def filtering( self, inputs, outputs, length, cortex_only ):
'''
Filter the mapped fibers.
'''
# check if we have all required input data
# we need at least:
# - outputs['fibers_mapped'] == Track file in T1 space with mapped scalars
if not os.path.exists( outputs['fibers_mapped'] ):
c.error( Colors.RED + 'Could not find ' + Colors.YELLOW + outputs['fibers_mapped'] + Colors.RED + ' but we really need it to start with stage 4!!' + Colors._CLEAR )
sys.exit( 2 )
# find the order of the mapped scalars
header = io.loadTrkHeaderOnly( outputs['fibers_mapped'] )
scalars = list( header['scalar_name'] )
# split the length range
length = length.split( ' ' )
min_length = int( length[0] )
max_length = int( length[1] )
# length filtering
c.info( Colors.YELLOW + ' Filtering ' + Colors.PURPLE + 'fiber length' + Colors.YELLOW + ' to be ' + Colors.PURPLE + '>' + str( min_length ) + ' and <' + str( max_length ) + Colors.YELLOW + ' for ' + Colors.PURPLE + os.path.split( outputs['fibers_mapped'] )[1] + Colors.YELLOW + ' and store as ' + Colors.PURPLE + os.path.split( outputs['fibers_mapped_length_filtered'] )[1] + Colors.YELLOW + '!' + Colors._CLEAR )
fyborg.fyborg( outputs['fibers_mapped'], outputs['fibers_mapped_length_filtered'], [FyFilterLengthAction( scalars.index( 'length' ), min_length, max_length )] )
header = io.loadTrkHeaderOnly( outputs['fibers_mapped_length_filtered'] )
new_count = header['n_count']
c.info( Colors.YELLOW + ' Number of tracks after ' + Colors.PURPLE + 'length filtering' + Colors.YELLOW + ': ' + str( new_count ) + Colors.YELLOW + Colors._CLEAR )
if cortex_only:
# special cortex filtering
c.info( Colors.YELLOW + ' Filtering for ' + Colors.PURPLE + 'valid cortex structures' + Colors.YELLOW + ' in ' + Colors.PURPLE + os.path.split( outputs['fibers_mapped_length_filtered'] )[1] + Colors.YELLOW + ' and store as ' + Colors.PURPLE + os.path.split( outputs['fibers_mapped_length_filtered_cortex_only'] )[1] + Colors.YELLOW + '!' + Colors._CLEAR )
c.info( Colors.PURPLE + ' Conditions for valid fibers:' + Colors._CLEAR )
c.info( Colors.PURPLE + ' 1.' + Colors.YELLOW + ' The fiber track has to pass through the cerebral white matter. (Label values: ' + Colors.PURPLE + '[2, 41]' + Colors.YELLOW + ')' + Colors._CLEAR )
c.info( Colors.PURPLE + ' 2.' + Colors.YELLOW + ' The fiber track shall only touch sub-cortical structures not more than ' + Colors.PURPLE + '5 times' + Colors.YELLOW + '. (Label values: ' + Colors.PURPLE + '[10, 49, 16, 28, 60, 4, 43]' + Colors.YELLOW + ')' + Colors._CLEAR )
c.info( Colors.PURPLE + ' 3.' + Colors.YELLOW + ' The track shall not pass through the corpus callosum (Labels: ' + Colors.PURPLE + '[251, 255]' + Colors.YELLOW + ') and end in the same hemisphere (Labels: ' + Colors.PURPLE + '[1000-1035]' + Colors.YELLOW + ' for left, ' + Colors.PURPLE + '[2000-2035]' + Colors.YELLOW + ' for right).' + Colors._CLEAR )
fyborg.fyborg( outputs['fibers_mapped_length_filtered'], outputs['fibers_mapped_length_filtered_cortex_only'], [FyFilterCortexAction( scalars.index( 'segmentation' ) )] )
header = io.loadTrkHeaderOnly( outputs['fibers_mapped_length_filtered_cortex_only'] )
new_count = header['n_count']
c.info( Colors.YELLOW + ' Number of tracks after ' + Colors.PURPLE + 'cortex filtering' + Colors.YELLOW + ': ' + str( new_count ) + Colors.YELLOW + Colors._CLEAR )
c.info( Colors.YELLOW + ' Copied filtered tracks from ' + Colors.PURPLE + os.path.split( outputs['fibers_mapped_length_filtered_cortex_only'] )[1] + Colors.YELLOW + ' to ' + Colors.PURPLE + os.path.split( outputs['fibers_final'] )[1] + Colors.YELLOW + '!' + Colors._CLEAR )
shutil.copyfile( outputs['fibers_mapped_length_filtered_cortex_only'], outputs['fibers_final'] )
else:
c.info( Colors.YELLOW + ' Info: ' + Colors.PURPLE + 'Cortical _and_ sub-cortical structures ' + Colors.YELLOW + 'will be included..' + Colors._CLEAR )
c.info( Colors.YELLOW + ' Copied filtered tracks from ' + Colors.PURPLE + os.path.split( outputs['fibers_mapped_length_filtered'] )[1] + Colors.YELLOW + ' to ' + Colors.PURPLE + os.path.split( outputs['fibers_final'] )[1] + Colors.YELLOW + '!' + Colors._CLEAR )
shutil.copyfile( outputs['fibers_mapped_length_filtered'], outputs['fibers_final'] )
def setupGrid( self, matrix ):
if self.__test:
self.__rows = 101
self.__cols = 101
self.__gridWidget = GridView( self, self.__rows, self.__cols, False )
