def get_parc_stats(fsDir, subjID, parcName, bVerbose=False):
import sys, os
import tempfile
from scai_utils import (
check_dir,
check_file,
check_bin_path,
info_log,
cmd_stdout,
saydo,
read_text_file,
remove_empty_strings,
)
# === Constants ===#
hemis = ["lh", "rh"]
segStatsBin = "mri_segstats"
# === Implementation ===#
check_bin_path(segStatsBin)
sDir = os.path.join(fsDir, subjID)
check_dir(sDir)
lblDir = os.path.join(sDir, "label")
check_dir(sDir)
if bVerbose:
info_log("Label directory = %s" % lblDir)
morphInfo = {}
rois = []
area_mm2 = []
for (i0, hemi) in enumerate(hemis):
if bVerbose:
info_log("Working on hemisphere: %s" % hemi)
annot = os.path.join(lblDir, "%s.%s.annot" % (hemi, parcName))
check_file(annot)
tmpSum = tempfile.mktemp()
cmd = "%s --annot %s %s %s --sum %s" % (segStatsBin, subjID, hemi, parcName, tmpSum)
saydo(cmd)
check_file(tmpSum)
if bVerbose:
print("Intermediate results saved at: %s" % tmpSum)
t = read_text_file(tmpSum)
for (i0, tline) in enumerate(t):
tline = tline.strip()
if len(tline) == 0:
continue
elif tline.startswith("#"):
continue
t_items = remove_empty_strings(tline.split())
if len(t_items) != 5:
continue
t_roi = t_items[4]
if t_roi.lower() == "none":
continue
rois.append("%s_%s" % (hemi, t_roi))
area_mm2.append(float(t_items[3]))
saydo("rm -rf %s" % tmpSum)
morphInfo["rois"] = rois
morphInfo["area_mm2"] = area_mm2
return morphInfo
ap.add_argument("--skipSubj", \
help="List of subjects to skip, separated by commas (e.g., S24,S40")
#ap.add_argument("groupMethod", \
# help="Method for getting the group identity of the subjects (e.g., get_qdec_info,diagnosis). Must be a function name followed by a field name. The field name will be the second input argument to the function. The first input argument will be the subject ID")
if len(sys.argv) == 1:
ap.print_help()
sys.exit(0)
args = ap.parse_args()
#=== Check the matching of environmental SUBJECTS_DIR ===#
check_env_var("SUBJECTS_DIR", args.fsDir)
#=== Discover subjects ===#
check_dir(args.fsDir)
ds = glob.glob(os.path.join(args.fsDir, args.sNameWC))
ds.sort()
skipIDs = []
if args.skipSubj != None:
skipIDs = args.skipSubj.split(",")
sIDs = []
grps = []
for (i0, d) in enumerate(ds):
t_sID = os.path.split(d)[1]
if skipIDs.count(t_sID) > 0:
info_log("Skipping subject: %s" % t_sID)
continue
help="Subject group (e.g., AWS). If not specified, will include all subjects and perform between-group analysis")
ap.add_argument("--fwhm-surf", dest="fwhmSurf", type=float, default=10,
help="FWHM for surface-based smoothing (default=0)")
ap.add_argument("--vol", action="store_true", dest="bVol",
help="Perform volumetric analysis, instead of surface-based analysis (default: false)")
ap.add_argument("--redo", dest="bRedo", action="store_true",
help="Force redo all steps (default: false)")
if len(sys.argv) == 1:
ap.print_help()
sys.exit(0)
args = ap.parse_args()
#=== Input sanity check ===#
check_dir(args.batchBase)
check_dir(args.fsDir)
#=== Check that SUBJECTS_DIR matches input argument ===#
envFSDir = os.path.abspath(os.getenv("SUBJECTS_DIR"))
if envFSDir != os.path.abspath(args.fsDir):
error_log("Input FreeSurfer SUBJECTS_DIR does not match environmental SUBECTS_DIR")
#=== Get the list of subjects ===#
