def DoIt_single(filenames, options):
fn_measure = basename(commonprefix(filenames))
fn_measure = myutils.strip_from_end(fn_measure, '.h5')
fn_measure = myutils.strip_from_end(fn_measure, '-type')
f_measure = h5files.open('adaption_common.h5', 'a', search = False)
files = [h5files.open(fn, 'r') for fn in filenames]
groups_without_adaption = [f['/adaption/recomputed'] for f in files]
groups_with_adaption = [f['/adaption/vessels_after_adaption'] for f in files]
import analyzeGeneral
dataman = myutils.DataManager(20, [ analyzeGeneral.DataBasicVessel(), analyzeGeneral.DataVesselSamples(), analyzeGeneral.DataVesselGlobal()])
with mpl_utils.PdfWriter(fn_measure+'flow_hist.pdf') as pdfpages:
if 1:
plotFlowHistogram(dataman, f_measure, filenames, options, pdfpages)
with mpl_utils.PdfWriter(fn_measure+'radius_hist.pdf') as pdfpages:
if 1:
plotRadiusHistogram(dataman, f_measure, filenames, options, pdfpages)
with mpl_utils.PdfWriter(fn_measure+'cap_flow_hist.pdf') as pdfpages:
if 1:
plotCapillaryFlowHistogram(dataman, f_measure, filenames, options, pdfpages)
with mpl_utils.PdfWriter(fn_measure+'cap_radius_hist.pdf') as pdfpages:
if 1:
plotCapillaryRadiusHistogram(dataman, f_measure, filenames, options, pdfpages)
def DoIt(filenames, options):
fn_measure = basename(commonprefix(filenames))
fn_measure = myutils.strip_from_end(fn_measure, '.h5')
fn_measure = myutils.strip_from_end(fn_measure, '-type')
f_measure = h5files.open('adaption_common.h5', 'a', search = False)
files = [h5files.open(fn, 'r') for fn in filenames]
groups_without_adaption = [f['/adaption/recomputed'] for f in files]
groups_with_adaption = [f['/adaption/vessels_after_adaption'] for f in files]
with mpl_utils.PdfWriter(fn_measure+'caps.pdf') as pdfpages:
import analyzeGeneral
dataman = myutils.DataManager(20, [ analyzeGeneral.DataBasicVessel(), analyzeGeneral.DataVesselSamples(), analyzeGeneral.DataVesselGlobal()])
# vesselgroups_without = groups_without_adaption
# vesselgroups_with = groups_with_adaption
# geometric_data_before = getGeometricData(groups_without_adaption)
# perfusion_data_before = getTotalPerfusion(groups_without_adaption)*60
# geometric_data_after = getGeometricData(groups_with_adaption)
# perfusion_data_after = getTotalPerfusion(groups_with_adaption)*60
if 1:
plotFlowHistogram(dataman, f_measure, filenames, options, pdfpages)
if 1:
plotRadiusHistogram(dataman, f_measure, filenames, options, pdfpages)
if 1:
plotCapillaryFlowHistogram(dataman, f_measure, filenames, options, pdfpages)
if 1:
plotCapillaryRadiusHistogram(dataman, f_measure, filenames, options, pdfpages)
def DoIt(inputFileNames, pattern, options):
for this_enlargement_factor in options.enlarge_factor:
inFiles = [h5files.open(fn, 'r') for fn in inputFileNames]
inGroups = list(
itertools.chain.from_iterable(
myutils.walkh5(f, pattern, return_h5objects=True)
for f in inFiles))
if len(inGroups) <= 0:
print 'no matching groups in hdf file(s)'
sys.exit(0)
for in_file in inputFileNames:
bloodflowparams = krebsutils.pyDictFromParamGroup(
h5files.open(in_file, 'r')['parameters/calcflow'])
#enlarge_vessels(float(this_enlargement_factor),in_file, bloodflowparams)
qsub.submit(
qsub.func(enlarge_vessels, float(this_enlargement_factor),
in_file, bloodflowparams),
name='job_modify_enlarge_' + str(this_enlargement_factor) +
'_',
num_cpus=6,
days=
5, # about one day per thread, the idea being that number of threads is proportional to systems size and runtime is
mem='%iMB' % (1000),
change_cwd=True)
def plotFlowHistogram(dataman, f_measure, filenames, options, pdfpages):
filenames = adaption.get_files_with_successful_adaption(filenames)
files = [h5files.open(fn, 'r') for fn in filenames]
destination_group = f_measure.require_group('allFlow')
if(options.two):
typelist = 'typeD- typeE- typeG- typeH-'
else:
typelist = 'typeA- typeB- typeC- typeF- typeI-'
if(options.all_types):
typelist = 'typeA- typeB- typeC- typeD- typeE- typeF- typeG- typeH- typeI-'
if(options.single):
typelist = 'typeF- '
fig = plt.figure()
#fig.suptitle('all Flows', fontsize=12)
#fig.set_title('all Flows')
ax1 = fig.add_subplot(211)
ax2 = fig.add_subplot(212)
ax1.set_title('M- network')
ax1.set_xlabel(r'flow/$\frac{\mu m^3}{s}$')
ax1.set_ylabel(r'$\rho$')
ax1.text(-0.1, 1.15, 'A', transform=ax1.transAxes,fontsize=16, fontweight='bold', va='top', ha='right')
ax2.set_title('Adapted- network')
ax2.set_xlabel(r'flow/$\frac{\mu m^3}{s}$')
ax2.set_ylabel(r'$\rho$')
ax2.text(-0.1, 1.15, 'B', transform=ax2.transAxes,fontsize=16, fontweight='bold', va='top', ha='right')
for (i,t) in enumerate(typelist.split()):
print('Capillary flow for type: %s' % t)
filteredFiles = filter( lambda fn: t in fn,filenames)
