def test_fit_sphere_to_headshape():
"""Test fitting a sphere to digitization points."""
# Create points of various kinds
rad = 0.09
big_rad = 0.12
center = np.array([0.0005, -0.01, 0.04])
dev_trans = np.array([0., -0.005, -0.01])
dev_center = center - dev_trans
dig = [
# Left auricular
{'coord_frame': FIFF.FIFFV_COORD_HEAD,
'ident': FIFF.FIFFV_POINT_LPA,
'kind': FIFF.FIFFV_POINT_CARDINAL,
'r': np.array([-1.0, 0.0, 0.0])},
# Nasion
{'coord_frame': FIFF.FIFFV_COORD_HEAD,
'ident': FIFF.FIFFV_POINT_NASION,
'kind': FIFF.FIFFV_POINT_CARDINAL,
'r': np.array([0.0, 1.0, 0.0])},
# Right auricular
{'coord_frame': FIFF.FIFFV_COORD_HEAD,
'ident': FIFF.FIFFV_POINT_RPA,
'kind': FIFF.FIFFV_POINT_CARDINAL,
'r': np.array([1.0, 0.0, 0.0])},
# Top of the head (extra point)
{'coord_frame': FIFF.FIFFV_COORD_HEAD,
'kind': FIFF.FIFFV_POINT_EXTRA,
'ident': 0,
'r': np.array([0.0, 0.0, 1.0])},
# EEG points
# Fz
{'coord_frame': FIFF.FIFFV_COORD_HEAD,
'kind': FIFF.FIFFV_POINT_EEG,
'ident': 0,
'r': np.array([0, .72, .69])},
# F3
{'coord_frame': FIFF.FIFFV_COORD_HEAD,
'kind': FIFF.FIFFV_POINT_EEG,
'ident': 1,
'r': np.array([-.55, .67, .50])},
# F4
{'coord_frame': FIFF.FIFFV_COORD_HEAD,
'kind': FIFF.FIFFV_POINT_EEG,
'ident': 2,
'r': np.array([.55, .67, .50])},
# Cz
{'coord_frame': FIFF.FIFFV_COORD_HEAD,
'kind': FIFF.FIFFV_POINT_EEG,
'ident': 3,
'r': np.array([0.0, 0.0, 1.0])},
# Pz
{'coord_frame': FIFF.FIFFV_COORD_HEAD,
'kind': FIFF.FIFFV_POINT_EEG,
'ident': 4,
'r': np.array([0, -.72, .69])},
]
for d in dig:
d['r'] *= rad
d['r'] += center
# Device to head transformation (rotate .2 rad over X-axis)
dev_head_t = Transform('meg', 'head', translation(*(dev_trans)))
info = Info(dig=dig, dev_head_t=dev_head_t)
# Degenerate conditions
pytest.raises(ValueError, fit_sphere_to_headshape, info,
dig_kinds=(FIFF.FIFFV_POINT_HPI,))
pytest.raises(ValueError, fit_sphere_to_headshape, info,
dig_kinds='foo', units='m')
info['dig'][0]['coord_frame'] = FIFF.FIFFV_COORD_DEVICE
pytest.raises(RuntimeError, fit_sphere_to_headshape, info, units='m')
info['dig'][0]['coord_frame'] = FIFF.FIFFV_COORD_HEAD
# # Test with 4 points that match a perfect sphere
dig_kinds = (FIFF.FIFFV_POINT_CARDINAL, FIFF.FIFFV_POINT_EXTRA)
with pytest.warns(RuntimeWarning, match='Only .* head digitization'):
r, oh, od = fit_sphere_to_headshape(info, dig_kinds=dig_kinds,
units='m')
kwargs = dict(rtol=1e-3, atol=1e-5)
assert_allclose(r, rad, **kwargs)
assert_allclose(oh, center, **kwargs)
assert_allclose(od, dev_center, **kwargs)
# Test with all points
dig_kinds = ('cardinal', FIFF.FIFFV_POINT_EXTRA, 'eeg')
kwargs = dict(rtol=1e-3, atol=1e-3)
with pytest.warns(RuntimeWarning, match='Only .* head digitization'):
r, oh, od = fit_sphere_to_headshape(info, dig_kinds=dig_kinds,
units='m')
assert_allclose(r, rad, **kwargs)
assert_allclose(oh, center, **kwargs)
assert_allclose(od, dev_center, **kwargs)