if args.group == None or len(args.group) == 0:
wc = "*_*"
bAll = True
else:
wc = args.group + "*"
bAll = False;
def get_dpath_track_data(trackDir):
# List of data fields to extract
dataFlds = [
"Count", "Volume", "Len_Min", "Len_Max", "Len_Avg", "Len_Center",
"AD_Avg", "AD_Avg_Weight", "AD_Avg_Center", "RD_Avg", "RD_Avg_Weight",
"RD_Avg_Center", "MD_Avg", "MD_Avg_Weight", "MD_Avg_Center", "FA_Avg",
"FA_Avg_Weight", "FA_Avg_Center"
]
dat = {}
import os
from scai_utils import check_dir, check_file, read_text_file, info_log
check_dir(trackDir)
# Read from pathstats.overall.txt
pso = os.path.join(trackDir, "pathstats.overall.txt")
check_file(pso)
txt = read_text_file(pso)
for (i0, t_line) in enumerate(txt):
t_line = t_line.strip()
if t_line.startswith("#"):
continue
t_items = t_line.split(" ")
if len(t_items) == 2:
if dataFlds.count(t_items[0]) == 1:
dat[t_items[0]] = float(t_items[1])
# Check for completeness
for (i0, t_fld) in enumerate(dataFlds):
if not (t_fld in dat):
info_log("WARNING: could not find data field %s in file %s" % \
(t_fld, pso), bWarn=True)
# Read from pathstats.byvoxel.txt
vso = os.path.join(trackDir, "pathstats.byvoxel.txt")
check_file(vso)
txt = read_text_file(vso)
import numpy as np
bvDat = {}
cols = {}
# Determine the number of data points
npts = 0
lidx0 = -1
bFoundHeader = False
for (i0, t_line) in enumerate(txt):
t_line = t_line.strip()
if len(t_line) == 0:
continue
if not (t_line.startswith("#") or t_line.startswith("x")):
npts += 1
if lidx0 == -1:
lidx0 = i0
if t_line.startswith("x"):
bFoundHeader = True
t_items = t_line.split(" ")
cols["x"] = t_items.index("x")
cols["y"] = t_items.index("y")
cols["z"] = t_items.index("z")
cols["AD"] = t_items.index("AD")
cols["RD"] = t_items.index("RD")
cols["MD"] = t_items.index("MD")
cols["FA"] = t_items.index("FA")
cols["AD_Avg"] = t_items.index("AD_Avg")
cols["RD_Avg"] = t_items.index("RD_Avg")
cols["MD_Avg"] = t_items.index("MD_Avg")
cols["FA_Avg"] = t_items.index("FA_Avg")
if not bFoundHeader:
raise Exception, "Cannot find header column in file %s" % vso
txt = txt[lidx0:lidx0 + npts]
#print(txt)
# Allocate space
bvDat["x"] = np.zeros(npts)
bvDat["y"] = np.zeros(npts)
bvDat["z"] = np.zeros(npts)
bvDat["AD"] = np.zeros(npts)
bvDat["RD"] = np.zeros(npts)
bvDat["MD"] = np.zeros(npts)
bvDat["FA"] = np.zeros(npts)
bvDat["AD_Avg"] = np.zeros(npts)
bvDat["RD_Avg"] = np.zeros(npts)
bvDat["MD_Avg"] = np.zeros(npts)
bvDat["FA_Avg"] = np.zeros(npts)
keys = bvDat.keys()
for (i0, t_line) in enumerate(txt):
t_items = t_line.split(" ")
for t_fld in keys:
bvDat[t_fld][i0] = float(t_items[cols[t_fld]])
dat["byVoxel"] = bvDat
return dat
def calculate_roi_tensor_measures(parcName, roiList, wmDepths, TENSOR_MEASURES,
HEMIS, dmriDir, annotDir, tensMeasMatFN,
logFileName):
#=== Load masks (gm and wm of different depths) ===#
import os
import nibabel as nb
import numpy as np
from scai_utils import check_file, check_dir, info_log, error_log
mask_shapes = []
nDepths = len(wmDepths)
roiNames = []
nzIdx = []
for i0 in range(nDepths):
nzIdx.append([])
bSpeech = []
parcDir = os.path.join(annotDir, parcName)
for (i0, wmDepth) in enumerate(wmDepths):
if wmDepth == -1:
parcTypeDir = os.path.join(parcDir, "gm")
