurtosis elements evals, evecs, kt = split_dki_param(dki_params) # Initializes RKT RTK = np.zeros(kt.shape[:-1]) # select relevant voxels to process rel_i = _positive_evals(evals[..., 0], evals[..., 1], evals[..., 2]) kt = kt[rel_i] evecs = evecs[rel_i] evals = evals[rel_i] # Rotate the kurtosis tensor from the standard Cartesian coordinate # system to another coordinate system in which the 3 orthonormal # eigenvectors of DT are the base coordinate Wyyyy = Wrotate_element(kt, 1, 1, 1, 1, evecs) Wzzzz = Wrotate_element(kt, 2, 2, 2, 2, evecs) Wyyzz = Wrotate_element(kt, 1, 1, 2, 2, evecs) # Compute radial kurtois tensor WTK = 3 / 8 * (Wyyyy + Wzzzz + 2 * Wyyzz) # Rescaling radial kurtosis tensor md = mean_diffusivity(evals) rd = radial_diffusivity(evals) RTK[rel_i] = WTK * md**2 / rd**2 if min_kurtosis is not None: RTK = RTK.clip(min=min_kurtosis) if max_kurtosis is not None: RTK = RTK.clip(max=max_kurtosis) return RTK.reshape(outshape)
[docs] def kurtosis_fractional_anisotropy(dki_params): r"""Compute the anisotropy of the kurtosis tensor (KFA). See :footcite:p:`Glenn2015` and :footcite:p:`NetoHenriques2021a` for further details about the method. Parameters ---------- dki_params : ndarray (x, y, z, 27) or (n, 27) All parameters estimated from the diffusion kurtosis model. Parameters are ordered as follows: 1) Three diffusion tensor's eigenvalues 2) Three lines of the eigenvector matrix each containing the first, second and third coordinates of the eigenvector 3) Fifteen elements of the kurtosis tensor Returns ------- kfa : array Calculated mean kurtosis tensor. Notes ----- The KFA is defined as :footcite:p:`Glenn2015`: .. math:: KFA \equiv \frac{||\mathbf{W} - MKT \mathbf{I}^{(4)}||_F}{||\mathbf{W}||_F} where $W$ is the kurtosis tensor, MKT the kurtosis tensor mean, $I^{(4)}$ is the fully symmetric rank 2 isotropic tensor and $||...||_F$ is the tensor's Frobenius norm :footcite:p:`Glenn2015`. References ---------- .. footbibliography:: """ Wxxxx = dki_params[..., 12] Wyyyy = dki_params[..., 13] Wzzzz = dki_params[..., 14] Wxxxy = dki_params[..., 15] Wxxxz = dki_params[..., 16] Wxyyy = dki_params[..., 17] Wyyyz = dki_params[..., 18] Wxzzz = dki_params[..., 19] Wyzzz = dki_params[..., 20] Wxxyy = dki_params[..., 21] Wxxzz = dki_params[..., 22] Wyyzz = dki_params[..., 23] Wxxyz = dki_params[..., 24] Wxyyz = dki_params[..., 25] Wxyzz = dki_params[..., 26] W = 1.0 / 5.0 * (Wxxxx + Wyyyy + Wzzzz + 2 * Wxxyy + 2 * Wxxzz + 2 * Wyyzz) # Compute's equation numerator A = ( (Wxxxx - W) ** 2 + (Wyyyy - W) ** 2 + (Wzzzz - W) ** 2 + 4 * (Wxxxy**2 + Wxxxz**2 + Wxyyy**2 + Wyyyz**2 + Wxzzz**2 + Wyzzz**2) + 6 * ((Wxxyy - W / 3) ** 2 + (Wxxzz - W / 3) ** 2 + (Wyyzz - W / 3) ** 2) + 12 * (Wxxyz**2 + Wxyyz**2 + Wxyzz**2) ) # Compute's equation denominator B = ( Wxxxx**2 + Wyyyy**2 + Wzzzz**2 + 4 * (Wxxxy**2 + Wxxxz**2 + Wxyyy**2 + Wyyyz**2 + Wxzzz**2 + Wyzzz**2) + 6 * (Wxxyy**2 + Wxxzz**2 + Wyyzz**2) + 12 * (Wxxyz**2 + Wxyyz**2 + Wxyzz**2) ) # Compute KFA KFA = np.zeros(A.shape) cond1 = B > 0 # Avoiding Singularity (if B = 0, KFA = 0) # Avoiding overestimating KFA for small MKT values (KFA=0, MKT < tol) cond2 = W > 1e-8 cond = np.logical_and(cond1, cond2) KFA[cond] = np.sqrt(A[cond] / B[cond]) return KFA
[docs] @warning_for_keywords() def dki_prediction(dki_params, gtab, *, S0=1.0): r"""Predict a signal given diffusion kurtosis imaging parameters The predicted signal is given by: .. math:: S=S_{0}e^{-bD+\frac{1}{6}b^{2}D^{2}K} See :footcite:p:`Jensen2005` and :footcite:p:`NetoHenriques2021a` for further details about the method. Parameters ---------- dki_params : ndarray (x, y, z, 27) or (n, 27) All parameters estimated from the diffusion kurtosis model. Parameters are ordered as follows: 1) Three diffusion tensor's eigenvalues 2) Three lines of the eigenvector matrix each containing the first, second and third coordinates of the eigenvector 3) Fifteen elements of the kurtosis tensor gtab : a GradientTable class instance The gradient table for this prediction S0 : float or ndarray, optional The non diffusion-weighted signal in every voxel, or across all voxels. Default: 1 Returns ------- pred_sig : (..., N) ndarray Simulated signal based on the DKI model. References ---------- .. footbibliography:: """ evals, evecs, kt = split_dki_param(dki_params) # Define DKI design matrix according to given gtab A = design_matrix(gtab) # Flat parameters and initialize pred_sig fevals = evals.reshape((-1, evals.shape[-1])) fevecs = evecs.reshape((-1,) + evecs.shape[-2:]) fkt = kt.reshape((-1, kt.shape[-1])) pred_sig = np.zeros((len(fevals), len(gtab.bvals))) if isinstance(S0, np.ndarray): S0_vol = np.reshape(S0, (len(fevals))) else: S0_vol = S0 # looping for all voxels for v in range(len(pred_sig)): DT = np.dot(np.dot(fevecs[v], np.diag(fevals[v])), fevecs[v].T) dt = lower_triangular(DT) MD = (dt[0] + dt[2] + dt[5]) / 3 if isinstance(S0_vol, np.ndarray): this_S0 = S0_vol[v] else: this_S0 = S0_vol X = np.concatenate((dt, fkt[v] * MD * MD, np.array([-np.log(this_S0)])), axis=0) pred_sig[v] = np.exp(np.dot(A, X)) # Reshape data according to the shape of dki_params pred_sig = pred_sig.reshape(dki_params.shape[:-1] + (pred_sig.shape[-1],)) return pred_sig