# Test with some noisy EEG points only.
dig_kinds = 'eeg'
with pytest.warns(RuntimeWarning, match='Only .* head digitization'):
r, oh, od = fit_sphere_to_headshape(info, dig_kinds=dig_kinds,
units='m')
kwargs = dict(rtol=1e-3, atol=1e-2)
assert_allclose(r, rad, **kwargs)
assert_allclose(oh, center, **kwargs)
assert_allclose(od, center, **kwargs)
# Test big size
dig_kinds = ('cardinal', 'extra')
info_big = deepcopy(info)
for d in info_big['dig']:
d['r'] -= center
d['r'] *= big_rad / rad
d['r'] += center
with pytest.warns(RuntimeWarning, match='Estimated head size'):
r, oh, od = fit_sphere_to_headshape(info_big, dig_kinds=dig_kinds,
units='mm')
assert_allclose(oh, center * 1000, atol=1e-3)
assert_allclose(r, big_rad * 1000, atol=1e-3)
del info_big
# Test offcenter
dig_kinds = ('cardinal', 'extra')
info_shift = deepcopy(info)
shift_center = np.array([0., -0.03, 0.])
for d in info_shift['dig']:
d['r'] -= center
d['r'] += shift_center
with pytest.warns(RuntimeWarning, match='from head frame origin'):
r, oh, od = fit_sphere_to_headshape(
info_shift, dig_kinds=dig_kinds, units='m')
assert_allclose(oh, shift_center, atol=1e-6)
assert_allclose(r, rad, atol=1e-6)
# Test "auto" mode (default)
# Should try "extra", fail, and go on to EEG
with pytest.warns(RuntimeWarning, match='Only .* head digitization'):
r, oh, od = fit_sphere_to_headshape(info, units='m')
kwargs = dict(rtol=1e-3, atol=1e-3)
assert_allclose(r, rad, **kwargs)
assert_allclose(oh, center, **kwargs)
assert_allclose(od, dev_center, **kwargs)
with pytest.warns(RuntimeWarning, match='Only .* head digitization'):
r2, oh2, od2 = fit_sphere_to_headshape(info, units='m')
assert_allclose(r, r2, atol=1e-7)
assert_allclose(oh, oh2, atol=1e-7)
assert_allclose(od, od2, atol=1e-7)
# this one should pass, 1 EXTRA point and 3 EEG (but the fit is terrible)
info = Info(dig=dig[:7], dev_head_t=dev_head_t)
with pytest.warns(RuntimeWarning, match='Only .* head digitization'):
r, oh, od = fit_sphere_to_headshape(info, units='m')
# this one should fail, 1 EXTRA point and 3 EEG (but the fit is terrible)
info = Info(dig=dig[:6], dev_head_t=dev_head_t)
pytest.raises(ValueError, fit_sphere_to_headshape, info, units='m')
pytest.raises(TypeError, fit_sphere_to_headshape, 1, units='m')
def _state_to_glm(glm):
"""
Convert state of GLM stored as dictionary to a MNE-NIRS GLM type.
Parameters
----------
glm : dict
State of GLM type.
Returns
-------
glm : _BaseGLM
RegressionResults or ContrastResults class
which stores the GLM results.
"""
if glm['classname'] == "<class 'mne_nirs.statistics._glm_level_first" \
".RegressionResults'>":
for channel in glm['data']:
# Recreate model type
if glm['data'][channel]['modelname'] == \
"<class 'nilearn.glm.regression.ARModel'>":
model = nilearn.glm.regression.ARModel(
glm['data'][channel]['model']['design'],
glm['data'][channel]['model']['rho'],
)
elif glm['data'][channel]['modelname'] == \
"<class 'nilearn.glm.regression.OLSModel'>":
model = nilearn.glm.regression.OLSModel(
glm['data'][channel]['model']['design'],
)
else:
raise IOError("Unknown model type "
f"{glm['data'][channel]['modelname']}")
for key in glm['data'][channel]['model']:
model.__setattr__(key, glm['data'][channel]['model'][key])
glm['data'][channel]['model'] = model
# Then recreate result type
res = nilearn.glm.regression.RegressionResults(
glm['data'][channel]['theta'],
glm['data'][channel]['Y'],
glm['data'][channel]['model'],
glm['data'][channel]['whitened_Y'],
glm['data'][channel]['whitened_residuals'],
cov=glm['data'][channel]['cov']
)
for key in glm['data'][channel]:
res.__setattr__(key, glm['data'][channel][key])
glm['data'][channel] = res
# Ensure order of dictionary matches info
data = {k: glm['data'][k] for k in glm['info']['ch_names']}
return RegressionResults(Info(glm['info']), data, glm['design'])
elif glm['classname'] == "<class 'mne_nirs.statistics._glm_level_first" \
".ContrastResults'>":
data = nilearn.glm.contrasts.Contrast(glm['data']['effect'],
glm['data']['variance'])
for key in glm['data']:
data.__setattr__(key, glm['data'][key])
return ContrastResults(Info(glm['info']), data, glm['design'])
else:
raise IOError('Unable to read data')