self.__layout.addWidget( self.__gridWidget, 0, 0 )
b_overwriteSpectralValue = True
maxEnergy = 255 / 3
automaton = C_spectrum_CAM_RGB( maxQuanta=maxEnergy )
automaton.component_add( 'R', maxEnergy / 3, b_overwriteSpectralValue )
automaton.component_add( 'G', maxEnergy / 3, b_overwriteSpectralValue )
automaton.component_add( 'B', maxEnergy / 3, b_overwriteSpectralValue )
automaton.updateRule_changeAmount(self.__updateAmount)
world = C_CAE( np.array( ( self.__rows, self.__cols ) ), automaton )
world.verbosity_set( 1 )
arr_world = np.zeros( ( self.__rows, self.__cols ) )
arr_world[0, 0] = 1
arr_world[50, 50] = maxEnergy / 3 + 1
arr_world[100, 100] = maxEnergy / 3 * 2 + 1
elif matrix:
maxEnergy = 255
arr_worldRaw = np.loadtxt( matrix, float, '#', '\t' )
arr_world = misc.arr_normalize( arr_worldRaw, scale=maxEnergy )
self.__rows, self.__cols = arr_world.shape
self.__gridWidget = GridView( self, self.__rows, self.__cols, False )
self.__layout.addWidget( self.__gridWidget, 0, 0 )
b_overwriteSpectralValue = True
automaton = C_spectrum_CAM_RGB( maxQuanta=maxEnergy )
automaton.component_add( 'R', maxEnergy / 3, b_overwriteSpectralValue )
automaton.component_add( 'G', maxEnergy / 3, b_overwriteSpectralValue )
automaton.component_add( 'B', maxEnergy / 3, b_overwriteSpectralValue )
print "Update amount = %d" % self.__updateAmount
automaton.updateRule_changeAmount(self.__updateAmount)
world = C_CAE( np.array( ( self.__rows, self.__cols ) ), automaton )
world.verbosity_set( 1 )
else:
c.error( 'No test mode and no matrix..' )
sys.exit()
print arr_world
world.initialize( arr_world )
self.__world = world
def mapping( self, inputs, outputs, radius ):
'''
Map all detected scalar volumes to each fiber.
'''
# check if we have all required input data
# we need at least:
# - outputs['fibers'] == Track file in T1 space
# - outputs['segmentation'] == Label Map
if not os.path.exists( outputs['fibers'] ):
c.error( Colors.RED + 'Could not find ' + Colors.YELLOW + outputs['fibers'] + Colors.RED + ' but we really need it to start with stage 3!!' + Colors._CLEAR )
sys.exit( 2 )
if not os.path.exists( outputs['segmentation'] ):
c.error( Colors.RED + 'Could not find ' + Colors.YELLOW + outputs['segmentation'] + Colors.RED + ' but we really need it to start with stage 3!!' + Colors._CLEAR )
sys.exit( 2 )
actions = []
for i in inputs:
if i == 'fibers' or i == 'segmentation' or i == 'T1' or i == 'b0':
# we do not map these
continue
if not os.path.exists( outputs[i + '_T1_space'] ):
# we can't map this since we didn't find the file
continue
# for normal scalars: append it to the actions
actions.append( FyMapAction( i, outputs[i + '_T1_space'] ) )
c.info( Colors.YELLOW + ' Configuring mapping of ' + Colors.PURPLE + os.path.split( outputs[i + '_T1_space'] )[1] + Colors.YELLOW + ' to ' + Colors.PURPLE + os.path.split( outputs['fibers'] )[1] + Colors.YELLOW + '!' + Colors._CLEAR )
# now the segmentation with the lookaround radius
actions.append( FyRadiusMapAction( 'segmentation', outputs['segmentation'], radius ) )
c.info( Colors.YELLOW + ' Configuring mapping of ' + Colors.PURPLE + os.path.split( outputs['segmentation'] )[1] + Colors.YELLOW + ' to ' + Colors.PURPLE + os.path.split( outputs['fibers'] )[1] + Colors.YELLOW + '!' + Colors._CLEAR )
# and also the fiber length
actions.append( FyLengthAction() )
c.info( Colors.YELLOW + ' Configuring mapping of ' + Colors.PURPLE + 'fiber length' + Colors.YELLOW + ' to ' + Colors.PURPLE + os.path.split( outputs['fibers'] )[1] + Colors.YELLOW + '!' + Colors._CLEAR )
# run, forest, run!!
c.info( Colors.YELLOW + ' Performing configured mapping for ' + Colors.PURPLE + os.path.split( outputs['fibers'] )[1] + Colors.YELLOW + ' and storing as ' + Colors.PURPLE + os.path.split( outputs['fibers_mapped'] )[1] + Colors.YELLOW + ' (~ 30 minutes)!' + Colors._CLEAR )
if self.__debug:
fyborg.fyborg( outputs['fibers'], outputs['fibers_mapped'], actions, 'debug' )
else:
fyborg.fyborg( outputs['fibers'], outputs['fibers_mapped'], actions )
def run(self, masterFile, inputFiles, outputDirectory, spacing, dimensions,
likefreesurfer, nii):
'''
Performs the equalization
'''
# sanity checks
outputDirectory = os.path.normpath(outputDirectory)
# prepare the output directory
if os.path.exists(outputDirectory):
c.error('The output directory already exists!')
c.error('Aborting..')