files = [h5files.open(fn, 'r+') for fn in filteredFiles]
if(len(files)==0):#that means no file of dedicated type is left after filter
continue
#h_nA, h_yA, x_edges_nA, x_edges_yA = generate_capillary_hist(dataman, files, destination_group, t[:-1])
#data = generate_murray_data_plot3(dataman, files, destination_group, t[:-1])
else:
h_nA,h_yA,xedges_nA,xedges_yA = generate_flow_hist(dataman, files, destination_group, t[:-1] + '_allFlow')
xedges_nA = np.asarray(xedges_nA)
xedges_yA = np.asarray(xedges_yA)
h1 =np.asarray(h_nA)
h2 =np.asarray(h_yA)
ax1.plot((xedges_nA[1:]+xedges_nA[:-1])/2, h1, label = type_to_label_dict[t], marker= vesselTypeMarkers[i], color=colors[i])
ax2.plot((xedges_yA[1:]+xedges_yA[:-1])/2, h2, label = type_to_label_dict[t], marker= vesselTypeMarkers[i], color=colors[i])
fontP = FontProperties()
fontP.set_size('x-small')
ax1.legend(prop=fontP)
if double_log:
ax2.loglog()
ax1.loglog()
else:
ax2.semilogx()
ax1.semilogx()
plt.tight_layout()
pdfpages.savefig(fig)
def render(self):
""" run povray. This should be called on the computing node. """
with h5files.open(self.fn, 'a') as f:
if 'po2vessels' in f[self.group_name]:
from krebs.povrayRenderOxygenDetailed import renderScene
renderScene(f[self.group_name], self.imageFilename,
self.params)
elif 'tumor' in f[self.group_name]:
from krebs.povrayRenderTumor import renderScene
self.params.timepoint = f[self.group_name].attrs.get('time')
renderScene(f[self.group_name]['vessels'],
f[self.group_name]['tumor'], self.imageFilename,
self.params)
# else:
# from krebs.povrayRenderVessels import renderScene
# renderScene(f[self.group_name],
# self.imageFilename,
# **self.params)
#renderScene(drug_grp, imagefn, parameters)
elif 'conc' in f[self.group_name]:
from krebs.povrayRenderIff import renderScene
renderScene(f[self.group_name], self.imageFilename,
self.params)
else:
from krebs.povrayRenderVessels import render_different_data_types
render_different_data_types(f[self.group_name], self.params)
def run(goodArguments):
print('starting with arguments: %s' % goodArguments)
no_files = len(goodArguments.oxygenFiles)
hypoxicVolumes = []
tumorVolumes = []
threshold = 15
test = myutils.MeanValueArray.empty()
for aFile in goodArguments.oxygenFiles:
with h5files.open(aFile.name) as f:
try:
if not 'po2' in f:
raise AssertionError('no proper oxygen file: %s!' % f)
except Exception, e:
print e.message
sys.exit(-1)
paths = myutils.walkh5(f, 'po2/out*')
print('found paths: %s' % paths)
hypoxicVolumes_per_time = []
tumorVolumes_per_time = []
timepoints = []
for path in paths:
hypoxicFraction, hypoxicTissueVolume = estimate_ratio_hypoxic(
f[path], threshold)
hypoxicVolumes_per_time.append(hypoxicTissueVolume)
t = f[path]['SOURCE'].attrs['time']
r = f[path]['SOURCE/tumor'].attrs['TUMOR_RADIUS']
volume = 4 / 3. * 3.1414 * r * r * r / 1e9
tumorVolumes_per_time.append(volume)
timepoints.append(t)
hypoxicVolumes.append(hypoxicVolumes_per_time)
tumorVolumes.append(tumorVolumes_per_time)
def __init__(self, dataman, filenames, pattern):
files = [h5files.open(fn, 'r+') for fn in filenames]
items = []
has_tumor = True
for f in files:
paths = myutils.walkh5(f['.'], pattern)
for path in paths:
g = f[path]
if g.attrs.get('CLASS', None) == 'GRAPH':
gvessels = g
gtumor = None
try:
source = h5files.openLink(g, 'SOURCE')
gtumor = source.parent['tumor']
g = source.parent
except Exception, e:
raise RuntimeError(
'tried to get tumor data but failed:' + str(e))
else:
gvessels, gtumor = g['vessels'], (g['tumor'] if 'tumor'
in g else None)
e = EnsembleItem(path=path,
gvessels=gvessels,
gtumor=gtumor,
group=g)
e.time = g.attrs['time']
has_tumor = has_tumor and gtumor is not None
e.vessel_system_length = dataman.obtain_data(
'vessel_system_length', gvessels)
items.append(e)
def DoReadIn(filenames, pattern, fn_measure,pdf,quantity):
#read in lots of stuff
files = [h5files.open(fn, 'r') for fn in filenames]
output_f = h5py.File(fn_measure)
if quantity == 'flow':
iflog=True
else:
iflog=False
if quantity == 'shearforce':
iflog=True
for t in 'typeA typeB typeC typeD typeE typeF typeG typeH typeI'.split():
#edges
fig1 = plt.figure()
plt.plot()
ax1 = plt.axes()
if 'all_flows_of_type_dist' in locals():
del all_flows_of_type_dist
for afile in files:
print(t)
if t in afile.filename:
print("where in")
if iflog:
flows_in_a_file = np.asarray(afile[pattern + '/edges/'+quantity])
flows_in_a_file = flows_in_a_file[np.nonzero(flows_in_a_file)]
flows_in_a_file = np.log10(flows_in_a_file)
if quantity == 'flow':
flows_in_a_file = np.asarray(afile[pattern + '/edges/'+quantity])
flows_in_a_file = flows_in_a_file[np.nonzero(flows_in_a_file)]
fac=60/100000.
