def write_scantsv(bids_dir, dicom_dir=None, live=False):
"""Make subject scan files (needed for NDA submission)"""
# TODO: improved with metadata
if not os.path.exists(dicom_dir):
logging.warning('Specify valid dicom directory with [-d] flag')
return
layout = BIDSLayout(bids_dir)
subs = sorted([x for x in layout.get_subjects()])
for sid in subs:
dcm = read_file(glob(op(dicom_dir, '*' + sid, '*'))[-1],
force=True).AcquisitionDate
date = '-'.join([dcm[:4], dcm[4:6], dcm[6:]])
logging.info("{0}'s scan date: {1}".format(sid, date))
scans = []
for scan in [
f.filename
for f in layout.get(subject=sid, extensions=['nii', 'nii.gz'])
]:
paths = scan.split(os.sep)
scans.append(os.sep.join(paths[-2:]))
outname = op(bids_dir, paths[-3], paths[-3] + '_scans.tsv')
if live:
with open(outname, 'wt') as tsvfile:
writer = csv.writer(tsvfile, delimiter='\t')
writer.writerow(['filename', 'acq_time'])
for scan in sorted(scans):
writer.writerow([scan, date])
logging.info('Wrote {0}'.format(outname))
def read_data(test=False):
""" Read train and test data
"""
df_train = pd.read_csv(op(PATH_RAW, 'train.csv'), sep=';')
df_test = pd.read_csv(op(PATH_RAW, 'test.csv'), sep=';')
if test:
return df_train, df_test
else:
return df_train
def split_geojson(gdf, outdir, groupby_col='type_id', remaps=None):
if remaps == None:
for name, group in gdf.groupby([groupby_col]):
group.to_file(op(outdir, "{}.geojson".format(name)),
driver="GeoJSON")
else:
gdf[groupby_col] = gdf[groupby_col].map(remaps).fillna(
gdf[groupby_col])
for name, group in gdf.groupby([groupby_col]):
group.to_file(op(outdir, "{}.geojson".format(name)),
driver="GeoJSON")
def add_sub(data_dir, subjpre, live=False):
"""Add BIDS sub- prefix to subjects converted with heudiconv"""
if not subjpre:
sys.exit('Specify subject prefix with [-p] flag')
subjs = sorted(
[x for x in os.listdir(data_dir) if subjpre in x and 'sub-' not in x])
for subj in subjs:
old = op(data_dir, subj)
new = op(data_dir, 'sub-' + subj)
logging.info(msg.format(old, new))
if live:
os.rename(old, new)
return subjs
def image_intercept(bbox_geojson,
image_folder,
out_folder,
image_id_field,
create_subset=False):
gdf = gpd.read_file(bbox_geojson)
unique = gdf[image_id_field].unique()
if create_subset == True:
for i in image_id_field:
src = op(image_folder, i)
dst = op(subset_folder, i)
copyfile(src, dst)
with open(op(out_folder,
sp(base(bbox_geojson))[0] + "_images.csv"), 'w') as f:
writer = csv.writer(f)
writer.writerows(zip(unique))
def get_fasttext():
""" Load fasttext french pretrained model
https://fasttext.cc/docs/en/pretrained-vectors.html
"""
filename = op(PATH_EXTERNAL, 'wiki.fr.bin')
model = KeyedVectors.load_word2vec_format(filename, binary=True)
return model
def drop_underscore(data_dir, live=False):
"""Remove underscore from subject id"""
# Change directories first, then files
subjs = sorted([x for x in os.listdir(data_dir) if x.startswith('sub-')])
for subj in subjs:
if subj.count('_') == 0:
continue
corr = subj.replace('_', '')
logging.info(msg.format(op(data_dir, subj), op(data_dir, corr)))
if not live:
return
os.rename(op(data_dir, subj), op(data_dir, corr))
# refresh after each rename
layout = BIDSLayout(data_dir)
files = [f.filename for f in layout.get() if subj in f.filename]
for file in files:
fix = file.replace(subj, corr)
os.rename(file, fix)
def get_stopwords():
""" Get french stopwords
"""
with open(op(PATH_EXTERNAL, 'fr-stopwords.txt')) as fp:
stopwords = fp.read().splitlines()
return stopwords
def main():
class MyParser(argparse.ArgumentParser):
# to run from commandline
def error(self, message):
sys.stderr.write('error: %s\n' % message)
self.print_help()
sys.exit(2)
defstr = ' (default %(default)s)'
parser = argparse.ArgumentParser(prog='makebids.py', description=__doc__)
parser.add_argument('datadir', help='''bids-like directory''')
parser.add_argument('-p',
dest='pre',
type=str,
help='''identifier across all subjects''')
parser.add_argument('-d',
'--dicoms',
type=str,
default=None,
help="""dicom directory""")
parser.add_argument('--live',
default=False,
action='store_true',
help="""WARNING: DON'T INCLUDE ON FIRST PASS""")
parser.add_argument('--full',
action='store_true',
default=False,
help="""Run through each option""")
parser.add_argument('-q',
'--quiet',
action='store_true',
default=False,
help="""Make the python logger only log errors""")
args = parser.parse_args()
bids_dir = os.path.abspath(args.datadir)
if not os.path.exists(bids_dir):
sys.exit('Specify valid BIDS data directory')
if args.dicoms:
dicom_dir = os.path.abspath(args.dicoms)
else:
dicom_dir = None
loglevel = logging.INFO
if args.quiet:
loglevel = logging.WARNING
# Set logging output
logging.basicConfig(filename=op(os.getcwd(), 'mbOUT.txt'),
format='%(asctime)s %(levelname)s:%(message)s',
level=loglevel)
# when files are renamed
refresh = lambda x=bids_dir: BIDSLayout(x)
if args.full:
add_sub(bids_dir, args.pre, args.live)
drop_underscore(bids_dir, args.live)
# using BIDS grabbids after renaming files
if dicom_dir:
write_scantsv(bids_dir, dicom_dir, args.live)
# set layout once no more file renamings
add_taskname(refresh(), args.live)
fix_fieldmaps(refresh(), args.live)
else:
choice = int(input(OPTIONS))
if choice == 1:
add_sub(bids_dir, args.pre, args.live)
elif choice == 2:
drop_underscore(bids_dir, args.live)
elif choice == 3:
write_scantsv(bids_dir, dicom_dir, args.live)
elif choice == 4:
add_taskname(refresh(), args.live)
elif choice == 5:
fix_fieldmaps(refresh(), args.live)
else:
sys.exit('Option not recognized')