def test_read_from_hdf5(self):
l = sim_struct.SimuList()
l.anisotropy_type = 'vn'
l.cond[3].value = 2
l.anisotropic_tissues = [3]
l.cond[0].distribution_type = 'beta'
v = np.zeros((255, 255, 255, 6))
for i in range(6):
v[:, :, :, i] = i
affine = np.array([[-1, 0, 0, 128], [0, 1, 0, -128], [0, 0, 1, -127],
[0, 0, 0, 1]])
l.anisotropy_vol = v
l.anisotropy_affine = affine
with tempfile.NamedTemporaryFile(suffix='.hdf5') as f:
fn_hdf5 = f.name
l._write_conductivity_to_hdf5(fn_hdf5)
l2 = sim_struct.SimuList()
l2._get_conductivity_from_hdf5(fn_hdf5)
assert np.isclose(l2.cond[3].value, 2)
assert l2.cond[46].value is None
assert l2.cond[0].name == 'WM'
assert l2.cond[46].name is None
assert np.all(l2.anisotropic_tissues == [3])
assert l2.anisotropy_type == 'vn'
assert np.allclose(l2.anisotropy_affine, affine)
assert np.allclose(l2.anisotropy_vol, v)
assert l2.cond[0].distribution_type == 'beta'
assert l2.cond[1].distribution_type is None
os.remove(fn_hdf5)
def test_list_cond_mat_struct(self, mat_session):
l = sim_struct.SimuList()
l.cond[0].distribution_type = 'uniform'
l.cond[0].distribution_parameters = [1.5, 3.0]
mat = l.cond_mat_struct()
scipy.io.savemat('tmp.mat', mat)
mat = scipy.io.loadmat('tmp.mat',
struct_as_record=True,
squeeze_me=False)
l2 = sim_struct.SimuList()
l2.read_cond_mat_struct(mat)
os.remove('tmp.mat')
assert l2.cond[0].distribution_type == 'uniform'
assert np.all(l2.cond[0].distribution_parameters == [1.5, 3.0])
def test_list_cond_array_aniso(self, sphere3_msh):
l = sim_struct.SimuList()
l.anisotropy_type = 'dir'
l.cond[3].value = 7
l.cond[4].value = 9
with tempfile.NamedTemporaryFile(suffix='.nii.gz') as f:
l.fn_tensor_nifti = f.name
v = np.zeros((255, 255, 255, 6))
for i in range(6):
v[:, :, :, i] = i
affine = np.array([[-1, 0, 0, 128], [0, 1, 0, -128], [0, 0, 1, -127],
[0, 0, 0, 1]])
img = nibabel.Nifti1Image(v, affine)
nibabel.save(img, l.fn_tensor_nifti)
msh = copy.deepcopy(sphere3_msh)
msh.elm.tag1[msh.elm.tag1 == 3] = 1
l.mesh = msh
elmcond = l.cond2elmdata()
assert np.allclose(elmcond.value[sphere3_msh.elm.tag1 == 1],
[0, -1, -2, -1, 3, 4, -2, 4, 5])
assert np.allclose(elmcond.value[sphere3_msh.elm.tag1 == 4],
[7, 0, 0, 0, 7, 0, 0, 0, 7])
assert np.allclose(elmcond.value[sphere3_msh.elm.tag1 == 5],
[9, 0, 0, 0, 9, 0, 0, 0, 9])
os.remove(l.fn_tensor_nifti)
def test_write_to_hdf5(self):
l = sim_struct.SimuList()
l.anisotropy_type = 'vn'
l.cond[3].value = 2
l.cond[0].distribution_type = 'beta'
l.anisotropic_tissues = [3]
v = np.zeros((255, 255, 255, 6))
for i in range(6):
v[:, :, :, i] = i
affine = np.array([[-1, 0, 0, 128], [0, 1, 0, -128], [0, 0, 1, -127],
[0, 0, 0, 1]])
l.anisotropy_vol = v
l.anisotropy_affine = affine
with tempfile.NamedTemporaryFile(suffix='.hdf5') as f:
fn_hdf5 = f.name