flows_in_a_file = np.multiply(fac,flows_in_a_file)
flows_in_a_file = np.log10(flows_in_a_file)
else:
flows_in_a_file = np.asarray(afile[pattern + '/edges/'+quantity])
flow_dist, bin_edges = np.histogram(flows_in_a_file, bins=50, density=True)
if 'all_flows_of_type_dist' in locals():
all_flows_of_type_dist = np.vstack((all_flows_of_type_dist,flow_dist))
else:
all_flows_of_type_dist = flow_dist
if 'all_flows_of_type_dist' in locals():
fig1.suptitle(t)
ax1.grid()
if quantity == 'flow':
ax1.set_xlabel(r'\log_{10}(flow)/nl/min')
elif quantity == 'radius':
ax1.set_xlabel(r'radius/\mu m')
ax1.set_ylabel('probability')
width = 1*(bin_edges[1]-bin_edges[0])
centers = (bin_edges[:-1]+bin_edges[1:])/2
if len(all_flows_of_type_dist.shape)==1:
ax1.plot(centers,all_flows_of_type_dist,'*')
else:
ys = np.mean(all_flows_of_type_dist,0)
yerr = np.std(all_flows_of_type_dist,0)
ax1.errorbar(centers,ys,yerr=yerr)
pdf.savefig()
def getDomainSizeFromVesselFile(fn):
with h5files.open(fn, 'r') as f:
ld = krebsutils.read_lattice_data_from_hdf(
krebsutils.find_lattice_group_(f['vessels']))
size = np.amax(ld.GetWorldSize())
# longest axis times the lattice spacing
return size
def create_mean_plot_velocities():
my_res_file = h5files.open('analyzeVesselTumor.h5', 'r')
count_hists = 0
the_keys = my_res_file['/global'].keys()
print(the_keys[0])
no_bins = 55
the_density_in = np.zeros(no_bins - 1)
the_density_out = np.zeros(no_bins - 1)
for key in my_res_file['/global'].keys():
print('working with %s' % key)
count_hists = count_hists + 1
count_in = my_res_file['/global/%s/flows_within_tumor/flows_inside' %
key]
count_out = my_res_file['/global/%s/flows_within_tumor/flows_outside' %
key]
radii_in = np.asarray(
my_res_file['/global/%s/radii_within_tumor/radii_inside' % key])
radii_out = np.asarray(
my_res_file['/global/%s/radii_within_tumor/radii_outside' % key])
count_in = np.asarray(count_in) / (np.pi * np.square(radii_in))
count_out = np.asarray(count_out) / (np.pi * np.square(radii_out))
density_in, bin_in_array = np.histogram(count_in,
np.logspace(1, 4, no_bins),
density=True)
density_out, bin_out_array = np.histogram(count_out,
np.logspace(1, 4, no_bins),
density=True)
the_density_in = the_density_in + density_in
the_density_out = the_density_out + density_out
print('we have %i histograms' % count_hists)
the_density_in = np.divide(the_density_in, count_hists)
the_density_out = np.divide(the_density_out, count_hists)
mu1 = np.sum(the_density_in * np.diff(bin_in_array) *
(0.5 * (bin_in_array[1:] + bin_in_array[:-1])))
quad1 = np.sum(the_density_in * np.diff(bin_in_array) * np.square(
(0.5 * (bin_in_array[1:] + bin_in_array[:-1]))))
std1 = np.sqrt(quad1 - mu1**2)
mu2 = np.sum(the_density_out * np.diff(bin_out_array) *
(0.5 * (bin_out_array[1:] + bin_out_array[:-1])))
quad2 = np.sum(the_density_out * np.diff(bin_out_array) * np.square(
(0.5 * (bin_out_array[1:] + bin_out_array[:-1]))))
std2 = np.sqrt(quad2 - mu2**2)
plt.title('Flow distribution')
ax = plt.subplot(211)
ax.bar(bin_out_array[:-1], the_density_out, np.diff(bin_out_array))
ax.set_xscale('log')
ax.text(10**3, 0.002, '$< v_{out} > = %f \pm %f$' % (mu2, std2))
plt.legend(['Outside'])
plt.ylabel('Propability')
ax = plt.subplot(212)
ax.bar(bin_in_array[:-1], the_density_in, np.diff(bin_in_array))
ax.set_xscale('log')
ax.text(10**2, 0.006, '$< v_{in} > = %f \pm %f$' % (mu1, std1))
plt.xlabel('Volumeflow through vessel segments/ $\mu m$')
plt.ylabel('Propability')
plt.legend(['Inside'])
plt.savefig('Flow_distribution.png')
plt.show()
def find_enlargement_factors(filenames):
factors = []
files = [h5files.open(fn, 'r+') for fn in filenames]
for f in files:
if 'enlargeFactor' in f.attrs.keys():
if not f.attrs.get('enlargeFactor') in factors:
factors.append(f.attrs.get('enlargeFactor'))
return factors
def from_vessel_file(filenames, grp_pattern):
dirs = set()
dataman = myutils.DataManager(20, [
krebs.plotIff.DataTissue(),
krebs.plotIff.DataGlobalIff(),
krebs.plotIff.DataRadialIff(),
krebs.analyzeGeneral.DataDistanceFromCenter(),
krebs.analyzeGeneral.DataVesselSamples(),
krebs.analyzeGeneral.DataBasicVessel(),
o2analysis.DataDetailedPO2()
])
f_measure = h5files.open('chache.h5', 'a', search=False)
def cachelocation(g):
path = posixpath.join(
'FileCS_' + myutils.checksum(basename(g.file.filename)),
g.name.strip(posixpath.sep))
return (f_measure, path)
#run with grp_pattern: iff/vessels
for fn in filenames:
with h5py.File(fn, 'r+') as f:
d = myutils.walkh5(f, grp_pattern)
assert len(d), 'you f****d up, pattern "%s" not found in "%s"!' % (
grp_pattern, fn)
dirs = set.union(dirs, d)
for group_path in dirs:
if 'vessel' in grp_pattern and not 'o2' in grp_pattern:
vesselgroup = f[group_path]
ldvessels = ku.read_lattice_data_from_hdf(
vesselgroup['lattice'])
fieldld = ku.SetupFieldLattice(ldvessels.worldBox, 3, 10,
0.)
phi_vessels = krebs.analyzeGeneral.CalcPhiVessels(
dataman,
vesselgroup,
fieldld,
scaling=1.,
samples_per_cell=5)
print('bla')
import nibabel as nib
new_image = nib.Nifti1Image(phi_vessels, affine=np.eye(4))
common_filename = os.path.splitext(os.path.basename(fn))[0]
new_image.to_filename(common_filename + '_vessels' +
'.nii')
if 'o2' in grp_pattern:
po2group = f[group_path]
#sample_length = 500.