[docs] class DiffusionKurtosisModel(ReconstModel): """Class for the Diffusion Kurtosis Model""" @warning_for_keywords() def __init__(self, gtab, *args, fit_method="WLS", return_S0_hat=False, **kwargs): """Diffusion Kurtosis Tensor Model. See :footcite:p:`Jensen2005` and :footcite:p:`NetoHenriques2021a` for further details about the model. Parameters ---------- gtab : GradientTable instance The gradient table for the data set. fit_method : str or callable, optional str be one of the following: - 'OLS' or 'ULLS' for ordinary least squares. - 'WLS', 'WLLS' or 'UWLLS' for weighted ordinary least squares. See func:`dki.ls_fit_dki`. - 'CLS' for LMI constrained ordinary least squares :footcite:p:`DelaHaije2020`. - 'CWLS' for LMI constrained weighted least squares :footcite:p:`DelaHaije2020`. See func:`dki.cls_fit_dki`. callable has to have the signature: ``fit_method(design_matrix, data, *args, **kwargs)`` return_S0_hat : bool Boolean to return (True) or not (False) the S0 values for the fit. *args Variable length argument list passed to the :func:`fit` method. **kwargs Arbitrary keyword arguments passed to the :func:`fit` method. References ---------- .. footbibliography:: """ ReconstModel.__init__(self, gtab) # Check if at least three b-values are given enough_b = check_multi_b(self.gtab, 3, non_zero=False) if not enough_b: msg = "DKI requires at least 3 b-values (which can include b=0)." raise ValueError(msg) self.common_fit_method = not callable(fit_method) if self.common_fit_method: try: self.fit_method = common_fit_methods[fit_method.upper()] except KeyError as e: msg = '"' + str(fit_method) + '" is not a known fit method. ' msg += " The fit method should either be a function or one of " msg += " thecommon fit methods." raise ValueError(msg) from e self.return_S0_hat = return_S0_hat self.args = args self.kwargs = kwargs self.min_signal = self.kwargs.pop("min_signal", None) if self.min_signal is None: self.min_signal = MIN_POSITIVE_SIGNAL elif self.min_signal <= 0: msg = "The `min_signal` key-word argument needs to be strictly" msg += " positive." raise ValueError(msg) self.design_matrix = design_matrix(self.gtab) self.inverse_design_matrix = np.linalg.pinv(self.design_matrix) tol = 1e-6 self.min_diffusivity = tol / -self.design_matrix.min() self.convexity_constraint = fit_method in {"CLS", "CWLS", "NLS"} if self.convexity_constraint: self.cvxpy_solver = self.kwargs.pop("cvxpy_solver", None) self.convexity_level = self.kwargs.pop("convexity_level", "full") msg = "convexity_level must be a positive, even number, or 'full'." if isinstance(self.convexity_level, str): if self.convexity_level == "full": self.sdp_constraints = load_sdp_constraints("dki") else: raise ValueError(msg) elif self.convexity_level < 0 or self.convexity_level % 2: raise ValueError(msg) else: if self.convexity_level > 4: msg = "Maximum convexity_level supported is 4." warnings.warn(msg, stacklevel=2) self.convexity_level = 4 self.sdp_constraints = load_sdp_constraints( "dki", order=self.convexity_level ) self.sdp = PositiveDefiniteLeastSquares(22, A=self.sdp_constraints) self.weights = fit_method in {"WLS", "WLLS", "CWLS"} self.is_multi_method = fit_method in [ "WLS", "WLLS", "LS", "LLS", "OLS", "OLLS", "CLS", "CWLS", ] self.is_iter_method = "weights_method" in self.kwargs if "num_iter" not in self.kwargs and self.is_iter_method: self.kwargs["num_iter"] = 4 self.extra = {}
[docs] @warning_for_keywords() def fit(self, data, *, mask=None): """Fit method of the DKI model. Parameters ---------- data : array The measured signal. mask : array A boolean array used to mark the coordinates in the data that should be analyzed that has the shape data.shape[-1] """ data_thres = np.maximum(data, self.min_signal) if self.is_multi_method and not self.is_iter_method: fit_result, extra = self.multi_fit( data_thres, mask=mask, weights=self.weights, **self.kwargs ) if extra is not None: self.extra = extra return fit_result if self.is_multi_method and self.is_iter_method: fit_result, extra = self.iterative_fit( data_thres, mask=mask, num_iter=self.kwargs["num_iter"], weights_method=self.kwargs["weights_method"], ) if extra is not None: self.extra = extra return fit_result S0_params = None if data.ndim == 1: # NOTE: doesn't use mask... params, extra = self.fit_method( self.design_matrix, data_thres, return_S0_hat=self.return_S0_hat, *self.args, **self.kwargs, ) if self.return_S0_hat: params, S0_params = params if extra is not None: for key in extra: self.extra[key] = extra[key] return DiffusionKurtosisFit(self, params, model_S0=S0_params) if mask is not None: # Check for valid shape of the mask if mask.shape != data.shape[:-1]: raise ValueError("Mask is not the same shape as data.") mask = np.asarray(mask, dtype=bool) data_in_mask = np.reshape(data[mask], (-1, data.shape[-1])) data_in_mask = np.maximum(data_in_mask, self.min_signal) params, extra = self.fit_method( self.design_matrix, data_in_mask, return_S0_hat=self.return_S0_hat, *self.args, **self.kwargs, ) if self.return_S0_hat: params, S0_params = params if mask is None: out_shape = data.shape[:-1] + (-1,) dki_params = params.reshape(out_shape) if self.return_S0_hat: S0_params = S0_params.reshape(out_shape).squeeze(axis=-1) if extra is not None: for key in extra: self.extra[key] = extra[key].reshape(data.shape) else: dki_params = np.zeros(data.shape[:-1] + (27,)) dki_params[mask, :] = params if self.return_S0_hat: S0_params_in_mask = np.zeros(data.shape[:-1]) S0_params_in_mask[mask] = S0_params.squeeze(axis=-1) S0_params = S0_params_in_mask if extra is not None: for key in extra: self.extra[key] = np.zeros(data.shape) self.extra[key][mask, :] = extra[key] return DiffusionKurtosisFit(self, dki_params, model_S0=S0_params)
