def test_coreg_class_init(drop_point_kind):
"""Test that Coregistration can be instantiated with various digs."""
fiducials, _ = read_fiducials(fid_fname)
info = read_info(raw_fname)
dig_list = []
eeg_chans = []
for pt in info['dig']:
if pt['kind'] != drop_point_kind:
dig_list.append(pt)
if pt['kind'] == FIFF.FIFFV_POINT_EEG:
eeg_chans.append(f"EEG {pt['ident']:03d}")
this_info = info.copy()
this_info.set_montage(DigMontage(dig=dig_list, ch_names=eeg_chans),
on_missing='ignore')
Coregistration(this_info,
subject='sample',
subjects_dir=subjects_dir,
fiducials=fiducials)
def test_coregistration(scale_mode, ref_scale, grow_hair, fiducials,
fid_match):
"""Test automated coregistration."""
subject = 'sample'
if fiducials is None:
fiducials, coord_frame = read_fiducials(fid_fname)
assert coord_frame == FIFF.FIFFV_COORD_MRI
info = read_info(raw_fname)
for d in info['dig']:
d['r'] = d['r'] * ref_scale
trans = read_trans(trans_fname)
coreg = Coregistration(info,
subject=subject,
subjects_dir=subjects_dir,
fiducials=fiducials)
assert np.allclose(coreg._last_parameters, coreg._parameters)
coreg.set_fid_match(fid_match)
default_params = list(coreg._default_parameters)
coreg.set_rotation(default_params[:3])
coreg.set_translation(default_params[3:6])
coreg.set_scale(default_params[6:9])
coreg.set_grow_hair(grow_hair)
coreg.set_scale_mode(scale_mode)
# Identity transform
errs_id = coreg.compute_dig_mri_distances()
is_scaled = ref_scale != [1., 1., 1.]
id_max = 0.03 if is_scaled and scale_mode == '3-axis' else 0.02
assert 0.005 < np.median(errs_id) < id_max
# Fiducial transform + scale
coreg.fit_fiducials(verbose=True)
assert coreg._extra_points_filter is None
coreg.omit_head_shape_points(distance=0.02)
assert coreg._extra_points_filter is not None
errs_fid = coreg.compute_dig_mri_distances()
assert_array_less(0, errs_fid)
if is_scaled or scale_mode is not None:
fid_max = 0.05
fid_med = 0.02
else:
fid_max = 0.03
fid_med = 0.01
assert_array_less(errs_fid, fid_max)
assert 0.001 < np.median(errs_fid) < fid_med
assert not np.allclose(coreg._parameters, default_params)
coreg.omit_head_shape_points(distance=-1)
coreg.omit_head_shape_points(distance=5. / 1000)
assert coreg._extra_points_filter is not None
# ICP transform + scale
coreg.fit_icp(verbose=True)
assert isinstance(coreg.trans, Transform)
errs_icp = coreg.compute_dig_mri_distances()
assert_array_less(0, errs_icp)
if is_scaled or scale_mode == '3-axis':
icp_max = 0.015
else:
icp_max = 0.01
assert_array_less(errs_icp, icp_max)
assert 0.001 < np.median(errs_icp) < 0.004
assert np.rad2deg(
_angle_between_quats(rot_to_quat(coreg.trans['trans'][:3, :3]),
rot_to_quat(trans['trans'][:3, :3]))) < 13
if scale_mode is None:
atol = 1e-7
else:
atol = 0.35
assert_allclose(coreg._scale, ref_scale, atol=atol)
coreg.reset()
assert_allclose(coreg._parameters, default_params)
def test_coreg_class_gui_match():
"""Test that using Coregistration matches mne coreg."""