#data = dataman.obtain_data('detailedPO2_global', 'po2_tissue', po2group, sample_length, cachelocation(po2group))
data = np.asarray(po2group['po2field'])
print('bla')
import nibabel as nib
new_image = nib.Nifti1Image(data, affine=np.eye(4))
common_filename = os.path.splitext(os.path.basename(fn))[0]
new_image.to_filename(common_filename + '_po2' + '.nii')
def __init__(self, dataman, filenames, pattern):
files = [h5files.open(fn, 'r+') for fn in filenames]
items = []
has_tumor = True
for f in files:
paths = myutils.walkh5(f['.'], pattern)
for path in paths:
po2group_w_a = f[path + '/vessels_after_adaption']
gvessels_w_a, gtumor = detailedo2.OpenVesselAndTumorGroups(
po2group_w_a)
po2group_no_a = f[path + '/recomputed']
gvessels_no_a, gtumor = detailedo2.OpenVesselAndTumorGroups(
po2group_no_a)
e = EnsembleItem(path=path,
po2group_w_a=po2group_w_a,
gvessels_w_a=gvessels_w_a,
po2group_no_a=po2group_no_a,
gvessels_no_a=gvessels_no_a,
gtumor=gtumor)
# if 'SOURCE' in po2group:
# source = h5files.openLink(po2group, 'SOURCE')
# if 'time' in source.attrs.keys():
# t = source.attrs['time']
# e.time = t
has_tumor = has_tumor and gtumor is not None
e.vessel_system_length = dataman.obtain_data(
'vessel_system_length', gvessels_no_a)
e.initialVesselType = GetVesselTypeLabel(po2group_no_a)
items.append(e)
if has_tumor:
d = collections.defaultdict(list) # path -> list of EnsembleItem
for e in items:
d[e.path].append(e)
tumor_snapshot_times = dict(
(k, np.average(map(lambda e: e.time, v)))
for k, v in d.items())
tumor_snapshot_order = sorted(
tumor_snapshot_times.keys(),
key=(lambda path: tumor_snapshot_times[path]))
tumor_snapshots = [(d[path], path, tumor_snapshot_times[path])
for path in tumor_snapshot_order]
self.files = files
self.items = items
if has_tumor:
self.tumor_snapshots = tumor_snapshots # list of tuple(items, path, time)
self.has_tumor = has_tumor
self.o2ConfigName = set(
item.po2group_no_a.attrs.get('O2_CONFIG_NAME', None)
for item in items)
if len(self.o2ConfigName) != 1:
raise RuntimeError(
"Detected different O2_CONFIG_NAMES %s. You don't want to mix configurations, do you?"
% self.o2ConfigName)
self.o2ConfigName = self.o2ConfigName.pop()
def enlarge_vessels(factor, gv_filename, bloodflowparams):
'''gvdst = group where the data is placed in, does not create a 'vesse' folder in it but writes nodes, edges directly;
gv = source vessel group
'''
inFile = h5files.open(gv_filename, 'r')
gv = inFile['vessels']
fac_as_percent = int(np.ceil(float(factor) * 100 - 100))
dest_file = h5files.open(
myutils.strip_from_end(gv_filename, '.h5') +
'_growth_by_%02i.h5' % fac_as_percent, 'a')
dest_file.attrs.create('enlargeFactor', data=fac_as_percent)
gvdst = dest_file.create_group('vessels')
gvdst.attrs['CLASS'] = 'GRAPH'
myutils.buildLink(gvdst, 'SOURCE', gv)
# first we need to copy some of the vessel data
gvedst = gvdst.create_group('edges')
gvndst = gvdst.create_group('nodes')
gvedst.attrs['COUNT'] = gv['edges'].attrs['COUNT']
gvndst.attrs['COUNT'] = gv['nodes'].attrs['COUNT']
gv.copy('lattice', gvdst)
for name in [
'lattice_pos', 'roots', 'nodeflags', 'gf', 'bc_conductivity_value',
'bc_node_index', 'bc_type', 'bc_value'
]:
gv['nodes'].copy(name, gvndst)
for name in ['radius', 'node_a_index', 'node_b_index', 'flags']:
if name == 'radius':
radii = gv['edges/' + name]
radii = factor * np.asarray(radii)
gvedst.create_dataset(name, data=radii)
else:
gv['edges'].copy(name, gvedst)
# then we recompute blood flow because the alorithm has changed and we may or may not want hematocrit
(pressure, flow, shearforce, hematocrit,
flags) = krebsutils.calc_vessel_hydrodynamics(
gvdst, return_flags=True, bloodflowparams=bloodflowparams)
# then we save the new data to complete the network copy
gvedst.create_dataset('flow', data=flow, compression=9)
gvedst.create_dataset('shearforce', data=shearforce, compression=9)
gvedst.create_dataset('hematocrit', data=hematocrit, compression=9)
gvndst.create_dataset('pressure', data=pressure, compression=9)
def doit(filenames):
dataman = myutils.DataManager(
50,
map(lambda x: x(), detailedo2Analysis.O2DataHandlers) + [
analyzeGeneral.DataTumorTissueSingle(),
analyzeGeneral.DataDistanceFromCenter(),
analyzeGeneral.DataBasicVessel(),
analyzeGeneral.DataVesselSamples(),
analyzeGeneral.DataVesselRadial(),
analyzeGeneral.DataVesselGlobal(),
analyzeBloodFlow.DataTumorBloodFlow()
])
ensemble = EnsembleFiles(dataman, filenames, 'po2/adaption/')
out_prefix, out_suffix = myutils.splitcommonpresuffix(
map(lambda s: basename(s), filenames))
output_base_filename = splitext(out_prefix + out_suffix)[0]
if ensemble.o2ConfigName:
fn_measure = 'detailedo2_%s_common.h5' % ensemble.o2ConfigName
else:
fn_measure = 'detailedo2_common.h5'
f_measure = h5files.open(fn_measure, 'a')
def cachelocation(g):
path = posixpath.join(
'FileCS_' + myutils.checksum(basename(g.file.filename)),
g.name.strip(posixpath.sep))
return (f_measure, path)
measurementinfo = MeasurementInfo(sample_length=30.,