@multi_voxel_fit @warning_for_keywords() def multi_fit(self, data, *, mask=None, **kwargs): """Convenience function for fitting multiple voxels.""" extra_args = ( {} if not self.convexity_constraint else { "cvxpy_solver": self.cvxpy_solver, "sdp": self.sdp, } ) params, extra = self.fit_method( self.design_matrix, data, self.inverse_design_matrix, return_S0_hat=self.return_S0_hat, min_diffusivity=self.min_diffusivity, **extra_args, **kwargs, ) S0_params = None if self.return_S0_hat: params, S0_params = params return DiffusionKurtosisFit(self, params, model_S0=S0_params), extra
[docs] def iterative_fit( self, data_thres, *, mask=None, num_iter=4, weights_method=weights_method_wls_m_est, ): """Iteratively Reweighted fitting for the DKI model. Parameters ---------- data_thres : array The measured signal. mask : array, optional A boolean array used to mark the coordinates in the data that should be analyzed that has the shape data.shape[-1] num_iter : int, optional Number of times to iterate. weights_method : callable, optional A function with args and returns as follows: (weights, robust) = weights_method(data, pred_sig, design_matrix, leverages, idx, num_iter, robust) Notes ----- On the first iteration, (C)WLS fit is performed. Subsequent iterations will be weighted according to weights_method. Outlier rejection should be handled within weights_method by setting the corresponding weights to zero (if weights_method implements outlier rejection). """ if num_iter < 2: # otherwise, weights_method will not be utilized raise ValueError("num_iter must be 2+") w, robust = True, None # w = True ensures WLS (not OLS) tmp, extra, leverages = None, None, None # initialize, for clarity TDX = num_iter for rdx in range(1, TDX + 1): if rdx > 1: # after first iteration, update weights # if using mask, define full-sized arrays for w and robust if rdx == 2 and mask is not None: cond = mask == 1 w = np.ones_like(data_thres, dtype=float) robust = np.zeros_like(data_thres, dtype=bool) robust[cond] = 1 # make prediction of the signal pred_sig = self.predict(tmp.model_params, S0=tmp.model_S0) # define weights for next fit if mask is not None: w[cond], robust_mask = weights_method( data_thres[cond], pred_sig[cond], self.design_matrix, leverages[cond], rdx, TDX, robust[cond], ) if robust_mask is not None: robust[cond] = robust_mask else: w, robust = weights_method( data_thres, pred_sig, self.design_matrix, leverages, rdx, TDX, robust, ) tmp, extra = self.multi_fit( data_thres, mask=mask, weights=w, return_leverages=True ) leverages = extra["leverages"] extra = {"robust": robust} return tmp, extra
[docs] @warning_for_keywords() def predict(self, dki_params, *, S0=1.0): """Predict a signal for this DKI model class instance given parameters See :footcite:p:`Jensen2005` and :footcite:p:`NetoHenriques2021a` for further details about the method. Parameters ---------- dki_params : ndarray (x, y, z, 27) or (n, 27) All parameters estimated from the diffusion kurtosis model. Parameters are ordered as follows: 1. Three diffusion tensor's eigenvalues 2. Three lines of the eigenvector matrix each containing the first, second and third coordinates of the eigenvector 3. Fifteen elements of the kurtosis tensor S0 : float or ndarray, optional The non diffusion-weighted signal in every voxel, or across all voxels. References ---------- .. footbibliography:: """ return dki_prediction(dki_params, self.gtab, S0=S0)
[docs] class DiffusionKurtosisFit(TensorFit): """Class for fitting the Diffusion Kurtosis Model""" @warning_for_keywords() def __init__(self, model, model_params, *, model_S0=None): """Initialize a DiffusionKurtosisFit class instance Since DKI is an extension of DTI, class instance is defined as subclass of the TensorFit from dti.py Parameters ---------- model : DiffusionKurtosisModel Class instance Class instance containing the Diffusion Kurtosis Model for the fit model_params : ndarray (x, y, z, 27) or (n, 27) All parameters estimated from the diffusion kurtosis model, not including S0. Parameters are ordered as follows: 1) Three diffusion tensor's eigenvalues 2) Three lines of the eigenvector matrix each containing the first, second and third coordinates of the eigenvector 3) Fifteen elements of the kurtosis tensor model_S0 : ndarray (x, y, z, 1) or (n, 1), optional S0 estimated from the diffusion kurtosis model. """ TensorFit.__init__(self, model, model_params, model_S0=model_S0) @property def kt(self): """ Return the 15 independent elements of the kurtosis tensor as an array """ return self.model_params[..., 12:27]