info_log("Loading gray-matter masks")
else:
parcTypeDir = os.path.join(parcDir, "wm%dmm" % wmDepth)
info_log("Loading white-matter masks of %d-mm depth" \
% wmDepth)
for (i1, troi) in enumerate(roiList):
for (i2, hemi) in enumerate(HEMIS):
maskFN = os.path.join(parcTypeDir, \
"%s_%s.diff.nii.gz" % (hemi, troi[0]))
check_file(maskFN, logFN=logFileName)
t_img = nb.load(maskFN)
t_img_dat = t_img.get_data()
mask_shapes.append(np.shape(t_img_dat))
t_img_dat = np.ndarray.flatten(t_img_dat)
nzIdx[i0].append(np.nonzero(t_img_dat)[0])
if wmDepth == -1:
if troi[2] == 'N':
bSpeech.append(0)
else:
bSpeech.append(1)
roiNames.append("%s_%s" % (hemi, troi[0]))
#=== Check that the dimensions of all mask images match ===#
if len(np.unique(mask_shapes)) != 1:
error_log("Non-unique matrix size among the mask files", logFN=logFN)
#=== Load the dtifit_* files and extract the measures ===#
nROIs = len(roiNames)
assert (len(bSpeech) == nROIs)
tensMeas = {}
mask_shape = np.unique(mask_shapes)
check_dir(dmriDir, logFN=logFileName)
for (i0, measName) in enumerate(TENSOR_MEASURES):
tensMeas[measName] = np.zeros([nROIs, nDepths])
for (i1, t_depth) in enumerate(wmDepths):
if t_depth == -1:
info_log("Extracting tensor measure %s from gray matter" \
% measName)
else:
info_log(
"Extracting tensor measure %s from %d-mm deep white matter"
% (measName, t_depth))
assert (len(nzIdx[i1]) == nROIs)
for (i2, troi) in enumerate(roiNames):
measImg = os.path.join(dmriDir, \
"dtifit_%s.nii.gz" % measName)
check_file(measImg, logFN=logFileName)
t_img = nb.load(measImg)
if not list(mask_shape[0]) == list(np.shape(t_img)):
error_log("The diffusion tensor measure volume %s (%s) does not have a dimension that matches those of the mask files" \
% (measImg, measName))
t_img_dat = np.ndarray.flatten(t_img.get_data())
tensMeas[measName][i2, i1] = \
np.mean(t_img_dat[nzIdx[i1][i2]])
#=== Write data to file ===#
from scipy.io import savemat
res = {
"roiNames": roiNames,
"bSpeech": bSpeech,
"parcName": parcName,
"maskShape": mask_shape,
"wmDepths": wmDepths,
"tensMeas": tensMeas
}
savemat(tensMeasMatFN, res)
check_file(tensMeasMatFN, logFN=logFileName)
info_log(
"Tensor measures (%d types) and associated data were saved at: %s" %
(len(TENSOR_MEASURES), tensMeasMatFN),
logFN=logFileName)
def get_dpath_track_data(trackDir):
# List of data fields to extract
dataFlds = ["Count", "Volume",
"Len_Min", "Len_Max", "Len_Avg", "Len_Center",
"AD_Avg", "AD_Avg_Weight", "AD_Avg_Center",
"RD_Avg", "RD_Avg_Weight", "RD_Avg_Center",
"MD_Avg", "MD_Avg_Weight", "MD_Avg_Center",
"FA_Avg", "FA_Avg_Weight", "FA_Avg_Center"]
dat = {}
import os
from scai_utils import check_dir, check_file, read_text_file, info_log
check_dir(trackDir)
# Read from pathstats.overall.txt
pso = os.path.join(trackDir, "pathstats.overall.txt")
check_file(pso)
txt = read_text_file(pso)
for (i0, t_line) in enumerate(txt):
t_line = t_line.strip()
if t_line.startswith("#"):
continue
t_items = t_line.split(" ")
if len(t_items) == 2:
if dataFlds.count(t_items[0]) == 1:
dat[t_items[0]] = float(t_items[1])
# Check for completeness
for (i0, t_fld) in enumerate(dataFlds):
if not (t_fld in dat):
info_log("WARNING: could not find data field %s in file %s" % \