cachelocation_callback=cachelocation,
distancemap_spec='radial')
with mpl_utils.PageWriter(output_base_filename + '.pdf',
fileformats=['pdf']) as pdfwriter:
if 0:
compare_tissue_saturation(dataman, ensemble, pdfwriter)
if 0:
#try:
# histogramGroupFinal = f_measure['combinedHistogramsFinal']
# histogramGroupInitial = f_measure['combinedHistogramsInitial']
#except KeyError:
#histogramGroupFinal = f_measure.recreate_group('combinedHistogramsFinal')
histogramGroupInitial = f_measure.recreate_group(
'combinedHistogramsInitial')
#ComputeHistogramsOfPo2Items(dataman, ensemble.items, measurementinfo, histogramGroupFinal)
ComputeHistogramsOfPo2Items(dataman, ensemble.items,
measurementinfo, histogramGroupInitial)
#PlotHistograms(pdfwriter, histogramGroupFinal, 'tum', 'Tumor')
PlotHistograms(pdfwriter, histogramGroupInitial, 'all', 'Initial')
def create_mean_plot_radii():
my_res_file = h5files.open('analyzeVesselTumor.h5', 'r')
count_hists = 0
the_keys = my_res_file['/global'].keys()
print(the_keys[0])
no_bins = 55
the_density_in = np.zeros(no_bins)
the_density_out = np.zeros(no_bins)
for key in my_res_file['/global'].keys():
print('working with %s' % key)
count_hists = count_hists + 1
count_in = my_res_file['/global/%s/radii_within_tumor/radii_inside' %
key]
count_out = my_res_file['/global/%s/radii_within_tumor/radii_outside' %
key]
density_in, bin_in_array = np.histogram(count_in,
no_bins,
density=True)
density_out, bin_out_array = np.histogram(count_out,
no_bins,
density=True)
the_density_in = the_density_in + density_in
the_density_out = the_density_out + density_out
print('we have %i histograms' % count_hists)
the_density_in = np.divide(the_density_in, count_hists)
the_density_out = np.divide(the_density_out, count_hists)
mu1 = np.sum(the_density_in * np.diff(bin_in_array) *
(0.5 * (bin_in_array[1:] + bin_in_array[:-1])))
quad1 = np.sum(the_density_in * np.diff(bin_in_array) * np.square(
(0.5 * (bin_in_array[1:] + bin_in_array[:-1]))))
std1 = np.sqrt(quad1 - mu1**2)
mu2 = np.sum(the_density_out * np.diff(bin_out_array) *
(0.5 * (bin_out_array[1:] + bin_out_array[:-1])))
quad2 = np.sum(the_density_out * np.diff(bin_out_array) * np.square(
(0.5 * (bin_out_array[1:] + bin_out_array[:-1]))))
std2 = np.sqrt(quad2 - mu2**2)
plt.title('Radii distribution')
ax = plt.subplot(211)
ax.bar(bin_out_array[:-1], the_density_out, np.diff(bin_out_array))
ax.text(30, 0.2, '$< r_{out} > = %f \pm %f$' % (mu2, std2))
plt.legend(['Outside'])
plt.ylabel('Propability')
ax = plt.subplot(212)
plt.bar(bin_in_array[:-1], the_density_in, np.diff(bin_in_array))
ax.text(20, 0.3, '$< r_{in} > = %f \pm %f$' % (mu1, std1))
plt.xlabel('radius of vessel segments/ $\mu m$')
plt.ylabel('Propability')
plt.legend(['Inside'])
plt.savefig('Radii_distribution.png')
plt.show()
def worker_on_client(fn, pattern, o2params):
print 'detailedo2 on %s / %s' % (fn, pattern)
h5files.search_paths = [
dirname(fn)
] # so the plotting and measurement scripts can find the original tumor files using the stored basename alone
#no, this is not working!!!!, hand num_threads over to the cpp side
#num_threads = o2params.pop('num_threads')
#krebsutils.set_num_threads(num_threads)
o2_refs = detailedo2.doit(fn, pattern, (o2params, o2params['name']))
if 0: #this is for data analysis on the clusters
for ref in o2_refs:
po2group = h5files.open(ref.fn)[ref.path]
detailedo2Analysis.WriteSamplesToDisk(po2group)
h5files.closeall() # just to be sure
def MakeVesselFilenamePart(fn):
with h5files.open(fn, mode='a') as f:
if 'parameters' in f:
if 'MESSAGE' in f['parameters'].attrs:
msg = f['parameters'].attrs['MESSAGE']
ensemble_index = f['parameters'].attrs['ENSEMBLE_INDEX']
if msg.startswith('vessels-'): msg=msg[len('vessels-'):]
if 'msg' not in locals():
msg = "hm"
ensemble_index = 1
f['parameters'].attrs['MESSAGE'] = msg
f['parameters'].attrs['ENSEMBLE_INDEX'] = ensemble_index
name = '%s-sample%02i' % (msg, ensemble_index)
return name
def create_nice_file_containing_all_the_data():
filenames = sys.argv[1:]
files = [h5files.open(fn, 'r') for fn in filenames]
fmeasure = h5files.open('analyzeVesselTumor.h5', 'a')
dataman = myutils.DataManager(10, [DataVesselLength(), DataVesselTumor()])
allgroups = defaultdict(list)
for f in files:
keys = filter(lambda k: k.startswith('out'), f.keys())
for k in keys:
allgroups[k].append(f[k])
allgroups = [(k, allgroups[k]) for k in sorted(allgroups.keys())]
groupslist = [allgroups[-1]]
outfn = 'Length-%s.pdf' % splitext(basename(filenames[0]))[0]
pprint(groupslist)
print '-> %s' % (outfn)
for key, groups in groupslist:
for group in groups:
cachelocation = (fmeasure, '%s_file_%s_FILEID%s' %
(group.name, group.file.filename,
myutils.checksum(group.file.filename)))
data = dataman.obtain_data('length_dist', 'length_dist',
group['vessels'], group['tumor'],
cachelocation)
data = dataman.obtain_data('within_fake_tumor',
'radii_within_tumor', group['vessels'],
group['tumor'], cachelocation)
