def main():
# Parse input arguments and get sanitised version
sargs = mm_parse_inputs.SanitisedArgs(mm_parse_inputs.parse_input_arguments())
# Enable calling NiftyReg and NiftySeg
nk = mm_niftk.MM_Niftk()
mmfn.check_create_directories([sargs.output_directory])
# Set default parameters
if sargs.number is not None:
iterations = sargs.number
else:
iterations = 200
fwhm = 0.15 # default 0.15
subsampleFactor = 4 # default 4; recommend 2 for in vivo mouse brains with resolution >100 micrometres.
nlevels = 4 # default 4
conv = 0.001 # default 0.001
nhistbins = 256 # default 200
# Get list of files (use file_name_filter)
nifti_files_list = sorted(glob.glob(os.path.join(sargs.input_directory, sargs.file_name_filter + '.nii*')))
print(" Processing {0} files: \n - - -\n \t{1}\n - - -\n ...".format(len(nifti_files_list), '\n\t'.join([ str(item) for item in nifti_files_list ])))
TASKS = [(nk, sargs, nf_path, iterations, fwhm, subsampleFactor, nlevels, conv, nhistbins) for nf_path in nifti_files_list]
# Set the number of parallel processes to use
pool = multiprocessing.Pool(np.int(multiprocessing.cpu_count() / 2))
# The _star function will unpack TASKS to use in the actual function
# Using pool.map because we do care about the order of the results.
all_output_paths = pool.map(nuc_n4_debias_star, TASKS)
pool.close()
pool.join()
all_output_paths = []
for counter, nf_path in enumerate(nifti_files_list):
outpath = nuc_n4_debias(nk, sargs, nf_path, iterations, fwhm, subsampleFactor, nlevels, conv, nhistbins)
all_output_paths.append(outpath)
print(" all_output_paths: {0}".format(all_output_paths))
print(" Copying headers from original images to bias-corrected images ...")
for counter, nifti_file in enumerate(nifti_files_list):
print " Processing {0} / {1}: {2} ...".format((counter + 1), len(nifti_files_list), nifti_file)
original_nf_name = os.path.basename(nifti_file).split(os.extsep)[0]
original_nifti = nib.load(nifti_file)
bias_corrected_path = mmfn.get_corresponding_file(sargs.output_directory, original_nf_name, path_only=True)
print("Bias corrected result path: '{0}'".format(bias_corrected_path))
updated_nii = mmfn.copy_header(nib.load(bias_corrected_path), original_nifti)
# This will overwrite the bias corrected files
nib.save(updated_nii, bias_corrected_path)
print(" Bias field correction completed; files saved to: \n{0}".format('\n\t'.join([ str(item) for item in all_output_paths ])))
def nuc_n4_debias(nk, sargs, nf_path, iterations=200, fwhm=0.15, subsampleFactor=4, nlevels=4, conv=0.001, nhistbins=256):
"""Runs niftkN4BiasFieldCorrection on NIfTI files.
Returns the path of the corrected output.
"""
# Get the input file and name, removing path and 1+ extensions
nf_path = os.path.normpath(os.path.join(nf_path))
nf_name = (os.path.basename(nf_path)).split(os.extsep)[0]
temp_dir = mmfn.create_temp_directories(sargs.output_directory)
outputBiasImage = os.path.normpath(os.path.join(temp_dir, nf_name + '_biasfield' + sargs.ext))
outputSubSampledMask = os.path.normpath(os.path.join(temp_dir, nf_name + '_ss_mask' + sargs.ext))
outputMask = os.path.normpath(os.path.join(temp_dir, nf_name + '_outmask' + sargs.ext))
outputSubSampledImage = os.path.normpath(os.path.join(temp_dir, nf_name + '_ss_img' + sargs.ext))
output_path = os.path.normpath(os.path.join(sargs.output_directory, sargs.out_name_prepend + nf_name + sargs.out_name_append + sargs.ext))
