def process_run(rundir, subid, logger):
"""
1. Find raw
2. Split func into xfm dir
3. Align func, create mean and 4d
4. Bias correct anat
5. Invxfm anat2func
6. Generate movement files
7. Clean up
"""
_, run = os.path.split(rundir)
rawdir = os.path.join(rundir, 'raw')
exists, mvdir, xfmdir = check_make_dirs(rundir, logger)
if exists:
return
anat, aparc, func = find_raw_files(rawdir, subid, logger)
if anat is None:
return
funcs = split_file(func, xfmdir, subid, run, logger)
if funcs is None:
return
movement, movedfuncs = align_funcs(funcs, logger)
exists, refunc = join_files(movedfuncs, rundir, subid, run)
if not exists:
return
#Write movement and displacement
logger.info('Write movement')
save_movement(movement, mvdir)
#Make mean from aligned
meanfunc = reg.make_mean(movedfuncs)
#Bias correct anat
b4anat = reg.n4_biascorrect(anat)
if b4anat is None:
logger.error('{0}: Bias correction failed'.format(anat))
return
#Bias correct mean
b4meanfunc = reg.n4_biascorrect(meanfunc)
if b4meanfunc is None:
logger.error('{0}: Bias correction failed'.format(meanfunc))
return
#Reslice mean to anat space
rb4meanfunc = reslice_mean(b4anat, b4meanfunc)
#Make mask by growing func (5x5x5 kernel) and mask with anat
mask = reg.grow_mask(rb4meanfunc, size= 3)
#Find mapping from func to anat
func2anat_xfm = reg.affine_register_mi(b4anat, b4meanfunc, mask= mask)
#Apply inverse mapping to anat
anat2func = reg.apply_transform(b4anat, '-i %s'%(func2anat_xfm),
target=meanfunc, use_nn = False)
#Apply transform to aparc nearest neighbor
aparc2func = reg.apply_transform(aparc, '-i %s'%(func2anat_xfm),
target=meanfunc, use_nn = True)
#Move
for item in (anat2func, aparc2func):
cmd = 'mv %s %s'%(item, rundir)
os.system(cmd)
def process_run3(rundir, subid, datadir, logger):
"""
1. Find raw
2. Split func into xfm dir
3. Center func
4. Align func, create mean and 4d
5. Bias correct anat
6. Invxfm anat2func
7. Generate movement files
8. Clean up
"""
_, run = os.path.split(rundir)
rawdir = os.path.join(rundir, 'raw')
exists, mvdir, xfmdir = check_make_dirs(rundir, logger)
if exists:
return
anat, aparc, func = find_raw_files(rawdir, subid, logger)
if anat is None:
return
funcs = split_file(func, xfmdir, subid, run, logger)
if funcs is None:
return
cfuncs = suzanne_center(funcs)
movement, movedfuncs = align_funcs(cfuncs, logger)
exists, refunc = join_files(movedfuncs, rundir, subid, run)
if not exists:
return
logger.info('Write movement')
save_movement(movement, mvdir)
meanfunc = reg.make_mean(movedfuncs)
threshold_mean(meanfunc, logger, thresh=275)
b4anat = bias_correct(anat, logger, xfmdir)
b4meanfunc = bias_correct(meanfunc, logger, xfmdir)
func2anat_xfm = reg.affine_register_mi(b4anat, b4meanfunc, mask= None)
anat2func = reg.apply_transform(b4anat, '-i %s'%(func2anat_xfm),
target=b4meanfunc, use_nn = False)
aparc2func = reg.apply_transform(aparc, '-i %s'%(func2anat_xfm),
target=b4meanfunc, use_nn = True)
pth, aparcnme = os.path.split(aparc2func)
aparc_mask = os.path.join(datadir, 'mask_{0}'.format(aparcnme))