l._write_conductivity_to_hdf5(fn_hdf5)
with h5py.File(fn_hdf5) as f:
assert np.isclose(f['cond/values'][3], 2)
assert np.isnan(f['cond/values'][46])
assert f['cond/names'][0] == b'WM'
assert np.all(f['cond'].attrs['anisotropic_tissues'] == [3])
assert f['cond'].attrs['anisotropy_type'] == 'vn'
assert np.allclose(f['cond/anisotropy_affine'], affine)
assert np.allclose(f['cond/anisotropy_vol'], v)
assert f['cond/distribution_types'][0] == b'beta'
os.remove(fn_hdf5)
def test_list_cond_array(self, sphere3_msh):
l = sim_struct.SimuList()
l.cond[0].value = None
l.cond[1].value = None
l.cond[2].value = 21
l.cond[3].value = 31
l.cond[4].value = 41
l.mesh = sphere3_msh
elmcond = l.cond2elmdata()
assert np.all(elmcond.value[sphere3_msh.elm.tag1 == 3] == 21)
assert np.all(elmcond.value[sphere3_msh.elm.tag1 == 4] == 31)
assert np.all(elmcond.value[sphere3_msh.elm.tag1 == 5] == 41)
def sampler_args(sphere3):
poslist = sim_struct.SimuList()
poslist.cond[2].name = 'inner'
poslist.cond[2].distribution_type = 'beta'
poslist.cond[2].distribution_parameters = [2, 3, 0.3, 0.4]
poslist.cond[3].name = 'middle'
poslist.cond[3].value = 1
poslist.cond[4].name = 'outer'
poslist.cond[4].value = 10
with tempfile.NamedTemporaryFile(suffix='.hdf5') as f:
fn_hdf5 = f.name
if os.path.isfile(fn_hdf5):
os.remove(fn_hdf5)
return sphere3, poslist, fn_hdf5, [3]
def test_list_cond_array_aniso_vn(self, sphere3_msh):
l = sim_struct.SimuList()
l.anisotropy_type = 'vn'
l.conductivity[3].value = 2
l.conductivity[4].value = 7
l.conductivity[5].value = 9
l.anisotropic_tissues = [3]
l.fn_tensor_nifti = '/tmp/test_aniso.nii.gz'
l.mesh = sphere3_msh
v = np.zeros((255, 255, 255, 6))
# set-up tensor
v1 = np.random.rand(3)
v1 /= np.linalg.norm(v1)
v2 = np.random.rand(3)
v2 = v2 - v1.dot(v2) * v1 / np.linalg.norm(v1)
v2 /= np.linalg.norm(v2)
v3 = np.random.rand(3)
v3 = v3 - v1.dot(v3) * v1 / np.linalg.norm(v1)
v3 = v3 - v2.dot(v3) * v2 / np.linalg.norm(v2)
v3 /= np.linalg.norm(v3)
t = np.outer(v1, v1) + 2 * np.outer(v2, v2) + 3 * np.outer(v3, v3)
v[:, :, :] = t.reshape(-1)[[0, 1, 2, 4, 5, 8]]
affine = np.array([[-1, 0, 0, 128], [0, 1, 0, -128], [0, 0, 1, -127],
[0, 0, 0, 1]])
l.anisotropy_vol = v
l.anisotropy_affine = affine
elmcond = l.cond2elmdata()
tensor = elmcond.value[sphere3_msh.elm.tag1 == 3].reshape(-1, 3, 3)
t = affine[:3, :3].dot(t).dot(affine[:3, :3])
assert np.allclose(np.linalg.det(tensor), 2**3)
assert np.allclose(
np.linalg.eig(tensor)[1][:, 0],
np.linalg.eig(t)[1][0])
assert np.allclose(
np.linalg.eig(tensor)[1][:, 1],
np.linalg.eig(t)[1][1])
assert np.allclose(
np.linalg.eig(tensor)[1][:, 2],
np.linalg.eig(t)[1][2])
assert np.allclose(elmcond.value[sphere3_msh.elm.tag1 == 4],
[7, 0, 0, 0, 7, 0, 0, 0, 7])
assert np.allclose(elmcond.value[sphere3_msh.elm.tag1 == 5],