sys.exit(2)
# create the output directory
os.mkdir(outputDirectory)
# MASTER
masterFile = os.path.normpath(masterFile)
# read the master
master = io.readImage(masterFile)
c.info('MASTER IMAGE: ' + str(masterFile))
# INPUTS
for i in range(len(inputFiles)):
inputFiles[i] = os.path.normpath(inputFiles[i])
c.info('INPUT IMAGE ' + str(i + 1) + ': ' + str(inputFiles[i]))
# print more info
c.info('OUTPUT DIRECTORY: ' + str(outputDirectory))
if likefreesurfer:
spacing = '1,1,1'
dimensions = '256,256,256'
if spacing != 'no':
c.info('SET SPACINGS: ' + str(spacing))
if dimensions != 'no':
c.info('SET DIMENSIONS: ' + str(dimensions))
# re-sample master to obtain an isotropic dataset
master = self.aniso2iso(master, spacing, dimensions)
masterFileBasename = os.path.split(masterFile)[1]
masterFileBasenameWithoutExt = os.path.splitext(masterFileBasename)[0]
if not nii:
masterOutputFileName = os.path.join(outputDirectory,
masterFileBasename)
else:
masterOutputFileName = os.path.join(
outputDirectory, masterFileBasenameWithoutExt) + '.nii'
io.saveImage(masterOutputFileName, master)
# equalize all images to the master
for i in range(len(inputFiles)):
currentInputFile = inputFiles[i]
c.info('Equalizing ' + str(currentInputFile) + ' to ' +
str(masterFile) + "...")
# load the image
currentImage = io.readImage(currentInputFile)
currentImageHeader = currentImage.header
c.info(' old spacing: ' + str(currentImageHeader.get_zooms()))
c.info(' old dimensions: ' + str(currentImage.shape[:3]))
# now resample
resampledImage = resampler.resample_img2img(currentImage, master)
# .. and save it
currentInputFileBasename = os.path.split(currentInputFile)[1]
currentInputFileBasenameWithoutExt = os.path.splitext(
currentInputFileBasename)[0]
if not nii:
outputFileName = os.path.join(outputDirectory,
currentInputFileBasename)
else:
outputFileName = os.path.join(
outputDirectory, currentInputFileBasenameWithoutExt)
savedImage = io.saveImage(outputFileName, resampledImage)
#c.info( ' new spacing: ' + str( savedImageHeader.get_zooms() ) )
c.info(' new dimensions: ' + str(savedImage.shape[:3]))
c.info('All done!')
def run(self, input, output, mode, verbose, jobs):
if len(input) < 2:
c.error("Please specify at least two *.trk files as input!")
sys.exit(2)
if os.path.exists(output):
# abort if file already exists
c.error("File " + str(output) + " already exists..")
c.error("Aborting..")
sys.exit(2)
jobs = int(jobs)
if jobs < 1 or jobs > 32:
jobs = 1
# load 'master'
mTracks = io.loadTrk(input[0])
# copy the tracks and the header from the 'master'
c.info("Master is " + input[0])
outputTracks = mTracks[0]
c.info("Number of tracks: " + str(len(outputTracks)))
header = mTracks[1]
# remove the first input
input.pop(0)
if mode == "add":
#
# ADD
#
for i in input:
iTracks = io.loadTrk(i)
# add the tracks
c.debug("Adding " + str(len(iTracks[0])) + " tracks from " + i + " to master..", verbose)
outputTracks = TrackvisCalcLogic.add(outputTracks, iTracks[0])
c.debug("Number of output tracks after final addition: " + str(len(outputTracks)), verbose)
elif mode == "sub":
#
# SUB
#
c.debug("Using " + str(jobs) + " threads..", verbose)
mergedOutputTracks = outputTracks[:]
for i in input:
iTracks = io.loadTrk(i)
# subtract the tracks
c.info("Subtracting " + i + " (" + str(len(iTracks[0])) + " tracks) from master..")
#
# THREADED COMPONENT
#
numberOfThreads = jobs
c.info("Splitting master into " + str(jobs) + " pieces..")
splittedOutputTracks = u.split_list(mergedOutputTracks, numberOfThreads)
# list of threads
t = [None] * numberOfThreads
# list of alive flags
a = [None] * numberOfThreads
# list of tempFiles
f = [None] * numberOfThreads
for n in xrange(numberOfThreads):
# mark thread as alive
a[n] = True
# fire the thread and give it a filename based on the number
tmpFile = tempfile.mkstemp(".trk", "t_calc")[1]
f[n] = tmpFile
t[n] = Process(
target=TrackvisCalcLogic.sub,
args=(splittedOutputTracks[n][:], iTracks[0][:], tmpFile, verbose, "Thread-" + str(n + 1)),
)
c.info("Starting Thread-" + str(n + 1) + "...")
t[n].start()
allDone = False
while not allDone:
time.sleep(1)
for n in xrange(numberOfThreads):
a[n] = t[n].is_alive()
if not any(a):
# if no thread is alive
allDone = True
#
# END OF THREADED COMPONENT
#
c.info("All Threads done!")
c.info("Merging output..")
# now read all the created tempFiles and merge'em to one
# first thread output is the master here
tmpMaster = f[0]
tMasterTracks = io.loadTrk(tmpMaster)
for tmpFileNo in xrange(1, len(f)):
tTracks = io.loadTrk(f[tmpFileNo])
# add them
mergedOutputTracks = TrackvisCalcLogic.add(tMasterTracks[0], tTracks[0])
c.info("Merging done!")
# some stats
c.info("Number of output tracks after final removal: " + str(len(mergedOutputTracks)))
outputTracks = mergedOutputTracks
# now save the outputTracks
io.saveTrk(output, outputTracks, header)
c.info("All done!")
def makeMatrix(inputs, outputs, no_cortex):
"""
Make 1/ADC, ADC, FA, FiberNumber, FiberLength, E1, E2, E3 connectivity matrices.
