def _tgramer(bundles, bundle_dict, affine):
tgram = nib.streamlines.Tractogram([], {'bundle': []})
for b in bundles:
print("Segmenting: %s" % b)
this_sl = list(bundles[b])
this_tgram = nib.streamlines.Tractogram(
this_sl,
data_per_streamline={
'bundle': (len(this_sl) * [bundle_dict[b]['uid']])
},
affine_to_rasmm=affine)
tgram = aus.add_bundles(tgram, this_tgram)
return tgram
def clean_bundles(subses_dict, bundles_file, bundle_dict, clean_params,
tracking_params, segmentation_params):
img = nib.load(subses_dict['dwi_file'])
sft = load_tractogram(bundles_file, img, Space.VOX)
img = nib.load(subses_dict['dwi_file'])
start_time = time()
tgram = nib.streamlines.Tractogram([], {'bundle': []})
if clean_params['return_idx']:
return_idx = {}
for b in bundle_dict.keys():
if b != "whole_brain":
idx = np.where(
sft.data_per_streamline['bundle'] == bundle_dict[b]['uid'])[0]
this_tg = StatefulTractogram(sft.streamlines[idx], img, Space.VOX)
this_tg = seg.clean_bundle(this_tg, **clean_params)
if clean_params['return_idx']:
this_tg, this_idx = this_tg
idx_file = bundles_file.split('.')[0] + '.json'
with open(idx_file) as ff:
bundle_idx = json.load(ff)["idx"][b]
return_idx[b] = np.array(bundle_idx)[this_idx].tolist()
this_tgram = nib.streamlines.Tractogram(
this_tg.streamlines,
data_per_streamline={
'bundle': (len(this_tg) * [bundle_dict[b]['uid']])
},
affine_to_rasmm=img.affine)
tgram = aus.add_bundles(tgram, this_tgram)
sft = StatefulTractogram(tgram.streamlines,
sft,
Space.VOX,
data_per_streamline=tgram.data_per_streamline)
seg_args = get_default_args(seg.clean_bundle)
for k in seg_args:
if callable(seg_args[k]):
seg_args[k] = seg_args[k].__name__
meta = dict(source=bundles_file, Parameters=seg_args)
if clean_params['return_idx']:
meta["idx"] = return_idx
meta["Timing"] = time() - start_time
return sft, meta
def test_add_bundles():
t1 = nib.streamlines.Tractogram([
np.array([[0, 0, 0], [0, 0, 0.5], [0, 0, 1], [0, 0, 1.5]]),
np.array([[0, 0, 0], [0, 0.5, 0.5], [0, 1, 1]])
])
t2 = nib.streamlines.Tractogram([
np.array([[0, 0, 0], [0, 0, 0.5], [0, 0, 1], [0, 0, 1.5]]),
np.array([[0, 0, 0], [0, 0.5, 0.5], [0, 1, 1]])
])
added = aus.add_bundles(t1, t2)
test_tgram = nib.streamlines.Tractogram([
np.array([[0, 0, 0], [0, 0, 0.5], [0, 0, 1], [0, 0, 1.5]]),
np.array([[0, 0, 0], [0, 0.5, 0.5], [0, 1, 1]]),
np.array([[0, 0, 0], [0, 0, 0.5], [0, 0, 1], [0, 0, 1.5]]),
np.array([[0, 0, 0], [0, 0.5, 0.5], [0, 1, 1]])
])
for sl1, sl2 in zip(added.streamlines, test_tgram.streamlines):
npt.assert_array_equal(sl1, sl2)
def _clean_bundles(row,
wm_labels,
bundle_dict,
reg_template,
odf_model="DTI",
directions="det",
n_seeds=2,
random_seeds=False,
force_recompute=False):
clean_bundles_file = _get_fname(
row, '%s_%s_clean_bundles.trk' % (odf_model, directions))
if not op.exists(clean_bundles_file) or force_recompute:
bundles_file = _bundles(row,
wm_labels,
bundle_dict,
reg_template,
odf_model=odf_model,
directions=directions,
n_seeds=n_seeds,
random_seeds=random_seeds,
force_recompute=False)
tg = nib.streamlines.load(bundles_file).tractogram
sl = tg.apply_affine(np.linalg.inv(row['dwi_affine'])).streamlines
tgram = nib.streamlines.Tractogram([], {'bundle': []})
for b in bundle_dict.keys():
idx = np.where(
tg.data_per_streamline['bundle'] == bundle_dict[b]['uid'])[0]
this_sl = sl[idx]
this_sl = seg.clean_fiber_group(this_sl)
this_tgram = nib.streamlines.Tractogram(
this_sl,
data_per_streamline={
'bundle': (len(this_sl) * [bundle_dict[b]['uid']])
},
affine_to_rasmm=row['dwi_affine'])
tgram = aus.add_bundles(tgram, this_tgram)
nib.streamlines.save(tgram, clean_bundles_file)
return clean_bundles_file
def _clean_bundles(self, row):
odf_model = self.tracking_params['odf_model']
directions = self.tracking_params['directions']
seg_algo = self.segmentation_params['seg_algo']
clean_bundles_file = self._get_fname(
row, f'_space-RASMM_model-{odf_model}_desc-{directions}-'
f'{seg_algo}-clean_tractography.trk')
if self.force_recompute or not op.exists(clean_bundles_file):
bundles_file = self._segment(row)
sft = load_tractogram(bundles_file, row['dwi_img'], Space.VOX)
tgram = nib.streamlines.Tractogram([], {'bundle': []})
if self.clean_params['return_idx']:
return_idx = {}
for b in self.bundle_dict.keys():
if b != "whole_brain":
idx = np.where(sft.data_per_streamline['bundle'] ==
self.bundle_dict[b]['uid'])[0]
this_tg = StatefulTractogram(sft.streamlines[idx],
row['dwi_img'], Space.VOX)
this_tg = seg.clean_bundle(this_tg, **self.clean_params)
if self.clean_params['return_idx']:
this_tg, this_idx = this_tg
idx_file = bundles_file.split('.')[0] + '_idx.json'
with open(idx_file) as ff:
bundle_idx = json.load(ff)[b]
return_idx[b] = \
np.array(bundle_idx)[this_idx].tolist()
this_tgram = nib.streamlines.Tractogram(
this_tg.streamlines,
data_per_streamline={
'bundle':
(len(this_tg) * [self.bundle_dict[b]['uid']])
},
affine_to_rasmm=row['dwi_affine'])
tgram = aus.add_bundles(tgram, this_tgram)
save_tractogram(
StatefulTractogram(
tgram.streamlines,
sft,
Space.VOX,
data_per_streamline=tgram.data_per_streamline),
clean_bundles_file)
seg_args = get_default_args(seg.clean_bundle)
for k in seg_args:
if callable(seg_args[k]):
seg_args[k] = seg_args[k].__name__
meta = dict(source=bundles_file, Parameters=seg_args)
meta_fname = clean_bundles_file.split('.')[0] + '.json'
afd.write_json(meta_fname, meta)
if self.clean_params['return_idx']:
afd.write_json(
clean_bundles_file.split('.')[0] + '_idx.json', return_idx)
return clean_bundles_file