def testdistancemap():
c = np.asarray((25, 25, 25))
a = make_test_array(c)
x = krebs.distancemap(a)
print "range: %f %f" % (x.min(), x.max())
grad = krebs.field_gradient(x, spacing=1.)
for i, g in enumerate(grad):
print "grad range %i: %f %f" % (i, g.min(), g.max())
gnorm = np.sum(g * g for g in grad)
for i in [c[2]]: #xrange(x.shape[2]):
plt.subplot(221)
p = plt.imshow(x[:, :, i],
vmin=x.min(),
vmax=6,
interpolation='nearest')
plt.colorbar(p)
plt.subplot(222)
p = plt.imshow(a[:, :, i], interpolation='nearest')
plt.colorbar(p)
plt.subplot(223)
p = plt.imshow(grad[0][:, :, i], interpolation='nearest')
plt.colorbar(p)
plt.subplot(224)
p = plt.imshow(gnorm[:, :, i], interpolation='nearest')
plt.colorbar(p)
plt.show()
def max_grad(self): if self.max_grad_ is not None: return self.max_grad_ import krebsutils dmin, dmax = self.value_bounds grad = krebsutils.field_gradient(self.data, spacing = 1.) g = np.sqrt(np.sum(tuple(g*g for g in grad), axis=0)) g = g.max() / (dmax-dmin + 1.e-24) # divide by data range because data range is mapped to 0, 1 g *= 1./self.spacing self.max_grad_ = g return g
def write(gmeasure, groupname):
graph = dataman.obtain_data(
'vessel_graph', vesselgroup,
['position', 'radius', 'flags', 'pressure'])
graph = graph.get_filtered(
myutils.bbitwise_and(graph['flags'],
krebsutils.CIRCULATED))
edgevalues = graph['pressure']
edgevalues = edgevalues[graph.edgelist]
# main calculation
thefield = krebsutils.CalcIntervascularInterpolationField(
graph.edgelist, graph['radius'], graph['position'],
edgevalues, fieldLdFine, 1.)
del edgevalues
# gradient of the interpolated blood pressure
thegrad = scipy.ndimage.gaussian_filter(thefield,
1.0,
0,
mode='nearest')
gradfield_ = krebsutils.field_gradient(thegrad)
thegrad = np.sum([np.square(g) for g in gradfield_], axis=0)
thegrad = np.sqrt(thegrad)
del gradfield_, thefield
# now we scale down the highres version, TODO: make it work with other than 3 dimensions
# first local average
m = self.fine_bin_subdivision
kernel = np.ones((m, m, m), dtype=np.float32)
thegrad = scipy.signal.fftconvolve(thegrad,
kernel,
mode='valid')
# then pick every m'th which contains the average of m finer boxes combined
thegrad = np.ascontiguousarray(thegrad[::m, ::m, ::m])
assert all(thegrad.shape == fieldLd.shape)
gmeasure.create_dataset(groupname,
data=thegrad,
compression=9,
dtype=np.float32)
graph = graph.get_filtered(
myutils.bbitwise_and(graph['flags'], krebsutils.CIRCULATED))
fieldld = krebsutils.SetupFieldLattice(wbbox, 3, 50., 0.)
print fieldld
edgevalues = graph['pressure']
edgevalues = edgevalues[graph.edgelist]
thefield = krebsutils.CalcIntervascularInterpolationField(
graph.edgelist, graph['radius'], graph['position'], edgevalues,
fieldld, 1.)
volfraction = krebsutils.make_vessel_volume_fraction_field(
graph['position'], graph.edgelist, graph['radius'], fieldld, 5)
thegrad = ndimage.gaussian_filter(thefield, 1.0, 0, mode='nearest')
gradfield_ = krebsutils.field_gradient(thegrad)
thegrad = np.sum([np.square(g) for g in gradfield_], axis=0)
thegrad = np.sqrt(thegrad)
del gradfield_
if SHOW_BLURRED_IMAGES:
kernel = (1.0 / 64.0) * np.ones((4, 4, 4), dtype=np.float32)
thefield = sig.fftconvolve(thefield, kernel)
volfraction = sig.fftconvolve(volfraction, kernel)
thegrad = sig.fftconvolve(thegrad, kernel)
z = fieldld.shape[2] / 2
fig, axes = pyplot.subplots(2, 2)
ax = axes[0, 0]
plt = ax.imshow(thefield[:, :, z])
def generate_data(statedata, dstgroup):
"""
main routine that does all the measurement.
