parser = argparse.ArgumentParser(description='Saves a file with feature sets extracted from NIFTI files. The format of this file can be selected to be used in different software packages, including Numpy binary format, Weka, Octave/Matlab and SVMPerf.')

parser.add_argument('-s', '--subjsf', dest='subjs', required=True, help='list file with the subjects for the analysis. Each line: <class_label>,<subject_file>')

parser.add_argument('-o', '--outdir', dest='outdir', required=True,

help='''name of the output directory where the results will be saved. \n

In this directory the following files will be created:

- included_subjects: list of full path to the subjects included in the feature set.

- excluded_subjects: list of full path to the subjects excluded from the feature set. if any.

- included_subjlabels: list of class labels of each subject in included_subjects.

- excluded_subjlabels: list of class labels of each subject in excluded_subjects, if any.

- features.*: file containing a NxM matrix with the features extracted from subjects (N: subj number, M: feat number).

''')

parser.add_argument('-m', '--mask', dest='mask', required=True,

help='Mask file to extract feature voxels, any voxel with values > 0 will be included in the extraction.')

parser.add_argument('-d', '--datadir', dest='datadir', required=False,

help='folder path where the subjects are, if the absolute path is not included in the subjects list file.', default='')

parser.add_argument('-p', '--prefix', dest='prefix', default='', required=False,

help='Prefix for the output filenames.')

parser.add_argument('-e', '--exclude', dest='exclude', default='', required=False,

help='subject list mask, i.e., text file where each line has 0 or 1 indicating with 1 which subject should be excluded in the measure. To help calculating measures for cross-validation folds.')

parser.add_argument('-t', '--type', dest='type', default='numpybin', choices=['numpybin','octave','arff', 'svmperf'], required=False,

help='type of the output file. Alloweds: numpybin (Numpy binary file), octave (Octave/Matlab binary file using Scipy.io.savemat), arff (Weka text file), svmperfdat (.dat for SVMPerf).')

parser.add_argument('-n', '--name', dest='dataname', default='aizkolari_extracted', required=False,

help='Name of the dataset. It is used for internal usage in SVMPerf and Weka.')

parser.add_argument('-k', '--scale', dest='scale', default=False, action='store_true', required=False,

help='This option will enable Range scaling of the non-excluded data and save a .range file with the max and min of the scaled dataset to scale other dataset with the same transformation.')

parser.add_argument('--scale_min', dest='scale_min', default=-1, type=int, required=False, help='Minimum value for the new scale range.')

parser.add_argument('--scale_max', dest='scale_max', default= 1, type=int, required=False, help='Maximum value for the new scale range.')

parser.add_argument('-r', '--thrP', dest='thresholdP', default='', required=False,

help='use following percentage (0-100) of ROBUST RANGE to threshold mask image (zero anything below the number). One or quoted list of floats separated by blank space.')

parser.add_argument('-b', '--thr', dest='lthreshold', default='', required=False,

help='use following number to threshold mask image (zero anything below the number).')

parser.add_argument('-u', '--uthr', dest='uthreshold', default='', required=False,

help='use following number to upper-threshold mask image (zero anything above the number).')

parser.add_argument('-a', '--abs', dest='absolute', action='store_true', required=False,

help='use absolute value of mask before thresholding.')

parser.add_argument('-l', '--leave', dest='leave', default=-1, required=False, type=int, help='index from subject list (counting from 0) indicating one subject to be left out of the training set. For leave-one-out measures.')

parser.add_argument('-v', '--verbosity', dest='verbosity', required=False, type=int, default=2, help='Verbosity level: Integer where 0 for Errors, 1 for Progression reports, 2 for Debug reports')

if otype == 'numpybin':

ext = au.numpyio_ext()

elif otype == 'octave':

ext = au.octaveio_ext()

elif otype == 'svmperf':

ext = au.svmperfio_ext()

elif otype == 'arff':

ext = au.wekaio_ext()

else:

err = 'get_out_extension: Extension type not supported: ' + otype

raise Exception(err)

filename = outdir + os.path.sep + filename

try:

filename += get_out_extension(otype)

except Exception, err:

au.log.error (str(err))

sys.exit(-1)

if np.isnan(data_min):

dmin = float(data.min())

else:

dmin = float(data_min)

if np.isnan(data_max):

dmax = float(data.max())

else:

dmax = float(data_max)

try:

factor = float(((range_max-range_min)/(dmax-dmin)) + ((range_min*dmax-range_max*dmin)/(dmax-dmin)))

d = data*factor

except Exception, err:

au.log.error (str(err))

sys.exit(-1)

f = open (fname, 'w')

f.write('#data_min, data_max, range_min, range_max')

f.write('\n')

f.write(str(dmin) + ',' + str(dmax) + ',' + str(scale_min) + ',' + str(scale_max))

#setting output file name

ofname = au.feats_str()

if leave > -1:

ofname += '.' + au.excluded_str() + str(leave)

if absolute: ofname += '.' + au.abs_str()

if lthr: ofname += '.lthr_' + str(lthr)

if uthr: ofname += '.uthr_' + str(uthr)

if thrp: ofname += '.thrP_' + str(thrp)

if scale: ofname += '.' + au.scaled_str()

if excluding:

excl_ofname = au.excluded_str() + '_' + ofname

exclfilename = get_filepath (outdir, excl_ofname , otype)

if prefix:

ofname = prefix + '_' + ofname

excl_ofname = prefix + '_' + excl_ofname

filename = get_filepath (outdir, ofname, otype)

#writing in a text file the scaling values of this training set

if scale:

write_scalingrange_file (outdir + os.path.sep + ofname + '.scaling_range', dmin, dmax, scale_min, scale_max)

#saving binary file depending on output type

if otype == 'numpybin':

np.save (filename, feats)

if excluding:

np.save (exclfilename, exclfeats)

elif otype == 'octave':

sio.savemat (filename, {au.feats_str(): feats, au.labels_str(): labels})

if excluding:

exclulabels[exclulabels == 0] = -1

sio.savemat (exclfilename, {au.feats_str(): exclfeats, au.labels_str(): exclulabels})

elif otype == 'svmperf':

labels[labels == 0] = -1

ae.write_svmperf_dat(filename, dataname, feats, labels)

if excluding:

exclulabels[exclulabels == 0] = -1

ae.write_svmperf_dat(exclfilename, dataname, exclfeats, exclulabels)

elif otype == 'arff':

featnames = np.arange(nfeats) + 1

ae.write_arff (filename, dataname, featnames, feats, labels)

if excluding:

ae.write_arff (exclfilename, dataname, featnames, exclfeats, exclulabels)

else:

err = 'Output method not recognised!'

au.log.error(err)

sys.exit(-1)

#population features

nsubjs = len(subjs)

s = nib.load(subjs[0])

subjsiz = np.prod (s.shape)

stype = s.get_data_dtype()

#loading subject data

data = np.empty([nsubjs, subjsiz], dtype=stype)

#number of voxels > 0 in mask

mask = mask.flatten()

nfeats = np.sum(mask > 0)

#reading each subject and saving the features in a vector

feats = np.empty([nsubjs, nfeats], dtype=stype)

#extracting features from non-excluded subjects

c = 0

for s in subjs:

au.log.debug("Reading " + s)

#check geometries

au.check_has_same_geometry (s, maskf)

#load subject

subj = nib.load(s).get_data().flatten()

#mask data and save it

feats[c,:] = subj[mask > 0]

c += 1

#scaling if asked

dmin = scale_min

dmax = scale_max

if scale:

au.log.info("Scaling data.")