# Run non-uniformity correction ...
n4_calling_list = [nk.niftkN4BiasFieldCorrection,
'--echo',
'-i', nf_path,
'-o', output_path,
'--outBiasField', outputBiasImage,
'--nlevels', str(nlevels),
'--sub', str(subsampleFactor),
'--niters', str(iterations),
'--FWHM', str(fwhm),
'--nbins', str(nhistbins),
'--convergence', str(conv),
'--outSubsampledMask', outputSubSampledMask,
'--outMask', outputMask,
'--outSubsampledImage', outputSubSampledImage]
if sargs.mask_directory is not None:
mask_path = mmfn.get_corresponding_file(sargs.mask_directory, nf_name, sargs.mask_name_filter, path_only=True)
n4_calling_list.extend(['--inMask', mask_path])
print(" Using mask: {0} ...".format(mask_path))
if mask_path is None:
return mask_path
# Check for output
if not os.path.isfile(output_path):
# subprocess.call(n4_calling_list)
mmfn.subprocess_call_and_log(n4_calling_list, temp_dir)
else:
print(" Output file, {0} already exists!".format(output_path))
return output_path
def go(mm):
args = mm.args
# Enable calling NiftyReg and NiftySeg
nk = mm_niftk.MM_Niftk()
# Get list of files
input_files_list = sorted(glob.glob(os.path.join(args.input_directory, args.input_name_filter + '.nii*')))
print(" Processing {0} files: \n - - -\n \t{1}\n - - -\n ...".format(len(input_files_list), '\n\t'.join([ str(item) for item in input_files_list ])))
TASKS = [(nk, args, f_path, args.iterations, args.fwhm, args.subsample, args.nlevels, args.convergence, args.nhistbins) for f_path in args.input_files_list]
if args.parallel:
# all_output_paths = []
# print("[nuc_n4_debias] + TASKS: {0}".format([nuc_n4_debias] + TASKS))
# for task in TASKS:
# all_output_paths.append(nuc_n4_debias(*task))
# mmfn.function_star([nuc_n4_debias] + task)
import multiprocessing
# Set the number of parallel processes to use
pool = multiprocessing.Pool(np.int(multiprocessing.cpu_count()))
# Prepend the function name to the list, so function_star() works
all_output_paths = pool.map(mmfn.function_star, [nuc_n4_debias] + TASKS)
pool.close()
pool.join()
else:
all_output_paths = []
for task in TASKS:
all_output_paths.append(nuc_n4_debias(*task))
print(" All output images: \n{0}".format(all_output_paths))
print(" Copying headers from original images to bias-corrected images ...")
for counter, nifti_file in enumerate(input_files_list):
print " Processing {0} / {1}: {2} ...".format((counter + 1), len(input_files_list), nifti_file)
original_nf_name = os.path.basename(nifti_file).split(os.extsep)[0]
original_nifti = nib.load(nifti_file)
bias_corrected_path = mmfn.get_corresponding_file(args.output_directory, original_nf_name, path_only=True)
updated_nii = mmfn.copy_header(nib.load(bias_corrected_path), original_nifti)
# This will overwrite the bias corrected files
nib.save(updated_nii, bias_corrected_path)
print(" Bias field correction completed; files saved to: \n{0}".format('\n\t'.join([ str(item) for item in all_output_paths ])))
def nuc_n4_debias(nk,
sargs,
nf_path,
iterations=200,
fwhm=0.15,
subsampleFactor=4,
nlevels=4,
conv=0.001,
nhistbins=256):
"""Runs niftkN4BiasFieldCorrection on NIfTI files.
Returns the path of the corrected output.
"""
# Get the input file and name, removing path and 1+ extensions
nf_path = os.path.normpath(os.path.join(nf_path))
nf_name = (os.path.basename(nf_path)).split(os.extsep)[0]
temp_dir = mmfn.create_temp_directories(sargs.output_directory)
outputBiasImage = os.path.normpath(
os.path.join(temp_dir, nf_name + '_biasfield' + sargs.ext))
outputSubSampledMask = os.path.normpath(
os.path.join(temp_dir, nf_name + '_ss_mask' + sargs.ext))
outputMask = os.path.normpath(
os.path.join(temp_dir, nf_name + '_outmask' + sargs.ext))
outputSubSampledImage = os.path.normpath(
os.path.join(temp_dir, nf_name + '_ss_img' + sargs.ext))
output_path = os.path.normpath(
os.path.join(
sargs.output_directory, sargs.out_name_prepend + nf_name +
sargs.out_name_append + sargs.ext))