[docs] def akc(self, sphere): r"""Calculate the apparent kurtosis coefficient (AKC) in each direction on the sphere for each voxel in the data Parameters ---------- sphere : Sphere class instance Sphere providing sample directions to compute the apparent kurtosis coefficient. Returns ------- akc : ndarray The estimates of the apparent kurtosis coefficient in every direction on the input sphere Notes ----- For each sphere direction with coordinates $(n_{1}, n_{2}, n_{3})$, the calculation of AKC is done using formula: .. math:: AKC(n)=\frac{MD^{2}}{ADC(n)^{2}}\sum_{i=1}^{3}\sum_{j=1}^{3} \sum_{k=1}^{3}\sum_{l=1}^{3}n_{i}n_{j}n_{k}n_{l}W_{ijkl} where $W_{ijkl}$ are the elements of the kurtosis tensor, MD the mean diffusivity and ADC the apparent diffusion coefficient computed as: .. math:: ADC(n)=\sum_{i=1}^{3}\sum_{j=1}^{3}n_{i}n_{j}D_{ij} where $D_{ij}$ are the elements of the diffusion tensor. """ return apparent_kurtosis_coef(self.model_params, sphere)
[docs] @warning_for_keywords() def mk(self, *, min_kurtosis=-3.0 / 7, max_kurtosis=10, analytical=True): r""" Compute mean kurtosis (MK) from the kurtosis tensor. See :footcite:t:`Tabesh2011` and :footcite:p:`NetoHenriques2021a` for further details about the method. Parameters ---------- min_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, mean kurtosis values that are smaller than `min_kurtosis` are replaced with `min_kurtosis`. Default = -3./7 (theoretical kurtosis limit for regions that consist of water confined to spherical pores :footcite:p:`Jensen2005`). max_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, mean kurtosis values that are larger than `max_kurtosis` are replaced with `max_kurtosis`. analytical : bool, optional If True, MK is calculated using its analytical solution, otherwise an exact numerical estimator is used (see Notes). Returns ------- mk : array Calculated MK. Notes ----- The MK is defined as the average of directional kurtosis coefficients across all spatial directions, which can be formulated by the following surface integral :footcite:p:`Tabesh2011`, :footcite:p:`NetoHenriques2021a`: .. math:: MK \equiv \frac{1}{4\pi} \int d\Omega_\mathbf{n} K(\mathbf{n}) This integral can be numerically solved by averaging directional kurtosis values sampled for directions of a spherical t-design :footcite:p:`Hardin1996`. Alternatively, MK can be solved from the analytical solution derived by :footcite:t:`Tabesh2011`. This solution is given by: .. math:: MK=F_1(\lambda_1,\lambda_2,\lambda_3)\hat{W}_{1111}+ F_1(\lambda_2,\lambda_1,\lambda_3)\hat{W}_{2222}+ F_1(\lambda_3,\lambda_2,\lambda_1)\hat{W}_{3333}+ \\ F_2(\lambda_1,\lambda_2,\lambda_3)\hat{W}_{2233}+ F_2(\lambda_2,\lambda_1,\lambda_3)\hat{W}_{1133}+ F_2(\lambda_3,\lambda_2,\lambda_1)\hat{W}_{1122} where $\hat{W}_{ijkl}$ are the components of the $W$ tensor in the coordinates system defined by the eigenvectors of the diffusion tensor $\mathbf{D}$ and .. math:: F_1(\lambda_1,\lambda_2,\lambda_3)= \frac{(\lambda_1+\lambda_2+\lambda_3)^2} {18(\lambda_1-\lambda_2)(\lambda_1-\lambda_3)} [\frac{\sqrt{\lambda_2\lambda_3}}{\lambda_1} R_F(\frac{\lambda_1}{\lambda_2},\frac{\lambda_1}{\lambda_3},1)+\\ \frac{3\lambda_1^2-\lambda_1\lambda_2-\lambda_2\lambda_3- \lambda_1\lambda_3} {3\lambda_1 \sqrt{\lambda_2 \lambda_3}} R_D(\frac{\lambda_1}{\lambda_2},\frac{\lambda_1}{\lambda_3},1)-1 ] F_2(\lambda_1,\lambda_2,\lambda_3)= \frac{(\lambda_1+\lambda_2+\lambda_3)^2} {3(\lambda_2-\lambda_3)^2} [\frac{\lambda_2+\lambda_3}{\sqrt{\lambda_2\lambda_3}} R_F(\frac{\lambda_1}{\lambda_2},\frac{\lambda_1}{\lambda_3},1)+\\ \frac{2\lambda_1-\lambda_2-\lambda_3}{3\sqrt{\lambda_2 \lambda_3}} R_D(\frac{\lambda_1}{\lambda_2},\frac{\lambda_1}{\lambda_3},1)-2] where $R_f$ and $R_d$ are the Carlson's elliptic integrals. References ---------- .. footbibliography:: """ return mean_kurtosis( self.model_params, min_kurtosis=min_kurtosis, max_kurtosis=max_kurtosis, analytical=analytical, )
[docs] @warning_for_keywords() def ak(self, *, min_kurtosis=-3.0 / 7, max_kurtosis=10, analytical=True): r""" Compute axial kurtosis (AK) of a diffusion kurtosis tensor. See :footcite:p:`Tabesh2011` and :footcite:p:`NetoHenriques2021a` for further details about the method. Parameters ---------- min_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, axial kurtosis values that are smaller than `min_kurtosis` are replaced with -3./7 (theoretical kurtosis limit for regions that consist of water confined to spherical pores :footcite:p:`Jensen2005`). max_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, axial kurtosis values that are larger than `max_kurtosis` are replaced with `max_kurtosis`. analytical : bool, optional If True, AK is calculated from rotated diffusion kurtosis tensor, otherwise it will be computed from the apparent diffusion kurtosis values along the principal axis of the diffusion tensor (see notes). Returns ------- ak : array Calculated AK. Notes ----- AK is defined as the directional kurtosis parallel to the fiber's main direction e1 :footcite:p:`Tabesh2011`, :footcite:p:`NetoHenriques2021a`. You can compute AK using to approaches: 1) AK is calculated from rotated diffusion kurtosis tensor, i.e.: .. math:: AK = \hat{W}_{1111} \frac{(\lambda_{1}+\lambda_{2}+\lambda_{3})^2}{(9 \lambda_{1}^2)} 2) AK can be sampled from the principal axis of the diffusion tensor: .. math:: AK = K(\mathbf{e}_1) Although both approaches leads to an exact calculation of AK, the first approach will be referred to as the analytical method while the second approach will be referred to as the numerical method based on their analogy to the estimation strategies for MK and RK. References ---------- .. footbibliography:: """ return axial_kurtosis( self.model_params, min_kurtosis=min_kurtosis, max_kurtosis=max_kurtosis, analytical=analytical, )
