def test_sphere_cart():
# test arrays of points
rs, thetas, phis = cart2sphere(*(sphere_points.T))
xyz = sphere2cart(rs, thetas, phis)
assert_array_almost_equal(xyz, sphere_points.T)
# test radius estimation
big_sph_pts = sphere_points * 10.4
rs, thetas, phis = cart2sphere(*big_sph_pts.T)
assert_array_almost_equal(rs, 10.4)
xyz = sphere2cart(rs, thetas, phis)
assert_array_almost_equal(xyz, big_sph_pts.T, decimal=6)
# test that result shapes match
x, y, z = big_sph_pts.T
r, theta, phi = cart2sphere(x[:1], y[:1], z)
assert_equal(r.shape, theta.shape)
assert_equal(r.shape, phi.shape)
x, y, z = sphere2cart(r[:1], theta[:1], phi)
assert_equal(x.shape, y.shape)
assert_equal(x.shape, z.shape)
# test a scalar point
pt = sphere_points[3]
r, theta, phi = cart2sphere(*pt)
xyz = sphere2cart(r, theta, phi)
assert_array_almost_equal(xyz, pt)
# Test full circle on x=1, y=1, z=1
x, y, z = sphere2cart(*cart2sphere(1.0, 1.0, 1.0))
assert_array_almost_equal((x, y, z), (1.0, 1.0, 1.0))
def test_sphere_cart():
# test arrays of points
rs, thetas, phis = cart2sphere(*(sphere_points.T))
xyz = sphere2cart(rs, thetas, phis)
yield assert_array_almost_equal, xyz, sphere_points.T
# test radius estimation
big_sph_pts = sphere_points * 10.4
rs, thetas, phis = cart2sphere(*big_sph_pts.T)
yield assert_array_almost_equal, rs, 10.4
xyz = sphere2cart(rs, thetas, phis)
yield assert_array_almost_equal, xyz, big_sph_pts.T, 6
# test that result shapes match
x, y, z = big_sph_pts.T
r, theta, phi = cart2sphere(x[:1], y[:1], z)
yield assert_equal, r.shape, theta.shape
yield assert_equal, r.shape, phi.shape
x, y, z = sphere2cart(r[:1], theta[:1], phi)
yield assert_equal, x.shape, y.shape
yield assert_equal, x.shape, z.shape
# test a scalar point
pt = sphere_points[3]
r, theta, phi = cart2sphere(*pt)
xyz = sphere2cart(r, theta, phi)
yield assert_array_almost_equal, xyz, pt
# Test full circle on x=0, y=0, z=0
x, y, z = sphere2cart(*cart2sphere(0., 0., 0.))
yield assert_array_equal, (x, y, z), (0., 0., 0.)
def test_sphere_cart():
# test arrays of points
rs, thetas, phis = cart2sphere(*(sphere_points.T))
xyz = sphere2cart(rs, thetas, phis)
yield assert_array_almost_equal(xyz, sphere_points.T)
# test radius estimation
big_sph_pts = sphere_points * 10.4
rs, thetas, phis = cart2sphere(*big_sph_pts.T)
yield assert_array_almost_equal(rs, 10.4)
xyz = sphere2cart(rs, thetas, phis)
yield assert_array_almost_equal(xyz, big_sph_pts.T, 6)
# test a scalar point
pt = sphere_points[3]
r, theta, phi = cart2sphere(*pt)
xyz = sphere2cart(r, theta, phi)
yield assert_array_almost_equal(xyz, pt)
def _pred_sig(self, theta, phi, beta, lambda1, lambda2):
"""
The predicted signal for a particular setting of the parameters
"""
Q = np.array([[lambda1, 0, 0], [0, lambda2, 0], [0, 0, lambda2]])
# If for some reason this is not symmetrical, then something is wrong
# (in all likelihood, this means that the optimization process is
# trying out some crazy value, such as nan). In that case, abort and
# return a nan:
if not np.allclose(Q.T, Q):
return np.nan
response_function = ozt.Tensor(Q, self.bvecs[:, self.b_idx],
self.bvals[:, self.b_idx])
# Convert theta and phi to cartesian coordinates:
x, y, z = geo.sphere2cart(1, theta, phi)
bvec = [x, y, z]
evals, evecs = response_function.decompose
rot_tensor = ozt.tensor_from_eigs(
evecs * ozu.calculate_rotation(bvec, evecs[0]), evals,
self.bvecs[:, self.b_idx], self.bvals[:, self.b_idx])
iso_sig = np.exp(-self.bvals[self.b_idx][0] * lambda1)
tensor_sig = rot_tensor.predicted_signal(1)
return beta * iso_sig + (1 - beta) * tensor_sig
def test_sphere_cart():
# test arrays of points
rs, thetas, phis = cart2sphere(*(sphere_points.T))
xyz = sphere2cart(rs, thetas, phis)
yield assert_array_almost_equal(xyz, sphere_points.T)
# test radius estimation
big_sph_pts = sphere_points * 10.4
rs, thetas, phis = cart2sphere(*big_sph_pts.T)
yield assert_array_almost_equal(rs, 10.4)
xyz = sphere2cart(rs, thetas, phis)
yield assert_array_almost_equal(xyz, big_sph_pts.T, 6)
# test a scalar point
pt = sphere_points[3]
r, theta, phi = cart2sphere(*pt)
xyz = sphere2cart(r, theta, phi)
yield assert_array_almost_equal(xyz, pt)
def callnonlin(x0,Sreal,S0,bvals,gradients):
d=x0[0]/10.**4
if len(x0)==1:
S=np.zeros(len(gradients))
for (i,g) in enumerate(gradients[1:]):
S[i+1]=S0*np.exp(-bvals[i+1]*d)
S[0]=S0
return Sreal-S
if len(x0)==4:
angles=[(x0[1],x0[2])]
fractions=[x0[3]/100.]
if len(x0)==7:
angles=[(x0[1],x0[2]),(x0[3],x0[4])]
fractions=[x0[5]/100.,x0[6]/100.]
if len(x0)==10:
angles=[(x0[1],x0[2]),(x0[3],x0[4]),(x0[5],x0[6])]
fractions=[x0[7]/100.,x0[8]/100.,x0[9]/100.]
f0=1-np.sum(fractions)
S=np.zeros(len(gradients))
sticks=[ sphere2cart(1,np.deg2rad(pair[0]),np.deg2rad(pair[1])) for pair in angles]
sticks=np.array(sticks)
for (i,g) in enumerate(gradients[1:]):
S[i+1]=f0*np.exp(-bvals[i+1]*d)+ np.sum([fractions[j]*np.exp(-bvals[i+1]*d*np.dot(s,g)**2) for (j,s) in enumerate(sticks)])
S[i+1]=S0*S[i+1]
S[0]=S0
return Sreal-S
def plot_ellipsoid_mpl(Tensor, n=60):
"""
Plot an ellipsoid from a tensor using matplotlib
Parameters
----------
Tensor: an ozt.Tensor class instance
n: optional. If an integer is provided, we will plot this for a sphere with
n (grid) equi-sampled points. Otherwise, we will plot the originally
provided Tensor's bvecs.
"""
x, y, z = sphere(n=n)
new_bvecs = np.vstack([x.ravel(), y.ravel(), z.ravel()])
Tensor = ozt.Tensor(Tensor.Q, new_bvecs,
np.ones(new_bvecs.shape[-1]) * Tensor.bvals[0])
v = Tensor.diffusion_distance.reshape(x.shape)
r, phi, theta = geo.cart2sphere(x, y, z)
x_plot, y_plot, z_plot = geo.sphere2cart(v, phi, theta)
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
ax.plot_surface(x_plot, y_plot, z_plot, rstride=2, cstride=2, shade=True)
return fig
def test_perpendicular_directions():
num = 35
vectors_v = np.zeros((4, 3))
for v in range(4):
theta = random.uniform(0, np.pi)
phi = random.uniform(0, 2*np.pi)
vectors_v[v] = sphere2cart(1., theta, phi)
vectors_v[3] = [1, 0, 0]
for vector_v in vectors_v:
pd = perpendicular_directions(vector_v, num=num, half=False)
# see if length of pd is equal to the number of intendend samples
assert_equal(num, len(pd))
# check if all directions are perpendicular to vector v
for d in pd:
cos_angle = np.dot(d, vector_v)
assert_almost_equal(cos_angle, 0)
# check if directions are sampled by multiples of 2*pi / num
delta_a = 2. * np.pi / num
for d in pd[1:]:
angle = np.arccos(np.dot(pd[0], d))
rest = angle % delta_a
if rest > delta_a * 0.99: # To correct cases of negative error
rest = rest - delta_a
assert_almost_equal(rest, 0)
def _pred_sig(self, theta, phi, beta, lambda1, lambda2):
"""
The predicted signal for a particular setting of the parameters
"""
Q = np.array([[lambda1, 0, 0],
[0, lambda2, 0],
[0, 0, lambda2]])
# If for some reason this is not symmetrical, then something is wrong
# (in all likelihood, this means that the optimization process is
# trying out some crazy value, such as nan). In that case, abort and
# return a nan:
if not np.allclose(Q.T, Q):
return np.nan
response_function = ozt.Tensor(Q,
self.bvecs[:,self.b_idx],
self.bvals[:,self.b_idx])
# Convert theta and phi to cartesian coordinates:
x,y,z = geo.sphere2cart(1, theta, phi)
bvec = [x,y,z]
evals, evecs = response_function.decompose
rot_tensor = ozt.tensor_from_eigs(
evecs * ozu.calculate_rotation(bvec, evecs[0]),
evals, self.bvecs[:,self.b_idx], self.bvals[:,self.b_idx])
iso_sig = np.exp(-self.bvals[self.b_idx][0] * lambda1)
tensor_sig = rot_tensor.predicted_signal(1)
return beta * iso_sig + (1-beta) * tensor_sig
def plot_ellipsoid_mpl(Tensor, n=60):
"""
Plot an ellipsoid from a tensor using matplotlib
Parameters
----------
Tensor: an ozt.Tensor class instance
n: optional. If an integer is provided, we will plot this for a sphere with
n (grid) equi-sampled points. Otherwise, we will plot the originally
provided Tensor's bvecs.
"""
x,y,z = sphere(n=n)
new_bvecs = np.vstack([x.ravel(), y.ravel(), z.ravel()])
Tensor = ozt.Tensor(Tensor.Q, new_bvecs,
np.ones(new_bvecs.shape[-1]) * Tensor.bvals[0])
v = Tensor.diffusion_distance.reshape(x.shape)
r, phi, theta = geo.cart2sphere(x,y,z)
x_plot, y_plot, z_plot = geo.sphere2cart(v, phi, theta)
fig = plt.figure()
ax = fig.add_subplot(111, projection='3d')
ax.plot_surface(x_plot, y_plot, z_plot, rstride=2, cstride=2, shade=True)
return fig
def _check_directions(angles):
"""
Helper function to check if direction ground truth have the right format
and are in cartesian coordinates
Parameters
----------
angles : array (K,2) or (K, 3)
List of K polar angles (in degrees) for the sticks or array of K
sticks as unit vectors.
Returns
-------
sticks : (K,3)
Sticks in cartesian coordinates.
"""
angles = np.array(angles)
if angles.shape[-1] == 3:
sticks = angles
else:
sticks = [
sphere2cart(1, np.deg2rad(pair[0]), np.deg2rad(pair[1]))
for pair in angles
]
sticks = np.array(sticks)
return sticks
def _check_directions(angles):
"""
Helper function to check if direction ground truth have the right format
and are in cartesian coordinates
Parameters
-----------
angles : array (K,2) or (K, 3)
List of K polar angles (in degrees) for the sticks or array of K
sticks as unit vectors.
Returns
--------
sticks : (K,3)
Sticks in cartesian coordinates.
"""
angles = np.array(angles)
if angles.shape[-1] == 3:
sticks = angles
else:
sticks = [sphere2cart(1, np.deg2rad(pair[0]), np.deg2rad(pair[1]))
for pair in angles]
sticks = np.array(sticks)
return sticks
def sticks_and_ball(gtab, d=0.0015, S0=100, angles=[(0, 0), (90, 0)],
fractions=[35, 35], snr=20):
""" Simulate the signal for a Sticks & Ball model.
Parameters
-----------
gtab : GradientTable
Signal measurement directions.
d : float
Diffusivity value.
S0 : float
Unweighted signal value.
angles : array (K,2) or (K, 3)
List of K polar angles (in degrees) for the sticks or array of K
sticks as unit vectors.
fractions : float
Percentage of each stick. Remainder to 100 specifies isotropic
component.
snr : float
Signal to noise ratio, assuming Rician noise. If set to None, no
noise is added.
Returns
--------
S : (N,) ndarray
Simulated signal.
sticks : (M,3)
Sticks in cartesian coordinates.
References
----------
.. [1] Behrens et al., "Probabilistic diffusion
tractography with multiple fiber orientations: what can we gain?",
Neuroimage, 2007.