(t_fld, pso), bWarn=True)
# Read from pathstats.byvoxel.txt
vso = os.path.join(trackDir, "pathstats.byvoxel.txt")
check_file(vso)
txt = read_text_file(vso)
import numpy as np
bvDat = {}
cols = {}
# Determine the number of data points
npts = 0
lidx0 = -1
bFoundHeader = False
for (i0, t_line) in enumerate(txt):
t_line = t_line.strip()
if len(t_line) == 0:
continue
if not (t_line.startswith("#") or t_line.startswith("x")):
npts += 1
if lidx0 == -1:
lidx0 = i0
if t_line.startswith("x"):
bFoundHeader = True
t_items = t_line.split(" ")
cols["x"] = t_items.index("x")
cols["y"] = t_items.index("y")
cols["z"] = t_items.index("z")
cols["AD"] = t_items.index("AD")
cols["RD"] = t_items.index("RD")
cols["MD"] = t_items.index("MD")
cols["FA"] = t_items.index("FA")
cols["AD_Avg"] = t_items.index("AD_Avg")
cols["RD_Avg"] = t_items.index("RD_Avg")
cols["MD_Avg"] = t_items.index("MD_Avg")
cols["FA_Avg"] = t_items.index("FA_Avg")
if not bFoundHeader:
raise Exception, "Cannot find header column in file %s" % vso
txt = txt[lidx0 : lidx0 + npts]
#print(txt)
# Allocate space
bvDat["x"] = np.zeros(npts)
bvDat["y"] = np.zeros(npts)
bvDat["z"] = np.zeros(npts)
bvDat["AD"] = np.zeros(npts)
bvDat["RD"] = np.zeros(npts)
bvDat["MD"] = np.zeros(npts)
bvDat["FA"] = np.zeros(npts)
bvDat["AD_Avg"] = np.zeros(npts)
bvDat["RD_Avg"] = np.zeros(npts)
bvDat["MD_Avg"] = np.zeros(npts)
bvDat["FA_Avg"] = np.zeros(npts)
keys = bvDat.keys()
for (i0, t_line) in enumerate(txt):
t_items = t_line.split(" ")
for t_fld in keys:
bvDat[t_fld][i0] = float(t_items[cols[t_fld]])
dat["byVoxel"] = bvDat
return dat
def generate_conn_mat(roiList,
sc_roiList,
parcTypeDir,
parcTracksDir,
hemi,
arg_bSpeech,
maskType,
connFN,
logFN=None):
import os
import sys
import numpy as np
import nibabel as nb
from scai_utils import check_file, check_dir, info_log, error_log
bSC = sc_roiList != None # Subcortical mode flag
# Process cortical ROIs (this is needed for both SC and C matrix types)
mask_shapes = []
roiNames = []
nzIdx = []
bSpeech = []
for (i0, troi) in enumerate(roiList):
targROI = troi[0]
maskFN = os.path.join(parcTypeDir, \
"%s_%s.diff.nii.gz" % (hemi, targROI))
check_file(maskFN, logFN=logFN)
t_img = nb.load(maskFN)
t_img_dat = t_img.get_data()
mask_shapes.append(np.shape(t_img_dat))
t_img_dat = np.ndarray.flatten(t_img_dat)
nzIdx.append(np.nonzero(t_img_dat)[0])
roiNames.append(troi[0])
if troi[2] == 'N':
bSpeech.append(0)
else:
bSpeech.append(1)
roiNames = np.array(roiNames)
bSpeech = np.array(bSpeech)
nzIdx = np.array(nzIdx)
if arg_bSpeech:
roiNames = roiNames[np.nonzero(bSpeech)[0]]
nzIdx = nzIdx[np.nonzero(bSpeech)[0]]
#print(roiNames) # DEBUG
#print(bSpeech) # DEBUG
# Process subcortical ROIs
if bSC:
parcSCDir = os.path.join(os.path.split(parcTypeDir)[0], "subcort")
check_dir(parcSCDir)
sc_roiNames = []
sc_nzIdx = []
for (i0, troi) in enumerate(sc_roiList):
if (hemi == "lh" and troi.startswith("Left-")) or \
(hemi == "rh" and troi.startswith("Right")):
sc_roiNames.append(troi)
maskFN = os.path.join(parcSCDir, \
"%s.diff.nii.gz" % (troi))
check_file(maskFN, logFN=logFN)
t_img = nb.load(maskFN)
t_img_dat = t_img.get_data()
mask_shapes.append(np.shape(t_img_dat))
t_img_dat = np.ndarray.flatten(t_img_dat)