[docs] @warning_for_keywords() def rk(self, *, min_kurtosis=-3.0 / 7, max_kurtosis=10, analytical=True): r"""Compute radial kurtosis (RK) of a diffusion kurtosis tensor. See :footcite:p:`Tabesh2011` for further details about the method. Parameters ---------- min_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, radial kurtosis values that are smaller than `min_kurtosis` are replaced with `min_kurtosis`. Default = -3./7 (theoretical kurtosis limit for regions that consist of water confined to spherical pores :footcite:p:`Jensen2005`). max_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, radial kurtosis values that are larger than `max_kurtosis` are replaced with `max_kurtosis`. analytical : bool, optional If True, RK is calculated using its analytical solution, otherwise an exact numerical estimator is used (see Notes). Returns ------- rk : array Calculated RK. Notes ----- RK is defined as the average of the directional kurtosis perpendicular to the fiber's main direction e1 :footcite:p:`Tabesh2011`, :footcite:p:`NetoHenriques2021a`: .. math:: RK \equiv \frac{1}{2\pi} \int d\Omega _\mathbf{\theta} K(\mathbf{\theta}) \delta (\mathbf{\theta}\cdot \mathbf{e}_1) This equation can be numerically computed by averaging apparent directional kurtosis samples for directions perpendicular to e1 :footcite:p:`NetoHenriques2021a`. Otherwise, RK can be calculated from its analytical solution :footcite:p:`Tabesh2011`: .. math:: K_{\bot} = G_1(\lambda_1,\lambda_2,\lambda_3)\hat{W}_{2222} + G_1(\lambda_1,\lambda_3,\lambda_2)\hat{W}_{3333} + G_2(\lambda_1,\lambda_2,\lambda_3)\hat{W}_{2233} where: .. math:: G_1(\lambda_1,\lambda_2,\lambda_3)= \frac{(\lambda_1+\lambda_2+\lambda_3)^2}{18\lambda_2(\lambda_2- \lambda_3)} \left (2\lambda_2 + \frac{\lambda_3^2-3\lambda_2\lambda_3}{\sqrt{\lambda_2\lambda_3}} \right) and .. math:: G_2(\lambda_1,\lambda_2,\lambda_3)= \frac{(\lambda_1+\lambda_2+\lambda_3)^2}{(\lambda_2-\lambda_3)^2} \left ( \frac{\lambda_2+\lambda_3}{\sqrt{\lambda_2\lambda_3}}- 2\right ) References ---------- .. footbibliography:: """ return radial_kurtosis( self.model_params, min_kurtosis=min_kurtosis, max_kurtosis=max_kurtosis, analytical=analytical, )
[docs] @warning_for_keywords() def kmax(self, *, sphere="repulsion100", gtol=1e-5, mask=None): """Compute the maximum value of a single voxel kurtosis tensor See :footcite:p:`NetoHenriques2021a` for further details about the method. Parameters ---------- sphere : Sphere class instance, optional The sphere providing sample directions for the initial search of the maximum value of kurtosis. gtol : float, optional This input is to refine kurtosis maximum under the precision of the directions sampled on the sphere class instance. The gradient of the convergence procedure must be less than gtol before successful termination. If gtol is None, fiber direction is directly taken from the initial sampled directions of the given sphere object Returns ------- max_value : float kurtosis tensor maximum value References ---------- .. footbibliography:: """ return kurtosis_maximum(self.model_params, sphere=sphere, gtol=gtol, mask=mask)
[docs] @warning_for_keywords() def mkt(self, *, min_kurtosis=-3.0 / 7, max_kurtosis=10): r"""Compute mean of the kurtosis tensor (MKT). See :footcite:p:`Hansen2013` for further details about the method. Parameters ---------- min_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, mean kurtosis values that are smaller than `min_kurtosis` are replaced with `min_kurtosis`. Default = -3./7 (theoretical kurtosis limit for regions that consist of water confined to spherical pores :footcite:p:`Jensen2005`). max_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, mean kurtosis values that are larger than `max_kurtosis` are replaced with `max_kurtosis`. Returns ------- mkt : array Calculated mean kurtosis tensor. Notes ----- The MKT is defined as :footcite:p:`Hansen2013`: .. math:: MKT \equiv \frac{1}{4\pi} \int d \Omega_{\mathnbf{n}} n_i n_j n_k n_l W_{ijkl} which can be directly computed from the trace of the kurtosis tensor: .. math:: MKT = \frac{1}{5} Tr(\mathbf{W}) = \frac{1}{5} (W_{1111} + W_{2222} + W_{3333} + 2W_{1122} + 2W_{1133} + 2W_{2233}) References ---------- .. footbibliography:: """ return mean_kurtosis_tensor( self.model_params, min_kurtosis=min_kurtosis, max_kurtosis=max_kurtosis )
[docs] def rtk(self, *, min_kurtosis=-3.0 / 7, max_kurtosis=10): r"""Compute the rescaled radial tensor kurtosis (RTK). See :footcite:p:`Hansen2016b` for further details about the method. Parameters ---------- min_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, radial kurtosis values that are smaller than `min_kurtosis` are replaced with `min_kurtosis`. max_kurtosis : float, optional To keep kurtosis values within a plausible biophysical range, radial kurtosis values that are larger than `max_kurtosis` are replaced with `max_kurtosis`. Returns ------- rtk : array Calculated escaled radial tensor kurtosis (RTK). Notes ----- Rescaled radial tensor kurtosis (RTK) is defined as :footcite:p:`Hansen2016b`: .. math:: RKT = \frac{3}{8}\frac{MD^2}{RD^2} (W_{2222}+ W_{3333}+2*W_{2233}) where W is the kurtosis tensor rotated to a coordinate system in which the 3 orthonormal eigenvectors of DT are the base coordinate, MD is the mean diffusivity, and RD is the radial diffusivity. References ---------- .. footbibliography:: """ return radial_tensor_kurtosis( self.model_params, min_kurtosis=min_kurtosis, max_kurtosis=max_kurtosis )