Enable Individual parts. Computed date overwrites old data.
"""
print statedata.file.filename, statedata.name
time = statedata.attrs['time']
tum_grp = statedata['tumor']
tum_ld = get_tumld(tum_grp)
cellvol = tum_ld.scale**3
distmap = calc_distmap(tum_grp)
radialmap = krebsutils.make_radial_field(tum_ld)
axial_max_rad = np.max(np.abs(tum_ld.worldBox.ravel()))
ptc = np.asarray(tum_grp['ptc'])
if 1:
## shape data: rim surface area, volume
vtkds, = extractVtkFields.Extractor(statedata['tumor'],
['ls']).asVtkDataSets()
area = integrate_surface(vtkds)
del vtkds
vol = np.sum(ptc) * cellvol
radius = np.average(radialmap[np.nonzero(
np.logical_and(distmap > -2 * tum_ld.scale,
distmap < 2 * tum_ld.scale))])
sphere_equiv_radius = math.pow(vol * 3. / (4. * math.pi), 1. / 3.)
sphere_equiv_area = 4. * math.pi * (sphere_equiv_radius**2)
sphericity = sphere_equiv_area / area
#cylinder vol = pi r^2 h
cylinder_equiv_radius = math.sqrt(vol / tum_ld.GetWorldSize()[2] /
math.pi)
cylinder_equiv_area = 2. * math.pi * cylinder_equiv_radius * tum_ld.GetWorldSize(
)[2]
cylindericity = cylinder_equiv_area / area
d = dict(time=time,
area=area,
volume=vol,
radius=radius,
sphericity=sphericity,
cylindericity=cylindericity,
sphere_equiv_radius=sphere_equiv_radius,
sphere_equiv_area=sphere_equiv_area,
cylinder_equiv_radius=cylinder_equiv_radius,
cylinder_equiv_area=cylinder_equiv_area)
pprint(d)
print 'geometry data for %s/%s' % (statedata.file.filename,
statedata.name)
g = dstgroup.recreate_group('geometry')
myutils.hdf_write_dict_hierarchy(g, '.', d)
if 1:
print 'regenerating radial for %s/%s' % (statedata.file.filename,
statedata.name)
dstdata = dstgroup.recreate_group('radial')
## radial data; vessels
vessels = krebsutils.read_vesselgraph(statedata['vessels'], [
'position', 'flags', 'shearforce', 'radius', 'flow', 'maturation'
])
vessels = vessels.get_filtered(
edge_indices=(vessels.edges['flags'] & krebsutils.CIRCULATED != 0))
sample_length = 50.
far = 1.e10
s = plotVessels.generate_samples(vessels, 'position', 'nodes',
sample_length)
dist = krebsutils.sample_field(s,
distmap,
tum_ld,
linear_interpolation=True,
extrapolation_value=far)
rad = krebsutils.sample_field(s,
radialmap,
tum_ld,
linear_interpolation=True,
extrapolation_value=far)
del s
dstdata.recreate_group('vs_r').create_dataset(
'bins', data=np.average((bins_rad[:-1], bins_rad[1:]), axis=0))
dstdata.recreate_group('vs_dr').create_dataset(
'bins', data=np.average((bins_dist[:-1], bins_dist[1:]), axis=0))
# dstdata['vs_r'] = dict(bins = np.average((bins_rad[:-1],bins_rad[1:]), axis=0))
# dstdata['vs_dr'] = dict(bins = np.average((bins_dist[:-1],bins_dist[1:]), axis=0))
# def add_histogram_data(dst, name, (s_avg, s_std, s_sqr)):
# dst[name] = dict(avg = s_avg, std = s_std, sqr = s_sqr)
for name in ['shearforce', 'radius', 'flow', 'maturation']:
s = plotVessels.generate_samples(vessels, name, 'edges',
sample_length)
myutils.MeanValueArray.fromHistogram1d(bins_rad, rad, s).write(
dstdata['vs_r'], name)
myutils.MeanValueArray.fromHistogram1d(bins_dist, dist, s).write(
dstdata['vs_dr'], name)