sc_nzIdx.append(np.nonzero(t_img_dat)[0])
#print(sc_nzIdx[-1]) # DEBUG
#print(maskFN) # DEBUG
sc_roiNames = np.array(sc_roiNames)
sc_nzIdx = np.array(sc_nzIdx)
#print(sc_roiNames) # DEBUG
nROIs = len(roiNames)
assert (len(nzIdx) == nROIs)
if len(np.unique(mask_shapes)) != 1:
error_log("Non-unique matrix size among the mask files", logFN=logFN)
imgShape = np.unique(mask_shapes)[0]
if bSC:
nROIs_sc = len(sc_roiNames)
#=== Check the completion of seed-only probtrackx ===#
#=== and calculate the conn matrix ===#
if not bSC:
d1_roiNames = roiNames
d2_roiNames = roiNames
else:
d1_roiNames = sc_roiNames
d2_roiNames = np.array(list(sc_roiNames) + list(roiNames))
connMat = np.zeros([len(d1_roiNames), len(d2_roiNames)])
#print(d2_roiNames) # DEBUG
#print(len(connMat)) # DEBUG
#print(len(connMat[0])) # DEBUG
#print(parcTracksDir) # DEBUG
if bSC:
tmp_dir = os.path.split(parcTracksDir)[1]
parcTracksSCDir = os.path.split(os.path.split(parcTracksDir)[0])[0]
parcTracksSCDir = os.path.join(parcTracksSCDir, "tracks_sc", tmp_dir)
#print(parcTracksSCDir) # DEBUG
check_dir(parcTracksSCDir)
for (i0, troi) in enumerate(d1_roiNames):
seedROI = troi
if not bSC:
trackResDir = os.path.join(parcTracksDir,
"%s_%s_%s" % \
(hemi, seedROI, maskType))
else:
trackResDir = os.path.join(parcTracksSCDir, seedROI)
check_probtrackx_complete(trackResDir,
"seedOnly",
doSeedNorm=True,
doSize=True,
logFN=logFN)
fdt_norm = os.path.join(trackResDir, "fdt_paths_norm.nii.gz")
t_img = nb.load(fdt_norm)
t_img_dat = t_img.get_data()
assert (list(np.shape(t_img_dat)) == list(imgShape))
t_img_dat = np.ndarray.flatten(t_img_dat)
for (i1, troi1) in enumerate(d2_roiNames):
if not bSC:
connMat[i0, i1] = np.mean(t_img_dat[nzIdx[i1]])
else:
if i1 < nROIs_sc:
connMat[i0, i1] = np.mean(t_img_dat[sc_nzIdx[i1]])
else:
connMat[i0, i1] = np.mean(t_img_dat[nzIdx[i1 - nROIs_sc]])
#=== Make symmetric ===#
if not bSC:
connMat = 0.5 * (connMat + connMat.T)
#print(connMat) ## DEBUG
#=== Write result .mat file ===#
from scipy.io import savemat
if not bSC:
res = {"roiNames": roiNames, "connMat": connMat}
else:
res = {
"d1_roiNames": d1_roiNames,
"d2_roiNames": d2_roiNames,
"connMat": connMat
}
savemat(connFN, res)
print("connFN = " + connFN)
check_file(connFN, logFN=logFN)
info_log("Connectivity matrix and associated data were saved at: %s" \
% (connFN),
logFN=logFN)
sID = args.subjID.replace("MRI_", "")
if machineSettings.keys().count(hostName) == 0:
info_log("Cannot find host name %s in machineSettings. _default_ settings will be used" % hostName)
hostName = "_default_"
#=== Set up some paths ===#
dataDir = machineSettings[hostName]["dataDir"]
batchDataDir = machineSettings[hostName]["batchDataDir"]
fsDataDir = machineSettings[hostName]["fsDataDir"]
sDataDir = os.path.join(dataDir, sID)
sBatchDataDir = os.path.join(batchDataDir, sID)
check_dir(dataDir)
check_dir(sDataDir)
check_dir(sBatchDataDir, bCreate=True)
check_dir(fsDataDir)
boldDir = os.path.join(sDataDir, "bold")
check_dir(boldDir)
#=== Look for the fmri_model.mat and fmri_contrasts.mat files ===#
modelMat = os.path.join(sDataDir, "fmri_model.mat")
check_file(modelMat)
info_log("modelMat = %s" % modelMat)
contrMat = os.path.join(sDataDir, "fmri_contrasts.mat")
