def test_expected_absolute_value_1d():
"""
Tests that expected absolute value works for the one dimensional case
"""
np.random.seed(0)
stddevs = np.array([0.0], dtype=np.float64)
stddevs[0] = 0.0
eav = e2d.expected_absolute_value(stddevs)
assert eav == 0.0
assert eav == _eav_by_brute_force(stddevs)
stddevs[0] = 1.0
eav = e2d.expected_absolute_value(stddevs)
arith_eav = np.linalg.norm(stddevs)
arith_eav *= arith_eav
assert np.isclose(eav, arith_eav, atol=0.0, rtol=0.0001)
#absolute(a - b) <= (atol + rtol * absolute(b))
#so rtol = 0.1 is around 10% rtol = 0.01 is 1 %
assert np.isclose(eav, _eav_by_brute_force(stddevs), atol=0.0, rtol=0.15)
for _ in range(10):
stddevs[0] = np.absolute(np.random.normal() * 10.0)
eav = e2d.expected_absolute_value(stddevs)
arith_eav = np.linalg.norm(stddevs)
arith_eav *= arith_eav
assert np.isclose(eav, arith_eav, atol=0.0, rtol=0.0001)
assert np.isclose(eav,
_eav_by_brute_force(stddevs),
atol=0.0,
rtol=0.15)
def test_expected_absolute_value_3d():
"""
Tests that expected absolute value works for the three dimensional case
"""
np.random.seed(0)
stddevs = np.array([0.0, 0.0, 0.0], dtype=np.float64)
eav = e2d.expected_absolute_value(stddevs)
assert eav == 0.0
assert eav == _eav_by_brute_force(stddevs)
stddevs = np.array([1.0, 1.0, 1.0], dtype=np.float64)
eav = e2d.expected_absolute_value(stddevs)
arith_eav = np.linalg.norm(stddevs)
arith_eav *= arith_eav
assert np.isclose(eav, arith_eav, atol=0.0, rtol=0.0001)
assert np.isclose(eav, _eav_by_brute_force(stddevs), atol=0.0, rtol=0.15)
for _ in range(10):
stddevs = np.absolute(np.random.normal((1, 1, 1)) * 10.0)
eav = e2d.expected_absolute_value(stddevs)
arith_eav = np.linalg.norm(stddevs)
arith_eav *= arith_eav
assert np.isclose(eav, arith_eav, atol=0.0, rtol=0.0001)
assert np.isclose(eav,
_eav_by_brute_force(stddevs),
atol=0.0,
rtol=0.15)
def init_reg(self):
"""
sets up the registration
"""
img = skimage.io.imread(self.image_file_name)
outline, _initial_guess = find_outer_contour(img)
target_point = make_target_point(outline)
self.plotter.initialise_new_reg(img, target_point, outline)
fle_sd = np.random.uniform(low=0.5, high=5.0)
moving_fle = np.zeros((1, 3), dtype=np.float64)
fixed_fle = np.array([fle_sd, fle_sd, fle_sd], dtype=np.float64)
fixed_fle_eavs = expected_absolute_value(fixed_fle)
moving_fle_eavs = expected_absolute_value(moving_fle)
if self.pbr is None:
self.pbr = PointBasedRegistration(target_point, fixed_fle_eavs,
moving_fle_eavs)
else:
self.pbr.reinit(target_point, fixed_fle_eavs, moving_fle_eavs)
if self.mouse_int is None:
self.mouse_int = AddFiducialMarker(self.fig, self.plotter,
self.pbr, self.logger,
fixed_fle, moving_fle)
self.mouse_int.reset_fiducials(fixed_fle_eavs)
return target_point
def test_pbr_3_fids():
"""
Tests for tre_from_fle_2d
"""
fixed_fle_std_dev = np.array([1.0, 1.0, 1.0], dtype=np.float64)
moving_fle_std_dev = np.array([0.0, 0.0, 0.0], dtype=np.float64)
fixed_fle_easv = expected_absolute_value(fixed_fle_std_dev)
moving_fle_easv = expected_absolute_value(moving_fle_std_dev)
target = np.array([[0.0, 0.0, 0.0]], dtype=np.float64)