# d = myutils.scatter_histogram(rad, s, bins_rad, 1.)
# add_histogram_data(dstdata['vs_r'], name, d)
# d = myutils.scatter_histogram(dist, s, bins_dist, 1.)
# add_histogram_data(dstdata['vs_dr'], name, d)
def make_mvd(flavor, s, bins, smap, mask_bound):
a = myutils.MeanValueArray.fromHistogram1d(bins, s,
np.ones_like(s))
b = myutils.MeanValueArray.fromHistogram1d(
bins, smap.ravel(), np.ones_like(smap.ravel()))
a.cnt = b.cnt.copy()
a.sum *= sample_length / cellvol
a.sqr *= a.sum**2
a.write(dstdata[flavor], 'mvd')
# d = myutils.scatter_histogram(xdata, np.ones_like(xdata), bins=bins, xdata2 = np.ravel(xdata2))
# m = dstdata[flavor]['bins'] > mask_bound
# for x in d[:2]:
# x *= sample_length/cellvol
# x.mask |= m
# for x in d[2:]:
# x *= (sample_length/cellvol)**2
# x.mask |= m
# add_histogram_data(dstdata[flavor], 'mvd', d)
make_mvd('vs_r', rad, bins_rad, radialmap, axial_max_rad)
make_mvd('vs_dr', dist, bins_dist, distmap, 1000.)
# pyplot.errorbar(dstdata['vs_dr']['bins'], dstdata['vs_dr']['mvd_avg'], yerr=dstdata['vs_dr']['mvd_std'])
# pyplot.show()
del rad, dist
for name in ['conc', 'sources', 'ptc', 'press']:
a = np.ravel(np.asarray(tum_grp[name]))
myutils.MeanValueArray.fromHistogram1d(
bins_rad, np.ravel(radialmap), a).write(dstdata['vs_r'], name)
myutils.MeanValueArray.fromHistogram1d(bins_dist,
np.ravel(distmap), a).write(
dstdata['vs_dr'], name)
# d = myutils.scatter_histogram(np.ravel(distmap), np.ravel(np.asarray(tum_grp[name])), bins_dist, 1.)
# add_histogram_data(dstdata['vs_dr'], name, d)
# d = myutils.scatter_histogram(np.ravel(radialmap), np.ravel(np.asarray(tum_grp[name])), bins_rad, 1.)
# add_histogram_data(dstdata['vs_r'], name, d)
# velocities, projected radially outward
vel_field = tuple(
np.asarray(tum_grp['vel'][:, :, :, i]) for i in range(3))
vnorm = np.zeros(distmap.shape, dtype=np.float32)
dgrad = krebsutils.field_gradient(radialmap, spacing=tum_ld.scale)
for vv, g in zip(vel_field, dgrad):
vnorm += vv * g
del vel_field, dgrad
phi_tumor = ptc * np.asarray(tum_grp['conc'])
oxy = np.asarray(statedata['fieldOxy'])
gf = np.array(statedata['fieldGf'])
for name, field in [('phi_tumor', phi_tumor), ('vel', vnorm),
('oxy', oxy), ('gf', gf)]:
a = np.ravel(field)
myutils.MeanValueArray.fromHistogram1d(
bins_rad, np.ravel(radialmap), a).write(dstdata['vs_r'], name)
myutils.MeanValueArray.fromHistogram1d(bins_dist,
np.ravel(distmap), a).write(
dstdata['vs_dr'], name)