"""
s = io.loadTrk(outputs["fibers_final"])
tracks = s[0]
header = s[1]
scalarNames = header["scalar_name"].tolist()
matrix = {}
# check if the segmentation is mapped
try:
scalarNames.index("segmentation")
except:
c.error(Colors.RED)
for s in scalarNames:
if not s:
continue
print s
return
for i in inputs:
if i == "fibers" or i == "segmentation" or i == "T1" or i == "b0":
# we do not map these
continue
# for tCounter, t in enumerate( tracks ):
try:
labelIndex = scalarNames.index("segmentation")
adcIndex = scalarNames.index("adc")
faIndex = scalarNames.index("fa")
e1Index = scalarNames.index("e1")
e2Index = scalarNames.index("e2")
e3Index = scalarNames.index("e3")
lengthIndex = scalarNames.index("length")
except:
c.error("Not all scalars were found: aparc_aseg_endlabel, adc, fa, length, e1, e2, e3")
sys.exit(2)
m_fn = numpy.zeros([68, 68])
m_fa = numpy.zeros([68, 68])
m_adc = numpy.zeros([68, 68])
m_adcinv = numpy.zeros([68, 68])
m_len = numpy.zeros([68, 68])
m_e1 = numpy.zeros([68, 68])
m_e2 = numpy.zeros([68, 68])
m_e3 = numpy.zeros([68, 68])
fslabel_vol = [
2012,
2019,
2032,
2014,
2020,
2018,
2027,
2028,
2003,
2024,
2017,
2026,
2002,
2023,
2010,
2022,
2031,
2029,
2008,
2025,
2005,
2021,
2011,
2013,
2007,
2016,
2006,
2033,
2009,
2015,
2001,
2030,
2034,
2035,
1012,
1019,
1032,
1014,
1020,
1018,
1027,
1028,
1003,
1024,
1017,
1026,
1002,
1023,
1010,
1022,
1031,
1029,
1008,
1025,
1005,
1021,
1011,
1013,
1007,
1016,
1006,
1033,
1009,
1015,
1001,
1030,
1034,
1035,
]
for tCounter, t in enumerate(tracks):
tCoordinates = t[0]
tScalars = t[1]
fa = numpy.mean(tScalars[:, faIndex])
adc = numpy.mean(tScalars[:, adcIndex])
e1 = numpy.mean(tScalars[:, e1Index])
e2 = numpy.mean(tScalars[:, e2Index])
e3 = numpy.mean(tScalars[:, e3Index])
len = tScalars[0, lengthIndex]
firstLabel = tScalars[0, labelIndex]
lastLabel = tScalars[-1, labelIndex]
try:
fIndex = fslabel_vol.index(firstLabel)
lIndex = fslabel_vol.index(lastLabel)
except:
continue
print "found", firstLabel, lastLabel
m_fn[fIndex, lIndex] += 1
m_fa[fIndex, lIndex] += fa
m_adc[fIndex, lIndex] += adc
m_e1[fIndex, lIndex] += e1
m_e2[fIndex, lIndex] += e2
m_e3[fIndex, lIndex] += e3
m_adcinv[fIndex, lIndex] += 1 / adc
m_len[fIndex, lIndex] += len
# symmetrize matrices
m_fn = m_fn + m_fn.T - numpy.diag(m_fn.diagonal())
m_fa = m_fa + m_fa.T - numpy.diag(m_fa.diagonal())
m_adc = m_adc + m_adc.T - numpy.diag(m_adc.diagonal())
m_e1 = m_e1 + m_e1.T - numpy.diag(m_e1.diagonal())
m_e2 = m_e2 + m_e2.T - numpy.diag(m_e2.diagonal())
m_e3 = m_e3 + m_e3.T - numpy.diag(m_e3.diagonal())
m_adcinv = m_adcinv + m_adcinv.T - numpy.diag(m_adcinv.diagonal())
m_len = m_len + m_len.T - numpy.diag(m_len.diagonal())
# normalize matrices
m_fa[:] /= m_fn[:]
m_adc[:] /= m_fn[:]
m_e1[:] /= m_fn[:]
m_e2[:] /= m_fn[:]
m_e3[:] /= m_fn[:]
m_adcinv[:] /= m_fn[:]
m_len[:] /= m_fn[:]
m_fa = numpy.nan_to_num(m_fa)
m_e1 = numpy.nan_to_num(m_e1)
m_e2 = numpy.nan_to_num(m_e2)
m_e3 = numpy.nan_to_num(m_e3)
m_adc = numpy.nan_to_num(m_adc)
m_adcinv = numpy.nan_to_num(m_adcinv)
m_len = numpy.nan_to_num(m_len)
# save as .mat and .csv
sio.savemat(
outputDirectory + "fibmap_all_cMatrix.mat",
{
"m_fiberNumber": m_fn,
"m_fa": m_fa,
"m_adc": m_adc,
"m_adcInverse": m_adcinv,
"m_fiberLength": m_len,
"m_e1": m_e1,
"m_e2": m_e2,
"m_e3": m_e3,
},
)
numpy.savetxt(outputDirectory + "fibmap_fibernumber_cMatrix.csv", m_fn, delimiter=",")
numpy.savetxt(outputDirectory + "fibmap_fa_cMatrix.csv", m_fa, delimiter=",")
numpy.savetxt(outputDirectory + "fibmap_e1_cMatrix.csv", m_e1, delimiter=",")
numpy.savetxt(outputDirectory + "fibmap_e2_cMatrix.csv", m_e2, delimiter=",")
numpy.savetxt(outputDirectory + "fibmap_e3_cMatrix.csv", m_e3, delimiter=",")
numpy.savetxt(outputDirectory + "fibmap_adc_cMatrix.csv", m_adc, delimiter=",")
numpy.savetxt(outputDirectory + "fibmap_adcinv_cMatrix.csv", m_adcinv, delimiter=",")
numpy.savetxt(outputDirectory + "fibmap_fiberlength_cMatrix.csv", m_len, delimiter=",")
c.info("Connectivity matrices generated and stored.")