@property def kfa(self): r"""Return the kurtosis tensor (KFA). See :footcite:p:`Glenn2015` for further details about the method. Notes ----- The KFA is defined as :footcite:p:`Glenn2015`: .. math:: KFA \equiv \frac{||\mathbf{W} - MKT \mathbf{I}^{(4)}||_F}{||\mathbf{W}||_F} where $W$ is the kurtosis tensor, MKT the kurtosis tensor mean, $I^{(4)}$ is the fully symmetric rank 2 isotropic tensor and $||...||_F$ is the tensor's Frobenius norm :footcite:p:`Glenn2015`. References ---------- .. footbibliography: """ return kurtosis_fractional_anisotropy(self.model_params)
[docs] @warning_for_keywords() def predict(self, gtab, *, S0=1.0): r"""Given a DKI model fit, predict the signal on the vertices of a gradient table See :footcite:p:`Jensen2005` and :footcite:p:`NetoHenriques2021a` for further details about the method. Parameters ---------- gtab : a GradientTable class instance The gradient table for this prediction S0 : float or ndarray, optional The non diffusion-weighted signal in every voxel, or across all voxels. Notes ----- The predicted signal is given by: .. math:: S(n,b)=S_{0}e^{-bD(n)+\frac{1}{6}b^{2}D(n)^{2}K(n)} $\mathbf{D(n)}$ and $\mathbf{K(n)}$ can be computed from the DT and KT using the following equations: .. math:: D(n)=\sum_{i=1}^{3}\sum_{j=1}^{3}n_{i}n_{j}D_{ij} and .. math:: K(n)=\frac{MD^{2}}{D(n)^{2}}\sum_{i=1}^{3}\sum_{j=1}^{3} \sum_{k=1}^{3}\sum_{l=1}^{3}n_{i}n_{j}n_{k}n_{l}W_{ijkl} where $D_{ij}$ and $W_{ijkl}$ are the elements of the second-order DT and the fourth-order KT tensors, respectively, and $MD$ is the mean diffusivity. References ---------- .. footbibliography:: """ return dki_prediction(self.model_params, gtab, S0=S0)
[docs] @warning_for_keywords() def params_to_dki_params(result, *, min_diffusivity=0): r"""Convert the 21 unique elements of the diffusion and kurtosis tensors to the parameter format adopted in DIPY Parameters ---------- results : array (21) Unique elements of the diffusion and kurtosis tensors in the following order: 1) six unique lower triangular DT elements; and 2) Fifteen unique elements of the kurtosis tensor. min_diffusivity : float, optional Because negative eigenvalues are not physical and small eigenvalues, much smaller than the diffusion weighting, cause quite a lot of noise in metrics such as fa, diffusivity values smaller than `min_diffusivity` are replaced with `min_diffusivity`. Returns ------- dki_params : array (27) All parameters estimated from the diffusion kurtosis model for all N voxels. Parameters are ordered as follows: 1) Three diffusion tensor eigenvalues. 2) Three blocks of three elements, containing the first second and third coordinates of the diffusion tensor eigenvectors. 3) Fifteen elements of the kurtosis tensor. """ # Extracting the diffusion tensor parameters from solution DT_elements = result[:6] evals, evecs = decompose_tensor( from_lower_triangular(DT_elements), min_diffusivity=min_diffusivity ) # Extracting kurtosis tensor parameters from solution MD_square = evals.mean(0) ** 2 KT_elements = result[6:21] / MD_square if MD_square else 0.0 * result[6:21] S0 = np.exp(-result[[-1]]) # Write output dki_params = np.concatenate( (evals, evecs[0], evecs[1], evecs[2], KT_elements, S0), axis=0 ) return dki_params
[docs] @warning_for_keywords() def ls_fit_dki( design_matrix, data, inverse_design_matrix, *, return_S0_hat=False, weights=True, min_diffusivity=0, return_lower_triangular=False, return_leverages=False, ): r"""Compute the diffusion and kurtosis tensors using an ordinary or weighted linear least squares approach. See :footcite:p:`Veraart2013` for further details about the method. Parameters ---------- design_matrix : array (g, 22) Design matrix holding the covariants used to solve for the regression coefficients. data : array (g) Data or response variables holding the data. inverse_design_matrix : array (22, g) Inverse of the design matrix. return_S0_hat : bool, optional Boolean to return (True) or not (False) the S0 values for the fit. weights : array ([X, Y, Z, ...], g), optional Weights to apply for fitting. These weights must correspond to the squared residuals such that $S = \sum_i w_i r_i^2$. If not provided, weights are estimated as the squared predicted signal from an initial OLS fit. min_diffusivity : float, optional Because negative eigenvalues are not physical and small eigenvalues, much smaller than the diffusion weighting, cause quite a lot of noise in metrics such as fa, diffusivity values smaller than `min_diffusivity` are replaced with `min_diffusivity`. return_lower_triangular : bool, optional Boolean to return (True) or not (False) the coefficients of the fit. return_leverages : bool, optional Boolean to return (True) or not (False) the fitting leverages. Returns ------- dki_params : array (27) All parameters estimated from the diffusion kurtosis model for all N voxels. Parameters are ordered as follows: 1) Three diffusion tensor eigenvalues. 2) Three blocks of three elements, containing the first second and third coordinates of the diffusion tensor eigenvectors. 