def remove_cci_outliers(streamlines):
s = Streamlines(streamlines)
cci = cluster_confidence(s,subsample=7)
keep_streamlines = Streamlines()
keep_streamlines_idx = list()
for i, sl in enumerate(s):
if cci[i] >= 1:
keep_streamlines.append(sl)
keep_streamlines_idx.append(i)
return keep_streamlines,keep_streamlines_idx
def clean_subject(ndata):
cci = []
streamlines_evl = Streamlines()
for j in range(np.shape(ndata)[0]):
tmp = ndata[j]
tmp = zero_remove(tmp)
streamlines_evl.append(tmp)
lengths = list(utils.length(streamlines_evl))
long_streamlines_evl = Streamlines()
for i, sl in enumerate(streamlines_evl):
if lengths[i] > 40:
long_streamlines_evl.append(sl)
streamlines_test = remove_indentical(10, long_streamlines_evl)
cci_idv = cluster_confidence(streamlines_test, subsample=64)
cci.append(cci_idv)
return gen_array(streamlines_test), np.array(cci[0])
def cal_cci(name_list, text):
cci = []
for i in range(len(name_list)):
path = os.path.join('np_data', name_list[i])
sample_data = np.load(path)
ndata = sample_data['arr_0']
streamlines_evl = Streamlines()
for j in range(np.shape(ndata)[0]):
tmp = ndata[j]
tmp = zero_remove(tmp)
streamlines_evl.append(tmp)
streamlines_test = remove_indentical(10, streamlines_evl)
cci_idv = cluster_confidence(streamlines_test, subsample=64)
cci.append(cci_idv)
print('the ' + str(i + 1) + ' finished')
path_save = os.path.join('data_4_model', text)
np.save(path_save, cci)
return cci
"""
lengths = list(length(streamlines))
long_streamlines = Streamlines()
for i, sl in enumerate(streamlines):
if lengths[i] > 40:
long_streamlines.append(sl)
"""
Now we calculate the Cluster Confidence Index using the corpus callosum
streamline bundle and visualize them.
"""
cci = cluster_confidence(long_streamlines)
# Visualize the streamlines, colored by cci
ren = window.Renderer()
hue = [0.5, 1]
saturation = [0.0, 1.0]
lut_cmap = actor.colormap_lookup_table(scale_range=(cci.min(), cci.max()/4),
hue_range=hue,
saturation_range=saturation)
bar3 = actor.scalar_bar(lut_cmap)
ren.add(bar3)
stream_actor = actor.line(long_streamlines, cci, linewidth=0.1,
"""
We do not want our results inflated by short streamlines, so we remove
streamlines shorter than 40mm prior to calculating the CCI.
"""
lengths = list(length(streamlines))
long_streamlines = Streamlines()
for i, sl in enumerate(streamlines):
if lengths[i] > 40:
long_streamlines.append(sl)
"""
Now we calculate the Cluster Confidence Index using the corpus callosum
streamline bundle and visualize them.
"""
cci = cluster_confidence(long_streamlines)
# Visualize the streamlines, colored by cci
ren = window.Renderer()
hue = [0.5, 1]
saturation = [0.0, 1.0]
lut_cmap = actor.colormap_lookup_table(scale_range=(cci.min(), cci.max() / 4),
hue_range=hue,
saturation_range=saturation)
bar3 = actor.scalar_bar(lut_cmap)
ren.add(bar3)
stream_actor = actor.line(long_streamlines,
def test_cluster_confidence():
mysl = np.array([np.arange(10)] * 3, 'float').T
# a short streamline (<20 mm) should raise an error unless override=True
test_streamlines = Streamlines()
test_streamlines.append(mysl)
assert_raises(ValueError, cluster_confidence, test_streamlines)
cci = cluster_confidence(test_streamlines, override=True)
# two identical streamlines should raise an error
test_streamlines = Streamlines()
test_streamlines.append(mysl, cache_build=True)
test_streamlines.append(mysl)
test_streamlines.finalize_append()
assert_raises(ValueError, cluster_confidence, test_streamlines)
# 3 offset collinear streamlines
test_streamlines = Streamlines()
test_streamlines.append(mysl, cache_build=True)