def run( self, input, output, matrix, jobs ):
'''
'''
if os.path.exists( output ):
# abort if file already exists
c.error( 'File ' + str( output ) + ' already exists..' )
c.error( 'Aborting..' )
sys.exit( 2 )
if not os.path.isfile( matrix ):
# abort if the matrix does not exist
c.error( 'Matrix-File ' + str( matrix ) + ' does not exist..' )
c.error( 'Aborting..' )
sys.exit( 2 )
jobs = int( jobs )
if jobs < 1 or jobs > 32:
jobs = 1
# read
c.info( 'Loading ' + input + '..' )
t = io.loadTrk( input )
tracks = t[0]
header = t[1]
#.. copy the current header
newHeader = numpy.copy( header )
# print old matrix in header
#
# WARNING: this matrix is actually never used by TrackVis (see email from Ruopeng).
# We still modify it to keep it in sync with the transformations which we apply point wise.
#
if hasattr( header, 'vox_to_ras' ):
oldMatrix = header['vox_to_ras']
c.info( 'Old transformation matrix:' )
c.info( ' ' + str( oldMatrix[0] ) )
c.info( ' ' + str( oldMatrix[1] ) )
c.info( ' ' + str( oldMatrix[2] ) )
c.info( ' ' + str( oldMatrix[3] ) )
#
# load our transformation Matrix
#
newMatrix = numpy.loadtxt( matrix, float, '#', ' ' )
#
# THREADED COMPONENT
#
numberOfThreads = jobs
c.info( 'Splitting the input into ' + str( jobs ) + ' pieces..' )
splittedOutputTracks = u.split_list( tracks, numberOfThreads )
# list of threads
t = [None] * numberOfThreads
# list of alive flags
a = [None] * numberOfThreads
# list of tempFiles
f = [None] * numberOfThreads
for n in xrange( numberOfThreads ):
# mark thread as alive
a[n] = True
# fire the thread and give it a filename based on the number
tmpFile = tempfile.mkstemp( '.trk', 't_transform' )[1]
f[n] = tmpFile
t[n] = Process( target=TrackvisTransformLogic.transform, args=( splittedOutputTracks[n][:], newMatrix, tmpFile, False, 'Thread-' + str( n + 1 ) ) )
c.info( "Starting Thread-" + str( n + 1 ) + "..." )
t[n].start()
allDone = False
while not allDone:
time.sleep( 1 )
for n in xrange( numberOfThreads ):
a[n] = t[n].is_alive()
if not any( a ):
# if no thread is alive
allDone = True
#
# END OF THREADED COMPONENT
#
c.info( "All Threads done!" )
c.info( "Merging output.." )
# now read all the created tempFiles and merge'em to one
# first thread output is the master here
tmpMaster = f[0]
tMasterTracks = io.loadTrk( tmpMaster )
for tmpFileNo in xrange( 1, len( f ) ):
tTracks = io.loadTrk( f[tmpFileNo] )
# add them
tracks = TrackvisCalcLogic.add( tMasterTracks[0], tTracks[0] )
c.info( "Merging done!" )
#
# replace the matrix in the header with a transformed one even if it will never be used by TrackVis
#
if hasattr( header, 'vox_to_ras' ):
result = numpy.dot( oldMatrix, newMatrix )
c.info( 'New transformation matrix:' )
c.info( ' ' + str( result[0] ) )
c.info( ' ' + str( result[1] ) )
c.info( ' ' + str( result[2] ) )
c.info( ' ' + str( result[3] ) )
newHeader['vox_to_ras'] = result
# write
c.info( 'Saving ' + output + '..' )
io.saveTrk( output, tracks, newHeader )
c.info( 'All done!' )
def run(self, input, output, mode, verbose, jobs):
if len(input) < 2:
c.error('Please specify at least two *.trk files as input!')
sys.exit(2)
if os.path.exists(output):
# abort if file already exists
c.error('File ' + str(output) + ' already exists..')
c.error('Aborting..')
sys.exit(2)
jobs = int(jobs)
if jobs < 1 or jobs > 32:
jobs = 1
# load 'master'
mTracks = io.loadTrk(input[0])
# copy the tracks and the header from the 'master'
c.info('Master is ' + input[0])
outputTracks = mTracks[0]
c.info('Number of tracks: ' + str(len(outputTracks)))
header = mTracks[1]
# remove the first input
input.pop(0)
if mode == 'add':
#
# ADD
#
for i in input:
iTracks = io.loadTrk(i)
# add the tracks
c.debug(
'Adding ' + str(len(iTracks[0])) + ' tracks from ' + i +
' to master..', verbose)
outputTracks = TrackvisCalcLogic.add(outputTracks, iTracks[0])
c.debug(
'Number of output tracks after final addition: ' +
str(len(outputTracks)), verbose)
elif mode == 'sub':
#
# SUB
#
c.debug('Using ' + str(jobs) + ' threads..', verbose)
mergedOutputTracks = outputTracks[:]
for i in input:
iTracks = io.loadTrk(i)
# subtract the tracks
c.info('Subtracting ' + i + ' (' + str(len(iTracks[0])) +
' tracks) from master..')