3) Fifteen elements of the kurtosis tensor. leverages : array (g) Leverages of the fitting problem (if return_leverages is True) References ---------- .. footbibliography:: """ # Set up least squares problem A = design_matrix y = np.log(data) # weights correspond to *squared* residuals # define W = sqrt(weights), so that W are weights for residuals # e.g. weights for WLS are diag(fn**2), so W are diag(fn) if weights is not False: if type(weights) is np.ndarray: # user supplied weights W = np.diag(np.sqrt(weights)) else: # Define weights as diag(fn**2) (fn = fitted signal from OLS) result = np.dot(inverse_design_matrix, y) W = np.diag(np.exp(np.dot(A, result))) # W = sqrt(diag(fn**2)) W_A = np.dot(W, A) inv_W_A_W = np.linalg.pinv(W_A).dot(W) result = np.dot(inv_W_A_W, y) if return_leverages: leverages = np.einsum("ij,ji->i", design_matrix, inv_W_A_W) else: # DKI ordinary linear least square solution result = np.dot(inverse_design_matrix, y) if return_leverages: leverages = np.einsum("ij,ji->i", design_matrix, inverse_design_matrix) if return_leverages: leverages = {"leverages": leverages} else: leverages = None if return_lower_triangular: return result, leverages # Write output dki_params = params_to_dki_params(result, min_diffusivity=min_diffusivity) if return_S0_hat: return (dki_params[..., 0:-1], dki_params[..., -1]), leverages else: return dki_params[..., 0:-1], leverages
[docs] @warning_for_keywords() def cls_fit_dki( design_matrix, data, inverse_design_matrix, sdp, *, return_S0_hat=False, weights=True, min_diffusivity=0, return_lower_triangular=False, return_leverages=False, cvxpy_solver=None, ): r"""Compute the diffusion and kurtosis tensors using a constrained ordinary or weighted linear least squares approach. See :footcite:p:`DelaHaije2020` for further details about the method. Parameters ---------- design_matrix : array (g, 22) Design matrix holding the covariants used to solve for the regression coefficients. data : array (g) Data or response variables holding the data. inverse_design_matrix : array (22, g) Inverse of the design matrix. sdp : PositiveDefiniteLeastSquares instance A CVXPY representation of a regularized least squares optimization problem. return_S0_hat : bool, optional Boolean to return (True) or not (False) the S0 values for the fit. weights : array ([X, Y, Z, ...], g), optional Weights to apply for fitting. These weights must correspond to the squared residuals such that $S = \sum_i w_i r_i^2$. If not provided, weights are estimated as the squared predicted signal from an initial OLS fit. weights : bool, optional Parameter indicating whether weights are used. min_diffusivity : float, optional Because negative eigenvalues are not physical and small eigenvalues, much smaller than the diffusion weighting, cause quite a lot of noise in metrics such as fa, diffusivity values smaller than `min_diffusivity` are replaced with `min_diffusivity`. return_lower_triangular : bool, optional Boolean to return (True) or not (False) the coefficients of the fit. return_leverages : bool, optional Boolean to return (True) or not (False) the fitting leverages. cvxpy_solver : str, optional cvxpy solver name. Optionally optimize the positivity constraint with a particular cvxpy solver. See https://www.cvxpy.org/ for details. Default: None (cvxpy chooses its own solver). Returns ------- dki_params : array (27) All parameters estimated from the diffusion kurtosis model for all N voxels. Parameters are ordered as follows: 1) Three diffusion tensor eigenvalues. 2) Three blocks of three elements, containing the first second and third coordinates of the diffusion tensor eigenvectors. 3) Fifteen elements of the kurtosis tensor. leverages : array (g) Leverages of the fitting problem (if return_leverages is True) References ---------- .. footbibliography:: """ # Set up least squares problem A = design_matrix y = np.log(data) # weights correspond to *squared* residuals # define W = sqrt(weights), so that W are weights for residuals # e.g. weights for WLS are diag(fn**2), so W are diag(fn) if weights is not False: if type(weights) is np.ndarray: # use supplied weights W = np.diag(np.sqrt(weights)) else: # Define weights as diag(fn**2) (fn = fitted signal from OLS) result = np.dot(inverse_design_matrix, y) W = np.diag(np.exp(np.dot(A, result))) # W = sqrt(diag(fn**2)) A = np.dot(W, A) y = np.dot(W, y) if return_leverages: inv_W_A_W = np.linalg.pinv(A).dot(W) leverages = np.einsum("ij,ji->i", design_matrix, inv_W_A_W) # Solve sdp result = sdp.solve(A, y, check=True, solver=cvxpy_solver) if return_leverages: leverages = {"leverages": leverages} else: leverages = None if return_lower_triangular: return result, leverages # Write output dki_params = params_to_dki_params(result, min_diffusivity=min_diffusivity) if return_S0_hat: return (dki_params[..., 0:-1], dki_params[..., -1]), leverages else: return dki_params[..., 0:-1], leverages
