def task_standardize_transform():
"""Standardize MRI images: transform phase."""
if MODE[0] != 'T2w':
return
# mode = MODE - 'std'
mode = MODE[:-1] if MODE[-1] == 'std' else MODE
sl = SAMPLELIST
paths = dwi.paths.Paths(mode)
outpaths = dwi.paths.Paths(tuple(mode) + ('std', ))
cfgpath = paths.std_cfg()
for case, scan in cases_scans(mode, sl):
inpath = paths.pmap(case=case, scan=scan)
mask = paths.mask('prostate', case, scan)
outpath = outpaths.pmap(case=case, scan=scan)
cmd = dwi.shell.standardize_transform(cfgpath,
inpath,
outpath,
mask=mask)
yield {
'name': taskname(mode, case, scan),
'actions': folders(outpath) + [cmd],
'file_dep': _files(cfgpath, inpath, mask),
'targets': _files(outpath),
'clean': True,
}
def get_task_grid_texture(mode,
c,
s,
ls,
mt,
tspec,
lt=None,
mbb=None,
nanbg=False,
fmt='txt'):
"""Grid classifier."""
paths = dwi.paths.Paths(mode)
pmap = paths.texture(c, s, None, mt, 'all', 0, tspec, voxel='all')
prostate = paths.mask('prostate', c, s)
lesion = [paths.mask('lesion', c, s, lesion=x) for x in ls]
out, target = paths.grid(c, s, mt, tspec, fmt=fmt)
if fmt == 'h5':
target = out
z = 5
# z = 1
d = dict(mbb=mbb,
voxelsize=None,
winsize=z,
voxelspacing=(z, 1, 1),
lesiontypes=lt,
use_centroid=False,
nanbg=nanbg)
cmd = dwi.shell.grid(pmap, None, prostate, lesion, out, **d)
return {
'name': taskname(mode, c, s, mt, tspec.method, tspec.winsize),
'actions': folders(out) + [cmd],
'file_dep': _files(pmap, prostate, *lesion),
'targets': _files(target),
}
def get_task_merge_textures(mode, mt, c, s, l, slices, portion, voxel):
"""Merge texture methods into singe file per case/scan/lesion."""
paths = dwi.paths.Paths(mode)
infiles = [paths.texture(c, s, l, mt, slices, portion, tspec, voxel=voxel)
for tspec in texture_methods_winsizes(mode, mt)]
outfile = paths.texture(c, s, l, mt+'_merged', slices, portion, None,
voxel=voxel)
cmd = dwi.shell.select_voxels(' '.join(map(str, infiles)), outfile)
return {
'name': taskname(mode, c, s, l, mt, slices, portion,
voxel),
'actions': folders(outfile) + [cmd],
'file_dep': _files(*infiles),
'targets': _files(outfile),
}
def get_task_select_roi_auto(mode, case, scan, algparams):
"""Select ROIs from the pmap DICOMs based on masks."""
paths = dwi.paths.Paths(mode)
ap_ = '_'.join(algparams)
mask = paths.mask('auto', case, scan, algparams=algparams)
roi = paths.roi('auto', case=case, scan=scan, algparams=algparams)
pmap = paths.pmap(case=case, scan=scan)
cmd = dwi.shell.select_voxels(pmap, roi, mask=mask, keepmasked=False)
return {
'name': taskname(mode, ap_, case, scan),
'actions': folders(roi) + [cmd],
'file_dep': _files(mask, pmap),
'targets': _files(roi),
'clean': True,
}
def task_make_subregion():
"""Make minimum bounding box + 10 voxel subregions from prostate masks."""
for mode, sl in product(MODES, SAMPLELISTS):
paths = dwi.paths.Paths(mode)
if str(mode) == 'DWI-Mono-ADCm':
for case, scan in cases_scans(mode, sl):
mask = paths.mask('prostate', case, scan)
subregion = paths.subregion(case=case, scan=scan)
cmd = dwi.shell.make_subregion(mask, subregion)
yield {
'name': taskname(mode, case, scan),
'actions': folders(subregion) + [cmd],
'file_dep': _files(mask),
'targets': _files(subregion),
'clean': True,
}
def task_make_subregion():
"""Make minimum bounding box + 10 voxel subregions from prostate masks."""