mask_from_aparc(aparc2func, aparc_mask)
for item in (anat2func, aparc2func):
cmd = 'mv %s %s'%(item, datadir)
os.system(cmd)
return b4meanfunc
#Make symbolic link of target/labels to subdir
_, targetnme = os.path.split(target)
cmd = 'ln -s {} {}'.format(target, subdir)
os.system(cmd)
target = os.path.join(subdir, targetnme)
_, labelnme = os.path.split(target_labels)
cmd = 'ln -s {} {}'.format(target_labels, subdir)
os.system(cmd)
target_labels = os.path.join(subdir, labelnme)
moving, moving_aparc = getdata(subdir)
moving_labels = aparc_to_mask(moving_aparc, labels) #Subject ROIs - Manually drawn
#Register subject to template
affine = register.affine_register_mi(target, moving)
#Apply to structural and labels
xfm_moving = register.apply_transform(moving, affine, outfile = None, target = target, use_nn = False)
xfm_moving_labels = register.apply_transform(moving_labels, affine, outfile = None, target = target, use_nn = True)
psepct = [0.5, .75, 1.0]
for pct in psepct:
outname = os.path.join(subdir, 'PSEtest_{}'.format(pct))
cmd = psewarp(target, xfm_moving, target_labels, xfm_moving_labels, pct, outname)
print cmd
os.system(cmd)
pseaffine = outname + 'Affine.txt'
psewarp = outname + 'Warp.nii.gz'
subjs = sorted(glob(os.path.join(datadirectory, 'B*')))
# try:
# taskid = os.environ['SGE_TASK_ID']
# except:
# raise IOError('Are you using SGE??')
# subj = dirs[int(taskid) - 1]
for subj in subjs:
jnk, subid = os.path.split(subj)
raw_dir = os.path.join(subj, 'run3/raw')
xfm_dir = os.path.join(subj, 'run3/xfm')
anat = os.path.join(raw_dir, '%s_brainmask.nii.gz'%(subid))
aparc = os.path.join(raw_dir, '%s_aparcaseg.nii.gz'%(subid))
b4meanfunc = glob(os.path.join(xfm_dir, 'n4_mean*'))[0]
b4anat = bias_correct(anat, xfm_dir)
func2anat_xfm = reg.affine_register_mi(b4anat, b4meanfunc, mask= None)
anat2func = reg.apply_transform(b4anat, '-i %s'%(func2anat_xfm), target=b4meanfunc, use_nn = False)
aparc2func = reg.apply_transform(aparc, '-i %s'%(func2anat_xfm), target=b4meanfunc, use_nn = True)
pth, aparcnme = os.path.split(aparc2func)
aparc_mask = os.path.join(subj, 'mask_{0}'.format(aparcnme))
mask_from_aparc(aparc2func, aparc_mask)
for item in (anat2func, aparc2func):
cmd = 'mv %s %s'%(item, subj)
os.system(cmd)
# ANTSPATH='/home/jagust/cindeem/bin/ANTs-1.9.v4-Linux/bin'
# def affine_register_mi_iter(target, moving):
# """ uses ANTS to create an affine mapping between different
import os, sys from glob import glob sys.path.insert(0, '/home/jagust/fmri-pstask/scripts/ANTS/ants_tools') import register as reg basedir = '/home/jagust/fmri-pstask/subjects' fdgdirs = sorted(glob(os.path.join(basedir, 'B*/fdg'))) for fdg in fdgdirs: testxfm = sorted(glob(os.path.join(fdg, 'xfm*'))) if not testxfm: scans = sorted(glob(os.path.join(fdg, '*nii*'))) fdg2anat_xfm = reg.affine_register_mi(scans[0], scans[1], mask=None) #0=Brainmask, 1=FDG fdg2anat = reg.apply_transform(scans[1], fdg2anat_xfm, target=scans[0], use_nn=False)
#basedir = '/home/jagust/fmri-pstask/subjects/'