[docs] def Wrotate(kt, Basis): r""" Rotate a kurtosis tensor from the standard Cartesian coordinate system to another coordinate system basis See :footcite:p:`Hui2008` for further details about the method. Parameters ---------- kt : (15,) Vector with the 15 independent elements of the kurtosis tensor Basis : array (3, 3) Vectors of the basis column-wise oriented inds : array(m, 4), optional Array of vectors containing the four indexes of m specific elements of the rotated kurtosis tensor. If not specified all 15 elements of the rotated kurtosis tensor are computed. Returns ------- Wrot : array (m,) or (15,) Vector with the m independent elements of the rotated kurtosis tensor. If 'indices' is not specified all 15 elements of the rotated kurtosis tensor are computed. Notes ----- The kurtosis tensor elements are assumed to be ordered as follows: .. math:: KT = \begin{pmatrix} W_{xxxx} & W_{yyyy} & W_{zzzz} & W_{xxxy} & W_{xxxz} \\ W_{xyyy} & W_{yyyz} & W_{xzzz} & W_{yzzz} & W_{xxyy} \\ W_{xxzz} & W_{yyzz} & W_{xxyz} & W_{xyyz} & W_{xyzz} \end{pmatrix} References ---------- .. footbibliography:: """ inds = np.array( [ [0, 0, 0, 0], [1, 1, 1, 1], [2, 2, 2, 2], [0, 0, 0, 1], [0, 0, 0, 2], [0, 1, 1, 1], [1, 1, 1, 2], [0, 2, 2, 2], [1, 2, 2, 2], [0, 0, 1, 1], [0, 0, 2, 2], [1, 1, 2, 2], [0, 0, 1, 2], [0, 1, 1, 2], [0, 1, 2, 2], ] ) Wrot = np.zeros(kt.shape) for e in range(len(inds)): Wrot[..., e] = Wrotate_element( kt, inds[e][0], inds[e][1], inds[e][2], inds[e][3], Basis ) return Wrot
# Defining keys to select a kurtosis tensor element with indexes (i, j, k, l) # on a kt vector that contains only the 15 independent elements of the kurtosis # tensor: Considering y defined by (i+1) * (j+1) * (k+1) * (l+1). Two elements # of the full 4D kurtosis tensor are equal if y obtain from the indexes of # these two element are equal. Therefore, the possible values of y (1, 16, 81, # 2, 3, 8, 24 27, 54, 4, 9, 36, 6, 12, 18) are used to point each element of # the kurtosis tensor on the format of a vector containing the 15 independent # elements. ind_ele = { 1: 0, 16: 1, 81: 2, 2: 3, 3: 4, 8: 5, 24: 6, 27: 7, 54: 8, 4: 9, 9: 10, 36: 11, 6: 12, 12: 13, 18: 14, }
[docs] def Wrotate_element(kt, indi, indj, indk, indl, B): r"""Compute the specified index element of a kurtosis tensor rotated to the coordinate system basis B See :footcite:p:`Hui2008` for further details about the method. Parameters ---------- kt : ndarray (x, y, z, 15) or (n, 15) Array containing the 15 independent elements of the kurtosis tensor indi : int Rotated kurtosis tensor element index i (0 for x, 1 for y, 2 for z) indj : int Rotated kurtosis tensor element index j (0 for x, 1 for y, 2 for z) indk : int Rotated kurtosis tensor element index k (0 for x, 1 for y, 2 for z) indl: int Rotated kurtosis tensor element index l (0 for x, 1 for y, 2 for z) B: array (x, y, z, 3, 3) or (n, 15) Vectors of the basis column-wise oriented Returns ------- Wre : float rotated kurtosis tensor element of index ind_i, ind_j, ind_k, ind_l Notes ----- It is assumed that initial kurtosis tensor elements are defined on the Cartesian coordinate system. References ---------- .. footbibliography:: """ Wre = 0 xyz = [0, 1, 2] for il in xyz: for jl in xyz: for kl in xyz: for ll in xyz: key = (il + 1) * (jl + 1) * (kl + 1) * (ll + 1) multiplyB = ( B[..., il, indi] * B[..., jl, indj] * B[..., kl, indk] * B[..., ll, indl] ) Wre = Wre + multiplyB * kt[..., ind_ele[key]] return Wre
[docs] def Wcons(k_elements): r""" Construct the full 4D kurtosis tensors from its 15 independent elements Parameters ---------- k_elements : (15,) elements of the kurtosis tensor in the following order: .. math:: KT = \begin{pmatrix} W_{xxxx} & W_{yyyy} & W_{zzzz} & W_{xxxy} & W_{xxxz} \\ W_{xyyy} & W_{yyyz} & W_{xzzz} & W_{yzzz} & W_{xxyy} \\ W_{xxzz} & W_{yyzz} & W_{xxyz} & W_{xyyz} & W_{xyzz} \end{pmatrix} Returns ------- W : array(3, 3, 3, 3) Full 4D kurtosis tensor """ W = np.zeros((3, 3, 3, 3)) xyz = [0, 1, 2] for ind_i in xyz: for ind_j in xyz: for ind_k in xyz: for ind_l in xyz: key = (ind_i + 1) * (ind_j + 1) * (ind_k + 1) * (ind_l + 1) W[ind_i][ind_j][ind_k][ind_l] = k_elements[ind_ele[key]] return W
[docs] def split_dki_param(dki_params): r"""Extract the diffusion tensor eigenvalues, the diffusion tensor eigenvector matrix, and the 15 independent elements of the kurtosis tensor from the model parameters estimated from the DKI model Parameters ---------- dki_params : ndarray (x, y, z, 27) or (n, 27) All parameters estimated from the diffusion kurtosis model. Parameters are ordered as follows: 1) Three diffusion tensor's eigenvalues 2) Three lines of the eigenvector matrix each containing the first, second and third coordinates of the eigenvector 3) Fifteen elements of the kurtosis tensor Returns ------- eigvals : array (x, y, z, 3) or (n, 3) Eigenvalues from eigen decomposition of the tensor. eigvecs : array (x, y, z, 3, 3) or (n, 3, 3) Associated eigenvectors from eigen decomposition of the tensor. Eigenvectors are columnar (e.g. eigvecs[:,j] is associated with eigvals[j]) kt : array (x, y, z, 15) or (n, 15) Fifteen elements of the kurtosis tensor """ evals = dki_params[..., :3] evecs = dki_params[..., 3:12].reshape(dki_params.shape[:-1] + (3, 3)) kt = dki_params[..., 12:27] return evals, evecs, kt
common_fit_methods = { "WLS": ls_fit_dki, "WLLS": ls_fit_dki, "LS": ls_fit_dki, "LLS": ls_fit_dki, "OLS": ls_fit_dki, "OLLS": ls_fit_dki, "NLS": nlls_fit_tensor, "NLLS": nlls_fit_tensor, "RESTORE": restore_fit_tensor, "IRLS": iterative_fit_tensor, "RWLS": robust_fit_tensor_wls, "RNLLS": robust_fit_tensor_nlls, "CLS": cls_fit_dki, "CWLS": cls_fit_dki, }