def test_path_length():
aoi = np.zeros((20, 20, 20), dtype=bool)
aoi[0, 0, 0] = 1
# A few tests for basic usage
x = np.arange(20)
streamlines = [np.array([x, x, x]).T]
pl = path_length(streamlines, np.eye(4), aoi)
expected = x.copy() * np.sqrt(3)
# expected[0] = np.inf
npt.assert_array_almost_equal(pl[x, x, x], expected)
aoi[19, 19, 19] = 1
pl = path_length(streamlines, np.eye(4), aoi)
expected = np.minimum(expected, expected[::-1])
npt.assert_array_almost_equal(pl[x, x, x], expected)
aoi[19, 19, 19] = 0
aoi[1, 1, 1] = 1
pl = path_length(streamlines, np.eye(4), aoi)
expected = (x - 1) * np.sqrt(3)
expected[0] = 0
npt.assert_array_almost_equal(pl[x, x, x], expected)
z = np.zeros(x.shape, x.dtype)
streamlines.append(np.array([x, z, z]).T)
pl = path_length(streamlines, np.eye(4), aoi)
npt.assert_array_almost_equal(pl[x, x, x], expected)
npt.assert_array_almost_equal(pl[x, 0, 0], x)
# Only streamlines that pass through aoi contribute to path length so if
# all streamlines are duds, plm will be all inf.
aoi[:] = 0
aoi[0, 0, 0] = 1
streamlines = []
for i in range(1000):
rando = np.random.random(size=(100, 3)) * 19 + .5
assert (rando > .5).all()
assert (rando < 19.5).all()
streamlines.append(rando)
pl = path_length(streamlines, np.eye(4), aoi)
npt.assert_array_almost_equal(pl, -1)
pl = path_length(streamlines, np.eye(4), aoi, fill_value=-12.)
npt.assert_array_almost_equal(pl, -12.)
def test_path_length():
aoi = np.zeros((20, 20, 20), dtype=bool)
aoi[0, 0, 0] = 1
# A few tests for basic usage
x = np.arange(20)
streamlines = [np.array([x, x, x]).T]
pl = path_length(streamlines, aoi, affine=np.eye(4))
expected = x.copy() * np.sqrt(3)
# expected[0] = np.inf
npt.assert_array_almost_equal(pl[x, x, x], expected)
aoi[19, 19, 19] = 1
pl = path_length(streamlines, aoi, affine=np.eye(4))
expected = np.minimum(expected, expected[::-1])
npt.assert_array_almost_equal(pl[x, x, x], expected)
aoi[19, 19, 19] = 0
aoi[1, 1, 1] = 1
pl = path_length(streamlines, aoi, affine=np.eye(4))
expected = (x - 1) * np.sqrt(3)
expected[0] = 0
npt.assert_array_almost_equal(pl[x, x, x], expected)
z = np.zeros(x.shape, x.dtype)
streamlines.append(np.array([x, z, z]).T)
pl = path_length(streamlines, aoi, affine=np.eye(4))
npt.assert_array_almost_equal(pl[x, x, x], expected)
npt.assert_array_almost_equal(pl[x, 0, 0], x)
# Only streamlines that pass through aoi contribute to path length so if
# all streamlines are duds, plm will be all inf.
aoi[:] = 0
aoi[0, 0, 0] = 1
streamlines = []
for i in range(1000):
rando = np.random.random(size=(100, 3)) * 19 + .5
assert (rando > .5).all()
assert (rando < 19.5).all()
streamlines.append(rando)
pl = path_length(streamlines, aoi, affine=np.eye(4))
npt.assert_array_almost_equal(pl, -1)
pl = path_length(streamlines, aoi, affine=np.eye(4), fill_value=-12.)
npt.assert_array_almost_equal(pl, -12.)
"""
Now we calculate the Path Length Map using the corpus callosum streamline
bundle and corpus callosum ROI.
NOTE: the mask used to seed the tracking does not have to be the Path
Length Map base ROI, as we do here, but it often makes sense for them to be the
same ROI if we want a map of the whole brain's distance back to our ROI.
(e.g. we could test a hypothesis about the motor system by making a streamline
bundle model of the cortico-spinal track (CST) and input a lesion mask as our
Path Length Map base ROI to restrict the analysis to the CST)
"""
path_length_map_base_roi = seed_mask
# calculate the WMPL
wmpl = path_length(streamlines, affine, path_length_map_base_roi)
# save the WMPL as a nifti
save_nifti('example_cc_path_length_map.nii.gz', wmpl.astype(np.float32),
affine)
# get the T1 to show anatomical context of the WMPL
t1_fname = get_fnames('stanford_t1')
t1_data = load_nifti_data(t1_fname)
fig = mpl.pyplot.figure()
fig.subplots_adjust(left=0.05, right=0.95)
ax = AxesGrid(fig, 111,
nrows_ncols=(1, 3),
cbar_location="right",
Now we calculate the Path Length Map using the corpus callosum streamline
bundle and corpus callosum ROI.
NOTE: the mask used to seed the tracking does not have to be the Path
Length Map base ROI, as we do here, but it often makes sense for them to be the
same ROI if we want a map of the whole brain's distance back to our ROI.
(e.g. we could test a hypothesis about the motor system by making a streamline
bundle model of the cortico-spinal track (CST) and input a lesion mask as our
Path Length Map base ROI to restrict the analysis to the CST)
"""
path_length_map_base_roi = seed_mask
# calculate the WMPL
wmpl = path_length(streamlines, path_length_map_base_roi, affine)
# save the WMPL as a nifti
path_length_img = nib.Nifti1Image(wmpl.astype(np.float32), affine)
nib.save(path_length_img, 'example_cc_path_length_map.nii.gz')
# get the T1 to show anatomical context of the WMPL
fetch_stanford_t1()
t1 = read_stanford_t1()
t1_data = t1.get_data()
fig = mpl.pyplot.figure()
fig.subplots_adjust(left=0.05, right=0.95)
ax = AxesGrid(fig, 111,
nrows_ncols=(1, 3),