def generate_samples(graph, data, association, scale):
'''A convenence function to call the c++ sampling routine for some data
stored in my graph structure. Additionaly it converts nodal to edge
properties because the sampling needs data associated with edges, but it
can still interpolate between data for the endpoints for each vessel.
Returns an array of the sampled data.
data = the name of the data, or the data itself,
association = 'edges' or 'nodes' or 'avg_edges'
scale = length which each sample represents approximately. Larger value = less samples.
'''
DATA_LINEAR = krebsutils.VesselSamplingFlags.DATA_LINEAR
DATA_CONST = krebsutils.VesselSamplingFlags.DATA_CONST
DATA_PER_NODE = krebsutils.VesselSamplingFlags.DATA_PER_NODE
if isinstance(data, str): data = getattr(graph, association)[data]
if association == 'edges':
#data = krebsutils.edge_to_node_property(int(np.amax(graph.edgelist)+1), graph.edgelist, graph.edges[name], 'avg')
return krebsutils.sample_edges(graph.nodes['position'], graph.edgelist, data, scale, DATA_CONST)
else:
return krebsutils.sample_edges(graph.nodes['position'], graph.edgelist, data, scale, DATA_LINEAR | DATA_PER_NODE)
def generate_samples(graph, name, association, scale):
DATA_LINEAR = krebsutils.VesselSamplingFlags.DATA_LINEAR
DATA_CONST = krebsutils.VesselSamplingFlags.DATA_CONST
DATA_PER_NODE = krebsutils.VesselSamplingFlags.DATA_PER_NODE
if not len(graph.edgelist):
return np.asarray([], dtype=float)
if association == 'edges':
data = krebsutils.edge_to_node_property(
int(np.amax(graph.edgelist) + 1), graph.edgelist,
graph.edges[name], 'avg')
else:
data = graph.nodes[name]
# return data
return krebsutils.sample_edges(graph.nodes['position'], graph.edgelist,
data, scale, DATA_LINEAR | DATA_PER_NODE)
print ldvessels
''' this splits splits space in lattices of 300.
the second 300 adds a 'safety layer of 100. mum
so we do not consider the outermost data for calculating the
actual mvd
'''
fieldld = krebsutils.SetupFieldLattice(wbbox, 3, 20., 100.)
wbbox = fieldld.worldBox
print 'field ld:'
print fieldld
z = fieldld.shape[2] / 2
weights = krebsutils.sample_edges_weights(graph.nodes['position'],
graph.edgelist, 30.)
positions = krebsutils.sample_edges(
graph.nodes['position'], graph.edgelist, graph.nodes['position'], 30.,
krebsutils.VesselSamplingFlags.DATA_PER_NODE
| krebsutils.VesselSamplingFlags.DATA_LINEAR)
eps = 1.0 - 1.e-15
x0, x1, y0, y1, z0, z1 = wbbox
ranges = [
np.arange(x0, x1, fieldld.scale * eps),
np.arange(y0, y1, fieldld.scale * eps),
np.arange(z0, z1, fieldld.scale * eps),
]
print 'histogram bin ends:', map(lambda r: (r.min(), r.max()), ranges)
mvd, _ = np.histogramdd(positions, bins=ranges, weights=weights)
mvd *= 1.e6 / (fieldld.scale**3)
print 'result shape:', mvd.shape
print('average mvd')
print(np.mean(mvd[1:-1, 1:-1, 1:-1]))
def sample_vessel_system(goodArguments):
filename = goodArguments.vesselFileNames
grouppath = goodArguments.grp_pattern
with h5py.File(filename, 'r') as file:
if ('vessels' in grouppath):
print('found vessels!')
vesselgroup = file[grouppath]
else:
if ('out' in grouppath):
outgroup = file[grouppath]
print('found tumor of type: %s' %
str(outgroup['tumor'].attrs.get('TYPE')))
vesselgroup = file[grouppath + '/vessels']
else:
print("unknown data structure!")
ldvessels = krebsutils.read_lattice_data_from_hdf(
vesselgroup['lattice'])
wbbox = ldvessels.worldBox
graph = krebsutils.read_vesselgraph(vesselgroup, ['position', 'flags'])
graph = graph.get_filtered(
myutils.bbitwise_and(graph['flags'], krebsutils.CIRCULATED))
print 'vessel ld:'
print ldvessels
''' this splits splits space in lattices of 300.
the second 300 adds a 'safety layer of 100. mum
so we do not consider the outermost data for calculating the
actual mvd
'''
sampling_lattice_spacing = goodArguments.sampling_lattice_spacing
fieldld = krebsutils.SetupFieldLattice(wbbox, 3, sampling_lattice_spacing,
100.)
wbbox = fieldld.worldBox
print 'field ld:'
print fieldld
z = fieldld.shape[2] / 2
longitudinal_sampling_distance = goodArguments.longitudinal
weights = krebsutils.sample_edges_weights(graph.nodes['position'],
graph.edgelist,
longitudinal_sampling_distance)
positions = krebsutils.sample_edges(
graph.nodes['position'], graph.edgelist, graph.nodes['position'],
longitudinal_sampling_distance,
krebsutils.VesselSamplingFlags.DATA_PER_NODE
| krebsutils.VesselSamplingFlags.DATA_LINEAR)
eps = 1.0 - 1.e-15
x0, x1, y0, y1, z0, z1 = wbbox
ranges = [
np.arange(x0, x1, fieldld.scale * eps),
np.arange(y0, y1, fieldld.scale * eps),
np.arange(z0, z1, fieldld.scale * eps),
]
print 'histogram bin ends:', map(lambda r: (r.min(), r.max()), ranges)
mvd, _ = np.histogramdd(positions, bins=ranges, weights=weights)
mvd *= 1.e6 / (fieldld.scale**3)
print 'result shape:', mvd.shape
print('average mvd')
print(np.mean(mvd[1:-1, 1:-1, 1:-1]))
''' new stuff '''
from scipy import ndimage
bool_field = mvd > 0
bool_field = np.logical_not(bool_field)
distance_map = ndimage.morphology.distance_transform_edt(bool_field)
distance_map = distance_map * sampling_lattice_spacing
# fig, ax = pyplot.subplots(1)
# plt = ax.imshow(mvd[:,:,z], interpolation = 'none')
# ax.set(title = 'MVD')
# divider = mpl_utils.make_axes_locatable(ax)
# cax = divider.append_axes("right", size = "5%", pad = 0.05)
# fig.colorbar(plt, cax = cax)
fig, ax = pyplot.subplots(1)
plt = ax.imshow(distance_map[:, :, z], interpolation='none')
#ax.set(title = 'Distance Map \n group: %s, file: %s' %(grouppath, filename))
ax.set(title='Distance Map \n smp_logitudinal: %s, lattice_const: %s' %
(longitudinal_sampling_distance, sampling_lattice_spacing))
divider = mpl_utils.make_axes_locatable(ax)
cax = divider.append_axes("right", size="5%", pad=0.05)
fig.colorbar(plt, cax=cax)
basename(filename)
with PdfPages('distmap_' + basename(filename) + '_' + grouppath +
'.pdf') as pdf:
pdf.savefig(fig)