data = dataman.obtain_data('within_fake_tumor',
'pressures_within_tumor',
group['vessels'], group['tumor'],
cachelocation)
data = dataman.obtain_data('within_fake_tumor',
'flows_within_tumor', group['vessels'],
group['tumor'], cachelocation)
print('you just created a big nice file')
def doit_optimize(vesselFileName, adaptParams, BfParams):
returns = adaption_cpp.doAdaptionOptimization(vesselFileName, adaptParams,
BfParams)
print("should be optimized vaules:")
print(returns)
print('from file: %s' % vesselFileName)
f_results = h5files.open("optimize_results.h5", 'a')
a_uuid = str(uuid.uuid4())
g = f_results.create_group(vesselFileName + '_' + a_uuid)
g.create_dataset('x_opt', data=returns)
myutils.hdf_write_dict_hierarchy(g, 'adaptParams', adaptParams)
myutils.hdf_write_dict_hierarchy(g, 'BfParams', BfParams)
g.file.flush()
#f_results.close()
return returns
def run_out_in_single_file(goodArguments):
print('starting with arguments: %s' % goodArguments)
no_files = len(goodArguments.oxygenFiles)
annoxicVolumes = []
hypoxicVolumes = []
normoxicVolumes = []
tumorVolumes = []
#threshold = 15
threshold1 = 0.1
threshold2 = 2.5
test = myutils.MeanValueArray.empty()
annoxicVolumes_per_time = []
hypoxicVolumes_per_time = []
normoxicVolumes_per_time = []
tumorVolumes_per_time = []
timepoints = []
for aFile in goodArguments.oxygenFiles:
with h5files.open(aFile.name) as f:
try:
if not 'po2' in f:
raise AssertionError('no proper oxygen file: %s!' % f)
except Exception, e:
print e.message
sys.exit(-1)
paths = myutils.walkh5(f, 'po2/out*')
print('found paths: %s' % paths)
for path in paths:
#hypoxicFraction,hypoxicTissueVolume = estimate_ratio_hypoxic(f[path], threshold)
t = f[path]['SOURCE'].attrs['time']
r = f[path]['SOURCE/tumor'].attrs['TUMOR_RADIUS']
volume = 4 / 3. * 3.1414 * r * r * r / 1e9
#volume=(2*r)*(2*r)*(2*r)/1e9
tumorVolumes_per_time.append(volume)
t = t / (3600.0 * 24) #days
timepoints.append(t)
annoxicTissueVolume, hypoxicTissueVolume, normoxicTissueVolume = estimate_annoxic_hypoxic_normoxic(
f[path], r, threshold1, threshold2)
annoxicVolumes_per_time.append(annoxicTissueVolume)
hypoxicVolumes_per_time.append(hypoxicTissueVolume)
normoxicVolumes_per_time.append(normoxicTissueVolume)
annoxicVolumes.append(annoxicVolumes_per_time)
hypoxicVolumes.append(hypoxicVolumes_per_time)
normoxicVolumes.append(normoxicVolumes_per_time)
tumorVolumes.append(tumorVolumes_per_time)
def create_mean_plot_length():
my_res_file = h5files.open('analyzeVesselTumor.h5', 'r')
count_hists = 0
the_keys = my_res_file['/global'].keys()
print(the_keys[0])
the_density_in = np.zeros_like(
my_res_file['/global/%s/length_dist/inside_density' % the_keys[0]])
the_density_out = np.zeros_like(
my_res_file['/global/%s/length_dist/outside_density' % the_keys[0]])
a_bin_in = my_res_file['/global/%s/length_dist/inside_bins' % the_keys[0]]
a_bin_out = my_res_file['/global/%s/length_dist/outside_bins' %
the_keys[0]]
for key in my_res_file['/global'].keys():
print('working with %s' % key)
count_hists = count_hists + 1
the_density_in = the_density_in + my_res_file[
'/global/%s/length_dist/inside_density' % key]
the_density_out = the_density_out + my_res_file[
'/global/%s/length_dist/outside_density' % key]
print('we have %i histograms' % count_hists)
the_density_in = np.divide(the_density_in, count_hists)
the_density_out = np.divide(the_density_out, count_hists)
mu1 = np.sum(the_density_in * np.diff(a_bin_in) *
(0.5 * (a_bin_in[1:] + a_bin_in[:-1])))
quad1 = np.sum(the_density_in * np.diff(a_bin_in) * np.square(
(0.5 * (a_bin_in[1:] + a_bin_in[:-1]))))
std1 = np.sqrt(quad1 - mu1**2)
mu2 = np.sum(the_density_out * np.diff(a_bin_out) *
(0.5 * (a_bin_out[1:] + a_bin_out[:-1])))
quad2 = np.sum(the_density_out * np.diff(a_bin_out) * np.square(
(0.5 * (a_bin_out[1:] + a_bin_out[:-1]))))
std2 = np.sqrt(quad2 - mu2**2)
plt.title('Length distribution')
ax = plt.subplot(211)
ax.bar(a_bin_out[:-1], the_density_out, np.diff(a_bin_out))
ax.text(300, 0.035, '$< l_{out} > = %f \pm %f$' % (mu2, std2))
plt.legend(['Outside'])
plt.ylabel('Propability')
ax = plt.subplot(212)
plt.bar(a_bin_in[:-1], the_density_in, np.diff(a_bin_in))
ax.text(300, 0.008, '$< l_{in} > = %f \pm %f$' % (mu1, std1))
plt.xlabel('length of vessel segments/ $\mu m$')
plt.ylabel('Propability')
plt.legend(['Inside'])
plt.savefig('Length_distribution.png')
def MeasurementFile(f, h5files, prefix='measure_'):
if not isinstance(f, (str, unicode)):
fn = f.filename
else:
fn = f
fnpath = dirname(fn)
fnbase = basename(fn).rsplit('.h5')[0]
fnmeasure = join(fnpath, prefix + fnbase + '.h5')
existed = os.path.isfile(fnmeasure)
fm = h5files.open(fnmeasure, 'a')
if not existed:
fm.attrs['TYPE'] = 'MEASUREMENT'
fm.attrs['SOURCE'] = str(fn)
fm['source'] = h5py.ExternalLink(fn, '/')
return fm
def plotSingleRun(fn):
out = splitext(basename(fn))[0]
f = h5files.open(fn, 'r')
rc = matplotlib.rc
rc('figure', figsize=(7, 7), dpi=100)
rc('font', size=8.)