#
# THREADED COMPONENT
#
numberOfThreads = jobs
c.info('Splitting master into ' + str(jobs) + ' pieces..')
splittedOutputTracks = u.split_list(mergedOutputTracks,
numberOfThreads)
# list of threads
t = [None] * numberOfThreads
# list of alive flags
a = [None] * numberOfThreads
# list of tempFiles
f = [None] * numberOfThreads
for n in xrange(numberOfThreads):
# mark thread as alive
a[n] = True
# fire the thread and give it a filename based on the number
tmpFile = tempfile.mkstemp('.trk', 't_calc')[1]
f[n] = tmpFile
t[n] = Process(target=TrackvisCalcLogic.sub,
args=(splittedOutputTracks[n][:],
iTracks[0][:], tmpFile, verbose,
'Thread-' + str(n + 1)))
c.info("Starting Thread-" + str(n + 1) + "...")
t[n].start()
allDone = False
while not allDone:
time.sleep(1)
for n in xrange(numberOfThreads):
a[n] = t[n].is_alive()
if not any(a):
# if no thread is alive
allDone = True
#
# END OF THREADED COMPONENT
#
c.info("All Threads done!")
c.info("Merging output..")
# now read all the created tempFiles and merge'em to one
# first thread output is the master here
tmpMaster = f[0]
tMasterTracks = io.loadTrk(tmpMaster)
for tmpFileNo in xrange(1, len(f)):
tTracks = io.loadTrk(f[tmpFileNo])
# add them
mergedOutputTracks = TrackvisCalcLogic.add(
tMasterTracks[0], tTracks[0])
c.info("Merging done!")
# some stats
c.info('Number of output tracks after final removal: ' +
str(len(mergedOutputTracks)))
outputTracks = mergedOutputTracks
# now save the outputTracks
io.saveTrk(output, outputTracks, header)
c.info('All done!')
def roi_extract( self, inputs, outputs ):
'''
'''
# check if we have all required input data
# we need at least:
# - outputs['fibers_mapped'] == Track file in T1 space with mapped scalars
if not os.path.exists( outputs['fibers_final'] ):
c.error( Colors.RED + 'Could not find ' + Colors.YELLOW + outputs['fibers_final'] + Colors.RED + ' but we really need it to start with stage 6!!' + Colors._CLEAR )
sys.exit( 2 )
s = io.loadTrk( outputs['fibers_final'] )
tracks = s[0]
header = s[1]
scalarNames = header['scalar_name'].tolist()
labels = {}
# check if the segmentation is mapped
try:
seg_index = scalarNames.index( 'segmentation' )
except:
c.error( Colors.RED + 'Could not find ' + Colors.YELLOW + 'segmentation' + Colors.RED + ' as a mapped scalar but we really need it!' )
sys.exit( 2 )
# create the roi subfolder
if not os.path.exists( outputs['roi'] ):
os.mkdir( outputs['roi'] )
# parse the color table
lut = colortable.freesurfer.split( '\n' )
colors = {}
for color in lut:
if not color or color[0] == '#':
continue
splitted_line = color.split( ' ' )
splitted_line = filter( None, splitted_line )
colors[splitted_line[0]] = splitted_line[1]
# loop through tracks
for i, t in enumerate( tracks ):
tCoordinates = t[0]
tScalars = t[1]
# grab the scalars for each point
for scalar in tScalars:
# but only the label value
label_value = str( int( scalar[seg_index] ) )
if not label_value in labels:
labels[label_value] = []
if not i in labels[label_value]:
# store the unique fiber id for this label
labels[label_value].append( i )
# now loop through all detected labels
for l in labels:
new_tracks = []
for t_id in labels[l]:
# grab the fiber + scalars
current_fiber = tracks[t_id]
new_tracks.append( current_fiber )
# now store the trk file
trk_outputfile = l + '_' + colors[l] + '.trk'
nii_outputfile = l + '_' + colors[l] + '.nii'
c.info( Colors.YELLOW + ' Creating fiber ROI ' + Colors.PURPLE + trk_outputfile + Colors.YELLOW + '!' + Colors._CLEAR )
io.saveTrk( os.path.join( outputs['roi'], trk_outputfile ), new_tracks, header, None, True )
# also create a roi label volume for this label value
c.info( Colors.YELLOW + ' Creating NII ROI ' + Colors.PURPLE + nii_outputfile + Colors.YELLOW + '!' + Colors._CLEAR )
cmd = 'ss;'
cmd += 'chb-fsstable;'
cmd += 'mri_binarize --i ' + outputs['segmentation'] + ' --o ' + os.path.join( outputs['roi'], nii_outputfile ) + ' --match ' + l + ' --binval ' + l + ';'
self.__logger.debug( cmd )
sp = subprocess.Popen( ["/bin/bash", "-i", "-c", cmd], stdout=sys.stdout )
sp.communicate()
def connectivity( self, inputs, outputs, cortex_only ):
'''
Generate connectivity matrices using mapped values.
'''
# check if we have all required input data
# we need at least:
# - outputs['fibers_mapped'] == Track file in T1 space with mapped scalars
if not os.path.exists( outputs['fibers_final'] ):
c.error( Colors.RED + 'Could not find ' + Colors.YELLOW + outputs['fibers_final'] + Colors.RED + ' but we really need it to start with stage 5!!' + Colors._CLEAR )
sys.exit( 2 )
s = io.loadTrk( outputs['fibers_final'] )
tracks = s[0]
header = s[1]
scalarNames = header['scalar_name'].tolist()
matrix = {}
indices = {}
# check if the segmentation is mapped
try:
indices['segmentation'] = scalarNames.index( 'segmentation' )
except:
c.error( Colors.RED + 'Could not find ' + Colors.YELLOW + 'segmentation' + Colors.RED + ' as a mapped scalar but we really need it!' )
sys.exit( 2 )
if cortex_only:
labels = [2012, 2019, 2032, 2014, 2020, 2018, 2027, 2028, 2003, 2024, 2017, 2026, 2002, 2023, 2010, 2022, 2031, 2029, 2008, 2025, 2005, 2021, 2011, 2013, 2007, 2016, 2006, 2033, 2009, 2015, 2001, 2030, 2034, 2035, 1012, 1019, 1032, 1014, 1020, 1018, 1027, 1028, 1003, 1024, 1017, 1026, 1002, 1023, 1010, 1022, 1031, 1029, 1008, 1025, 1005, 1021, 1011, 1013, 1007, 1016, 1006, 1033, 1009, 1015, 1001, 1030, 1034, 1035]
else:
labels = [2012, 2019, 2032, 2014, 2020, 2018, 2027, 2028, 2003, 2024, 2017, 2026, 2002, 2023, 2010, 2022, 2031, 2029, 2008, 2025, 2005, 2021, 2011, 2013, 2007, 2016, 2006, 2033, 2009, 2015, 2001, 2030, 2034, 2035, 49, 50, 51, 52, 58, 53, 54, 1012, 1019, 1032, 1014, 1020, 1018, 1027, 1028, 1003, 1024, 1017, 1026, 1002, 1023, 1010, 1022, 1031, 1029, 1008, 1025, 1005, 1021, 1011, 1013, 1007, 1016, 1006, 1033, 1009, 1015, 1001, 1030, 1034, 1035, 10, 11, 12, 13, 26, 17, 18, 16]
c.info( Colors.YELLOW + ' Getting ready to create connectivity matrices for the following labels: ' + Colors.PURPLE + str( labels ) + Colors._CLEAR )
c.info( Colors.YELLOW + ' Note: Mapped scalar values along the points will be averaged for each fiber track.' + Colors._CLEAR )
# create matrices for the attached scalars
for i, s in enumerate( scalarNames ):
if i >= header['n_scalars']:
break
if not s or s == 'segmentation':
continue
# this is a scalar value for which a matrix will be created
matrix[s] = np.zeros( [len( labels ), len( labels )] )
indices[s] = scalarNames.index( s )
c.info( Colors.YELLOW + ' Preparing matrix (' + Colors.PURPLE + '[' + str( len( labels ) ) + 'x' + str( len( labels ) ) + ']' + Colors.YELLOW + ') for ' + Colors.PURPLE + s + Colors.YELLOW + ' values!' + Colors._CLEAR )
if s == 'adc':
s = 'inv_adc'
matrix[s] = np.zeros( [len( labels ), len( labels )] )
indices[s] = scalarNames.index( 'adc' )
c.info( Colors.YELLOW + ' Preparing matrix (' + Colors.PURPLE + '[' + str( len( labels ) ) + 'x' + str( len( labels ) ) + ']' + Colors.YELLOW + ') for ' + Colors.PURPLE + s + Colors.YELLOW + ' values!' + Colors._CLEAR )
# always create one for the fiber counts
matrix['fibercount'] = np.zeros( [len( labels ), len( labels )] )
indices['fibercount'] = 0
c.info( Colors.YELLOW + ' Preparing matrix (' + Colors.PURPLE + '[' + str( len( labels ) ) + 'x' + str( len( labels ) ) + ']' + Colors.YELLOW + ') for ' + Colors.PURPLE + 'fibercount' + Colors.YELLOW + ' values!' + Colors._CLEAR )
c.info( Colors.YELLOW + ' Analyzing fibers of ' + Colors.PURPLE + os.path.split( outputs['fibers_final'] )[1] + Colors.YELLOW + '..' + Colors._CLEAR )
for tCounter, t in enumerate( tracks ):
tCoordinates = t[0]
tScalars = t[1]
# find the segmentation labels for the start and end points
start_label = tScalars[0, indices['segmentation']]
end_label = tScalars[-1, indices['segmentation']]
try:
# now grab the index of the labels in our label list
start_index = labels.index( start_label )
end_index = labels.index( end_label )
except:
# this label is not monitored, so ignore this track
continue
# loop through all different scalars
for m in matrix:
# calculate the mean for each track
value = np.mean( tScalars[:, indices[m]] )
if m == 'inv_adc':
# invert the value since it is 1-ADC
value = 1 / value
elif m == 'fibercount':
# in the case of fibercount, add 1
value = 1
# store value in the matrix
matrix[m][start_index, end_index] += value
if not start_index == end_index:
matrix[m][end_index, start_index] += value
# fiber loop is done, all values are stored
# now normalize the matrices
np.seterr( all='ignore' ) # avoid div by 0 warnings
cbar = None
for m in matrix:
if not m == 'fibercount':
# normalize it
matrix[m][:] /= matrix['fibercount']
matrix[m] = np.nan_to_num( matrix[m] )
# store the matrix
c.info( Colors.YELLOW + ' Storing ' + Colors.PURPLE + m + Colors.YELLOW + ' connectivity matrix as ' + Colors.PURPLE + os.path.split( outputs['matrix_' + m] )[1] + Colors.YELLOW + '!' + Colors._CLEAR )
np.savetxt( outputs['matrix_' + m], matrix[m], delimiter='\t' )
# store a picture
picture_path = os.path.splitext( os.path.split( outputs['matrix_' + m] )[1] )[0] + '.png'
c.info( Colors.YELLOW + ' Generating ' + Colors.PURPLE + m + ' image' + Colors.YELLOW + ' as ' + Colors.PURPLE + picture_path + Colors.YELLOW + '!' + Colors._CLEAR )