"""
fractions = [f / 100. for f in fractions]
f0 = 1 - np.sum(fractions)
S = np.zeros(len(gtab.bvals))
angles = np.array(angles)
if angles.shape[-1] == 3:
sticks = angles
else:
sticks = [sphere2cart(1, np.deg2rad(pair[0]), np.deg2rad(pair[1]))
for pair in angles]
sticks = np.array(sticks)
for (i, g) in enumerate(gtab.bvecs[1:]):
S[i + 1] = f0 * np.exp(-gtab.bvals[i + 1] * d) + \
np.sum([fractions[j] * np.exp(-gtab.bvals[i + 1] * d * np.dot(s, g) ** 2)
for (j, s) in enumerate(sticks)])
S[i + 1] = S0 * S[i + 1]
S[gtab.b0s_mask] = S0
S = add_noise(S, snr, S0)
return S, sticks
def plot_signal(bvecs,
signal,
origin=[0, 0, 0],
maya=True,
cmap='jet',
file_name=None,
colorbar=False,
figure=None,
vmin=None,
vmax=None,
offset=0,
azimuth=60,
elevation=90,
roll=0,
non_neg=False):
"""
Interpolate a measured signal, using RBF interpolation.
Parameters
----------
signal:
bvecs: array (3,n)
the x,y,z locations where the signal was measured
offset : float
where to place the plotted voxel (on the z axis)
"""
s0 = Sphere(xyz=bvecs.T)
vertices = s0.vertices
faces = s0.faces
x, y, z = vertices.T
r, phi, theta = geo.cart2sphere(x, y, z)
x_plot, y_plot, z_plot = geo.sphere2cart(signal, phi, theta)
if non_neg:
signal[np.where(signal < 0)] = 0
# Call and return straightaway:
return _display_maya_voxel(x_plot,
y_plot,
z_plot,
faces,
signal,
origin,
cmap=cmap,
colorbar=colorbar,
figure=figure,
vmin=vmin,
vmax=vmax,
file_name=file_name,
azimuth=azimuth,
elevation=elevation)
def multi_tensor_odf(odf_verts, mevals, angles, fractions):
r'''Simulate a Multi-Tensor ODF.
Parameters
----------
odf_verts : (N,3) ndarray
Vertices of the reconstruction sphere.
mevals : sequence of 1D arrays,
Eigen-values for each tensor.
angles : sequence of 2d tuples,
Sequence of principal directions for each tensor in polar angles
or cartesian unit coordinates.
fractions : sequence of floats,
Percentages of the fractions for each tensor.
Returns
-------
ODF : (N,) ndarray
Orientation distribution function.
Examples
--------
Simulate a MultiTensor ODF with two peaks and calculate its exact ODF.
>>> import numpy as np
>>> from dipy.sims.voxel import multi_tensor_odf, all_tensor_evecs
>>> from dipy.data import get_sphere
>>> sphere = get_sphere('symmetric724')
>>> vertices, faces = sphere.vertices, sphere.faces
>>> mevals = np.array(([0.0015, 0.0003, 0.0003],[0.0015, 0.0003, 0.0003]))
>>> angles = [(0, 0), (90, 0)]
>>> odf = multi_tensor_odf(vertices, mevals, angles, [50, 50])
'''
mf = [f / 100. for f in fractions]
angles = np.array(angles)
if angles.shape[-1] == 3:
sticks = angles
else:
sticks = [
sphere2cart(1, np.deg2rad(pair[0]), np.deg2rad(pair[1]))
for pair in angles
]
sticks = np.array(sticks)
odf = np.zeros(len(odf_verts))
mevecs = []
for s in sticks:
mevecs += [all_tensor_evecs(s).T]
for (j, f) in enumerate(mf):
odf += f * single_tensor_odf(
odf_verts, evals=mevals[j], evecs=mevecs[j])
return odf
def test_invert_transform():
n = 100.
theta = np.arange(n)/n * np.pi # Limited to 0,pi
phi = (np.arange(n)/n - .5) * 2 * np.pi # Limited to 0,2pi
x, y, z = sphere2cart(1, theta, phi) # Let's assume they're all unit vecs
r, new_theta, new_phi = cart2sphere(x, y, z) # Transform back
yield assert_array_almost_equal, theta, new_theta
yield assert_array_almost_equal, phi, new_phi
def SticksAndBall(bvals,
gradients,
d=0.0015,
S0=100,
angles=[(0, 0), (90, 0)],
fractions=[35, 35],
snr=20):
""" Simulating the signal for a Sticks & Ball model
Based on the paper by Tim Behrens, H.J. Berg, S. Jbabdi, "Probabilistic Diffusion Tractography with multiple fiber orientations
what can we gain?", Neuroimage, 2007.
Parameters
-----------
bvals : array, shape (N,)
gradients : array, shape (N,3) also known as bvecs
d : diffusivity value
S0 : unweighted signal value
angles : array (K,2) list of polar angles (in degrees) for the sticks
or array (K,3) with sticks as Cartesian unit vectors and K the number of sticks
fractions : percentage of each stick
snr : signal to noise ration assuming gaussian noise. Provide None for no noise.
Returns
--------
S : simulated signal
sticks : sticks in cartesian coordinates
"""
fractions = [f / 100. for f in fractions]
f0 = 1 - np.sum(fractions)
S = np.zeros(len(gradients))
angles = np.array(angles)
if angles.shape[-1] == 3:
sticks = angles
if angles.shape[-1] == 2:
sticks = [
sphere2cart(1, np.deg2rad(pair[0]), np.deg2rad(pair[1]))
for pair in angles
]
sticks = np.array(sticks)
for (i, g) in enumerate(gradients[1:]):
S[i + 1] = f0 * np.exp(-bvals[i + 1] * d) + np.sum([
fractions[j] * np.exp(-bvals[i + 1] * d * np.dot(s, g)**2)
for (j, s) in enumerate(sticks)
])
S[i + 1] = S0 * S[i + 1]
S[0] = S0
if snr != None:
std = S0 / snr
S = S + np.random.randn(len(S)) * std
return S, sticks
def multi_tensor_pdf(pdf_points,
mevals,
angles,
fractions,
tau=1 / (4 * np.pi**2)):
r'''Simulate a Multi-Tensor ODF.
Parameters
----------
pdf_points : (N, 3) ndarray
Points to evaluate the PDF.
mevals : sequence of 1D arrays,
Eigen-values for each tensor. By default, values typical for prolate
white matter are used.
angles : sequence,
Sequence of principal directions for each tensor in polar angles
or cartesian unit coordinates.
fractions : sequence of floats,
Percentages of the fractions for each tensor.
tau : float,
diffusion time. By default the value that makes q=sqrt(b).
Returns
-------
pdf : (N,) ndarray,
Probability density function of the water displacement.
References
----------
.. [1] Cheng J., "Estimation and Processing of Ensemble Average Propagator
and its Features in Diffusion MRI", PhD Thesis, 2012.
'''
mf = [f / 100. for f in fractions]
angles = np.array(angles)
if angles.shape[-1] == 3:
sticks = angles
else:
sticks = [
sphere2cart(1, np.deg2rad(pair[0]), np.deg2rad(pair[1]))
for pair in angles
]
sticks = np.array(sticks)
pdf = np.zeros(len(pdf_points))
mevecs = []
for s in sticks:
mevecs += [all_tensor_evecs(s).T]
for j, f in enumerate(mf):
pdf += f * single_tensor_pdf(
pdf_points, evals=mevals[j], evecs=mevecs[j], tau=tau)
return pdf
def multi_tensor_odf(odf_verts, mevals, angles, fractions):
r'''Simulate a Multi-Tensor ODF.
Parameters
----------
odf_verts : (N,3) ndarray
Vertices of the reconstruction sphere.
mevals : sequence of 1D arrays,
Eigen-values for each tensor.
angles : sequence of 2d tuples,
Sequence of principal directions for each tensor in polar angles
or cartesian unit coordinates.
fractions : sequence of floats,
Percentages of the fractions for each tensor.
Returns
-------
ODF : (N,) ndarray
Orientation distribution function.
Examples
--------
Simulate a MultiTensor ODF with two peaks and calculate its exact ODF.
>>> import numpy as np
>>> from dipy.sims.voxel import multi_tensor_odf, all_tensor_evecs
>>> from dipy.data import get_sphere
>>> sphere = get_sphere('symmetric724')
>>> vertices, faces = sphere.vertices, sphere.faces
>>> mevals = np.array(([0.0015, 0.0003, 0.0003],[0.0015, 0.0003, 0.0003]))
>>> angles = [(0, 0), (90, 0)]
>>> odf = multi_tensor_odf(vertices, mevals, angles, [50, 50])
'''
mf = [f / 100. for f in fractions]
angles = np.array(angles)
if angles.shape[-1] == 3:
sticks = angles
else:
sticks = [sphere2cart(1, np.deg2rad(pair[0]), np.deg2rad(pair[1]))
for pair in angles]
sticks = np.array(sticks)
odf = np.zeros(len(odf_verts))
mevecs = []
for s in sticks:
mevecs += [all_tensor_evecs(s).T]
for (j, f) in enumerate(mf):
odf += f * single_tensor_odf(odf_verts,
evals=mevals[j], evecs=mevecs[j])
return odf
def radon_params(self, ang_res=64):
#calculate radon integration parameters
phis = np.linspace(0, 2 * np.pi, ang_res)[:-1]
planars = []
for phi in phis:
planars.append(sphere2cart(1, np.pi / 2, phi))
planars = np.array(planars)
planarsR = []
for v in self.odf_vertices:
R = vec2vec_rotmat(np.array([0, 0, 1]), v)
planarsR.append(np.dot(R, planars.T).T)
self.equators = planarsR
self.equatorn = len(phis)
def radon_params(self,ang_res=64):
#calculate radon integration parameters
phis=np.linspace(0,2*np.pi,ang_res)[:-1]
planars=[]
for phi in phis:
planars.append(sphere2cart(1,np.pi/2,phi))
planars=np.array(planars)
planarsR=[]
for v in self.odf_vertices:
R=vec2vec_rotmat(np.array([0,0,1]),v)
planarsR.append(np.dot(R,planars.T).T)
self.equators=planarsR
self.equatorn=len(phis)
def multi_tensor_pdf(pdf_points, mevals, angles, fractions,
tau=1 / (4 * np.pi ** 2)):
r'''Simulate a Multi-Tensor ODF.
Parameters
----------
pdf_points : (N, 3) ndarray
Points to evaluate the PDF.
mevals : sequence of 1D arrays,
Eigen-values for each tensor. By default, values typical for prolate
white matter are used.
angles : sequence,
Sequence of principal directions for each tensor in polar angles
or cartesian unit coordinates.
fractions : sequence of floats,
Percentages of the fractions for each tensor.
tau : float,
diffusion time. By default the value that makes q=sqrt(b).
Returns
-------
pdf : (N,) ndarray,
Probability density function of the water displacement.
References
----------
.. [1] Cheng J., "Estimation and Processing of Ensemble Average Propagator
and its Features in Diffusion MRI", PhD Thesis, 2012.
'''
mf = [f / 100. for f in fractions]
angles = np.array(angles)
if angles.shape[-1] == 3:
sticks = angles
else:
sticks = [sphere2cart(1, np.deg2rad(pair[0]), np.deg2rad(pair[1]))
for pair in angles]
sticks = np.array(sticks)
pdf = np.zeros(len(pdf_points))
mevecs = []
for s in sticks:
mevecs += [all_tensor_evecs(s).T]
for j, f in enumerate(mf):
pdf += f * single_tensor_pdf(pdf_points,
evals=mevals[j], evecs=mevecs[j], tau=tau)
return pdf
def plot_tensor_3d(Tensor, cmap='jet', mode='ADC', file_name=None,
origin=[0,0,0], colorbar=False, figure=None, vmin=None,
vmax=None, offset=0, azimuth=60, elevation=90, roll=0,
scale_factor=1.0, rgb_pdd=False):
"""
mode: either "ADC", "ellipse" or "pred_sig"
"""
Q = Tensor.Q
sphere = create_unit_sphere(5)
vertices = sphere.vertices
faces = sphere.faces
x,y,z = vertices.T
new_bvecs = np.vstack([x.ravel(), y.ravel(), z.ravel()])
Tensor = ozt.Tensor(Q, new_bvecs,
Tensor.bvals[0] * np.ones(new_bvecs.shape[-1]))
if mode == 'ADC':
v = Tensor.ADC * scale_factor
elif mode == 'ellipse':
v = Tensor.diffusion_distance * scale_factor
elif mode == 'pred_sig':
v = Tensor.predicted_signal(1) * scale_factor
else:
raise ValueError("Mode not recognized")
r, phi, theta = geo.cart2sphere(x,y,z)
x_plot, y_plot, z_plot = geo.sphere2cart(v, phi, theta)
if rgb_pdd:
evals, evecs = Tensor.decompose
xyz = evecs[0]
r = np.abs(xyz[0])/np.sum(np.abs(xyz))
g = np.abs(xyz[1])/np.sum(np.abs(xyz))
b = np.abs(xyz[2])/np.sum(np.abs(xyz))
color = (r, g, b)
else:
color = None
# Call and return straightaway:
return _display_maya_voxel(x_plot, y_plot, z_plot, faces, v, origin,
cmap=cmap, colorbar=colorbar, color=color,
figure=figure,
vmin=vmin, vmax=vmax, file_name=file_name,
azimuth=azimuth, elevation=elevation)
def get_mask(vert, hori):
delta_v = 45 / 240 # vertical leads to row
delta_h = 60 / 320
ds = math.pi / 180 # degree scale
mask = np.zeros((vert, hori, 3))
for i in range(vert): #first scanning row
for k in range(hori): #scanning column
# above convert one pixel into sphere coordinate
phi = (67.5 + delta_v / 2 + delta_v *
i) * ds #relate to row, which is the latitude, phi, 0 to pi
lmd = (-30 + delta_h / 2 + delta_h *
k) * ds #relate to column, which is the longitude, xy,lmd
v = sphere2cart(1, phi, lmd)
mask[i, k, :] = [v[0].tolist(), v[1].tolist(), v[2].tolist()]
return mask #in cartesian
def scale_bvecs_by_sig(bvecs, sig):
"""
Helper function to rescale your bvecs according to some signal, so that
they don't fall on the unit sphere, but instead are represented in space as
distance from the origin.