for mode, sl in product(MODES, SAMPLELISTS):
paths = dwi.paths.Paths(mode)
if str(mode) == 'DWI-Mono-ADCm':
for case, scan in cases_scans(mode, sl):
mask = paths.mask('prostate', case, scan)
subregion = paths.subregion(case=case, scan=scan)
cmd = dwi.shell.make_subregion(mask, subregion)
yield {
'name': taskname(mode, case, scan),
'actions': folders(subregion) + [cmd],
'file_dep': _files(mask),
'targets': _files(subregion),
'clean': True,
}
def get_task_histogram(mode, masktype, samplelist):
"""Plot image histograms."""
paths = dwi.paths.Paths(mode)
if masktype == 'lesion':
it = (dict(case=c, scan=s, lesion=l) for c, s, l in
lesions(mode, samplelist))
else:
it = (dict(case=c, scan=s) for c, s in cases_scans(mode, samplelist))
inpaths = [paths.roi(masktype, **x) for x in it]
figpath = paths.histogram(masktype, samplelist)
cmd = dwi.shell.histogram(inpaths, figpath, params=None)
return {
'name': taskname(mode, masktype, samplelist),
'actions': folders(figpath) + [cmd],
'file_dep': _files(inpaths),
'targets': _files(figpath),
}
def get_task_select_roi_manual(mode, case, scan, masktype):
"""Select ROIs from the pmap DICOMs based on masks."""
paths = dwi.paths.Paths(mode)
mask = paths.mask(masktype, case, scan)
roi = paths.roi(masktype, case=case, scan=scan)
pmap = paths.pmap(case=case, scan=scan)
d = dict(mask=mask, keepmasked=(masktype == 'prostate'))
if 'std' in mode:
d['astype'] = 'float32' # Integers cannot have nans.
cmd = dwi.shell.select_voxels(pmap, roi, **d)
return {
'name': taskname(mode, masktype, case, scan),
'actions': folders(roi) + [cmd],
'file_dep': _files(mask, pmap),
'targets': _files(roi),
'clean': True,
}
def get_task_histogram(mode, masktype, samplelist):
"""Plot image histograms."""
paths = dwi.paths.Paths(mode)
if masktype == 'lesion':
it = (dict(case=c, scan=s, lesion=l)
for c, s, l in lesions(mode, samplelist))
else:
it = (dict(case=c, scan=s) for c, s in cases_scans(mode, samplelist))
inpaths = [paths.roi(masktype, **x) for x in it]
figpath = paths.histogram(masktype, samplelist)
cmd = dwi.shell.histogram(inpaths, figpath, params=None)
return {
'name': taskname(mode, masktype, samplelist),
'actions': folders(figpath) + [cmd],
'file_dep': _files(inpaths),
'targets': _files(figpath),
}
def get_task_grid_texture(mode, c, s, ls, mt, tspec, lt=None, mbb=None,
nanbg=False, fmt='txt'):
"""Grid classifier."""
paths = dwi.paths.Paths(mode)
pmap = paths.texture(c, s, None, mt, 'all', 0, tspec, voxel='all')
prostate = paths.mask('prostate', c, s)
lesion = [paths.mask('lesion', c, s, lesion=x) for x in ls]
out, target = paths.grid(c, s, mt, tspec, fmt=fmt)
if fmt == 'h5':
target = out
z = 5
# z = 1
d = dict(mbb=mbb, voxelsize=None, winsize=z, voxelspacing=(z, 1, 1),
lesiontypes=lt, use_centroid=False, nanbg=nanbg)
cmd = dwi.shell.grid(pmap, None, prostate, lesion, out, **d)
return {
'name': taskname(mode, c, s, mt, tspec.method, tspec.winsize),
'actions': folders(out) + [cmd],
'file_dep': _files(pmap, prostate, *lesion),
'targets': _files(target),
}
def task_standardize_train():
"""Standardize MRI images: training phase.
Pay attention to the sample list: all samples should be used.
