def test_all_models_dispersable():
scheme = wu_minn_hcp_acquisition_scheme()
dispersable_models = [
[cylinder_models.C1Stick()],
[cylinder_models.C2CylinderStejskalTannerApproximation()],
[cylinder_models.C3CylinderCallaghanApproximation()],
[cylinder_models.C4CylinderGaussianPhaseApproximation()],
[gaussian_models.G1Ball(),
gaussian_models.G2Zeppelin()], [gaussian_models.G3TemporalZeppelin()],
[sphere_models.S1Dot(),
gaussian_models.G2Zeppelin()],
[
sphere_models.S2SphereStejskalTannerApproximation(),
gaussian_models.G2Zeppelin()
]
]
spherical_distributions = [
distribute_models.SD1WatsonDistributed,
distribute_models.SD2BinghamDistributed
]
for model in dispersable_models:
for distribution in spherical_distributions:
dist_mod = distribution(model)
params = {}
for param, card in dist_mod.parameter_cardinality.items():
params[param] = np.random.rand(
card) * dist_mod.parameter_scales[param]
assert_equal(isinstance(dist_mod(scheme, **params), np.ndarray),
True)
def test_C3_watson_gamma_equals_gamma_watson():
scheme = wu_minn_hcp_acquisition_scheme()
cylinder = cylinder_models.C3CylinderCallaghanApproximation()
watsoncyl = distribute_models.SD1WatsonDistributed([cylinder])
gammawatsoncyl = distribute_models.DD1GammaDistributed(
[watsoncyl],
target_parameter='C3CylinderCallaghanApproximation_1_diameter')
params1 = {
'SD1WatsonDistributed_1_C3CylinderCallaghanApproximation_1_lambda_par':
1.7e-9,
'DD1Gamma_1_alpha': 2.,
'DD1Gamma_1_beta': 4e-6,
'SD1WatsonDistributed_1_SD1Watson_1_odi': 0.4,
'SD1WatsonDistributed_1_SD1Watson_1_mu': [0., 0.]
}
gammacyl = distribute_models.DD1GammaDistributed([cylinder])
watsongammacyl = distribute_models.SD1WatsonDistributed([gammacyl])
params2 = {
'DD1GammaDistributed_1_C3CylinderCallaghanApproximation_1_lambda_par':
1.7e-9,
'DD1GammaDistributed_1_DD1Gamma_1_alpha': 2.,
'DD1GammaDistributed_1_DD1Gamma_1_beta': 4e-6,
'SD1Watson_1_odi': 0.4,
'SD1Watson_1_mu': [0., 0.]
}
assert_array_almost_equal(watsongammacyl(scheme, **params2),
gammawatsoncyl(scheme, **params1), 5)
def test_soderman_equivalent_to_callaghan_with_one_root_and_function(
samples=100):
mu = [0, 0]
lambda_par = .1
diameter = 10e-5
diffusion_perpendicular = 1.7e-09
delta = np.tile(1e-3, samples) # delta towards zero
Delta = np.tile(.15e-2, samples) # Delta towards infinity
qvals_perp = np.linspace(0, 1e5, samples)
n_perp = np.tile(np.r_[1., 0., 0.], (samples, 1))
scheme = acquisition_scheme_from_qvalues(qvals_perp, n_perp, delta, Delta)
soderman = cylinder_models.C2CylinderSodermanApproximation(
mu=mu, lambda_par=lambda_par, diameter=diameter)
callaghan = cylinder_models.C3CylinderCallaghanApproximation(
number_of_roots=1,
number_of_functions=1,
mu=mu,
lambda_par=lambda_par,
diameter=diameter,
diffusion_perpendicular=diffusion_perpendicular)
E_soderman = soderman(scheme)
E_callaghan = callaghan(scheme)
assert_array_almost_equal(E_soderman, E_callaghan)
def test_callaghan_profile_narrow_pulse_not_restricted(samples=100):
# in short diffusion times the model should approach a Gaussian
# profile as np.exp(-b * lambda_perp)
mu = [0, 0]
lambda_par = .1e-9
diameter = 10e-5
diffusion_perpendicular = 1.7e-09
delta = np.tile(1e-3, samples) # delta towards zero
Delta = np.tile(.15e-2, samples) # Delta towards infinity
qvals_perp = np.linspace(0, 3e5, samples)
n_perp = np.tile(np.r_[1., 0., 0.], (samples, 1))
scheme = acquisition_scheme_from_qvalues(qvals_perp, n_perp, delta, Delta)
# needed to increase the number of roots and functions to approximate
# the gaussian function.
callaghan = cylinder_models.C3CylinderCallaghanApproximation(
number_of_roots=20,
number_of_functions=50,
mu=mu,
lambda_par=lambda_par,
diameter=diameter,
diffusion_perpendicular=diffusion_perpendicular)
E_callaghan = callaghan(scheme)
E_free_diffusion = np.exp(-scheme.bvalues * diffusion_perpendicular)
assert_equal(np.max(np.abs(E_callaghan - E_free_diffusion)) < 0.01, True)
def test_RTAP_to_diameter_callaghan(samples=10000):
mu = [0, 0]
lambda_par = 1.7
diameter = 10e-6
delta = np.tile(1e-10, samples) # delta towards zero
Delta = np.tile(1e10, samples) # Delta towards infinity
qvals_perp = np.linspace(0, 10e6, samples)
n_perp = np.tile(np.r_[1., 0., 0.], (samples, 1))
scheme = acquisition_scheme_from_qvalues(qvals_perp, n_perp, delta, Delta)
callaghan = cylinder_models.C3CylinderCallaghanApproximation(
mu=mu, lambda_par=lambda_par, diameter=diameter)
E_callaghan = callaghan(scheme)
rtap_callaghan = 2 * np.pi * np.trapz(E_callaghan * qvals_perp,
x=qvals_perp)
diameter_callaghan = 2 / np.sqrt(np.pi * rtap_callaghan)
assert_almost_equal(diameter_callaghan, diameter, 7)
wu_minn_hcp_acquisition_scheme)
from dmipy.core.modeling_framework import (MultiCompartmentSphericalMeanModel)
from dmipy.core.acquisition_scheme import acquisition_scheme_from_bvalues
import numpy as np
from dmipy.utils.spherical_mean import (estimate_spherical_mean_shell)
from numpy.testing import assert_equal, assert_almost_equal
from dipy.data import get_sphere
sphere = get_sphere().subdivide()
scheme = wu_minn_hcp_acquisition_scheme()
models = [
cylinder_models.C1Stick(),
cylinder_models.C2CylinderSodermanApproximation(),
cylinder_models.C3CylinderCallaghanApproximation(),
cylinder_models.C4CylinderGaussianPhaseApproximation(),
gaussian_models.G1Ball(),
gaussian_models.G2Zeppelin(),
gaussian_models.G3RestrictedZeppelin(),
sphere_models.S2SphereSodermanApproximation()
]
distributable_models = [
cylinder_models.C2CylinderSodermanApproximation(),
cylinder_models.C3CylinderCallaghanApproximation(),
cylinder_models.C4CylinderGaussianPhaseApproximation(),
]
def test_model_spherical_mean_analytic_vs_numerical(bvalue=1e9,