rc('axes', titlesize=10., labelsize=10.)
rc('pdf', compression=6, fonttype=42)
rc(
'figure', **{
'subplot.left': 0.02,
'subplot.right': 1. - 0.05,
'subplot.bottom': 0.01,
'subplot.top': 1. - 0.05,
'subplot.wspace': 0.1,
'subplot.hspace': 0.1
})
#rc('savefig', facecolor='none', edgecolor='none', dpi=100)
rc('savefig', dpi=100)
rc('font', **{'family': 'sans-serif'}) #,'sans-serif':['Helvetica']})
rc('path', simplify_threshold=0.01)
#rc('text', usetex=True, **{ 'latex.unicode' : True })
with mpl_utils.PdfWriter(out + ".pdf") as pdfpages:
dataman = myutils.DataManager(100, [
DataTumorTissueSingle(),
DataTumorMeasureCurvature(),
DataTumor3dRendering(),
DataTissueRadial(),
DataTissueRadialAveraged()
])
resultfile = ResultFile(f, dataman)
is3d = resultfile.obtain_data('ld').shape[2] > 1
if is3d:
plot3dRendering(resultfile,
pdfpages,
showVessels=True,
showTumor=False)
plot3dRendering(resultfile, pdfpages)
plot3dRendering(resultfile, pdfpages, showVessels=True)
plotSnapshots(resultfile, pdfpages)
def worker_on_client(vessel_fn, tumor_parameters, o2_parameter_set_name,
num_threads):
krebsutils.set_num_threads(num_threads)
tumor_fn = tumor_parameters['fn_out']
tend = tumor_parameters['tend']
pattern1 = 'out0000'
pattern2 = 'out%04i' % int(round(tend / tumor_parameters['out_intervall']))
pattern = '|'.join([pattern1, pattern2])
print tumor_fn, pattern
#os.system("%s -s '%s'" % (krebsjobs.submitTumOnlyVessels.exe,dicttoinfo(tumor_parameters)))
os.system("%s -s '%s'" %
(krebs.tumors.run_faketum, dicttoinfo(tumor_parameters)))
o2_refs = detailedo2.doit(
tumor_fn, pattern,
(getattr(krebs.detailedo2Analysis.parameterSetsO2,
o2_parameter_set_name), o2_parameter_set_name))
for ref in o2_refs:
po2group = h5files.open(ref.fn)[ref.path]
krebs.analyzeTissueOxygenDetailed.WriteSamplesToDisk(po2group)
krebs.povrayRenderOxygenDetailed.doit(o2_refs[1].fn, o2_refs[1].path)
h5files.closeall()
def createParamFile(files, label):
writepath = '/localdisk/thierry/tmp/%s_parms.py' % label
#mode = 'a' if os.path.exists(writepath) else 'w'
mode = 'w'
if os.path.exists(writepath):
os.remove(writepath)
with open(writepath,mode) as f:
f.write('# -*- coding: utf-8 -*-\n')
for t in typelist:
for pso_file in files:
if t in pso_file:
f_opt_data = h5files.open(pso_file, 'r', search = False)
xopt = f_opt_data.attrs.get('xopt')
f.write('pso_param_%s_%s_vary1=deepcopy(adaption_master)\n' % (label,t))
f.write('pso_param_%s_%s_vary1[\'adaption\'].update(\n' % (label,t))
f.write(' k_m = %f,\n' % xopt[0])
f.write(' k_c = %f,\n' % xopt[1])
f.write(' k_s = %f,\n' % xopt[2])
f.write(' cond_length = %f,\n' % xopt[3])
f.write(' Q_refdot = %f,\n' % xopt[4])
f.write(')\n')
print("Created file: %s"%writepath)
def ProduceData(fitParameters, filename):
from krebs.analyzeGeneral import DataBasicVessel, DataVesselSamples, DataVesselGlobal
from krebs.detailedo2Analysis import DataDetailedPO2
import krebs.detailedo2Analysis.singleVesselCases as singleVesselCases
paramspo2Override = dict(
massTransferCoefficientModelNumber = 1,
conductivity_coeff1 = fitParameters[0],
conductivity_coeff2 = fitParameters[1],
conductivity_coeff3 = fitParameters[2],
)
f = h5files.open(filename,'a')
krebsutils.set_num_threads(2)
dataman = myutils.DataManager(20, [DataDetailedPO2(),DataBasicVessel(), DataVesselSamples(), DataVesselGlobal()])
for k, params in fitCases:
params = deepcopy(params)
params.paramspo2.update(paramspo2Override)
singleVesselCases.GenerateSingleCapillaryWPo2(dataman, f, k, 16, params)
return f
distancemap_spec)
plot_data(ax, ranges, mvd.avg * 1e6, mvd.std * 1e6, 'rg'[j],
't = %i h' % time)
lim = ax.get_ylim()
ax.vlines(ranges[1:-1], lim[0], lim[1])
#ax.set(ylim = lim, xlim = (ax.get_xlim()[0], 1000))
ax.set(xlabel=r'distance from rim [$\mu m$]',
ylabel=r'mvd [1/mm$^2$]')
ax.legend()
pdfpages.savefig(fig)
if __name__ == '__main__':
filenames = sys.argv[1:]
files = [h5files.open(fn, 'r') for fn in filenames]
fmeasure = h5files.open('analyzeBloodVolumeMeasurements.h5', 'a')
dataman = myutils.DataManager(10, [
DataTumorTissueSingle(),
DataVesselRadial(),
DataVesselSamples(),
DataBasicVessel(),
DataDistanceFromCenter()
])
#allgroups = sorted(filter(lambda k: k.startswith('out'), files[0].keys()))