target = '/home/jagust/fmri-pstask/subjects/test_labelalign/template/rois/r0p5-mni_template.nii.gz'
brainmask = glob(os.path.join(dir, 'r*brainmask.nii.gz'))[0]
affine = glob(os.path.join(dir, 'r*Affine*'))[0]
pseaffine = glob(os.path.join(dir, 'PSE-0.75', '*_Affine*'))[0]
psewarp = glob(os.path.join(dir, 'PSE-0.75', '*_Warp*'))[0]
full_transform = '{} {} {}'.format(psewarp, pseaffine, affine)
funcstr = ['cope*', 'zstat*']
subjdir, _ = os.path.split(dir)
meanfunc = glob(os.path.join(subjdir, '*run3', 'xfm', 'mean*nii.gz'))[0]
_, mfuncnme = os.path.split(meanfunc)
if not os.path.isfile(os.path.join(dir, mfuncnme)):
cmd = 'ln -s {} {}'.format(meanfunc, dir)
os.system(cmd)
meanfunc = os.path.join(dir, mfuncnme)
mfuncaffine = register.affine_register_mi(brainmask, meanfunc)
xfm_mfunc = register.apply_transform(meanfunc, mfuncaffine, outfile = None, target = brainmask, use_nn = False)
statdir = glob(os.path.join(subjdir, '*uncorr*', '*feat', 'stats'))[0]
funcs = []
for globstr in funcstr:
funcs.extend(glob(os.path.join(statdir, globstr)))
for func in funcs:
xfm_func = register.apply_transform(func, mfuncaffine, outfile = None, target = brainmask, use_nn = False)
final_func = xfm_func.replace('xfm_', 'finalxfm_')
final_func = register.apply_transform(xfm_func, full_transform, outfile = final_func, target = target, use_nn = False)
def process_run3(rundir, subid, logger):
"""
1. Find raw
2. Split func into xfm dir
3. Align func, create mean and 4d
4. Bias correct anat
5. Invxfm anat2func
6. Generate movement files
7. Clean up
"""
_, run = os.path.split(rundir)
rawdir = os.path.join(rundir, 'raw')
exists, mvdir, xfmdir = check_make_dirs(rundir, logger)
if exists:
return
anat, aparc, func = find_raw_files(rawdir, subid, logger)
if anat is None:
return
funcs = split_file(func, xfmdir, subid, run, logger)
if funcs is None:
return
movement, movedfuncs = align_funcs(funcs, logger)
exists, refunc = join_files(movedfuncs, rundir, subid, run)
if not exists:
return
#Write movement and displacement
logger.info('Write movement')
save_movement(movement, mvdir)
#Make mean from aligned
meanfunc = reg.make_mean(movedfuncs)
#Bias correct anat
b4anat = reg.n4_biascorrect(anat)
if b4anat is None:
logger.error('{0}: Bias correction failed'.format(anat))
return
_, b4anatname = os.path.split(b4anat)
cmd = 'mv %s %s'%(b4anat, xfmdir)
os.system(cmd)
b4anat = os.path.join(xfmdir, b4anatname)
# threshold mean to remove "cruft" current value == 300
threshold_mean(meanfunc, logger, thresh=300)
#Bias correct mean
#b4meanfunc = reg.n4_biascorrect(meanfunc)
#if b4meanfunc is None:
# logger.error('{0}: Bias correction failed'.format(meanfunc))
# return
#Reslice mean to anat space
b4meanfunc = meanfunc
rb4meanfunc = reslice_mean(b4anat, b4meanfunc)
#Make mask by growing func (5x5x5 kernel) and mask with anat
# mask = reg.grow_mask(rb4meanfunc, size= 3)
#Find mapping from func to anat
func2anat_xfm = reg.affine_register_mi(b4anat, b4meanfunc, mask= None)
#Apply inverse mapping to anat
anat2func = reg.apply_transform(b4anat, '-i %s'%(func2anat_xfm),