check_file(contrMat)
info_log("contrMat = %s" % contrMat)
def run_probtrackx(seedMask,
targMask,
bedpBase,
brainMask,
outDir,
doSeedNorm=True,
doSize=True,
doTargMaskedFDT=True,
ccStop=False,
bRedo=False,
logFN=None):
#=========================================================#
# Mode 1: from seed to targ
# Specify both seedMask and targMask
#
# Mode 2: from seed to all
# Specify only seedMask; set targMask=None
#
# Options:
# ccStop: Use corpus callosum stop mask
#
#=========================================================#
import os
from scai_utils import check_file, check_dir, check_bin_path, \
saydo, cmd_stdout, info_log, error_log
from mri_utils import nz_voxels
#== Get seed and targ nvox ==#
check_file(seedMask, logFN=logFN)
(seed_nVoxels, seed_mm3) = nz_voxels(seedMask)
seed_nVoxels = float(seed_nVoxels)
#assert(seed_nVoxels > 0)
if targMask != None:
check_file(targMask, logFN=logFN)
(targ_nVoxels, targ_mm3) = nz_voxels(targMask)
targ_nVoxels = float(targ_nVoxels)
assert (targ_nVoxels > 0)
check_bin_path("probtrackx", logFN=logFN)
check_dir(outDir, logFN=logFN)
if targMask != None:
#= Prepare waypoint file =#
wpfn = os.path.join(outDir, "waypoints.txt")
wptext = os.path.abspath(targMask) + "\n"
wpf = open(wpfn, "w")
wpf.write(wptext)
wpf.close()
check_file(wpfn, logFN=logFN)
cmd = 'probtrackx --mode=seedmask -x %s ' % seedMask + \
'-s %s ' % bedpBase + \
'-m %s ' % brainMask + \
'-l -c 0.2 -S 2000 --steplength=0.5 ' + \
'-P 5000 ' + \
'--forcedir --opd --pd --dir=%s ' % outDir
if targMask != None:
cmd += "--stop=%s --waypoints=%s " % (targMask, wpfn)
fdt_paths_fn = os.path.join(outDir, "fdt_paths.nii.gz")
#== Get the size of fdt_paths.nii.gz. If the size is zero, start over. ==#
if not os.path.isfile(fdt_paths_fn) \
or os.path.getsize(fdt_paths_fn) <= 1 \
or bRedo:
saydo(cmd, logFN=logFN)
#== Check for probtrackx completion ==#
check_file(fdt_paths_fn, logFN=logFN)
#== Save probtrackx command ==#
cmd_fn = os.path.join(outDir, "command.txt")
cmd_f = open(cmd_fn, "wt")
cmd_f.write("%s\n" % cmd)
cmd_f.close()
check_file(cmd_fn, logFN=logFN)
#== Generate seed size-normalized fdt_paths ==#
fdt_paths_norm_fn = os.path.join(outDir, "fdt_paths_norm.nii.gz")
check_bin_path("fslmaths", logFN=logFN)
norm_cmd = "fslmaths -dt float %s -div %d %s -odt float" % \
(fdt_paths_fn, seed_nVoxels, fdt_paths_norm_fn)
if not os.path.isfile(fdt_paths_norm_fn) or bRedo:
saydo(norm_cmd, logFN=logFN)
check_file(fdt_paths_norm_fn, logFN=logFN)
if doSize:
#== Write to seed size file ==#
seed_size_fn = os.path.join(outDir, 'seed_size.txt')
seed_size_f = open(seed_size_fn, 'w')
seed_size_f.write("%d %f" % (int(seed_nVoxels), seed_mm3))
seed_size_f.close()
check_file(seed_size_fn, logFN=logFN)
info_log("INFO: Saved seed size data to file: %s" % seed_size_fn,
logFN=logFN)
if targMask != None:
#== Write to targ size file ==#
targ_size_fn = os.path.join(outDir, 'targ_size.txt')
targ_size_f = open(targ_size_fn, 'w')
targ_size_f.write("%d %f" % (int(targ_nVoxels), targ_mm3))
targ_size_f.close()
check_file(targ_size_fn, logFN=logFN)
info_log("INFO: Saved targ size data to file: %s" % targ_size_fn,
logFN=logFN)
if (targMask != None) and doTargMaskedFDT:
#== Get target masked tract density ==#