test_streamlines.append(mysl + 1)
test_streamlines.append(mysl + 2)
test_streamlines.finalize_append()
cci = cluster_confidence(test_streamlines, override=True)
assert_equal(cci[0], cci[2])
assert_true(cci[1] > cci[0])
# 3 parallel streamlines
mysl = np.zeros([10, 3])
mysl[:, 0] = np.arange(10)
mysl2 = mysl.copy()
mysl2[:, 1] = 1
mysl3 = mysl.copy()
mysl3[:, 1] = 2
mysl4 = mysl.copy()
mysl4[:, 1] = 4
mysl5 = mysl.copy()
mysl5[:, 1] = 5000
test_streamlines_p1 = Streamlines()
test_streamlines_p1.append(mysl, cache_build=True)
test_streamlines_p1.append(mysl2)
test_streamlines_p1.append(mysl3)
test_streamlines_p1.finalize_append()
test_streamlines_p2 = Streamlines()
test_streamlines_p2.append(mysl, cache_build=True)
test_streamlines_p2.append(mysl3)
test_streamlines_p2.append(mysl4)
test_streamlines_p2.finalize_append()
test_streamlines_p3 = Streamlines()
test_streamlines_p3.append(mysl, cache_build=True)
test_streamlines_p3.append(mysl2)
test_streamlines_p3.append(mysl3)
test_streamlines_p3.append(mysl5)
test_streamlines_p3.finalize_append()
cci_p1 = cluster_confidence(test_streamlines_p1, override=True)
cci_p2 = cluster_confidence(test_streamlines_p2, override=True)
# test relative distance
assert_array_equal(cci_p1, cci_p2 * 2)
# test simple cci calculation
expected_p1 = np.array([1. / 1 + 1. / 2, 1. / 1 + 1. / 1, 1. / 1 + 1. / 2])
expected_p2 = np.array([1. / 2 + 1. / 4, 1. / 2 + 1. / 2, 1. / 2 + 1. / 4])
assert_array_equal(expected_p1, cci_p1)
assert_array_equal(expected_p2, cci_p2)
# test power variable calculation (dropoff with distance)
cci_p1_pow2 = cluster_confidence(test_streamlines_p1,
power=2,
override=True)
expected_p1_pow2 = np.array([
np.power(1. / 1, 2) + np.power(1. / 2, 2),
np.power(1. / 1, 2) + np.power(1. / 1, 2),
np.power(1. / 1, 2) + np.power(1. / 2, 2)
])
assert_array_equal(cci_p1_pow2, expected_p1_pow2)
# test max distance (ignore distant sls)
cci_dist = cluster_confidence(test_streamlines_p3,
max_mdf=5,
override=True)
expected_cci_dist = np.concatenate([cci_p1, np.zeros(1)])
assert_array_equal(cci_dist, expected_cci_dist)
score_tmp = cal_dist_score(avg_dist, cc_uniques)
scores.append(score_tmp)
return np.array(scores)
result_dis = cal_score_subject(bundle)
#%%
streamlines_evl = Streamlines()
for j in range(np.shape(bundle)[0]):
tmp = bundle[j]
tmp = zero_remove(tmp)
streamlines_evl.append(tmp)
result_cci = cluster_confidence(streamlines_evl, subsample=64)
#%%
preds_cci = 1 * (result_cci > results.cci_thre)
preds_dis = 1 * (result_dis < results.dis_thre)
pred_if = 1 * (pred_if > 0)
pred_1 = 1 * ((preds_cci + preds_dis + pred_if) > 2)
bundle_str = Streamlines()
for i in range(np.shape(bundle)[0]):
tmp = bundle[i]
tmp = zero_remove(tmp)
#tmp = tmp[~np.all(tmp == 0, axis=-1)]
for i in range(id.__len__()): #
for j in range(i + 1): #
edge_s_list = []
# print(i,j)
if (i + 1, j + 1) in streamline_dict and mat_medians[i, j] > 0:
edge_s_list += streamline_dict[(i + 1, j + 1)]
if (j + 1, i + 1) in streamline_dict and mat_medians[i, j] > 0:
edge_s_list += streamline_dict[(j + 1, i + 1)]
edge_vec_vols = [mat_medians[i, j]] * edge_s_list.__len__()
s_list = s_list + edge_s_list
vec_vols = vec_vols + edge_vec_vols
s = Streamlines(s_list)
cci = cluster_confidence(s)
keep_streamlines = Streamlines()
for i, sl in enumerate(s):
if cci[i] >= 1:
keep_streamlines.append(sl)
# Visualize the streamlines we kept
ren = window.Renderer()
keep_streamlines_actor = actor.line(keep_streamlines, linewidth=0.1)
ren.add(keep_streamlines_actor)
interactive = True
if interactive: