def main(args=None):
if args is None:
args = sys.argv[1:]
# create param objects
param_seg = ParamSeg()
param_data = ParamData()
param_model = ParamModel()
param = Param()
# get parser
parser = get_parser()
arguments = parser.parse(args)
# set param arguments ad inputted by user
param_seg.fname_im = arguments["-i"]
param_seg.fname_im_original = arguments["-i"]
param_seg.fname_seg = arguments["-s"]
if '-vertfile' in arguments:
if extract_fname(arguments['-vertfile'])[1].lower() == "none":
param_seg.fname_level = None
elif os.path.isfile(arguments['-vertfile']):
param_seg.fname_level = arguments['-vertfile']
else:
param_seg.fname_level = None
printv('WARNING: -vertfile input file: "' + arguments['-vertfile'] + '" does not exist.\nSegmenting GM without using vertebral information', 1, 'warning')
if '-denoising' in arguments:
param_data.denoising = bool(int(arguments['-denoising']))
if '-normalization' in arguments:
param_data.normalization = bool(int(arguments['-normalization']))
if '-p' in arguments:
param_data.register_param = arguments['-p']
if '-w-levels' in arguments:
param_seg.weight_level = arguments['-w-levels']
if '-w-coordi' in arguments:
param_seg.weight_coord = arguments['-w-coordi']
if '-thr-sim' in arguments:
param_seg.thr_similarity = arguments['-thr-sim']
if '-model' in arguments:
param_model.path_model_to_load = os.path.abspath(arguments['-model'])
if '-res-type' in arguments:
param_seg.type_seg = arguments['-res-type']
if '-ref' in arguments:
param_seg.fname_manual_gmseg = arguments['-ref']
if '-ofolder' in arguments:
param_seg.path_results = os.path.abspath(arguments['-ofolder'])
param_seg.qc = arguments.get("-qc", None)
if '-r' in arguments:
param.rm_tmp = bool(int(arguments['-r']))
param.verbose = int(arguments.get('-v'))
sct.init_sct(log_level=param.verbose, update=True) # Update log level
start_time = time.time()
seg_gm = SegmentGM(param_seg=param_seg, param_data=param_data, param_model=param_model, param=param)
seg_gm.segment()
elapsed_time = time.time() - start_time
printv('\nFinished! Elapsed time: ' + str(int(np.round(elapsed_time))) + 's', param.verbose)
# save quality control and sct.printv(info)
if param_seg.type_seg == 'bin':
wm_col = 'red'
gm_col = 'blue'
b = '0,1'
else:
wm_col = 'blue-lightblue'
gm_col = 'red-yellow'
b = '0.4,1'
if param_seg.qc is not None:
generate_qc(param_seg.fname_im_original, seg_gm.fname_res_gmseg,
seg_gm.fname_res_wmseg, param_seg, args, os.path.abspath(param_seg.qc))
if param.rm_tmp:
# remove tmp_dir
sct.rmtree(seg_gm.tmp_dir)
sct.display_viewer_syntax([param_seg.fname_im_original, seg_gm.fname_res_gmseg, seg_gm.fname_res_wmseg], colormaps=['gray', gm_col, wm_col], minmax=['', b, b], opacities=['1', '0.7', '0.7'], verbose=param.verbose)
def get_parser():
# Initialize the parser
parser = Parser(__file__)
parser.usage.set_description('Segmentation of the white and gray matter.'
' The segmentation is based on a multi-atlas method that uses a dictionary of pre-segmented gray matter images (already included in SCT) and finds the most similar images for identifying the gray matter using label fusion approach. The model used by this method contains: a template of the white/gray matter segmentation along the cervical spinal cord, and a PCA reduced space to describe the variability of intensity in that template.'
' This method was inspired from [Asman et al., Medical Image Analysis 2014] and features the following additions:\n'
'- possibility to add information from vertebral levels for improved accuracy\n'
'- intensity normalization of the image to segment (allows the segmentation of any kind of contrast)\n'
'- pre-registration based on non-linear transformations')
parser.add_option(name="-i",
type_value="file",
description="Image to segment",
mandatory=True,
example='t2star.nii.gz')
parser.add_option(name="-s",
type_value="file",
description="Spinal cord segmentation",
mandatory=True,
example='sc_seg.nii.gz')
parser.add_option(name="-vertfile",
type_value="str",
description='Labels of vertebral levels. This could either be an image (e.g., label/template/PAM50_levels.nii.gz) or a text file that specifies "slice,level" at each line. Example:\n'
"0,3\n"
"1,3\n"
"2,4\n"
"3,4\n"
"4,4\n",
mandatory=False,
default_value=ParamSeg().fname_level)
parser.add_option(name="-vert",
mandatory=False,
deprecated_by='-vertfile')
parser.add_option(name="-l",
mandatory=False,
deprecated_by='-vertfile')
parser.usage.addSection('SEGMENTATION OPTIONS')
parser.add_option(name="-denoising",
type_value='multiple_choice',
description="1: Adaptative denoising from F. Coupe algorithm, 0: no WARNING: It affects the model you should use (if denoising is applied to the target, the model should have been computed with denoising too)",
mandatory=False,
default_value=int(ParamData().denoising),
example=['0', '1'])
parser.add_option(name="-normalization",
type_value='multiple_choice',
description="Normalization of the target image's intensity using median intensity values of the WM and the GM, recomended with MT images or other types of contrast than T2*",
mandatory=False,
default_value=int(ParamData().normalization),
example=['0', '1'])
parser.add_option(name="-p",
type_value='str',
description="Registration parameters to register the image to segment on the model data. Use the same format as for sct_register_to_template and sct_register_multimodal.",
mandatory=False,
default_value=ParamData().register_param,
example='step=1,type=seg,algo=centermassrot,metric=MeanSquares,smooth=2,iter=1:step=2,type=seg,algo=columnwise,metric=MeanSquares,smooth=3,iter=1:step=3,type=seg,algo=bsplinesyn,metric=MeanSquares,iter=3')
parser.add_option(name="-w-levels",
type_value='float',
description="Weight parameter on the level differences to compute the similarities",
mandatory=False,
default_value=ParamSeg().weight_level,
example=2.0)
parser.add_option(name="-w-coordi",
type_value='float',
description="Weight parameter on the euclidean distance (based on images coordinates in the reduced sapce) to compute the similarities ",
mandatory=False,
default_value=ParamSeg().weight_coord,
example=0.005)
parser.add_option(name="-thr-sim",
type_value='float',
description="Threshold to select the dictionary slices most similar to the slice to segment (similarities are normalized to 1)",
mandatory=False,
default_value=ParamSeg().thr_similarity,
example=0.6)
parser.add_option(name="-model",
type_value="folder",
description="Path to the computed model",
mandatory=False,
example='/home/jdoe/gm_seg_model/')
parser.usage.addSection('\nOUTPUT OTIONS')
parser.add_option(name="-res-type",
type_value='multiple_choice',
description="Type of result segmentation : binary or probabilistic",
mandatory=False,
default_value=ParamSeg().type_seg,
example=['bin', 'prob'])
parser.add_option(name="-ratio",
type_value='multiple_choice',
description="Compute GM/WM CSA ratio by slice or by vertebral level (average across levels)",
mandatory=False,
default_value=ParamSeg().ratio,
example=['0', 'slice', 'level'])
parser.add_option(name="-ref",
type_value="file",
description="Reference segmentation of the gray matter for segmentation validation --> output Dice coefficient and Hausdorff's and median distances)",
mandatory=False,
example='manual_gm_seg.nii.gz')
parser.add_option(name="-ofolder",
type_value="folder_creation",
description="Output folder",
mandatory=False,
default_value=ParamSeg().path_results,
example='gm_segmentation_results/')
parser.usage.addSection('MISC')
parser.add_option(name='-qc',
type_value='multiple_choice',
description='Output images for quality control.',
mandatory=False,
example=['0', '1'],
default_value=str(int(ParamSeg().qc)))
parser.add_option(name="-r",
type_value="multiple_choice",
description='Remove temporary files.',
mandatory=False,
default_value=str(int(Param().rm_tmp)),
example=['0', '1'])
parser.add_option(name="-v",
type_value='multiple_choice',
description="Verbose: 0 = nothing, 1 = classic, 2 = expended",
mandatory=False,
example=['0', '1', '2'],
default_value=str(Param().verbose))
return parser
def main(args=None):
if args is None:
args = sys.argv[1:]
# create param objects
param_seg = ParamSeg()
param_data = ParamData()
param_model = ParamModel()
param = Param()
# get parser
parser = get_parser()
arguments = parser.parse(args)
# set param arguments ad inputted by user
param_seg.fname_im = arguments["-i"]
param_seg.fname_im_original = arguments["-i"]
param_seg.fname_seg = arguments["-s"]
if '-vertfile' in arguments:
if sct.extract_fname(arguments['-vertfile'])[1].lower() == "none":
param_seg.fname_level = None
elif os.path.isfile(arguments['-vertfile']):
param_seg.fname_level = arguments['-vertfile']
else:
sct.printv(parser.usage.generate(error='ERROR: -vertfile input file: "'+arguments['-vertfile']+'" does not exist.'))
if '-denoising' in arguments:
param_data.denoising = bool(int(arguments['-denoising']))
if '-normalization' in arguments:
param_data.normalization = arguments['-normalization']
if '-p' in arguments:
param_data.register_param = arguments['-p']
if '-w-levels' in arguments:
param_seg.weight_level = arguments['-w-levels']
if '-w-coordi' in arguments:
param_seg.weight_coord = arguments['-w-coordi']
if '-thr-sim' in arguments:
param_seg.thr_similarity = arguments['-thr-sim']
if '-model' in arguments:
param_model.path_model_to_load = os.path.abspath(arguments['-model'])
if '-res-type' in arguments:
param_seg.type_seg= arguments['-res-type']
if '-ratio' in arguments:
param_seg.ratio = arguments['-ratio']
if '-ref' in arguments:
param_seg.fname_manual_gmseg = arguments['-ref']
if '-ofolder' in arguments:
param_seg.path_results= arguments['-ofolder']
if '-qc' in arguments:
param_seg.qc = bool(int(arguments['-qc']))
if '-r' in arguments:
param.rm_tmp= bool(int(arguments['-r']))
if '-v' in arguments:
param.verbose= arguments['-v']
seg_gm = SegmentGM(param_seg=param_seg, param_data=param_data, param_model=param_model, param=param)
start = time.time()
seg_gm.segment()
end = time.time()
t = end - start
printv('Done in ' + str(int(round(t / 60))) + ' min, ' + str(round(t % 60,1)) + ' sec', param.verbose, 'info')