"""
if MODE[0] != 'T2w':
return
# mode = MODE - 'std'
mode = MODE[:-1] if MODE[-1] == 'std' else MODE
paths = dwi.paths.Paths(mode)
std_cfg = paths.std_cfg()
inpaths = [paths.roi('prostate', case=c, scan=s) for c, s in
cases_scans(mode, 'all')]
cmd = dwi.shell.standardize_train(inpaths, std_cfg, 'none')
yield {
'name': taskname(mode),
'actions': [cmd],
'file_dep': _files(*inpaths),
'targets': _files(std_cfg),
'clean': True,
}
def get_task_texture(mode, masktype, case, scan, lesion, slices, portion,
tspec, voxel):
"""Generate texture features."""
paths = dwi.paths.Paths(mode)
inpath = paths.pmap(case=case, scan=scan)
deps = [inpath]
mask = paths.mask(masktype, case, scan, lesion=lesion)
if mask is not None:
deps.append(mask)
outfile = paths.texture(case, scan, lesion, masktype, slices, portion,
tspec, voxel=voxel)
cmd = dwi.shell.get_texture(mode, inpath, tspec, slices, portion, outfile,
voxel, mask=mask)
return {
'name': taskname(mode, masktype, slices, portion, case, scan, lesion,
tspec.method, tspec.winsize, voxel),
'actions': folders(outfile) + [cmd],
'file_dep': _files(*deps),
'targets': _files(outfile),
'clean': True,
}
def get_task_merge_textures(mode, mt, c, s, l, slices, portion, voxel):
"""Merge texture methods into singe file per case/scan/lesion."""
paths = dwi.paths.Paths(mode)
infiles = [
paths.texture(c, s, l, mt, slices, portion, tspec, voxel=voxel)
for tspec in texture_methods_winsizes(mode, mt)
]
outfile = paths.texture(c,
s,
l,
mt + '_merged',
slices,
portion,
None,
voxel=voxel)
cmd = dwi.shell.select_voxels(' '.join(map(str, infiles)), outfile)
return {
'name': taskname(mode, c, s, l, mt, slices, portion, voxel),
'actions': folders(outfile) + [cmd],
'file_dep': _files(*infiles),
'targets': _files(outfile),
}
def task_standardize_transform():
"""Standardize MRI images: transform phase."""
if MODE[0] != 'T2w':
return
# mode = MODE - 'std'
mode = MODE[:-1] if MODE[-1] == 'std' else MODE
sl = SAMPLELIST
paths = dwi.paths.Paths(mode)
outpaths = dwi.paths.Paths(tuple(mode) + ('std',))
cfgpath = paths.std_cfg()
for case, scan in cases_scans(mode, sl):
inpath = paths.pmap(case=case, scan=scan)
mask = paths.mask('prostate', case, scan)
outpath = outpaths.pmap(case=case, scan=scan)
cmd = dwi.shell.standardize_transform(cfgpath, inpath, outpath,
mask=mask)
yield {
'name': taskname(mode, case, scan),
'actions': folders(outpath) + [cmd],
'file_dep': _files(cfgpath, inpath, mask),
'targets': _files(outpath),
'clean': True,
}
"""
if MODE[0] != 'T2w':
return
# mode = MODE - 'std'
mode = MODE[:-1] if MODE[-1] == 'std' else MODE
paths = dwi.paths.Paths(mode)
std_cfg = paths.std_cfg()
inpaths = [
paths.roi('prostate', case=c, scan=s)
for c, s in cases_scans(mode, 'all')
]
cmd = dwi.shell.standardize_train(inpaths, std_cfg, 'none')
yield {
'name': taskname(mode),
'actions': [cmd],
'file_dep': _files(*inpaths),
'targets': _files(std_cfg),
'clean': True,
}
def task_standardize_transform():
"""Standardize MRI images: transform phase."""
if MODE[0] != 'T2w':
return
# mode = MODE - 'std'
mode = MODE[:-1] if MODE[-1] == 'std' else MODE
sl = SAMPLELIST
paths = dwi.paths.Paths(mode)
outpaths = dwi.paths.Paths(tuple(mode) + ('std', ))
cfgpath = paths.std_cfg()