target=meanfunc, use_nn = False)
#Apply transform to aparc nearest neighbor
aparc2func = reg.apply_transform(aparc, '-i %s'%(func2anat_xfm),
target=meanfunc, use_nn = True)
#Skull strip shit
pth, aparcnme = os.path.split(aparc2func)
aparc_mask = os.path.join(datadir, 'mask_{0}'.format(aparcnme))
mask_from_aparc(aparc2func, aparc_mask)
#masked_b4meanfunc = b4meanfunc.replace('n4_', 'masked_n4_')
#mask_img(b4meanfunc, aparc_mask, masked_b4meanfunc)
#Move
for item in (anat2func, aparc2func):
cmd = 'mv %s %s'%(item, rundir)
os.system(cmd)
return b4meanfunc
def process_runs(rundir, subid, run3meanfunc, logger):
_, run = os.path.split(rundir)
rawdir = os.path.join(rundir, 'raw')
exists, mvdir, xfmdir = check_make_dirs(rundir, logger)
if exists:
return
_, run3 = os.path.split(run3meanfunc)
cmd = 'cp %s %s'%(run3meanfunc, xfmdir)
os.system(cmd)
run3meanfunc = os.path.join(xfmdir, run3)
anat, aparc, func = find_raw_files(rawdir, subid, logger)
if anat is None:
return
funcs = split_file(func, xfmdir, subid, run, logger)
if funcs is None:
return
movement, affines, moved = make_affines(funcs, logger)
logger.info('Write movement')
save_movement(movement, mvdir)
meanfunc = reg.make_mean(moved)
#Remove old movedfuncs
newdir = os.path.join(xfmdir, 'first_realign')
os.mkdir(newdir)
for f in moved[1:]:
cmd = 'mv %s %s'%(f, newdir)
os.system(cmd)
#Bias correct anat
b4anat = reg.n4_biascorrect(anat)
if b4anat is None:
logger.error('{0}: Bias correction failed'.format(anat))
return
_, b4anatname = os.path.split(b4anat)
cmd = 'mv %s %s'%(b4anat, xfmdir)
os.system(cmd)
b4anat = os.path.join(xfmdir, b4anatname)
# threshold mean to remove "cruft" current value == 300
threshold_mean(meanfunc, logger, thresh=300)
#Bias correct mean
#b4meanfunc = reg.n4_biascorrect(meanfunc)
#if b4meanfunc is None:
# logger.error('{0}: Bias correction failed'.format(meanfunc))
# return
b4meanfunc = meanfunc
#Reslice mean to anat space
rb4meanfunc = reslice_mean(b4anat, b4meanfunc)
#Make mask by growing func (5x5x5 kernel) and mask with anat
# mask = reg.grow_mask(rb4meanfunc, size= 3)
#Find mapping from func to anat
func2anat_xfm = reg.affine_register_mi(b4anat, b4meanfunc, mask= None)
#Apply inverse mapping to anat
anat2func = reg.apply_transform(b4anat, '-i %s'%(func2anat_xfm),
target=meanfunc, use_nn = False)
#Apply transform to aparc nearest neighbor
aparc2func = reg.apply_transform(aparc, '-i %s'%(func2anat_xfm),
target=meanfunc, use_nn = True)
#Skull strip shit
pth, aparcnme = os.path.split(aparc2func)
aparc_mask = os.path.join(rundir, 'mask_{0}'.format(aparcnme))
mask_from_aparc(aparc2func, aparc_mask)
#masked_b4meanfunc = b4meanfunc.replace('n4_', 'masked_n4_')
#mask_img(b4meanfunc, aparc_mask, masked_b4meanfunc)
#Make new moved funcs based on 2 affines
movedfuncs = align_means(funcs, b4meanfunc, run3meanfunc, affines, logger)
exists, refunc = join_files(movedfuncs, rundir, subid, run)
if not exists:
return
#Move
for item in (anat2func, aparc2func):
cmd = 'mv %s %s'%(item, rundir)
os.system(cmd)