allgroups = defaultdict(list)
for f in files:
keys = filter(lambda k: k.startswith('out'), f.keys())
for k in keys:
allgroups[k].append(f[k])
parser.add_option(
"--write-flow",
dest="write_flow",
help=
"writes the recomputed flow; note: hematocrit is completely ignored!",
default=False,
action="store_true")
options, args = parser.parse_args()
if options.write_flow:
options.force_flow_recompute = True
dataman = myutils.DataManager(100, [DataBasicVessel()])
filenames, pattern = args[:-1], args[-1]
files = [
h5files.open(fn, 'a' if options.add_resistor_bc else 'r+')
for fn in filenames
]
groups = list(
itertools.chain.from_iterable(
myutils.walkh5(f, pattern, return_h5objects=True) for f in files))
for vesselgroup in groups:
if options.force_flow_recompute and not options.add_resistor_bc:
(pressure, flow,
shearforce) = krebsutils.calc_vessel_hydrodynamics(vesselgroup)
vessels = dataman.obtain_data('vessel_graph', vesselgroup,
['flags', 'radius'])
vessels.edges['flow'] = flow
vessels.nodes['pressure'] = pressure
else:
def DoIt(filenames, pattern, with_o2):
fn_measure = basename(commonprefix(filenames))
fn_measure = myutils.strip_from_end(fn_measure, '.h5')
fn_measure = myutils.strip_from_end(fn_measure, '-type')
def cachelocation(g):
path = posixpath.join(
'FileCS_' + myutils.checksum(basename(g.file.filename)),
g.name.strip(posixpath.sep))
return (f_measure, path)
if with_o2:
fn_measure = myutils.strip_from_end(fn_measure, '_detailedpo2')
files = [h5files.open(fn, 'a') for fn in filenames]
f_measure = h5files.open('plotVessels_chache.h5', 'a', search=False)
groups = list(
itertools.chain.from_iterable(
myutils.walkh5(f, pattern, return_h5objects=True) for f in files))
if len(groups) <= 0:
print 'no matching groups in hdf file(s)'
sys.exit(0)
if with_o2:
name = posixpath.commonprefix(map(lambda g: g.name, groups))
name = myutils.strip_from_start(name, '/po2/vessels').replace('/', '-')
fn_measure += name
with mpl_utils.PdfWriter(fn_measure + '.pdf') as pdfpages:
rc = matplotlib.rc
rc('font', size=8.)
rc('axes', titlesize=10., labelsize=8.)
if with_o2:
import detailedo2Analysis as o2analysis
import detailedo2Analysis.plotsForPaper
import detailedo2
dataman = myutils.DataManager(20, [
o2analysis.DataDetailedPO2(),
analyzeGeneral.DataTumorTissueSingle(),
analyzeGeneral.DataDistanceFromCenter(),
analyzeGeneral.DataBasicVessel(),
analyzeGeneral.DataVesselSamples(),
analyzeBloodFlow.DataTumorBloodFlow(),
analyzeGeneral.DataVesselRadial(),
analyzeGeneral.DataVesselGlobal()
])
vesselgroups = list(
detailedo2.OpenVesselAndTumorGroups(g)[0] for g in groups)
#original_vesselgroups = list(h5files.openLink(g, 'SOURCE') for g in vesselgroups)
if 1:
PrintGlobalDataWithOxygen(pdfpages, groups, vesselgroups,
f_measure, dataman)
'''FormatParameters makes the network creation parameters
that does not work, if we have an o2 file'''
#text = FormatParameters(original_vesselgroups[0].file)
text = [' ']
text += detailedo2Analysis.plotsForPaper.FormatParameters(
groups[0])
fig, _ = mpl_utils.MakeTextPage(text,
figsize=(mpl_utils.a4size[0],
mpl_utils.a4size[0]))
pdfpages.savefig(fig, postfix='_vesselsparams')
if 1:
res = getMultiScatter(300. * len(filenames), vesselgroups)
plotMultiScatterBeauty(res, pdfpages)
else:
dataman = myutils.DataManager(20, [
analyzeGeneral.DataTumorTissueSingle(),
analyzeGeneral.DataVesselRadial(),
analyzeGeneral.DataDistanceFromCenter(),
analyzeBloodFlow.DataTumorBloodFlow(),
analyzeGeneral.DataBasicVessel(),
analyzeGeneral.DataVesselSamples(),
analyzeGeneral.DataVesselGlobal()
])
#dataman = myutils.DataManager(20, [ analyzeGeneral.DataBasicVessel(), analyzeGeneral.DataVesselSamples(), analyzeGeneral.DataVesselGlobal()])
vesselgroups = groups
if 0:
res = getMultiScatter(300. * len(filenames), vesselgroups)
plotMultiScatterBeauty(res, pdfpages)
if 0:
PlotRadiusHistogram2(dataman, vesselgroups, pdfpages)
if 0 and all(map(lambda g: 'data' in g.parent, vesselgroups)):
data = VesselData()
for g in vesselgroups:
data.add(g.parent['data'])
plot_topological_stats_avg(data, pdfpages)
if 0: #reproduce swine
plot_geometric_stuff_on_RC(dataman, f_measure, filenames,
options, pdfpages)
if 1:
PrintGlobalData(pdfpages, vesselgroups, f_measure, dataman)