"""
x, y, z = bvecs
r, theta, phi = geo.cart2sphere(x, y, z)
# Simply replace r with sig:
x, y, z = geo.sphere2cart(sig, theta, phi)
return x, y, z
def scale_bvecs_by_sig(bvecs, sig):
"""
Helper function to rescale your bvecs according to some signal, so that
they don't fall on the unit sphere, but instead are represented in space as
distance from the origin.
"""
x,y,z = bvecs
r, theta, phi = geo.cart2sphere(x, y, z)
# Simply replace r with sig:
x,y,z = geo.sphere2cart(sig, theta, phi)
return x,y,z
def random_uniform_on_sphere(n=1, coords='xyz'):
r'''Random unit vectors from a uniform distribution on the sphere.
Parameters
-----------
n : int
Number of random vectors
coords : {'xyz', 'radians', 'degrees'}
'xyz' for cartesian form
'radians' for spherical form in rads
'degrees' for spherical form in degrees
Notes
------
The uniform distribution on the sphere, parameterized by spherical
coordinates $(\theta, \phi)$, should verify $\phi\sim U[0,2\pi]$, while
$z=\cos(\theta)\sim U[-1,1]$.
References
-----------
.. [1] http://mathworld.wolfram.com/SpherePointPicking.html.
Returns
--------
X : array, shape (n,3) if coords='xyz' or shape (n,2) otherwise
Uniformly distributed vectors on the unit sphere.
Examples
---------
>>> from dipy.core.sphere_stats import random_uniform_on_sphere
>>> X = random_uniform_on_sphere(4, 'radians')
>>> X.shape == (4, 2)
True
>>> X = random_uniform_on_sphere(4, 'xyz')
>>> X.shape == (4, 3)
True
'''
z = np.random.uniform(-1, 1, n)
theta = np.arccos(z)
phi = np.random.uniform(0, 2*np.pi, n)
if coords == 'xyz':
r = np.ones(n)
return np.vstack(geometry.sphere2cart(r, theta, phi)).T
angles = np.vstack((theta, phi)).T
if coords == 'radians':
return angles
if coords == 'degrees':
return np.rad2deg(angles)
def random_uniform_on_sphere(n=1, coords='xyz'):
r"""Random unit vectors from a uniform distribution on the sphere.
Parameters
-----------
n : int
Number of random vectors
coords : {'xyz', 'radians', 'degrees'}
'xyz' for cartesian form
'radians' for spherical form in rads
'degrees' for spherical form in degrees
Notes
------
The uniform distribution on the sphere, parameterized by spherical
coordinates $(\theta, \phi)$, should verify $\phi\sim U[0,2\pi]$, while
$z=\cos(\theta)\sim U[-1,1]$.
References
-----------
.. [1] http://mathworld.wolfram.com/SpherePointPicking.html.
Returns
--------
X : array, shape (n,3) if coords='xyz' or shape (n,2) otherwise
Uniformly distributed vectors on the unit sphere.
Examples
---------
>>> from dipy.core.sphere_stats import random_uniform_on_sphere
>>> X = random_uniform_on_sphere(4, 'radians')
>>> X.shape == (4, 2)
True
>>> X = random_uniform_on_sphere(4, 'xyz')
>>> X.shape == (4, 3)
True
"""
z = np.random.uniform(-1, 1, n)
theta = np.arccos(z)
phi = np.random.uniform(0, 2 * np.pi, n)
if coords == 'xyz':
r = np.ones(n)
return np.vstack(geometry.sphere2cart(r, theta, phi)).T
angles = np.vstack((theta, phi)).T
if coords == 'radians':
return angles
if coords == 'degrees':
return np.rad2deg(angles)
def _tensor_helper(self, theta, phi):
"""
This code is used in all three error functions, so we write it out
only once here
"""
# Convert to cartesian coordinates:
x, y, z = geo.sphere2cart(1, theta, phi)
bvec = [x, y, z]
# decompose is an auto-attr of the tensor object, so is only run once
# and then cached:
evals, evecs = self.response_function.decompose
rot = ozu.calculate_rotation(bvec, evecs[0])
rot_evecs = evecs * rot
rot_tensor = ozt.tensor_from_eigs(rot_evecs, evals,
self.bvecs[:, self.b_idx],
self.bvals[self.b_idx])
return rot_tensor
def simulations_dipy(bvals,gradients,d=0.0015,S0=100,angles=[(0,0),(90,0)],fractions=[35,35],snr=20):
fractions=[f/100. for f in fractions]
f0=1-np.sum(fractions)
S=np.zeros(len(gradients))
sticks=[ sphere2cart(1,np.deg2rad(pair[0]),np.deg2rad(pair[1])) for pair in angles]
sticks=np.array(sticks)
for (i,g) in enumerate(gradients[1:]):
S[i+1]=f0*np.exp(-bvals[i+1]*d)+ np.sum([fractions[j]*np.exp(-bvals[i+1]*d*np.dot(s,g)**2) for (j,s) in enumerate(sticks)])
S[i+1]=S0*S[i+1]
S[0]=S0
if snr!=None:
std=S0/snr
S=S+np.random.randn(len(S))*std
return S,sticks
def _tensor_helper(self, theta, phi):
"""
This code is used in all three error functions, so we write it out
only once here
"""
# Convert to cartesian coordinates:
x,y,z = geo.sphere2cart(1, theta, phi)
bvec = [x,y,z]
# decompose is an auto-attr of the tensor object, so is only run once
# and then cached:
evals, evecs = self.response_function.decompose
rot = ozu.calculate_rotation(bvec, evecs[0])
rot_evecs = evecs * rot
rot_tensor = ozt.tensor_from_eigs(rot_evecs,
evals,
self.bvecs[:,self.b_idx],
self.bvals[:,self.b_idx])
return rot_tensor
def SticksAndBall(bvals,gradients,d=0.0015,S0=100,angles=[(0,0),(90,0)],fractions=[35,35],snr=20):
""" Simulating the signal for a Sticks & Ball model
Based on the paper by Tim Behrens, H.J. Berg, S. Jbabdi, "Probabilistic Diffusion Tractography with multiple fiber orientations
what can we gain?", Neuroimage, 2007.
Parameters
-----------
bvals : array, shape (N,)
gradients : array, shape (N,3) also known as bvecs
d : diffusivity value
S0 : unweighted signal value
angles : array (K,2) list of polar angles (in degrees) for the sticks
or array (K,3) with sticks as Cartesian unit vectors and K the number of sticks
fractions : percentage of each stick
snr : signal to noise ration assuming gaussian noise. Provide None for no noise.
Returns
--------
S : simulated signal
sticks : sticks in cartesian coordinates
"""
fractions=[f/100. for f in fractions]
f0=1-np.sum(fractions)
S=np.zeros(len(gradients))
angles=np.array(angles)
if angles.shape[-1]==3:
sticks=angles
if angles.shape[-1]==2:
sticks=[ sphere2cart(1,np.deg2rad(pair[0]),np.deg2rad(pair[1])) for pair in angles]
sticks=np.array(sticks)
for (i,g) in enumerate(gradients[1:]):
S[i+1]=f0*np.exp(-bvals[i+1]*d)+ np.sum([fractions[j]*np.exp(-bvals[i+1]*d*np.dot(s,g)**2) for (j,s) in enumerate(sticks)])
S[i+1]=S0*S[i+1]
S[0]=S0
if snr!=None:
std=S0/snr
S=S+np.random.randn(len(S))*std
return S,sticks
def rotate_needles(needles=[(0,0),(90,0),(90,90)],angles=(90,0,90)):
theta,phi,psi=angles
rot_theta = np.deg2rad(theta)
rot_phi = np.deg2rad(phi)
rot_psi = np.deg2rad(psi)
#ANGLES FOR ROTATION IN EULER
v = np.array([np.cos(rot_phi)*np.sin(rot_theta),np.sin(rot_phi)*np.sin(rot_theta),np.cos(rot_theta)])
#print v
k_cross = np.array([[0,-v[2],v[1]],[v[2],0,-v[0]],[-v[1],v[0],0]])
#print k_cross
#rot = rotation_vec2mat(r)
rot = np.eye(3)+np.sin(rot_psi)*k_cross+(1-np.cos(rot_psi))*np.dot(k_cross,k_cross)
#print rot
#needles=[(0,0),(90,0),(90,90)]
#INITIAL NEEDLES in (theta, phi)
needle_rads = []
for n in needles:
needle_rads+=[[1]+list(np.deg2rad(n))]
#print needle_rads
#initial needles in (r, theta, phi)
rad_needles=[]
for n in needle_rads:
rad_needles+=[cart2sphere(*tuple(np.dot(rot,sphere2cart(*n))))[1:3]]
#print rad_needles
#ROTATED NEEDLES IN SPHERICAL COORDS (RADIANS)
deg_angles = []
for n in rad_needles:
a = []
for c in n:
a+=[np.rad2deg(c)]
deg_angles += [a]
#print deg_angles
#ROTATED NEEDLES IN SPHERICAL COORDS (DEGREES)
return deg_angles
def plot_signal(bvecs, signal, origin=[0,0,0],
maya=True, cmap='jet', file_name=None,
colorbar=False, figure=None, vmin=None, vmax=None,
offset=0, azimuth=60, elevation=90, roll=0):
"""
Interpolate a measured signal, using RBF interpolation.
Parameters
----------
signal:
bvecs: array (3,n)
the x,y,z locations where the signal was measured
offset : float
where to place the plotted voxel (on the z axis)
"""
s0 = Sphere(xyz=bvecs.T)
vertices = s0.vertices
faces = s0.faces
x,y,z = vertices.T
r, phi, theta = geo.cart2sphere(x,y,z)
x_plot, y_plot, z_plot = geo.sphere2cart(signal, phi, theta)
# Call and return straightaway:
return _display_maya_voxel(x_plot, y_plot, z_plot, faces,
signal, origin,
cmap=cmap, colorbar=colorbar, figure=figure,
vmin=vmin, vmax=vmax, file_name=file_name,
azimuth=azimuth, elevation=elevation)
def compare_dni_dsi_gqi_gqi2_eit():
#stop
btable=np.loadtxt(get_data('dsi515btable'))
bvals=btable[:,0]
bvecs=btable[:,1:]
type=3
axesn=200
SNR=20
data,gold,angs,anglesn,axesn,angles=generate_gold_data(bvals,bvecs,fibno=type,axesn=axesn,SNR=SNR)
gq=GeneralizedQSampling(data,bvals,bvecs,1.2,odf_sphere='symmetric642',squared=False,save_odfs=True)
gq2=GeneralizedQSampling(data,bvals,bvecs,3.,odf_sphere='symmetric642',squared=True,save_odfs=True)
ds=DiffusionSpectrum(data,bvals,bvecs,odf_sphere='symmetric642',auto=False,save_odfs=True)
ds.filter_width=32.