fiducials, _ = read_fiducials(fid_fname)
info = read_info(raw_fname)
coreg = Coregistration(info,
subject='sample',
subjects_dir=subjects_dir,
fiducials=fiducials)
assert_allclose(coreg.trans['trans'], np.eye(4), atol=1e-6)
# mne coreg -s sample -d subjects -f MEG/sample/sample_audvis_trunc_raw.fif
# then "Fit Fid.", Save... to get trans, read_trans:
want_trans = [
[9.99428809e-01, 2.94733196e-02, 1.65350307e-02, -8.76054692e-04],
[-1.92420650e-02, 8.98512006e-01, -4.38526988e-01, 9.39774036e-04],
[-2.77817696e-02, 4.37958330e-01, 8.98565888e-01, -8.29207990e-03],
[0, 0, 0, 1]
]
coreg.set_fid_match('matched')
coreg.fit_fiducials(verbose=True)
assert_allclose(coreg.trans['trans'], want_trans, atol=1e-6)
# Set ICP iterations to one, click "Fit ICP"
want_trans = [
[9.99512792e-01, 2.80128177e-02, 1.37659665e-02, 6.08855276e-04],
[-1.91694051e-02, 8.98992002e-01, -4.37545270e-01, 9.66848747e-04],
[-2.46323701e-02, 4.37068194e-01, 8.99091005e-01, -1.44129358e-02],
[0, 0, 0, 1]
]
coreg.fit_icp(1, verbose=True)
assert_allclose(coreg.trans['trans'], want_trans, atol=1e-6)
# Set ICP iterations to 20, click "Fit ICP"
with catch_logging() as log:
coreg.fit_icp(20, verbose=True)
log = log.getvalue()
want_trans = [
[9.97582495e-01, 2.12266613e-02, 6.61706254e-02, -5.07694029e-04],
[1.81089472e-02, 8.39900672e-01, -5.42437911e-01, 7.81218382e-03],
[-6.70908988e-02, 5.42324841e-01, 8.37485850e-01, -2.50057746e-02],
[0, 0, 0, 1]
]
assert_allclose(coreg.trans['trans'], want_trans, atol=1e-6)
assert 'ICP 19' in log
assert 'ICP 20' not in log # converged on 19
# Change to uniform scale mode, "Fit Fiducials" in scale UI
coreg.set_scale_mode('uniform')
coreg.fit_fiducials()
want_scale = [0.975] * 3
want_trans = [
[9.99428809e-01, 2.94733196e-02, 1.65350307e-02, -9.25998494e-04],
[-1.92420650e-02, 8.98512006e-01, -4.38526988e-01, -1.03350170e-03],
[-2.77817696e-02, 4.37958330e-01, 8.98565888e-01, -9.03170835e-03],
[0, 0, 0, 1]
]
assert_allclose(coreg.scale, want_scale, atol=5e-4)
assert_allclose(coreg.trans['trans'], want_trans, atol=1e-6)
# Click "Fit ICP" in scale UI
with catch_logging() as log:
coreg.fit_icp(20, verbose=True)
log = log.getvalue()
assert 'ICP 18' in log
assert 'ICP 19' not in log
want_scale = [1.036] * 3
want_trans = [
[9.98992383e-01, 1.72388796e-02, 4.14364934e-02, 6.19427126e-04],
[6.80460501e-03, 8.54430079e-01, -5.19521892e-01, 5.58008114e-03],
[-4.43605632e-02, 5.19280374e-01, 8.53451848e-01, -2.03358755e-02],
[0, 0, 0, 1]
]
assert_allclose(coreg.scale, want_scale, atol=5e-4)
assert_allclose(coreg.trans['trans'], want_trans, atol=1e-6)
# Change scale mode to 3-axis, click "Fit ICP" in scale UI
coreg.set_scale_mode('3-axis')
with catch_logging() as log:
coreg.fit_icp(20, verbose=True)
log = log.getvalue()
assert 'ICP 7' in log
assert 'ICP 8' not in log
want_scale = [1.025, 1.010, 1.121]
want_trans = [
[9.98387098e-01, 2.04762165e-02, 5.29526398e-02, 4.97257097e-05],
[1.13287698e-02, 8.42087150e-01, -5.39222538e-01, 7.09863892e-03],
[-5.56319728e-02, 5.38952649e-01, 8.40496957e-01, -1.46372067e-02],
[0, 0, 0, 1]
]
assert_allclose(coreg.scale, want_scale, atol=5e-4)
assert_allclose(coreg.trans['trans'], want_trans, atol=1e-6)
subjects_dir=subjects_dir,
surfaces="head-dense",
dig=True,
eeg=[],
meg='sensors',
show_axes=True,
coord_frame='meg')
view_kwargs = dict(azimuth=45,
elevation=90,
distance=0.6,
focalpoint=(0., 0., 0.))
# %%
# Set up the coregistration model
fiducials = "estimated" # get fiducials from fsaverage
coreg = Coregistration(info, subject, subjects_dir, fiducials=fiducials)
fig = mne.viz.plot_alignment(info, trans=coreg.trans, **plot_kwargs)
# %%
# Do first a coregistration fit using only 3 fiducial points. This allows
# to find a good initial solution before further optimization using
# head shape points. This can also be useful to detect outlier head shape
# points which are too far from the skin surface. One can see for example
# that on this dataset there is one such point and we will omit it from
# the subsequent fit.
coreg.fit_fiducials(verbose=True)
fig = mne.viz.plot_alignment(info, trans=coreg.trans, **plot_kwargs)
# %%
# Next we refine the transformation using a few iteration of the
# Iterative Closest Point (ICP) algorithm. As the initial fiducials