[9, 0, 0, 0, 9, 0, 0, 0, 9])
def test_list_read_mat_cond(self, mat_session):
l = sim_struct.SimuList()
l.read_cond_mat_struct(mat_session['poslist'][0][0][0][0])
assert l.cond[1].name == 'GM'
def test_list_postprocess(self):
l = sim_struct.SimuList()
with pytest.raises(ValueError):
l.postprocess = 'Y'
def test_list_anisotropy(self):
l = sim_struct.SimuList()
with pytest.raises(ValueError):
l.anisotropy_type = None
def test_set_conductivity(self):
l = sim_struct.SimuList()
l.conductivity[1].value = 5
assert l.cond[0].value == 5
def test_postprocessing(self, gpc_regression_instance, sphere3):
fn = 'gpc_testing.hdf5'
if os.path.isfile(fn):
os.remove(fn)
msh = sphere3.crop_mesh(elm_type=4)
msh.write_hdf5(fn, 'mesh_roi/')
# "normalized" coordinates
coords_norm = np.array([[-.5], [0], [.7]])
# Define regression object.
# The random variable is the conductivity of the layer 3
# Uniform distribution, between 0 and 10
gpc_reg = \
simnibs_gpc.gPC_regression([3],
['beta'], [[1], [1]],
[[0], [10]], [[0], [1]],
coords_norm, 'TCS',
data_file=fn)
gpc_reg.regularization_factors = [0]
# Create potentials
pot = np.tile(msh.nodes.node_coord[None, :, 0],
(coords_norm.shape[0], 1))
lst = sim_struct.SimuList()
lst.cond[3].value = 1 # Layer 4 conductivity
lst.cond[4].value = 10 # Layer 5 conductivity
lst._write_conductivity_to_hdf5(fn)
for i, c in enumerate(gpc_reg.grid.coords):
pot[i] *= c # Linear relationship between potential and random variable
# Record potentials
with h5py.File(fn, 'a') as f:
f.create_group('mesh_roi/data_matrices')
f['mesh_roi/data_matrices'].create_dataset('v_samples', data=pot)
# Postprocess
gpc_reg.postprocessing('eEjJ')
with h5py.File(fn, 'r') as f:
# Linear relationship between random variable and potential, uniform
# distribution
mean_E = f['mesh_roi/elmdata/E_mean'][()]
mean_E_expected = -np.array([5, 0, 0]) * 1e3
assert np.allclose(mean_E, mean_E_expected)
mean_e = f['mesh_roi/elmdata/normE_mean'][()]
assert np.allclose(mean_e, 5e3)
# J is a bit more complicated
mean_J = f['mesh_roi/elmdata/J_mean'][()]
assert np.allclose(mean_J[msh.elm.tag1==4],
mean_E_expected)
assert np.allclose(mean_J[msh.elm.tag1==5],
10 * mean_E_expected)
coeffs = gpc_reg.expand(gpc_reg.grid.coords ** 2, return_error=False)
mean = gpc_reg.mean(coeffs)
assert np.allclose(mean_J[msh.elm.tag1==3],
mean * -np.array([1, 0, 0]) * 1e3)
dsets = f['mesh_roi/elmdata/'].keys()
std_E = f['mesh_roi/elmdata/E_std'][()]
assert np.allclose(std_E, np.array([np.sqrt(1e8 / 12), 0., 0.]))
assert 'normE_std' in dsets
assert 'J_std' in dsets
assert 'E_sobol_3' in dsets
assert 'normE_sobol_3' in dsets
assert 'J_sobol_3' in dsets
assert 'E_sensitivity_3' in dsets
assert 'normE_sensitivity_3' in dsets
assert 'J_sensitivity_3' in dsets
os.remove(fn)