ds.update()
ds.fit()
ei=EquatorialInversion(data,bvals,bvecs,odf_sphere='symmetric642',auto=False,save_odfs=True,fast=True)
ei.radius=np.arange(0,5,0.1)
ei.gaussian_weight=None#0.01
ei.set_operator('laplacian')
ei.update()
ei.fit()
ei2=EquatorialInversion(data,bvals,bvecs,odf_sphere='symmetric642',auto=False,save_odfs=True,fast=True)
ei2.radius=np.arange(0,5,0.1)
ei2.gaussian_weight=None
ei2.set_operator('laplap')
ei2.update()
ei2.fit()
ei3=EquatorialInversion(data,bvals,bvecs,odf_sphere='symmetric642',auto=False,save_odfs=True,fast=True)
ei3.radius=np.arange(0,5,0.1)
ei3.gaussian_weight=None
ei3.set_operator('signal')
ei3.update()
ei3.fit()
"""
blobs=np.zeros((2,4,642))
no=200
blobs[0,0,:]=gq.ODF[no]
blobs[0,1,:]=gq2.ODF[no]
blobs[0,2,:]=ds.ODF[no]
blobs[0,3,:]=ei.ODF[no]
no=399
blobs[1,0,:]=gq.ODF[no]
blobs[1,1,:]=gq2.ODF[no]
blobs[1,2,:]=ds.ODF[no]
blobs[1,3,:]=ei.ODF[no]
show_blobs(blobs[None,:,:,:],ei.odf_vertices,ei.odf_faces,1.2)
"""
#stop
vts=gq.odf_vertices
def simple_peaks(ODF,faces,thr):
x,g=ODF.shape
PK=np.zeros((x,5))
IN=np.zeros((x,5))
for (i,odf) in enumerate(ODF):
peaks,inds=peak_finding(odf,faces)
ibigp=np.where(peaks>thr*peaks[0])[0]
l=len(ibigp)
if l>3:
l=3
PK[i,:l]=peaks[:l]
IN[i,:l]=inds[:l]
return PK,IN
thresh=0.5
PK,IN=simple_peaks(ds.ODF,ds.odf_faces,thresh)
res,me,st =do_compare(gold,vts,PK,IN,0,anglesn,axesn,type)
PK,IN=simple_peaks(gq.ODF,ds.odf_faces,thresh)
res2,me2,st2 =do_compare(gold,vts,PK,IN,0,anglesn,axesn,type)
PK,IN=simple_peaks(gq2.ODF,ds.odf_faces,thresh)
res3,me3,st3 =do_compare(gold,vts,PK,IN,0,anglesn,axesn,type)
PK,IN=simple_peaks(ei.ODF,ds.odf_faces,thresh)
res4,me4,st4 =do_compare(gold,vts,PK,IN,0,anglesn,axesn,type)
PK,IN=simple_peaks(ei2.ODF,ds.odf_faces,thresh)
res5,me5,st5 =do_compare(gold,vts,PK,IN,0,anglesn,axesn,type)
PK,IN=simple_peaks(ei3.ODF,ei3.odf_faces,thresh)
res6,me6,st6 =do_compare(gold,vts,PK,IN,0,anglesn,axesn,type)
#res7,me7,st7 =do_compare(ei2.PK,ei2.IN,0,anglesn,axesn,type)
if type==1:
plt.figure(1)
plt.plot(res,'r',label='DSI')
plt.plot(res2,'g',label='GQI')
plt.plot(res3,'b',label='GQI2')
plt.plot(res4,'k',label='EITL')
plt.plot(res5,'k--',label='EITL2')
plt.plot(res6,'k-.',label='EITS')
#plt.xlabel('angle')
#plt.ylabel('resolution')
#plt.title('Angular accuracy')
plt.legend()
plt.show()
else:
if type==3:
x,y,z=sphere2cart(np.ones(len(angles)),np.deg2rad(angles),np.zeros(len(angles)))
x2,y2,z2=sphere2cart(np.ones(len(angles)),np.deg2rad(angles),np.deg2rad(120*np.ones(len(angles))))
angles2=[]
for i in range(len(x)):
angles2.append(np.rad2deg(np.arccos(np.dot([x[i],y[i],z[i]],[x2[i],y2[i],z2[i]]))))
angles=angles2
plt.figure(1)
plt.plot(angles,me,'r',linewidth=3.,label='DSI')
plt.plot(angles,me2,'g',linewidth=3.,label='GQI')
plt.plot(angles,me3,'b',linewidth=3.,label='GQI2')
plt.plot(angles,me4,'k',linewidth=3.,label='EITL')
plt.plot(angles,me5,'k--',linewidth=3.,label='EITL2')
plt.plot(angles,me6,'k-.',linewidth=3.,label='EITS')
#plt.plot(angles,me7,'r--',linewidth=3.,label='EITL2')
plt.xlabel('angle')
plt.ylabel('similarity')
title='Angular similarity of ' + str(type) + '-fibres crossing with SNR ' + str(SNR)
plt.title(title)
plt.legend(loc='center right')
plt.savefig('/tmp/test.png',dpi=300)
plt.show()
def sticks_and_ball(bvals,
gradients,
d=0.0015,
S0=100,
angles=[(0, 0), (90, 0)],
fractions=[35, 35],
snr=20):
""" Simulate the signal for a Sticks & Ball model.
Parameters
-----------
bvals : (N,) ndarray
B-values for measurements.
gradients : (N,3) ndarray
Also known as b-vectors.
d : float
Diffusivity value.
S0 : float
Unweighted signal value.
angles : array (K,2) or (M,3)
List of K polar angles (in degrees) for the sticks or array of M
sticks as Cartesian unit vectors.
fractions : float
Percentage of each stick.
snr : float
Signal to noise ratio, assuming gaussian noise. If set to None, no
noise is added.
Returns
--------
S : (N,) ndarray
Simulated signal.
sticks : (M,3)
Sticks in cartesian coordinates.
References
----------
.. [1] Behrens et al., "Probabilistic diffusion
tractography with multiple fiber orientations: what can we gain?",
Neuroimage, 2007.
"""
fractions = [f / 100. for f in fractions]
f0 = 1 - np.sum(fractions)
S = np.zeros(len(gradients))
angles = np.array(angles)
if angles.shape[-1] == 3:
sticks = angles
else:
sticks = [
sphere2cart(1, np.deg2rad(pair[0]), np.deg2rad(pair[1]))
for pair in angles
]
sticks = np.array(sticks)
for (i, g) in enumerate(gradients[1:]):
S[i + 1] = f0 * np.exp(-bvals[i+1] * d) + \
np.sum([
fractions[j] * np.exp(-bvals[i + 1] * d * np.dot(s, g)**2)
for (j,s) in enumerate(sticks)
])
S[i + 1] = S0 * S[i + 1]
S[0] = S0
if snr is not None:
std = S0 / snr
S = S + np.random.randn(len(S)) * std
return S, sticks
def sticks_and_ball(bvals, gradients, d=0.0015, S0=100, angles=[(0,0), (90,0)],
fractions=[35,35], snr=20):
""" Simulate the signal for a Sticks & Ball model.
Parameters
-----------
bvals : (N,) ndarray
B-values for measurements.
gradients : (N,3) ndarray
Also known as b-vectors.
d : float
Diffusivity value.
S0 : float
Unweighted signal value.
angles : array (K,2) or (M,3)
List of K polar angles (in degrees) for the sticks or array of M
sticks as Cartesian unit vectors.
fractions : float
Percentage of each stick.
snr : float
Signal to noise ratio, assuming gaussian noise. If set to None, no
noise is added.
Returns
--------
S : (N,) ndarray
Simulated signal.
sticks : (M,3)
Sticks in cartesian coordinates.
References
----------
.. [1] Behrens et al., "Probabilistic diffusion
tractography with multiple fiber orientations: what can we gain?",
Neuroimage, 2007.
"""
fractions = [f / 100. for f in fractions]
f0 = 1 - np.sum(fractions)
S = np.zeros(len(gradients))
angles=np.array(angles)
if angles.shape[-1] == 3:
sticks = angles
else:
sticks = [sphere2cart(1, np.deg2rad(pair[0]), np.deg2rad(pair[1]))
for pair in angles]
sticks = np.array(sticks)
for (i, g) in enumerate(gradients[1:]):
S[i + 1] = f0 * np.exp(-bvals[i+1] * d) + \
np.sum([
fractions[j] * np.exp(-bvals[i + 1] * d * np.dot(s, g)**2)
for (j,s) in enumerate(sticks)
])
S[i + 1] = S0 * S[i + 1]
S[0] = S0
if snr is not None:
std = S0 / snr
S = S + np.random.randn(len(S)) * std
return S, sticks
def test_kurtosis_maximum():
# TEST 1
# simulate a crossing fibers interserting at 70 degrees. The first fiber
# is aligned to the x-axis while the second fiber is aligned to the x-z
# plane with an angular deviation of 70 degrees from the first one.
# According to Neto Henriques et al, 2015 (NeuroImage 111: 85-99), the
# kurtosis tensor of this simulation will have a maxima aligned to axis y
angles = [(90, 0), (90, 0), (20, 0), (20, 0)]
signal_70, dt_70, kt_70 = multi_tensor_dki(gtab_2s, mevals_cross, S0=100,
angles=angles,
fractions=frac_cross, snr=None)
# prepare inputs
dkiM = dki.DiffusionKurtosisModel(gtab_2s, fit_method="WLS")
dkiF = dkiM.fit(signal_70)
MD = dkiF.md
kt = dkiF.kt
R = dkiF.evecs
evals = dkiF.evals
dt = lower_triangular(np.dot(np.dot(R, np.diag(evals)), R.T))
sphere = get_sphere('symmetric724')
# compute maxima
k_max_cross, max_dir = dki._voxel_kurtosis_maximum(dt, MD, kt, sphere,
gtol=1e-5)
yaxis = np.array([0., 1., 0.])
cos_angle = np.abs(np.dot(max_dir[0], yaxis))
assert_almost_equal(cos_angle, 1.)
# TEST 2
# test the function on cases of well aligned fibers oriented in a random
# defined direction. According to Neto Henriques et al, 2015 (NeuroImage
# 111: 85-99), the maxima of kurtosis is any direction perpendicular to the
# fiber direction. Moreover, according to multicompartmetal simulations,
# kurtosis in this direction has to be equal to:
fie = 0.49
ADi = 0.00099
ADe = 0.00226
RDi = 0
RDe = 0.00087
RD = fie*RDi + (1-fie)*RDe
RK = 3 * fie * (1-fie) * ((RDi-RDe) / RD) ** 2
# prepare simulation:
theta = random.uniform(0, 180)
phi = random.uniform(0, 320)
angles = [(theta, phi), (theta, phi)]
mevals = np.array([[ADi, RDi, RDi], [ADe, RDe, RDe]])
frac = [fie*100, (1 - fie)*100]
signal, dt, kt = multi_tensor_dki(gtab_2s, mevals, angles=angles,
fractions=frac, snr=None)
# prepare inputs
dkiM = dki.DiffusionKurtosisModel(gtab_2s, fit_method="WLS")
dkiF = dkiM.fit(signal)
MD = dkiF.md
kt = dkiF.kt
R = dkiF.evecs
evals = dkiF.evals
dt = lower_triangular(np.dot(np.dot(R, np.diag(evals)), R.T))
# compute maxima
k_max, max_dir = dki._voxel_kurtosis_maximum(dt, MD, kt, sphere, gtol=1e-5)
# check if max direction is perpendicular to fiber direction
fdir = np.array([sphere2cart(1., np.deg2rad(theta), np.deg2rad(phi))])
cos_angle = np.abs(np.dot(max_dir[0], fdir[0]))
assert_almost_equal(cos_angle, 0., decimal=5)
# check if max direction is equal to expected value
assert_almost_equal(k_max, RK)
# According to Neto Henriques et al., 2015 (NeuroImage 111: 85-99),
# e.g. see figure 1 of this article, kurtosis maxima for the first test is
# also equal to the maxima kurtosis value of well-aligned fibers, since
# simulations parameters (apart from fiber directions) are equal
assert_almost_equal(k_max_cross, RK)
# Test 3 - Test performance when kurtosis is spherical - this case, can be
# problematic since a spherical kurtosis does not have an maximum
k_max, max_dir = dki._voxel_kurtosis_maximum(dt_sph, np.mean(evals_sph),
kt_sph, sphere, gtol=1e-2)
assert_almost_equal(k_max, Kref_sphere)
# Test 4 - Test performance when kt have all elements zero - this case, can
# be problematic this case does not have an maximum
k_max, max_dir = dki._voxel_kurtosis_maximum(dt_sph, np.mean(evals_sph),
np.zeros(15), sphere,
gtol=1e-2)
assert_almost_equal(k_max, 0.0)
def plot_tensor_3d(Tensor,
cmap='jet',
mode='ADC',
file_name=None,
origin=[0, 0, 0],
colorbar=False,
figure=None,
vmin=None,
vmax=None,
offset=0,
azimuth=60,
elevation=90,
roll=0,
scale_factor=1.0,
rgb_pdd=False):
"""
mode: either "ADC", "ellipse" or "pred_sig"
"""
Q = Tensor.Q
sphere = create_unit_sphere(5)
vertices = sphere.vertices
faces = sphere.faces
x, y, z = vertices.T
new_bvecs = np.vstack([x.ravel(), y.ravel(), z.ravel()])
Tensor = ozt.Tensor(Q, new_bvecs,
Tensor.bvals[0] * np.ones(new_bvecs.shape[-1]))
if mode == 'ADC':
v = Tensor.ADC * scale_factor
elif mode == 'ellipse':
v = Tensor.diffusion_distance * scale_factor
elif mode == 'pred_sig':
v = Tensor.predicted_signal(1) * scale_factor
else:
raise ValueError("Mode not recognized")
r, phi, theta = geo.cart2sphere(x, y, z)
x_plot, y_plot, z_plot = geo.sphere2cart(v, phi, theta)
if rgb_pdd:
evals, evecs = Tensor.decompose
xyz = evecs[0]
r = np.abs(xyz[0]) / np.sum(np.abs(xyz))
g = np.abs(xyz[1]) / np.sum(np.abs(xyz))
b = np.abs(xyz[2]) / np.sum(np.abs(xyz))
color = (r, g, b)
else:
color = None
# Call and return straightaway:
return _display_maya_voxel(x_plot,
y_plot,
z_plot,
faces,
v,
origin,
cmap=cmap,
colorbar=colorbar,
color=color,
figure=figure,
vmin=vmin,
vmax=vmax,
file_name=file_name,
azimuth=azimuth,
elevation=elevation)
def sig_on_sphere(bvecs, val, fig=None, sphere_dim=1000, r_from_val=False,
**kwargs):
"""
Presente values on a sphere.