# Run non-uniformity correction ...
n4_calling_list = [
nk.niftkN4BiasFieldCorrection, '--echo', '-i', nf_path, '-o',
output_path, '--outBiasField', outputBiasImage, '--nlevels',
str(nlevels), '--sub',
str(subsampleFactor), '--niters',
str(iterations), '--FWHM',
str(fwhm), '--nbins',
str(nhistbins), '--convergence',
str(conv), '--outSubsampledMask', outputSubSampledMask, '--outMask',
outputMask, '--outSubsampledImage', outputSubSampledImage
]
if sargs.mask_directory is not None:
mask_path = mmfn.get_corresponding_file(sargs.mask_directory,
nf_name,
sargs.mask_name_filter,
path_only=True)
n4_calling_list.extend(['--inMask', mask_path])
print(" Using mask: {0} ...".format(mask_path))
if mask_path is None:
return mask_path
# Check for output
if not os.path.isfile(output_path):
# subprocess.call(n4_calling_list)
mmfn.subprocess_call_and_log(n4_calling_list, temp_dir)
else:
print(" Output file, {0} already exists!".format(output_path))
return output_path
def go(mm):
args = mm.args
# Enable calling NiftyReg and NiftySeg
nk = mm_niftk.MM_Niftk()
# Get list of files
input_files_list = sorted(
glob.glob(
os.path.join(args.input_directory,
args.input_name_filter + '.nii*')))
print(
" Processing {0} files: \n - - -\n \t{1}\n - - -\n ...".format(
len(input_files_list),
'\n\t'.join([str(item) for item in input_files_list])))
TASKS = [(nk, args, f_path, args.iterations, args.fwhm, args.subsample,
args.nlevels, args.convergence, args.nhistbins)
for f_path in args.input_files_list]
if args.parallel:
# all_output_paths = []
# print("[nuc_n4_debias] + TASKS: {0}".format([nuc_n4_debias] + TASKS))
# for task in TASKS:
# all_output_paths.append(nuc_n4_debias(*task))
# mmfn.function_star([nuc_n4_debias] + task)
import multiprocessing
# Set the number of parallel processes to use
pool = multiprocessing.Pool(np.int(multiprocessing.cpu_count()))
# Prepend the function name to the list, so function_star() works
all_output_paths = pool.map(mmfn.function_star,
[nuc_n4_debias] + TASKS)
pool.close()
pool.join()
else:
all_output_paths = []
for task in TASKS:
all_output_paths.append(nuc_n4_debias(*task))
print(" All output images: \n{0}".format(all_output_paths))
print(
" Copying headers from original images to bias-corrected images ..."
)
for counter, nifti_file in enumerate(input_files_list):
print " Processing {0} / {1}: {2} ...".format((counter + 1),
len(input_files_list),
nifti_file)
original_nf_name = os.path.basename(nifti_file).split(os.extsep)[0]
original_nifti = nib.load(nifti_file)
bias_corrected_path = mmfn.get_corresponding_file(
args.output_directory, original_nf_name, path_only=True)
updated_nii = mmfn.copy_header(nib.load(bias_corrected_path),
original_nifti)
# This will overwrite the bias corrected files
nib.save(updated_nii, bias_corrected_path)
print(" Bias field correction completed; files saved to: \n{0}".format(
'\n\t'.join([str(item) for item in all_output_paths])))
def main():
# Parse input arguments and get sanitised version
sargs = mm_parse_inputs.SanitisedArgs(
mm_parse_inputs.parse_input_arguments())
# Enable calling NiftyReg and NiftySeg
nk = mm_niftk.MM_Niftk()
mmfn.check_create_directories([sargs.output_directory])
# Set default parameters
if sargs.number is not None:
iterations = sargs.number
else:
iterations = 200
fwhm = 0.15 # default 0.15
subsampleFactor = 4 # default 4; recommend 2 for in vivo mouse brains with resolution >100 micrometres.
nlevels = 4 # default 4
conv = 0.001 # default 0.001
nhistbins = 256 # default 200
# Get list of files (use file_name_filter)
nifti_files_list = sorted(
glob.glob(
os.path.join(sargs.input_directory,
sargs.file_name_filter + '.nii*')))
print(
" Processing {0} files: \n - - -\n \t{1}\n - - -\n ...".format(
len(nifti_files_list),
'\n\t'.join([str(item) for item in nifti_files_list])))
TASKS = [(nk, sargs, nf_path, iterations, fwhm, subsampleFactor, nlevels,
conv, nhistbins) for nf_path in nifti_files_list]
# Set the number of parallel processes to use
pool = multiprocessing.Pool(np.int(multiprocessing.cpu_count() / 2))
# The _star function will unpack TASKS to use in the actual function
# Using pool.map because we do care about the order of the results.
all_output_paths = pool.map(nuc_n4_debias_star, TASKS)
pool.close()
pool.join()
all_output_paths = []
for counter, nf_path in enumerate(nifti_files_list):
outpath = nuc_n4_debias(nk, sargs, nf_path, iterations, fwhm,
subsampleFactor, nlevels, conv, nhistbins)
all_output_paths.append(outpath)
print(" all_output_paths: {0}".format(all_output_paths))
print(
" Copying headers from original images to bias-corrected images ...")
for counter, nifti_file in enumerate(nifti_files_list):
print " Processing {0} / {1}: {2} ...".format((counter + 1),
len(nifti_files_list),
nifti_file)
original_nf_name = os.path.basename(nifti_file).split(os.extsep)[0]
original_nifti = nib.load(nifti_file)
bias_corrected_path = mmfn.get_corresponding_file(
sargs.output_directory, original_nf_name, path_only=True)
print("Bias corrected result path: '{0}'".format(bias_corrected_path))
updated_nii = mmfn.copy_header(nib.load(bias_corrected_path),
original_nifti)
# This will overwrite the bias corrected files
nib.save(updated_nii, bias_corrected_path)
print(" Bias field correction completed; files saved to: \n{0}".format(
'\n\t'.join([str(item) for item in all_output_paths])))