[feats, dmin, dmax] = rescale(feats, scale_min, scale_max)

#extracting features from excluded subjects

exclfeats = []

if exclusubjs:

au.log.info("Processing excluded subjects.")

nexcl = len(exclusubjs)

exclfeats = np.empty([nexcl, nfeats], dtype=stype)

c = 0

for s in exclusubjs:

au.log.debug("Reading " + s)

#check geometries

au.check_has_same_geometry (s, maskf)

#load subject

subj = nib.load(s).get_data().flatten()

#mask data and save it

exclfeats[c,:] = subj[mask > 0]

c += 1

if scale:

[exclfeats, emin, emax] = rescale(exclfeats, scale_min, scale_max, dmin, dmax)

#parsing arguments

parser = set_parser()

try:

args = parser.parse_args ()

except argparse.ArgumentError, exc:

au.log.error (exc.message + '\n' + exc.argument)

parser.error(str(msg))

return -1

subjsf = args.subjs.strip ()

outdir = args.outdir.strip ()

datadir = args.datadir.strip ()

excluf = args.exclude.strip ()

otype = args.type.strip ()

dataname = args.dataname.strip()

maskf = args.mask.strip()

prefix = args.prefix.strip()

leave = args.leave

scale = args.scale

scale_min = args.scale_min

scale_max = args.scale_max

thrps = args.thresholdP.strip().split()

lthr = args.lthreshold.strip()

uthr = args.uthreshold.strip()

absolute = args.absolute

au.setup_logger(args.verbosity)

#checking number of files processed

if not os.path.exists(maskf):

err = 'Mask file not found: ' + maskf

au.log.error(err)

sys.exit(-1)

#number of subjects

subjsnum = au.file_len(subjsf)

#reading subjects list

subjlabels = np.zeros(subjsnum, dtype=int)

subjslist = {}

subjfile = open(subjsf, 'r')

c = 0

for s in subjfile:

line = s.strip().split(',')

subjlabels[c] = int(line[0])

subjfname = line[1].strip()

if not os.path.isabs(subjfname) and datadir:

subjslist[c] = datadir + os.path.sep + subjfname

else:

subjslist[c] = subjfname

c += 1

subjfile.close()

#excluding if excluf or leave > -1

subjmask = []

excluding = False

if excluf:

excluding = True

subjmask = np.loadtxt(excluf, dtype=int)

else:

subjmask = np.zeros(subjsnum, dtype=int)

if leave > -1:

excluding = True

subjmask[leave] = 1

subjs = [ subjslist[elem] for elem in subjslist if subjmask[elem] == 0]

labels = subjlabels[subjmask == 0]

exclusubjs = [ subjslist[elem] for elem in subjslist if subjmask[elem] == 1]

exclulabels = subjlabels[subjmask == 1]

if not excluding:

exclusubjs = []

#mask process

au.log.info('Processing ' + maskf)

#loading mask and masking it with globalmask

mask = nib.load(maskf).get_data()

#thresholding

if absolute: mask = np.abs(mask)

if lthr: mask[mask < lthr] = 0

if uthr: mask[mask > uthr] = 0

if thrps:

for t in thrps:

au.log.info ("Thresholding " + maskf + " with robust range below " + str(t) + " percent.")

thrm = au.threshold_robust_range (mask, t)

au.log.info ("Extracting features.")

[feats, exclfeats, dmin, dmax] = extract_features (subjs, exclusubjs, thrm, maskf, scale, scale_min, scale_max)

au.log.info ("Saving data files.")

[filename, exclfilename] = save_data (outdir, prefix, dataname, otype, excluding, leave, feats, labels, exclfeats, exclulabels, dmin, dmax, scale, scale_min, scale_max, lthr, uthr, t, absolute)

else:

au.log.info ("Extracting features.")

[feats, exclfeats, dmin, dmax] = extract_features (subjs, exclusubjs, mask, maskf, scale, scale_min, scale_max)

au.log.info ("Saving data files.")

[filename, exclfilename] = save_data (outdir, prefix, dataname, otype, excluding, leave, feats, labels, exclfeats, exclulabels, dmin, dmax, scale, scale_min, scale_max, lthr, uthr, thrps, absolute)

au.log.info ("Saved " + filename)

if excluding:

au.log.info ("Saved " + exclfilename)

#saving description files

np.savetxt(filename + '.' + au.subjectfiles_str(), subjs, fmt='%s')

np.savetxt(filename + '.' + au.labels_str(), labels, fmt='%i')

if excluding:

np.savetxt(exclfilename + '.' + au.subjectfiles_str(), exclusubjs, fmt='%s')

np.savetxt(exclfilename + '.' + au.labels_str(), exclulabels, fmt='%i')