Parameters
----------
bvecs: co-linear to the array with data, the theta, phi on the circle
from which the data was taken
val: array with data
fig: matplotlib figure, optional. Default: make new figure
sphere_dim: The data will be interpolated into a sphere_dim by sphere_dim
grid
r_from_val: Whether to double-code the values both by color and by the
distance from the center
cmap: Specify a matplotlib colormap to use for coloring the data.
Additional kwargs can be passed to the matplotlib.pyplot.plot3D command.
"""
# We don't need the r output
_, theta, phi = geo.cart2sphere(bvecs[0], bvecs[1], bvecs[2])
if fig is None:
fig = plt.figure()
ax = fig.add_subplot(1,1,1, projection='3d')
# Get the cmap argument out of your kwargs, or use the default:
cmap = kwargs.pop('cmap', matplotlib.cm.RdBu)
u = np.linspace(0, 2 * np.pi, sphere_dim)
v = np.linspace(0, np.pi, sphere_dim)
x_inter = 10 * np.outer(np.cos(u), np.sin(v))
y_inter = 10 * np.outer(np.sin(u), np.sin(v))
inter_val = np.zeros(x_inter.shape).ravel()
z_inter = np.outer(np.ones(np.size(u)), np.cos(v))
grid_r, grid_theta, grid_phi = geo.cart2sphere(x_inter, y_inter, z_inter)
for idx, this in enumerate(zip(grid_theta.ravel(), grid_phi.ravel())):
this_theta = np.abs(theta - np.array(this[0]))
this_phi = np.abs(phi - np.array(this[1]))
# The closest theta and phi:
min_idx = np.argmin(this_theta + this_phi)
inter_val[idx] = val[min_idx]
# Calculate the values from the colormap that will be used for the
# face-colors:
c = np.array(color_from_val(inter_val,
min_val=np.min(val),
max_val=np.max(val),
cmap_or_lut=cmap))
new_shape = (x_inter.shape + (3,))
c = np.reshape(c, new_shape)
if r_from_val:
# Use the interpolated values to set z:
new_x, new_y, new_z = geo.sphere2cart(inter_val.reshape(x_inter.shape),
grid_theta, grid_phi)
ax.plot_surface(new_x, new_y, new_z, rstride=4, cstride=4,
facecolors=c, **kwargs)
else:
ax.plot_surface(x_inter, y_inter, z_inter, rstride=4, cstride=4,
facecolors=c, **kwargs)
return fig
def sig_on_sphere(bvecs,
val,
fig=None,
sphere_dim=1000,
r_from_val=False,
**kwargs):
"""
Presente values on a sphere.
Parameters
----------
bvecs: co-linear to the array with data, the theta, phi on the circle
from which the data was taken
val: array with data
fig: matplotlib figure, optional. Default: make new figure
sphere_dim: The data will be interpolated into a sphere_dim by sphere_dim
grid
r_from_val: Whether to double-code the values both by color and by the
distance from the center
cmap: Specify a matplotlib colormap to use for coloring the data.
Additional kwargs can be passed to the matplotlib.pyplot.plot3D command.
"""
# We don't need the r output
_, theta, phi = geo.cart2sphere(bvecs[0], bvecs[1], bvecs[2])
if fig is None:
fig = plt.figure()
ax = fig.add_subplot(1, 1, 1, projection='3d')
# Get the cmap argument out of your kwargs, or use the default:
cmap = kwargs.pop('cmap', matplotlib.cm.RdBu)
u = np.linspace(0, 2 * np.pi, sphere_dim)
v = np.linspace(0, np.pi, sphere_dim)
x_inter = 10 * np.outer(np.cos(u), np.sin(v))
y_inter = 10 * np.outer(np.sin(u), np.sin(v))
inter_val = np.zeros(x_inter.shape).ravel()
z_inter = np.outer(np.ones(np.size(u)), np.cos(v))
grid_r, grid_theta, grid_phi = geo.cart2sphere(x_inter, y_inter, z_inter)
for idx, this in enumerate(zip(grid_theta.ravel(), grid_phi.ravel())):
this_theta = np.abs(theta - np.array(this[0]))
this_phi = np.abs(phi - np.array(this[1]))
# The closest theta and phi:
min_idx = np.argmin(this_theta + this_phi)
inter_val[idx] = val[min_idx]
# Calculate the values from the colormap that will be used for the
# face-colors:
c = np.array(
color_from_val(inter_val,
min_val=np.min(val),
max_val=np.max(val),
cmap_or_lut=cmap))
new_shape = (x_inter.shape + (3, ))
c = np.reshape(c, new_shape)
if r_from_val:
# Use the interpolated values to set z:
new_x, new_y, new_z = geo.sphere2cart(inter_val.reshape(x_inter.shape),
grid_theta, grid_phi)
ax.plot_surface(new_x,
new_y,
new_z,
rstride=4,
cstride=4,
facecolors=c,
**kwargs)
else:
ax.plot_surface(x_inter,
y_inter,
z_inter,
rstride=4,
cstride=4,
facecolors=c,
**kwargs)
return fig
def spkm(data,
k,
weights=None,
seeds=None,
antipodal=True,
max_iter=1000,
calc_sse=True):
"""
Spherical k means.
Parameters
----------
data : 2d float array
Unit vectors on the hyper-sphere. This array has n data points rows, by
m feature columns.
k : int
The number of clusters
weights : 1d float array
Some data-points may be more important than others, so they will receive
more weighting in determining the centroids
seeds : float array (optional).
If n by k array is provided, these are used as centroids to initialize
the algorithm. Otherwise, random centroids are chosen
antipodal : bool
In cases in which antipodal symmetry can be assumed, we want to cluster
together points that are pointing close to *opposite* directions. In that
case, correlations between putative centroids and each data point treats
correlation and anti-correlation in equal vein.
max_iter : int
If you run this many iterations without convergence, warn and exit.
calc_sse : bool
Whether to calculate SSE or not.
Returns
-------
mu : the estimated centroid
y_n : assignments of each data point to a centroid
SSE : the sum of squared error in centroid-to-data-point assignment
"""
# 0. Preliminaries:
# For the calculation of the centroids, we want to make sure that the data
# are all pointing into the same hemisphere (expects 3 by n):
data = ozu.vecs2hemi(data.T).T
# If no weights are provided treat all data points equally:
if weights is None:
weights = np.ones(data.shape[0])
# 1. Initialization:
if seeds is None:
# Choose random seeds.
# thetas are uniform [0,pi]:
theta = np.random.rand(k) * np.pi
# phis are uniform [0, 2pi]
phi = np.random.rand(k) * 2 * np.pi
# They're all unit vectors:
r = np.ones(k)
# et voila:
seeds = np.array(geo.sphere2cart(theta, phi, r)).T
mu = seeds.copy()
is_changing = True
last_y_n = False
iter = 0
while is_changing:
# Make sure they're all unit vectors, so that correlation below is scaled
# properly:
mu = np.array([ozu.unit_vector(x) for x in mu])
data = np.array([ozu.unit_vector(x) for x in data])
# 2. Data assignment:
# Calculate all the correlations in one swoop:
corr = np.dot(data, mu.T)
# In cases where antipodal symmetry is assumed,
if antipodal == True:
corr = np.abs(corr)
# This chooses the centroid for each one:
y_n = np.argmax(corr, -1)
# 3. Centroid estimation:
for this_k in range(k):
idx = np.where(y_n == this_k)
if len(idx[0]) > 0:
# The average will be based on the data points that are considered
# in this centroid with a weighted average:
this_sum = np.dot(weights[idx], data[idx])
# This goes into the volume of the sphere, so we renormalize to the
# surface (or to the origin, if it's 0):
this_norm = ozu.l2_norm(this_sum)
if this_norm > 0:
# Scale by the mean of the weights
mu[this_k] = (this_sum / this_norm) * np.mean(weights[idx])
elif this_norm < 0:
mu[this_k] = np.array([0, 0, 0])
# Did it change?
if np.all(y_n == last_y_n):
# 4. Stop if there's no change in assignment:
is_changing = False
else:
last_y_n = y_n
# Another stopping condition is if this has gone on for a while
iter += 1
if iter > max_iter:
is_changing = False
# Once you are done computing 'em all, calculate the resulting SSE:
SSE = 0
if calc_sse:
for this_k in range(k):
idx = np.where(y_n == this_k)
len_idx = len(idx[0])
if len_idx > 0:
scaled_data = data[idx] * weights[idx].reshape(len_idx, 1)
SSE += np.sum((mu[this_k] - scaled_data)**2)
return mu, y_n, SSE
def vertices(self):
return np.column_stack(sphere2cart(1, self.theta, self.phi))
def main():
# object_settings_file = Path(os.path.join(IMAGE_DIR_LIST[0], "_object_settings.json"))
# camera_settings_file = Path(os.path.join(IMAGE_DIR_LIST[0], "_camera_settings.json"))
if object_settings_file.is_file():
with open(object_settings_file) as file:
object_settings_data = json.load(file)
if SELECTED_OBJECTS is None:
CLASSES = object_settings_data['exported_object_classes']
else:
CLASSES = SELECTED_OBJECTS
CATEGORIES = [{
'id': i,
'name': CLASSES[i].replace('_16k', '').replace('_16K', ''),
'supercategory': 'shape',
} for i in range(0,len(CLASSES))]
FIXED_TRANSFORMS = {}
for i in range(0,len(object_settings_data['exported_object_classes'])):
class_name = object_settings_data['exported_objects'][i]['class']
transform = object_settings_data['exported_objects'][i]['fixed_model_transform']
if class_name in CLASSES:
class_name = class_name.replace('_16k', '').replace('_16K', '')
FIXED_TRANSFORMS[class_name] = transform
# print(FIXED_TRANSFORMS)
# SEGMENTATION_DATA = object_settings_data['exported_objects']
else:
raise Exception("Object settings file not found")
if camera_settings_file.is_file():
with open(camera_settings_file) as file:
camera_settings_data = json.load(file)
CAMERA_INTRINSICS = camera_settings_data['camera_settings'][0]['intrinsic_settings']
else:
raise Exception("Camera settings file not found")
VIEWPOINTS = [viewpoints_xyz[i].tolist() for i in range(0, len(viewpoints_xyz))]
INPLANE_ROTATIONS = [inplane_rot_angles[i] for i in range(0, len(inplane_rot_angles))]
coco_output = {
"info": INFO,
"licenses": LICENSES,
"categories": CATEGORIES,
"viewpoints" : VIEWPOINTS,
"inplane_rotations" : INPLANE_ROTATIONS,
"camera_intrinsic_settings": CAMERA_INTRINSICS,
"fixed_transforms": FIXED_TRANSFORMS,
"images": [],
"annotations": []
}
image_global_id = 1
segmentation_global_id = 1
# filter for jpeg images
for IMAGE_DIR_T in IMAGE_DIR_LIST:
# for root, _, files in os.walk(IMAGE_DIR):
for SCENE in SCENES:
if IMAGE_DIR_T == "":
IMAGE_DIR = os.path.join(ROOT_DIR, SCENE)
else:
IMAGE_DIR = os.path.join(ROOT_DIR, IMAGE_DIR_T, SCENE)
all_dir_files = os.listdir(IMAGE_DIR)
image_files = filter_for_jpeg(IMAGE_DIR, all_dir_files)
# dir_name = os.path.basename(IMAGE_DIR)
SEGMENTATION_DATA = get_segmentation_data_for_scene(IMAGE_DIR)
print(SEGMENTATION_DATA)
# go through each image
for ii in trange(len(image_files)):
image_filename = image_files[ii]
if IMAGE_DIR_T == "":
image_out_filename = os.path.join(SCENE, os.path.basename(image_filename))
else:
image_out_filename = os.path.join(IMAGE_DIR_T, SCENE, os.path.basename(image_filename))
img_size = (960,540)
image_info = pycococreatortools.create_image_info(
image_global_id, image_out_filename, img_size
)
# plt.figure()
# skimage.io.imshow(skimage.io.imread(image_filename))
# plt.show()
# filter for associated png annotations
# for root, _, files in os.walk(IMAGE_DIR):
segmentation_image_files = filter_for_annotations(IMAGE_DIR, all_dir_files, image_filename)
label_files = filter_for_labels(IMAGE_DIR, all_dir_files, image_filename)
boxes = []
labels = []
segmentation_ids = []
label_filename = label_files[0]
# go through each associated json file containing objects data
# for label_filename in label_files:
# print("File %d - %s"% (image_global_id, label_filename))
my_file = Path(label_filename)
segmentation_image = skimage.io.imread(segmentation_image_files[0])
# print("File %d - %s"% (image_global_id, segmentation_image_files[0]))
if my_file.is_file():
with open(label_filename) as file:
label_data = json.load(file)
# all_objects_yaw_only = True
for i in range(0, len(label_data['objects'])):
class_name = label_data['objects'][i]['class']
if class_name not in SELECTED_OBJECTS:
continue
# print(class_name)
class_bounding_box = label_data['objects'][i]['bounding_box']
quat = label_data['objects'][i]['quaternion_xyzw']
angles = RT_transform.quat2euler(get_wxyz_quaternion(quat))
# angles = RT_transform.quat2euler(get_wxyz_quaternion(quat), 'syxz')
# angles = apply_angle_symmetry(angles, SYMMETRY_INFO[class_name])
# This function gives angles with this convention of euler - https://en.wikipedia.org/wiki/Euler_angles#Signs_and_ranges (geometric definition)
# if np.isclose(angles[1], 0):
# print("Test")
theta, phi = euler2sphere(angles[1], angles[0])
actual_angles = np.array([1, theta, phi])
xyz_coord = sphere2cart(1, theta, phi)
viewpoint_id = find_viewpoint_id(viewpoints_xyz, xyz_coord)
r_xyz = get_viewpoint_from_id(viewpoints_xyz, viewpoint_id)
recovered_angles = np.array(cart2sphere(r_xyz[0], r_xyz[1], r_xyz[2]))
inplane_rotation_id = find_inplane_rotation_id(inplane_rot_angles, angles[2])
# inplate_rotation_angle = get_inplane_rotation_from_id(INPLANE_ROTATIONS, inplane_rotation_id)
if np.all(np.isclose(actual_angles, recovered_angles, atol=0.4)) == False:
print("Mismatch in : {}".format(label_filename))
print("sphere2cart angles : {}".format(actual_angles))
print("cart2sphere angles : {}".format(recovered_angles))
# elif np.all(np.isclose(actual_angles, recovered_angles, atol=0.4)) == True:
# print("Match")
# print(inplate_rotation_angle)
class_label = [x['id'] for x in CATEGORIES if x['name'] in class_name][0]
segmentation_id = [x['segmentation_class_id'] for x in SEGMENTATION_DATA if x['class'] in class_name][0]
boxes.append(class_bounding_box['top_left'] + class_bounding_box['bottom_right'])
labels.append(class_label)
segmentation_ids.append([x['segmentation_class_id'] for x in SEGMENTATION_DATA if x['class'] in class_name][0])
# Create binary masks from segmentation image for every object
# for segmentation_image_file in segmentation_image_files:
# segmentation_image = skimage.io.imread(segmentation_image_file)
binary_mask = np.copy(segmentation_image)
binary_mask[binary_mask != segmentation_id] = 0
binary_mask[binary_mask == segmentation_id] = 1
# skimage.io.imshow(binary_mask, cmap=plt.cm.gray)
# plt.show()
# TODO : check if its actually a crowd in case of multiple instances of one object type
# class_label = [x['class'] for x in SEGMENTATION_DATA if x['segmentation_class_id'] in segmentation_id][0]
category_info = {'id': class_label, 'is_crowd': 0}
annotation_info = pycococreatortools.create_annotation_info(
segmentation_global_id, image_global_id, category_info, binary_mask,
img_size, tolerance=2)
# print(annotation_info)
if annotation_info is not None:
annotation_info['viewpoint_id'] = int(viewpoint_id)
annotation_info['inplane_rotation_id'] = int(inplane_rotation_id)
annotation_info['camera_pose'] = label_data['camera_data']
annotation_info['location'] = label_data['objects'][i]['location']
annotation_info['quaternion_xyzw'] = quat
coco_output["annotations"].append(annotation_info)
coco_output["images"].append(image_info)
else:
tqdm.write("File %s doesn't have boxes or labels in json file" % image_filename)
segmentation_global_id = segmentation_global_id + 1
else:
tqdm.write("File %s doesn't have a label file" % image_filename)
image_global_id = image_global_id + 1
with open('{}/{}.json'.format(ROOT_DIR, OUTFILE_NAME), 'w') as output_json_file:
json.dump(coco_output, output_json_file)
def vertices(self):
return np.column_stack(sphere2cart(1, self.theta, self.phi))
def generate_gold_data(bvals,bvecs,fibno=1,axesn=10,SNR=None):
#1 fiber case
if fibno==1:
data=np.zeros((axesn,len(bvals)))
gold=np.zeros((axesn,3))
X=random_uniform_on_sphere(axesn,'xyz')
for (j,x) in enumerate(X):
S,sticks=SticksAndBall(bvals,bvecs,d=0.0015,S0=100,angles=np.array([x]),fractions=[100],snr=SNR)
data[j,:]=S[:]
gold[j,:3]=x
return data,gold,None,None,axesn,None
#2 fibers case
if fibno==2:
angles=np.arange(0,92,2.5)#5
u=np.array([0,0,1])
anglesn=len(angles)
data=np.zeros((anglesn*axesn,len(bvals)))
gold=np.zeros((anglesn*axesn,6))
angs=[]
for (i,a) in enumerate(angles):
v=sphere2cart(1,np.deg2rad(a),0)
X=random_uniform_on_sphere(axesn,'xyz')
for (j,x) in enumerate(X):
R=rodriguez_axis_rotation(x,360*np.random.rand())
ur=np.dot(R,u)
vr=np.dot(R,v)
S,sticks=SticksAndBall(bvals,bvecs,d=0.0015,S0=100,angles=np.array([ur,vr]),fractions=[50,50],snr=SNR)
data[i*axesn+j,:]=S[:]
gold[i*axesn+j,:3]=ur
gold[i*axesn+j,3:]=vr
angs.append(a)
return data,gold,angs,anglesn,axesn,angles
#3 fibers case
if fibno==3:
#phis=np.array([0.,120.,240.])
#angles=np.arange(0,92,5)
angles=np.linspace(0,54.73,40)#20
anglesn=len(angles)
data=np.zeros((anglesn*axesn,len(bvals)))
gold=np.zeros((anglesn*axesn,9))
angs=[]
for (i,a) in enumerate(angles):
u=sphere2cart(1,np.deg2rad(a),np.deg2rad(0))
v=sphere2cart(1,np.deg2rad(a),np.deg2rad(120))
w=sphere2cart(1,np.deg2rad(a),np.deg2rad(240))
X=random_uniform_on_sphere(axesn,'xyz')
for (j,x) in enumerate(X):
R=rodriguez_axis_rotation(x,360*np.random.rand())
ur=np.dot(R,u)
vr=np.dot(R,v)
wr=np.dot(R,w)
S,sticks=SticksAndBall(bvals,bvecs,d=0.0015,S0=100,angles=np.array([ur,vr,wr]),fractions=[33,33,33],snr=SNR)
print i,j
data[i*axesn+j,:]=S[:]
gold[i*axesn+j,:3]=ur
gold[i*axesn+j,3:6]=vr
gold[i*axesn+j,6:]=wr
angs.append(a)
return data,gold,angs,anglesn,axesn,angles
def plot_signal_interp(bvecs,
signal,
origin=[0, 0, 0],
maya=True,
cmap='jet',
file_name=None,
colorbar=False,
figure=None,
vmin=None,
vmax=None,
offset=0,
azimuth=60,
elevation=90,
roll=0,
points=False,
cmap_points=None,
scale_points=False,
non_neg=False,
interp_kwargs=dict(function='thin_plate', smooth=0)):
"""
Interpolate a measured signal, using RBF interpolation.
Parameters
----------
signal:
bvecs: array (3,n)
the x,y,z locations where the signal was measured
offset : float
where to place the plotted voxel (on the z axis)
points : bool
whether to show the sampling points on the interpolated
surface. Default: False.
"""
bvecs_new = np.hstack([bvecs, -bvecs])
new_signal = np.hstack([signal, signal])
s0 = Sphere(xyz=bvecs_new.T)
s1 = create_unit_sphere(7)
signal[np.isnan(signal)] = 0
interp_signal = interp_rbf(new_signal, s0, s1, **interp_kwargs)
vertices = s1.vertices
if non_neg:
interp_signal[interp_signal < 0] = 0
faces = s1.faces
x, y, z = vertices.T
r, phi, theta = geo.cart2sphere(x, y, z)
x_plot, y_plot, z_plot = geo.sphere2cart(interp_signal, phi, theta)
if points:
r, phi, theta = geo.cart2sphere(s0.x, s0.y, s0.z)
x_p, y_p, z_p = geo.sphere2cart(signal, phi, theta)
figure = _display_maya_voxel(x_p,
y_p,
z_p,
faces,
signal,
origin,
cmap=cmap,
colorbar=colorbar,
figure=figure,
vmin=vmin,
vmax=vmax,
file_name=file_name,
azimuth=azimuth,
elevation=elevation,
points=True,
cmap_points=cmap_points,
scale_points=scale_points)
# Call and return straightaway:
return _display_maya_voxel(x_plot,
y_plot,
z_plot,
faces,
interp_signal,
origin,
cmap=cmap,
colorbar=colorbar,
figure=figure,
vmin=vmin,
vmax=vmax,
file_name=file_name,
azimuth=azimuth,
elevation=elevation)
def multi_tensor(gtab,
mevals,
S0=100,
angles=[(0, 0), (90, 0)],
fractions=[50, 50],
snr=20):
r"""Simulate a Multi-Tensor signal.
Parameters
-----------
gtab : GradientTable
mevals : array (K, 3)
each tensor's eigenvalues in each row
S0 : float
Unweighted signal value (b0 signal).
angles : array (K,2) or (K,3)
List of K tensor directions in polar angles (in degrees) or unit vectors
fractions : float
Percentage of the contribution of each tensor. The sum of fractions
should be equal to 100%.
snr : float
Signal to noise ratio, assuming Rician noise. If set to None, no
noise is added.
Returns
--------
S : (N,) ndarray
Simulated signal.
sticks : (M,3)
Sticks in cartesian coordinates.
Examples
--------
>>> import numpy as np
>>> from dipy.sims.voxel import multi_tensor
>>> from dipy.data import get_data
>>> from dipy.core.gradients import gradient_table
>>> from dipy.io.gradients import read_bvals_bvecs
>>> fimg, fbvals, fbvecs = get_data('small_101D')
>>> bvals, bvecs = read_bvals_bvecs(fbvals, fbvecs)
>>> gtab = gradient_table(bvals, bvecs)
>>> mevals=np.array(([0.0015, 0.0003, 0.0003],[0.0015, 0.0003, 0.0003]))
>>> e0 = np.array([1, 0, 0.])
>>> e1 = np.array([0., 1, 0])
>>> S = multi_tensor(gtab, mevals)
"""
if np.round(np.sum(fractions), 2) != 100.0:
raise ValueError('Fractions should sum to 100')
fractions = [f / 100. for f in fractions]
S = np.zeros(len(gtab.bvals))
angles = np.array(angles)
if angles.shape[-1] == 3:
sticks = angles
else:
sticks = [
sphere2cart(1, np.deg2rad(pair[0]), np.deg2rad(pair[1]))
for pair in angles
]
sticks = np.array(sticks)
for i in range(len(fractions)):
S = S + fractions[i] * single_tensor(gtab,
S0=S0,
evals=mevals[i],
evecs=all_tensor_evecs(
sticks[i]).T,
snr=None)
return add_noise(S, snr, S0), sticks
def compare_sdni_eitl():
btable=np.loadtxt(get_data('dsi515btable'))
bvals=btable[:,0]
bvecs=btable[:,1:]
type=3
axesn=200
SNR=20
data,gold,angs,anglesn,axesn,angles=generate_gold_data(bvals,bvecs,fibno=type,axesn=axesn,SNR=SNR)
ei=EquatorialInversion(data,bvals,bvecs,odf_sphere='symmetric642',
auto=False,save_odfs=True,fast=True)
ei.radius=np.arange(0,5,0.1)
ei.gaussian_weight=None
ei.set_operator('laplacian')
#{'reflect','constant','nearest','mirror', 'wrap'}
ei.set_mode(order=1,zoom=1,mode='constant')
ei.update()
ei.fit()
dn=DiffusionNabla(data,bvals,bvecs,odf_sphere='symmetric642',
auto=False,save_odfs=True,fast=False)
dn.radius=np.arange(0,5,0.1)
dn.gaussian_weight=None
dn.update()
dn.fit()
vts=ei.odf_vertices
def simple_peaks(ODF,faces,thr):
x,g=ODF.shape
PK=np.zeros((x,5))
IN=np.zeros((x,5))
for (i,odf) in enumerate(ODF):
peaks,inds=peak_finding(odf,faces)
ibigp=np.where(peaks>thr*peaks[0])[0]
l=len(ibigp)
if l>3:
l=3
PK[i,:l]=peaks[:l]
IN[i,:l]=inds[:l]
return PK,IN
print "test0"
thresh=0.5
PK,IN=simple_peaks(ei.ODF,ei.odf_faces,thresh)
res,me,st =do_compare(gold,vts,PK,IN,0,anglesn,axesn,type)
#PKG,ING=simple_peaks(eig.ODF,ei.odf_faces,thresh)
#resg,meg,stg =do_compare(gold,vts,PKG,ING,0,anglesn,axesn,type)
PK,IN=simple_peaks(dn.ODF,dn.odf_faces,thresh)
res2,me2,st2 =do_compare(gold,vts,PK,IN,0,anglesn,axesn,type)
if type==3:
x,y,z=sphere2cart(np.ones(len(angles)),np.deg2rad(angles),np.zeros(len(angles)))
x2,y2,z2=sphere2cart(np.ones(len(angles)),np.deg2rad(angles),np.deg2rad(120*np.ones(len(angles))))
angles2=[]
for i in range(len(x)):
angles2.append(np.rad2deg(np.arccos(np.dot([x[i],y[i],z[i]],[x2[i],y2[i],z2[i]]))))
angles=angles2
print "test1"
plt.figure(1)
plt.plot(angles,me,'k:',linewidth=3.,label='EITL')
plt.plot(angles,me2,'k--',linewidth=3.,label='sDNI')
#plt.plot(angles,me7,'r--',linewidth=3.,label='EITL2')
plt.xlabel('angle')
plt.ylabel('similarity')
title='Angular similarity of ' + str(type) + '-fibres crossing with SNR ' + str(SNR)
plt.title(title)
plt.legend(loc='center right')
plt.savefig('/tmp/test.png',dpi=300)
plt.show()
#needles = np.array([-np.pi/4,np.pi/4])
needles2d = np.array([0,np.pi/4.,np.pi/2.])
