def test_append_file(self):
# Create test data
data = OrderedDict()
data['FileName'] = 'filename.txt'
data['A'] = 2
data['B'] = 'b'
# First write
write_csv(data, self.csv_file_name)
self.assertTrue(os.path.exists(self.csv_file_name),
'Could not create file')
# Second write
write_csv(data, self.csv_file_name)
# Create the expected file
correct_file_name = os.path.join(self.test_dir, 'test_correct.csv')
with open(correct_file_name, 'w') as f:
f.write('FileName,A,B' + os.linesep)
f.write('filename.txt,2,b' + os.linesep)
f.write('filename.txt,2,b' + os.linesep)
# Test
self.assertTrue(filecmp.cmp(correct_file_name, self.csv_file_name),
'Files are not the same')
def test_special_characters(self):
# Create test data
data = OrderedDict()
data['Spacing.X [mm]'] = 3
data['C^asdf'] = 'Hello'
# Create the expected file
correct_file_name = os.path.join(self.test_dir, 'test_correct.csv')
with open(correct_file_name, 'w') as f:
f.write('Spacing.X [mm],C^asdf' + os.linesep)
f.write('3,Hello' + os.linesep)
# Print result
write_csv(data, self.csv_file_name)
# Test
self.assertTrue(os.path.exists(self.csv_file_name),
'Could not create file')
self.assertTrue(filecmp.cmp(correct_file_name, self.csv_file_name),
'Files are not the same')
def test_delimiter(self):
# Create test data
data = OrderedDict()
data['FileName'] = 'filename.txt'
data['A'] = 2
data['B'] = 'b'
# Create the expected file
correct_file_name = os.path.join(self.test_dir, 'test_correct.csv')
with open(correct_file_name, 'w') as f:
f.write('FileName;A;B' + os.linesep)
f.write('filename.txt;2;b' + os.linesep)
# Print result
write_csv(data, self.csv_file_name, delimiter=';')
# Test
self.assertTrue(os.path.exists(self.csv_file_name),
'Could not create file')
self.assertTrue(filecmp.cmp(correct_file_name, self.csv_file_name),
'Files are not the same')
def Muscle(input_filename, converted_filename, segmentation_filename, bone_threshold, smoothing_iterations, segmentation_iterations, segmentation_multiplier, initial_neighborhood_radius, closing_radius, csv_filename='', tiff_filename='', histogram_filename=''):
# Python 2/3 compatible input
from six.moves import input
# Input must be an AIM
if not input_filename.lower().endswith('.aim'):
os.sys.exit('[ERROR] Input \"{}\" must be an AIM'.format(input_filename))
# Read input
if not os.path.isfile(input_filename):
os.sys.exit('[ERROR] Cannot find file \"{}\"'.format(input_filename))
# Internal constants
bone_label = 1
muscle_label = 2
# Compute calibration constants
print('Computing calibration constants')
reader = vtkbone.vtkboneAIMReader()
reader.SetFileName(input_filename)
reader.DataOnCellsOff()
reader.Update()
m,b = get_aim_density_equation(reader.GetProcessingLog())
del reader
print(' m: {}'.format(m))
print(' b: {}'.format(b))
print('')
# Converting image
print('Converting {} to {}'.format(input_filename, converted_filename))
ImageConverter(input_filename, converted_filename, overwrite=True)
print('')
print('Reading in converted image')
image = sitk.ReadImage(converted_filename)
# Segment bone
print('Segmenting bone')
seg_bone = segment_bone(image, (bone_threshold - b)/m)
seg_bone = (seg_bone>0)*bone_label
print('')
# Find centroid
print('Finding centroid of the two largest bones')
stat_filter = sitk.LabelShapeStatisticsImageFilter()
stat_filter.Execute(sitk.RelabelComponent(sitk.ConnectedComponent(seg_bone)))
centroids = [
stat_filter.GetCentroid(1),
stat_filter.GetCentroid(2)
]
seed = [0 for i in range(len(centroids[0]))]
for i in range(len(seed)):
seed[i] = 0.5*(centroids[0][i] + centroids[1][i])
seed_index = image.TransformPhysicalPointToIndex(seed)
print(' Centroid1: {}'.format(centroids[0]))
print(' Centroid2: {}'.format(centroids[1]))
print(' Seed (physical): {}'.format(seed))
print(' Seed (index): {}'.format(seed_index))
print('')
# Smooth image
print('Performing anisotropic smoothing')
timeStep = image.GetSpacing()[0] / 2.0**4
smooth_image = sitk.GradientAnisotropicDiffusion(
sitk.Cast(image, sitk.sitkFloat32),
timeStep=timeStep,
numberOfIterations=smoothing_iterations
)
print('')
# Segment muscle
print('Segmenting muscle')
radius = int(max(1, initial_neighborhood_radius/image.GetSpacing()[0]))
print(' initialNeighborhoodRadius [vox]: {}'.format(radius))
seg_muscle = sitk.ConfidenceConnected(
smooth_image,
seedList=[seed_index],
numberOfIterations=segmentation_iterations,
multiplier=segmentation_multiplier,
initialNeighborhoodRadius=radius,
replaceValue=1
)
# Take largest component
seg_muscle = (sitk.RelabelComponent(sitk.ConnectedComponent(seg_muscle>0))==1)*muscle_label
print('')
# One, solid peice of background
print('Cleaning up segmentation')
vector_radius = [int(max(1, closing_radius//s)) for s in image.GetSpacing()]
print(' Closing radius [mm]: {}'.format(closing_radius))
print(' Vector radius [voxels]: {}'.format(vector_radius))
seg = (seg_bone+seg_muscle)>0
seg = sitk.BinaryDilate(seg, vector_radius)
background = sitk.RelabelComponent(sitk.ConnectedComponent(seg<1))==1
seg_muscle = sitk.BinaryErode(background<1, vector_radius)*muscle_label
print('')
# Join segmentation
seg_muscle = sitk.Mask(seg_muscle, 1-(seg_bone>0))
seg = seg_bone + seg_muscle
seg = sitk.Cast(seg, sitk.sitkInt8)
# Write segmentation
print('Writing segmentation to ' + segmentation_filename)
sitk.WriteImage(seg, segmentation_filename)
print('')
print('Performing quantification')
# Quantify Density
intensity_filter = sitk.LabelIntensityStatisticsImageFilter()
intensity_filter.Execute(seg, image)
muscle_density = intensity_filter.GetMean(muscle_label)
# Quantify cross sectional area
# Note that since
stat_filter = sitk.LabelShapeStatisticsImageFilter()
stat_filter.Execute(seg)
ave_cross_area = float(stat_filter.GetNumberOfPixels(muscle_label)) / seg.GetSize()[2]
print(' density: {}'.format(muscle_density))
print(' cross section: {}'.format(ave_cross_area))
print('')
# Write results
entry = OrderedDict()
entry['Filename'] = input_filename
entry['Spacing.X [mm]'] = image.GetSpacing()[0]
entry['Spacing.Y [mm]'] = image.GetSpacing()[1]
entry['Spacing.Z [mm]'] = image.GetSpacing()[2]
entry['density_slope'] = m
entry['density_intercept'] = b
entry['muscle density [native]'] = muscle_density
entry['A.Cross [vox^2]'] = ave_cross_area
print(echo_arguments('Muscle Outcomes:', entry))
# Write CSV
if len(csv_filename)>0:
print(' Writing to csv file ' + csv_filename)
write_csv(entry, csv_filename)
# Write TIFF
if len(tiff_filename)>0:
overlay = sitk.LabelOverlay(
sitk.Cast(sitk.IntensityWindowing(
sitk.Cast(image, sitk.sitkFloat32),
windowMinimum = 0,
windowMaximum = (bone_threshold - b)/m,
outputMinimum = 0.0,
outputMaximum = 255.0
), sitk.sitkUInt8),
seg,
opacity=0.3
)
size = list(overlay.GetSize())
index = [0, 0, int(size[2]//2)]
size[2]=0
#Step 5: Extract that specific slice using the Extract Image Filter
tiff_image = sitk.Extract(
overlay,
size=size,
index=index
)
print(' Save single slice to ' + tiff_filename)
sitk.WriteImage(tiff_image, tiff_filename)
# Create histogram
if len(histogram_filename)>0:
import matplotlib.pyplot as plt
print('Saving histogram to ' + histogram_filename)
data = m*sitk.GetArrayFromImage(image)+b
mask = sitk.GetArrayFromImage(seg)
data = data.ravel()
mask = mask.ravel()
plt.figure(figsize=(8, 6))
plt.hist(data[mask==muscle_label], bins=1000, density=True)
plt.xlabel('Density [mg HA/ccm]')
plt.ylabel('Normalized Count')
plt.pause(0.1)
plt.savefig(histogram_filename)