def _plot_subnetwork_graph(Theta, coords, network_ix):
(x, y, z) = coords
p = Theta.shape[0]
network_ix = network_ix * np.ones((p,))
# 3D glass image of brain
fig = mlab.figure(bgcolor=(1, 1, 1), size=(900, 769))
mlab.clf()
fig.scene.disable_render = True
vmin = np.min(Theta[np.abs(Theta) != 0])
vmax = np.max(np.abs(Theta))
tubes, nodes = plot_graph(-Theta, x, y, z,
node_size=.6,
edge_vmin=vmin,
edge_vmax=vmax,
node_colormap='spectral',
node_color=(0.2, 0.2, 0.2),
node_scalar=network_ix,
tube_radius=.15)
lut = tubes.module_manager.scalar_lut_manager.lut.table.to_array()
lut = 255 * plt.cm.hot_r(np.linspace(0, 1, 256))
tubes.module_manager.scalar_lut_manager.lut.table = lut
tubes.update_pipeline()
#nodes.module_manager.scalar_lut_manager.lut.table = color
nodes.update_pipeline()
viz3d.plot_anat_3d(outline_color=(0, 0, 0), gyri_opacity=0.15)
fig.scene.disable_render = False
return fig
def plot_connectivity_graph(Theta=None, atlas_name="HarvardOxford",
fig_name=None, partial=None, retain=.1,
cluster_labels=None, plot_networks=False):
""" plot the connectivity graph inside a glass brain
"""
if Theta is None:
Theta = np.identity(96, dtype=np.float)
Theta[10, 0] = .9
Theta[86, 3] = -.7
Theta_ = Theta.copy()
p = Theta_.shape[0]
# 3D positions of regions + labels
regions = get_regions(atlas_name)
X, Y, Z, roi_labels = zip(
*[(int(region["x"]),
int(region["y"]),
int(region["z"]),
region["label"])
for region in regions])
(x, y, z) = map(np.array, (X, Y, Z))
# 3D glass image of brain
fig = mlab.figure(bgcolor=(1, 1, 1), size=(900, 769))
mlab.clf()
fig.scene.disable_render = True
# 2mm cortical map of Harvard_Oxford --> voxel to MNI coordinates (mm)
if atlas_name in {"HarvardOxford", "HarvardOxfordExt"}:
affine = np.identity(4, dtype=np.float)
affine[0, 0] = -2.
affine[1:3, 1:3] = 2. * affine[1:3, 1:3]
affine[:3, 3] = np.array([90., -126., -72.])
affine /= 1.1
elif atlas_name == "Juelich":
affine = np.identity(4, dtype=np.float)
affine[0, 0] = -2.
affine[1:3, 1:3] = 2. * affine[1:3, 1:3]
affine[:3, 3] = np.array([90., -126., -72.])
affine /= 1.1
(x, y, z) = maps_3d.coord_transform(x, y, z, affine)
partial_var = 1. / Theta_.flat[::p + 1]
if partial:
Theta_ = -Theta_.dot(np.diag(partial_var))
pctl = (1 - retain) * 100
thr = stats.scoreatpercentile(
np.abs(Theta_[np.triu_indices(Theta_.shape[0], k=1)]), pctl)
Theta_[np.abs(Theta_) < thr] = 0
#rois = mlab.points3d(x, y, z, partial_var, scale_factor=10, figure=fig)
# construct edges as a list of tuples [(p1,p2),(p3,p4),...] which means
# that an edge exists between vertices p1 and p2, p3 and p4, and so on
#Theta_ = np.tril(Theta, -1)
#edges_n1, edges_n2 = np.nonzero(Theta_)
if cluster_labels is not None:
n_clusters = np.unique(cluster_labels).size
n_members = [cluster_labels[cluster_labels == c].size
for c in np.arange(n_clusters - 1)]
n_large_clusters = 5
large_cluster_ix = np.argsort(
n_members, kind="mergesort")[-1:-n_large_clusters - 1:-1]
n_networks = n_large_clusters + 1
network = [mlab2.find(large_cluster_ix == cluster_labels[ix])[0]
if cluster_labels[ix] in large_cluster_ix
else n_large_clusters for ix in np.arange(p)]
else:
network = None
vmin = np.min(Theta_[np.abs(Theta_) != 0])
vmax = np.max(np.abs(Theta_))
tubes, nodes = plot_graph(-Theta_, x, y, z,
node_size=.6,
edge_vmin=vmin,
edge_vmax=vmax,
node_colormap='spectral',
node_color=(0.2, 0.2, 0.2),
node_scalar=network,
tube_radius=.15)
#tubes.parent.parent.parent.filter.vary_radius = \
# 'vary_radius_by_absolute_scalar'
#tubes.module_manager.scalar_lut_manager.reverse_lut = True
# Make points of the lut transparent
lut = tubes.module_manager.scalar_lut_manager.lut.table.to_array()
lut = 255 * plt.cm.hot_r(np.linspace(0, 1, 256))
tubes.module_manager.scalar_lut_manager.lut.table = lut
tubes.update_pipeline()
nodes.update_pipeline()
viz3d.plot_anat_3d(outline_color=(0, 0, 0), gyri_opacity=0.15)
fig.scene.disable_render = False
if fig_name:
mlab_save_views(fig_name, fig)
return fig
if plot_networks:
ix_range = np.arange(p)
for nw_ix in np.arange(n_networks - 1):
nw_members = ix_range[cluster_labels == large_cluster_ix[nw_ix]]
fig2 = _plot_subnetwork_graph(
Theta[np.ix_(nw_members, nw_members)],
(x[nw_members], y[nw_members], z[nw_members]), nw_ix)
if fig_name:
mlab_save_views(fig_name + "_nw%i" % nw_ix, fig2)
return large_cluster_ix