angles2d = np.linspace(-np.pi, np.pi,100)
def signal2d(S0,f,b,d,needles2d,angles2d):
return S0*np.array([(1-f[0]-f[1]-f[2])*np.exp(-b*d) + \
f[0]*np.exp(-b*d*(np.cos(needles2d[0]-gradient))**2) + \
f[1]*np.exp(-b*d*(np.cos(needles2d[1]-gradient))**2) + \
f[2]*np.exp(-b*d*(np.cos(needles2d[2]-gradient))**2) \
for gradient in angles2d])
#needles3d = [sphere2cart(1,0,0),sphere2cart(1,np.pi/2.,0),sphere2cart(1,np.pi/2.,np.pi/2.)]
needles3d = np.array([sphere2cart(1,0,0),sphere2cart(1,np.pi/2.,0),sphere2cart(1,np.pi/2.,np.pi/2.)])
angles2d = np.linspace(-np.pi, np.pi,100)
def signal3d(S0,f,b0,d,needles3d,bvals,gradients):
S = S0*np.array([(1-f[0]-f[1]-f[2])*np.exp(-b0*d) + \
f[0]*np.exp(-bvals[i]*d*(np.dot(needles3d[0],gradient))**2) + \
f[1]*np.exp(-bvals[i]*d*(np.dot(needles3d[1],gradient))**2) + \
f[2]*np.exp(-bvals[i]*d*(np.dot(needles3d[2],gradient))**2) \
for (i,gradient) in enumerate(gradients)])
return S, np.dot(np.diag(S),gradients)
#dsi202 = np.row_stack((dsi102[1:,:],-dsi102[1:,:]))
#print dsi202.shape
#print bvals102.shape
def test_kurtosis_maximum():
# TEST 1
# simulate a crossing fibers interserting at 70 degrees. The first fiber
# is aligned to the x-axis while the second fiber is aligned to the x-z
# plane with an angular deviation of 70 degrees from the first one.
# According to Neto Henriques et al, 2015 (NeuroImage 111: 85-99), the
# kurtosis tensor of this simulation will have a maxima aligned to axis y
angles = [(90, 0), (90, 0), (20, 0), (20, 0)]
signal_70, dt_70, kt_70 = multi_tensor_dki(gtab_2s, mevals_cross, S0=100,
angles=angles,
fractions=frac_cross, snr=None)
# prepare inputs
dkiM = dki.DiffusionKurtosisModel(gtab_2s, fit_method="WLS")
dkiF = dkiM.fit(signal_70)
MD = dkiF.md
kt = dkiF.kt
R = dkiF.evecs
evals = dkiF.evals
dt = lower_triangular(np.dot(np.dot(R, np.diag(evals)), R.T))
sphere = get_sphere('symmetric724')
# compute maxima
k_max_cross, max_dir = dki._voxel_kurtosis_maximum(dt, MD, kt, sphere,
gtol=1e-5)
yaxis = np.array([0., 1., 0.])
cos_angle = np.abs(np.dot(max_dir[0], yaxis))
assert_almost_equal(cos_angle, 1.)
# TEST 2
# test the function on cases of well aligned fibers oriented in a random
# defined direction. According to Neto Henriques et al, 2015 (NeuroImage
# 111: 85-99), the maxima of kurtosis is any direction perpendicular to the
# fiber direction. Moreover, according to multicompartmetal simulations,
# kurtosis in this direction has to be equal to:
fie = 0.49
ADi = 0.00099
ADe = 0.00226
RDi = 0
RDe = 0.00087
RD = fie*RDi + (1-fie)*RDe
RK = 3 * fie * (1-fie) * ((RDi-RDe) / RD) ** 2
# prepare simulation:
theta = random.uniform(0, 180)
phi = random.uniform(0, 320)
angles = [(theta, phi), (theta, phi)]
mevals = np.array([[ADi, RDi, RDi], [ADe, RDe, RDe]])
frac = [fie*100, (1 - fie)*100]
signal, dt, kt = multi_tensor_dki(gtab_2s, mevals, angles=angles,
fractions=frac, snr=None)
# prepare inputs
dkiM = dki.DiffusionKurtosisModel(gtab_2s, fit_method="WLS")
dkiF = dkiM.fit(signal)
MD = dkiF.md
kt = dkiF.kt
R = dkiF.evecs
evals = dkiF.evals
dt = lower_triangular(np.dot(np.dot(R, np.diag(evals)), R.T))
# compute maxima
k_max, max_dir = dki._voxel_kurtosis_maximum(dt, MD, kt, sphere, gtol=1e-5)
# check if max direction is perpendicular to fiber direction
fdir = np.array([sphere2cart(1., np.deg2rad(theta), np.deg2rad(phi))])
cos_angle = np.abs(np.dot(max_dir[0], fdir[0]))
assert_almost_equal(cos_angle, 0., decimal=5)
# check if max direction is equal to expected value
assert_almost_equal(k_max, RK)
# According to Neto Henriques et al., 2015 (NeuroImage 111: 85-99),
# e.g. see figure 1 of this article, kurtosis maxima for the first test is
# also equal to the maxima kurtosis value of well-aligned fibers, since
# simulations parameters (apart from fiber directions) are equal
assert_almost_equal(k_max_cross, RK)
# Test 3 - Test performance when kurtosis is spherical - this case, can be
# problematic since a spherical kurtosis does not have an maximum
k_max, max_dir = dki._voxel_kurtosis_maximum(dt_sph, np.mean(evals_sph),
kt_sph, sphere, gtol=1e-2)
assert_almost_equal(k_max, Kref_sphere)
# Test 4 - Test performance when kt have all elements zero - this case, can
# be problematic this case does not have an maximum
k_max, max_dir = dki._voxel_kurtosis_maximum(dt_sph, np.mean(evals_sph),
np.zeros(15), sphere,
gtol=1e-2)
assert_almost_equal(k_max, 0.0)
def test():
# img=nib.load('/home/eg309/Data/project01_dsi/connectome_0001/tp1/RAWDATA/OUT/mr000001.nii.gz')
btable = np.loadtxt(get_data("dsi515btable"))
# volume size
sz = 16
# shifting
origin = 8
# hanning width
filter_width = 32.0
# number of signal sampling points
n = 515
# odf radius
radius = np.arange(2.1, 6, 0.2)
# create q-table
bv = btable[:, 0]
bmin = np.sort(bv)[1]
bv = np.sqrt(bv / bmin)
qtable = np.vstack((bv, bv, bv)).T * btable[:, 1:]
qtable = np.floor(qtable + 0.5)
# copy bvals and bvecs
bvals = btable[:, 0]
bvecs = btable[:, 1:]
# S=img.get_data()[38,50,20]#[96/2,96/2,20]
S, stics = SticksAndBall(
bvals, bvecs, d=0.0015, S0=100, angles=[(0, 0), (60, 0), (90, 90)], fractions=[0, 0, 0], snr=None
)
S2 = S.copy()
S2 = S2.reshape(1, len(S))
dn = DiffusionNabla(S2, bvals, bvecs, auto=False)
pR = dn.equators
odf = dn.odf(S)
# Xs=dn.precompute_interp_coords()
peaks, inds = peak_finding(odf.astype("f8"), dn.odf_faces.astype("uint16"))
print peaks
print peaks / peaks.min()
# print dn.PK
dn.fit()
print dn.PK
# """
ren = fvtk.ren()
colors = fvtk.colors(odf, "jet")
fvtk.add(ren, fvtk.point(dn.odf_vertices, colors, point_radius=0.05, theta=8, phi=8))
fvtk.show(ren)
# """
stop
# ds=DiffusionSpectrum(S2,bvals,bvecs)
# tpr=ds.pdf(S)
# todf=ds.odf(tpr)
"""
#show projected signal
Bvecs=np.concatenate([bvecs[1:],-bvecs[1:]])
X0=np.dot(np.diag(np.concatenate([S[1:],S[1:]])),Bvecs)
ren=fvtk.ren()
fvtk.add(ren,fvtk.point(X0,fvtk.yellow,1,2,16,16))
fvtk.show(ren)
"""
# qtable=5*matrix[:,1:]
# calculate radius for the hanning filter
r = np.sqrt(qtable[:, 0] ** 2 + qtable[:, 1] ** 2 + qtable[:, 2] ** 2)
# setting hanning filter width and hanning
hanning = 0.5 * np.cos(2 * np.pi * r / filter_width)
# center and index in q space volume
q = qtable + origin
q = q.astype("i8")
# apply the hanning filter
values = S * hanning
"""
#plot q-table
ren=fvtk.ren()
colors=fvtk.colors(values,'jet')
fvtk.add(ren,fvtk.point(q,colors,1,0.1,6,6))
fvtk.show(ren)
"""
# create the signal volume
Sq = np.zeros((sz, sz, sz))
for i in range(n):
Sq[q[i][0], q[i][1], q[i][2]] += values[i]
# apply fourier transform
Pr = fftshift(np.abs(np.real(fftn(fftshift(Sq), (sz, sz, sz)))))
# """
ren = fvtk.ren()
vol = fvtk.volume(Pr)
fvtk.add(ren, vol)
fvtk.show(ren)
# """
"""
from enthought.mayavi import mlab
mlab.pipeline.volume(mlab.pipeline.scalar_field(Sq))
mlab.show()
"""
# vertices, edges, faces = create_unit_sphere(5)
vertices, faces = sphere_vf_from("symmetric362")
odf = np.zeros(len(vertices))
for m in range(len(vertices)):
xi = origin + radius * vertices[m, 0]
yi = origin + radius * vertices[m, 1]
zi = origin + radius * vertices[m, 2]
PrI = map_coordinates(Pr, np.vstack((xi, yi, zi)), order=1)
for i in range(len(radius)):
odf[m] = odf[m] + PrI[i] * radius[i] ** 2
"""
ren=fvtk.ren()
colors=fvtk.colors(odf,'jet')
fvtk.add(ren,fvtk.point(vertices,colors,point_radius=.05,theta=8,phi=8))
fvtk.show(ren)
"""
"""
#Pr[Pr<500]=0
ren=fvtk.ren()
#ren.SetBackground(1,1,1)
fvtk.add(ren,fvtk.volume(Pr))
fvtk.show(ren)
"""
peaks, inds = peak_finding(odf.astype("f8"), faces.astype("uint16"))
Eq = np.zeros((sz, sz, sz))
for i in range(n):
Eq[q[i][0], q[i][1], q[i][2]] += S[i] / S[0]
LEq = laplace(Eq)
# Pr[Pr<500]=0
ren = fvtk.ren()
# ren.SetBackground(1,1,1)
fvtk.add(ren, fvtk.volume(Eq))
fvtk.show(ren)
phis = np.linspace(0, 2 * np.pi, 100)
planars = []
for phi in phis:
planars.append(sphere2cart(1, np.pi / 2, phi))
planars = np.array(planars)
planarsR = []
for v in vertices:
R = vec2vec_rotmat(np.array([0, 0, 1]), v)
planarsR.append(np.dot(R, planars.T).T)
"""
ren=fvtk.ren()
fvtk.add(ren,fvtk.point(planarsR[0],fvtk.green,1,0.1,8,8))
fvtk.add(ren,fvtk.point(2*planarsR[1],fvtk.red,1,0.1,8,8))
fvtk.show(ren)
"""
azimsums = []
for disk in planarsR:
diskshift = 4 * disk + origin
# Sq0=map_coordinates(Sq,diskshift.T,order=1)
# azimsums.append(np.sum(Sq0))
# Eq0=map_coordinates(Eq,diskshift.T,order=1)
# azimsums.append(np.sum(Eq0))
LEq0 = map_coordinates(LEq, diskshift.T, order=1)
azimsums.append(np.sum(LEq0))
azimsums = np.array(azimsums)
# """
ren = fvtk.ren()
colors = fvtk.colors(azimsums, "jet")
fvtk.add(ren, fvtk.point(vertices, colors, point_radius=0.05, theta=8, phi=8))
fvtk.show(ren)
# """
# for p in planarsR[0]:
"""
def test():
K.set_image_data_format(
'channels_last') # TF dimension ordering in this code
featdef = {
#'magfld': tf.FixedLenSequenceFeature(shape=[YMagfld*XMagfld*ZMagfld], dtype=tf.string, allow_missing=True),
#'stokes': tf.FixedLenSequenceFeature(shape=[YStokes*XStokes*ZStokes], dtype=tf.string, allow_missing=True)
'magfld':
tf.FixedLenSequenceFeature(shape=[],
dtype=tf.float32,
allow_missing=True),
'stokes':
tf.FixedLenSequenceFeature(shape=[],
dtype=tf.float32,
allow_missing=True),
# 'name': tf.FixedLenFeature(shape=[], dtype=tf.string)
}
def _parse_record(example_proto, clip=False):
"""Parse a single record into x and y images"""
example = tf.parse_single_example(example_proto, featdef)
#x = tf.decode_raw(example['stokes'], tf.float32)
x = example['stokes']
# normalize Stokes Parameters
if doNormalize:
x = tf.reshape(x, (WStokes * YStokes * XStokes, ZStokes))
x = tf.slice(x, (0, 0), (WStokes * YDim * XDim, ZStokes))
x = x - nm
x = x / ns
x = tf.reshape(x, (WStokes, YStokes, XStokes, ZStokes))
x = tf.slice(x, (0, 0, 0, 0), (WStokes, YDim, XDim, ZStokes))
if doWLConvolution == False:
# unstack wavelength dimension (axis=0)
# now we have a list of WDim tensors of shape (YDim*XDim, ZStokes)
wl = tf.unstack(x, axis=0)
# concatenate the wavelengths in the channel (Stokes Param) dimension (axis=2)
x = tf.concat(wl, axis=2)
x = tf.slice(x, (0, 0, 0), (YDim, XDim, WStokes * ZStokes))
y = example['magfld']
y = tf.reshape(y, (YMagfld, XMagfld, ZMagfld))
y = tf.slice(y, (0, 0, 0), (YDim, XDim, ZMagfld))
# name = example['name']
# return x, y, name
return x, y
print('-' * 30)
print('Opening test dataset ...')
print('-' * 30)
dsTest = tf.data.TFRecordDataset(pathTest).map(_parse_record,
num_parallel_calls=8)
dsTest = dsTest.prefetch(10)
dsTest = dsTest.repeat(1)
dsTest = dsTest.batch(sizeBatch)
print('-' * 30)
print('Creating and compiling model...')
print('-' * 30)
model = UNet()
#print(model.summary())
print('-' * 30)
print('Testing the model...')
print('-' * 30)
sess = K.get_session()
n = 0
scores = model.evaluate(dsTest, steps=nStep, verbose=1)
for m in range(0, len(scores)):
print('%s=%f' % (model.metrics_names[m], scores[m]))
if doPlot:
plt.gray()
r = np.zeros((ZMagfld, (nStep + 1) * sizeBatch))
mae = np.zeros((ZMagfld, (nStep + 1) * sizeBatch))
mse = np.zeros((ZMagfld, (nStep + 1) * sizeBatch))
# add an iterator on the test set to the graph
diTest = dsTest.make_one_shot_iterator().get_next()
start = time.time()
# for each batch in the test set
for t in range(0, nStep + 1):
#for t in range(0, 1):
# get the next batch of test examples
examples = sess.run(diTest)
x = examples[0]
print(x.shape)
y = examples[1]
print(y.shape)
#print(y)
if False:
plt.imshow(x[1, 56, :, :, 0])
plt.show()
plt.imshow(y[1, :, :, 0])
plt.show()
plt.imshow(x[1, 56, :, :, 1])
plt.show()
plt.imshow(y[1, :, :, 1])
plt.show()
plt.imshow(x[1, 56, :, :, 2])
plt.show()
plt.imshow(y[1, :, :, 2])
plt.show()
plt.imshow(x[1, 56, :, :, 3])
plt.show()
y_pred = model.predict(x, steps=1)
print(y_pred.shape)
#print(y_pred)
if False:
plt.imshow(y_pred[1, :, :, 0])
plt.show()
plt.imshow(y_pred[1, :, :, 1])
plt.show()
plt.imshow(y_pred[1, :, :, 2])
plt.show()
# convert Intensity units from Gauss to kG
#y[:,:,:,0] = np.true_divide(y[:,:,:,0], 1000.0)
#y_pred[:,:,:,0] = np.true_divide(y_pred[:,:,:,0], 1000.0)
if useBquv:
# transform from spherical coords to Ramos style Bq,Bu,Bv Cartesian coords
y[:, :, :, 1:] = np.radians(y[:, :, :, 1:])
Bx, By, Bz = sphere2cart(y[:, :, :, 0], y[:, :, :, 1], y[:, :, :,
2])
Bxsq = np.square(Bx)
Bysq = np.square(By)
Bxy = np.multiply(Bx, By)
Bq = np.sign(Bxsq - Bysq) * np.sqrt(np.abs(Bxsq - Bysq))
Bu = np.sign(Bxy) * np.sqrt(np.abs(Bxy))
Bv = Bz
y[:, :, :, 0], y[:, :, :, 1], y[:, :, :, 2] = Bq, Bu, Bv
y_pred[:, :, :, 1:] = np.radians(y_pred[:, :, :, 1:])
Bx, By, Bz = sphere2cart(y_pred[:, :, :, 0], y_pred[:, :, :, 1],
y_pred[:, :, :, 2])
Bxsq = np.square(Bx)
Bysq = np.square(By)
Bxy = np.multiply(Bx, By)
Bq = np.sign(Bxsq - Bysq) * np.sqrt(np.abs(Bxsq - Bysq))
Bu = np.sign(Bxy) * np.sqrt(np.abs(Bxy))
Bv = Bz
y_pred[:, :, :, 0], y_pred[:, :, :, 1], y_pred[:, :, :,
2] = Bq, Bu, Bv
# for each test example in this batch
for i in range(x.shape[0]):
#for i in range(28, 29):
n += 1
print('.', end='', flush=True)
# for each output dimension
for d in range(0, ZMagfld):
yt = y[i, 0:64, 0:64, d]
yp = y_pred[i, 0:64, 0:64, d]
# sum the Pearson Rs to take an average per dimension after the main loop
pr, pp = stats.pearsonr(yt.flatten(), yp.flatten())
r[d, n - 1] = pr
mae[d, n - 1] = metrics.mean_absolute_error(yt, yp)
mse[d, n - 1] = metrics.mean_squared_error(yt, yp)
yUnits = units[useBquv]
yChannel = channel[useBquv]
ySym = symbol[useBquv]
if doPlot:
#xt = x[i, 52, 0:64, 0:64, 0]
#plt.imshow(xt)
#plt.show()
plt.imshow(yt)
plt.show()
plt.imshow(yp)
plt.show()
if False:
title = '%s: R=%.6f' % (yChannel[d], pr)
plt.title(title)
plt.imshow(corr(yt, yp))
plt.show()
if True:
title = '%s KDE' % (yChannel[d])
print('%s [%s]' % (yChannel[d], yUnits[d]))
yleg = "%s [%s]" % (ySym[d], yUnits[d])
xleg = "%s [%s]" % (ySym[d], yUnits[d])
kde_error(yt, yp, title, yleg, xleg, yUnits[d])
print("")
#imI = normalize(level1[i,:,0:64,0:64,:])
#imI = level1[i,:,0:64,0:64,:]
#print('%f,'%(imI[56,32,32,0]), end='')
#print('%f,'%(imI[56,32,32,1]), end='')
#print('%f,'%(imI[56,32,32,2]), end='')
#print('%f,'%(imI[56,32,32,3]), end='')
#print('%s,'%(fname[i]))
#break
#for j in range(X.shape[1]):
#if False: #for j in range(0, 1):
# #print(fname[i],end='')
# imI = level1[i,j,0:64,0:64,0]
# #nz = np.nonzero(imI)
# #print('%d,'%(j), end='')
# #print('%f,'%(np.percentile(imI, 95.0)), end='')
# #print('%f,'%(np.percentile(imI, 10.0)), end='')
# #print('%f,'%(np.min(imI)), end='')
# #print('%f,'%(np.max(imI)), end='')
# #print('%f,'%(np.mean(imI)), end='')
# #print('%f,'%(np.std(imI)), end='')
# plt.imshow(imI)
# #plt.title(fname[i])
# plt.show()
# imQ = level1[i,j,0:64,0:864,1]
# plt.gray()
# plt.imshow(imQ)
# #plt.title(fname[i])
# plt.show()
# #print('%f,'%(np.percentile(imQ[nz], 95.0)), end='')
# #print('%f,'%(np.percentile(imQ[nz], 10.0)), end='')
# #print('%f,'%(np.min(imQ[nz])), end='')
# #print('%f,'%(np.max(imQ[nz])), end='')
# #print('%f,'%(np.mean(imQ[nz])), end='')
# #print('%f,'%(np.std(imQ[nz])), end='')
# imU = level1[i,j,0:64,0:864,2]
# #nz = np.nonzero(imU)
# plt.gray()
# plt.imshow(imU)
# #plt.title(fname[i])
# plt.show()
# #print('%f,'%(np.percentile(imU[nz], 95.0)), end='')
# #print('%f,'%(np.percentile(imU[nz], 10.0)), end='')
# #print('%f,'%(np.min(imU[nz])), end='')
# #print('%f,'%(np.max(imU[nz])), end='')
# #print('%f,'%(np.mean(imU[nz])), end='')
# #print('%f,'%(np.std(imU[nz])), end='')
# imV = level1[i,j,0:64,0:864,3]
# #nz = np.nonzero(imV)
# plt.gray()
# plt.imshow(imV)
# #plt.title(fname[i])
# plt.show()
# #print('%f,'%(np.percentile(imV[nz], 95.0)), end='')
# #print('%f,'%(np.percentile(imV[nz], 10.0)), end='')
# #print('%f,'%(np.min(imV[nz])), end='')
# #print('%f,'%(np.max(imV[nz])), end='')
# #print('%f,'%(np.mean(imV[nz])), end='')
# #print('%f'%(np.std(imV[nz])))
#scores = model.evaluate(diTest, steps=1, verbose=1)
#print(scores)
#scores = model.evaluate(dsTest.take(2), steps=1, verbose=1)
#y_pred = model.predict(dsTest, steps=1)
#print(y_pred)
#print('\nevaluate result: loss={}, mae={}'.format(*scores))
#image = Y[0,:,:,0]
#fig = plt.figure(num='Level 2 - Predicted')
#plt.gray()
#plt.imshow(image)
#plt.show()
#plt.close(fig)
end = time.time()
elapsed = end - start
#print(r)
#print('%d test examples in %d seconds: %.3f'%(n,elapsed, elapsed/n))
print('\n%d test examples' % (n))
print('%d seconds elapsed' % (elapsed))
print('Average inversion time %f.2 seconds' % (elapsed / n))
rmean = np.mean(r, axis=0)
rstd = np.std(r, axis=0)
for m in range(0, ZMagfld):
print('%s: 2D correlation R mean %f' % (yChannel[m], rmean[m]))
print('%s: 2D correlation R stdev %f' % (yChannel[m], rstd[m]))
mmean = np.mean(mae, axis=0)
mrms = np.mean(mse, axis=0)
mstd = np.std(mae, axis=0)
for m in range(0, ZMagfld):
print('%s: MAE mean %f' % (yChannel[m], mmean[m]))
print('%s: MAE stdev %f' % (yChannel[m], mstd[m]))
print('%s: MSE mean %f' % (yChannel[m], mrms[m]))
if doPlot:
plt.plot(r[0],
marker='o',
color='blue',
linewidth=2,
label="intensity")
plt.plot(r[1],
marker='*',
color='red',
linewidth=2,
label="inclination")
plt.plot(r[2], marker='x', color='green', linewidth=2, label="azimuth")
plt.legend()
plt.show()
K.clear_session()
def multi_tensor(gtab, mevals, S0=100, angles=[(0, 0), (90, 0)],
fractions=[50, 50], snr=20):
r"""Simulate a Multi-Tensor signal.
Parameters
-----------
gtab : GradientTable
mevals : array (K, 3)
each tensor's eigenvalues in each row
S0 : float
Unweighted signal value (b0 signal).
angles : array (K,2) or (K,3)
List of K tensor directions in polar angles (in degrees) or unit vectors
fractions : float
Percentage of the contribution of each tensor. The sum of fractions
should be equal to 100%.
snr : float
Signal to noise ratio, assuming Rician noise. If set to None, no
noise is added.
Returns
--------
S : (N,) ndarray
Simulated signal.
sticks : (M,3)
Sticks in cartesian coordinates.
Examples
--------
>>> import numpy as np
>>> from dipy.sims.voxel import multi_tensor
>>> from dipy.data import get_data
>>> from dipy.core.gradients import gradient_table
>>> from dipy.io.gradients import read_bvals_bvecs
>>> fimg, fbvals, fbvecs = get_data('small_101D')
>>> bvals, bvecs = read_bvals_bvecs(fbvals, fbvecs)
>>> gtab = gradient_table(bvals, bvecs)
>>> mevals=np.array(([0.0015, 0.0003, 0.0003],[0.0015, 0.0003, 0.0003]))
>>> e0 = np.array([1, 0, 0.])
>>> e1 = np.array([0., 1, 0])
>>> S = multi_tensor(gtab, mevals)
"""
if np.round(np.sum(fractions), 2) != 100.0:
raise ValueError('Fractions should sum to 100')
fractions = [f / 100. for f in fractions]
S = np.zeros(len(gtab.bvals))
angles = np.array(angles)
if angles.shape[-1] == 3:
sticks = angles
else:
sticks = [sphere2cart(1, np.deg2rad(pair[0]), np.deg2rad(pair[1]))
for pair in angles]
sticks = np.array(sticks)
for i in range(len(fractions)):
S = S + fractions[i] * single_tensor(gtab, S0=S0, evals=mevals[i],
evecs=all_tensor_evecs(
sticks[i]).T,
snr=None)
return add_noise(S, snr, S0), sticks