img = plot.imshow( matrix[m], interpolation='nearest' )
img.set_cmap( 'jet' )
img.set_norm( LogNorm() )
img.axes.get_xaxis().set_visible( False )
img.axes.get_yaxis().set_visible( False )
if not cbar:
cbar = plot.colorbar()
cbar.set_label( m )
cbar.set_ticks( [] )
plot.savefig( os.path.join( os.path.split( outputs['matrix_' + m] )[0], picture_path ) )
np.seterr( all='warn' ) # reactivate div by 0 warnings
# now store the matlab version as well
c.info( Colors.YELLOW + ' Storing ' + Colors.PURPLE + 'matlab data bundle' + Colors.YELLOW + ' containing ' + Colors.PURPLE + 'all matrices' + Colors.YELLOW + ' as ' + Colors.PURPLE + os.path.split( outputs['matrix_all'] )[1] + Colors.YELLOW + '!' + Colors._CLEAR )
scipy.io.savemat( outputs['matrix_all'], matrix, oned_as='row' )
parser.add_argument( '-co', '--cortex_only', action='store_true', dest='cortex_only', help='Perform filtering for cortex specific analysis and skip sub-cortical structures.' )
parser.add_argument( '-s', '--stage', action='store', dest='stage', default=0, type=int, help='Start with a specific stage while skipping the ones before. E.g. --stage 3 directly starts the mapping without preprocessing, --stage 4 starts with the filtering' )
parser.add_argument( '-overwrite', '--overwrite', action='store_true', dest='overwrite', help='Overwrite any existing output. DANGER!!' )
parser.add_argument( '-v', '--verbose', action='store_true', dest='verbose', help='Show verbose output' )
# always show the help if no arguments were specified
if len( sys.argv ) == 1:
parser.print_help()
sys.exit( 1 )
options = parser.parse_args()
# validate the inputs here
if not os.path.isdir( options.input ):
c.error( Colors.RED + 'Could not find the input directory! Specify a valid directory using -i $PATH.' + Colors._CLEAR )
sys.exit( 2 )
if os.path.exists( options.output ) and int( options.stage ) == 0:
if not options.overwrite:
c.error( Colors.RED + 'The output directory exists! Add --overwrite to erase previous content!' + Colors._CLEAR )
c.error( Colors.RED + 'Or use --stage > 2 to start with a specific stage which re-uses the previous content..' + Colors._CLEAR )
sys.exit( 2 )
else:
# silently delete the existing output
shutil.rmtree( options.output )
if options.stage > 0 and not os.path.exists( options.output ):
# we start with a specific stage so we need the output stuff
c.error( Colors.RED + 'The output directory does not exist! We need it when using -s/--stage to resume the process!' + Colors._CLEAR )
def fyborg( trkFile, outputTrkFile, actions, *args ):
if not actions:
c.error( "We gotta do something.." )
return
showDebug = 'debug' in args
singleThread = 'singlethread' in args
c.debug( "trkFile:" + str( trkFile ), showDebug )
c.debug( "outputTrkFile:" + str( outputTrkFile ), showDebug )
c.debug( "args:" + str( args ), showDebug )
# load trk file
s = io.loadTrk( trkFile )
tracks = s[0]
origHeader = s[1]
tracksHeader = numpy.copy( s[1] )
numberOfScalars = origHeader['n_scalars']
scalars = origHeader['scalar_name'].tolist()
numberOfTracks = origHeader['n_count']
# show some file informations
printTrkInfo( tracksHeader, trkFile )
# grab the scalarNames
scalarNames = []
for a in actions:
if a.scalarName() != FyAction.NoScalar:
scalarNames.append( a.scalarName() )
# increase the number of scalars
tracksHeader['n_scalars'] += len( scalarNames )
# .. attach the new scalar names
for i in range( len( scalarNames ) ):
tracksHeader['scalar_name'][numberOfScalars + i] = scalarNames[i]
#
# THREADED COMPONENT
#
if singleThread:
numberOfThreads = 1
else:
numberOfThreads = multiprocessing.cpu_count()
c.info( 'Splitting master into ' + str( numberOfThreads ) + ' pieces..' )
splittedOutputTracks = u.split_list( tracks[:], numberOfThreads )
# list of threads
t = [None] * numberOfThreads
# list of alive flags
a = [None] * numberOfThreads
# list of tempFiles
f = [None] * numberOfThreads
for n in xrange( numberOfThreads ):
# configure actions
__actions = []
for act in actions:
__actions.append( act )
# mark thread as alive
a[n] = True
# fire the thread and give it a filename based on the number
tmpFile = tempfile.mkstemp( '.trk', 'fyborg' )[1]
f[n] = tmpFile
t[n] = Process( target=fyborgLooper_, args=( splittedOutputTracks[n][:], tracksHeader, tmpFile, __actions, showDebug, n + 1 ) )
c.info( "Starting Thread-" + str( n + 1 ) + "..." )
t[n].start()
allDone = False
while not allDone:
time.sleep( 1 )
for n in xrange( numberOfThreads ):
a[n] = t[n].is_alive()
if not any( a ):
# if no thread is alive
allDone = True
#
# END OF THREADED COMPONENT
#
c.info( "All Threads done!" )
#
# Merging stage
#
c.info( "Merging tracks.." )
outputTracks = []
# now read all the created tempFiles and merge'em to one
for tmpFileNo in xrange( 0, len( f ) ):
tTracks = io.loadTrk( f[tmpFileNo] )
# add them
outputTracks.extend( tTracks[0] )
c.info( "Merging done!" )
io.saveTrk( outputTrkFile, outputTracks, tracksHeader, None, True )
c.info( "All done!" )