pbr = pbreg.PointBasedRegistration(target, fixed_fle_easv, moving_fle_easv)
centre = np.array([0.0, 0.0, 0.0], dtype=np.float64)
radius = 20.0
expected_tre_squared = 0
expected_fre = 0
repeats = 100
no_fids = 3
ave_tresq, ave_fresq, expected_tre_squared, expected_fre, p_value = \
_run_registrations(pbr, no_fids, centre, radius,
fixed_fle_std_dev,
moving_fle_std_dev, repeats)
assert np.isclose(ave_tresq, expected_tre_squared, atol=0.0, rtol=0.10)
assert np.isclose(ave_fresq, expected_fre, atol=0.0, rtol=0.05)
assert p_value > 0.05
def test_init_with_moving_fle():
"""
Init pbr with moving fle should yield non implemented error
"""
fixed_fle_std_dev = np.array([1.0, 1.0, 1.0], dtype=np.float64)
moving_fle_std_dev = np.array([1.0, 1.0, 1.0], dtype=np.float64)
fixed_fle_easv = expected_absolute_value(fixed_fle_std_dev)
moving_fle_easv = expected_absolute_value(moving_fle_std_dev)
target = np.array([[0.0, 0.0, 0.0]], dtype=np.float64)
with pytest.raises(NotImplementedError):
pbreg.PointBasedRegistration(target, fixed_fle_easv, moving_fle_easv)
def test_pbr_20_fids_offset_target():
"""
Tests for tre_from_fle_2d
"""
fixed_fle_std_dev = np.array([1.0, 1.0, 1.0], dtype=np.float64)
moving_fle_std_dev = np.array([0.0, 0.0, 0.0], dtype=np.float64)
fixed_fle_easv = expected_absolute_value(fixed_fle_std_dev)
moving_fle_easv = expected_absolute_value(moving_fle_std_dev)
target = np.array([[2.0, 1.0, 0.0]], dtype=np.float64)
pbr = pbreg.PointBasedRegistration(target, fixed_fle_easv, moving_fle_easv)
centre = np.array([0.0, 0.0, 0.0], dtype=np.float64)
radius = 20.0
repeats = 200
no_fids = 20
#test get transformed target before registration
status, transformed_target = pbr.get_transformed_target()
assert not status
assert transformed_target is None
ave_tresq, ave_fresq, expected_tre_squared, expected_fre, p_value = \
_run_registrations(pbr, no_fids, centre, radius,
fixed_fle_std_dev,
moving_fle_std_dev, repeats)
assert np.isclose(ave_tresq, expected_tre_squared, atol=0.0, rtol=0.10)
assert np.isclose(ave_fresq, expected_fre, atol=0.0, rtol=0.05)
assert p_value > 0.05
#test get transformed target after registration
status, transformed_target = pbr.get_transformed_target()
assert status
assert np.allclose(np.transpose(transformed_target), target, atol=1.0)
def getfle():
"""
Returns values for fiducial localisation errors
Values are randomly selected from a uniform
distribution from 0.5 to 5.0 pixels
"""
fle_sd = np.random.uniform(low=0.5, high=5.0)
#change fle_ratio if you want anisotropic fle
fle_ratio = np.array([1.0, 1.0, 1.0], dtype=np.float64)
anis_scale = math.sqrt(3.0 / (np.linalg.norm(fle_ratio)**2))
fixed_fle = fle_ratio * fle_sd * anis_scale
moving_fle = np.array([0., 0., 0.], dtype=np.float64)
fixed_fle_eavs = expected_absolute_value(fixed_fle)
moving_fle_eavs = expected_absolute_value(moving_fle)
returnjson = jsonify({
'fixed_fle_sd': fixed_fle.tolist(),
'moving_fle_sd': moving_fle.tolist(),
'fixed_fle_eav': fixed_fle_eavs.tolist(),
'moving_fle_eav': moving_fle_eavs.tolist()
})
return returnjson