check_bin_path("fslstats", logFN=logFN)
(so, se) = cmd_stdout("fslstats %s -k %s -m" \
% (fdt_paths_norm_fn, targMask))
assert (len(se) == 0)
so = so.split()
assert (len(so) >= 1)
targ_masked_norm_fdt = float(so[0])
targ_masked_norm_fdt_fn = \
os.path.join(outDir, "targ_masked_norm_fdt.txt")
tmnff = open(targ_masked_norm_fdt_fn, "wt")
tmnff.write("%f" % targ_masked_norm_fdt)
tmnff.close()
check_file(targ_masked_norm_fdt_fn, logFN=logFN)
info_log("INFO: Saved target-masked normalized FDT value tofile: %s" \
% targ_masked_norm_fdt_fn,
logFN=logFN)
help="Functional activation (sig) mid value (e.g., 5)")
if len(sys.argv) == 1:
ap.print_help()
sys.exit(1)
args = ap.parse_args()
if args.type == "both":
types = ["sf10", "vol"]
else:
types = [args.type]
check_bin_path(CONVERT_BIN)
check_dir(CON_IMG_DIR)
l2Base = os.path.abspath(L2_BATCH_BASE)
check_dir(l2Base)
stitchRes = {}
for t_type in types:
conDir = os.path.join(l2Base,
"%s_%s_%s" % (args.grp, args.contr, t_type))
check_dir(conDir)
stitchRes[t_type] = {}
if t_type.startswith("sf"):
#=== OSGM and BGC ===#
if args.grp == "ALL":
def calculate_roi_tensor_measures(parcName, roiList, wmDepths,
TENSOR_MEASURES, HEMIS,
dmriDir, annotDir,
tensMeasMatFN, logFileName):
#=== Load masks (gm and wm of different depths) ===#
import os
import nibabel as nb
import numpy as np
from scai_utils import check_file, check_dir, info_log, error_log
mask_shapes = []
nDepths = len(wmDepths)
roiNames = []
nzIdx = []
for i0 in range(nDepths):
nzIdx.append([])
bSpeech = []
parcDir = os.path.join(annotDir, parcName)
for (i0, wmDepth) in enumerate(wmDepths):
if wmDepth == -1:
parcTypeDir = os.path.join(parcDir, "gm")
info_log("Loading gray-matter masks")
else:
parcTypeDir = os.path.join(parcDir, "wm%dmm" % wmDepth)
info_log("Loading white-matter masks of %d-mm depth" \
% wmDepth)
for (i1, troi) in enumerate(roiList):
for (i2, hemi) in enumerate(HEMIS):
maskFN = os.path.join(parcTypeDir, \
"%s_%s.diff.nii.gz" % (hemi, troi[0]))
check_file(maskFN, logFN=logFileName)
t_img = nb.load(maskFN)
t_img_dat = t_img.get_data()
mask_shapes.append(np.shape(t_img_dat))
t_img_dat = np.ndarray.flatten(t_img_dat)
nzIdx[i0].append(np.nonzero(t_img_dat)[0])
if wmDepth == -1:
if troi[2] == 'N':
bSpeech.append(0)
else:
bSpeech.append(1)
roiNames.append("%s_%s" % (hemi, troi[0]))
#=== Check that the dimensions of all mask images match ===#
if len(np.unique(mask_shapes)) != 1:
error_log("Non-unique matrix size among the mask files",
logFN=logFN)
#=== Load the dtifit_* files and extract the measures ===#
nROIs = len(roiNames)
assert(len(bSpeech) == nROIs)
tensMeas = {}
mask_shape = np.unique(mask_shapes)
check_dir(dmriDir, logFN=logFileName)
for (i0, measName) in enumerate(TENSOR_MEASURES):
tensMeas[measName] = np.zeros([nROIs, nDepths])
for (i1, t_depth) in enumerate(wmDepths):
if t_depth == -1:
info_log("Extracting tensor measure %s from gray matter" \
% measName)
else:
info_log("Extracting tensor measure %s from %d-mm deep white matter" % (measName, t_depth))
assert(len(nzIdx[i1]) == nROIs)
for (i2, troi) in enumerate(roiNames):
measImg = os.path.join(dmriDir, \
"dtifit_%s.nii.gz" % measName)
check_file(measImg, logFN=logFileName)
t_img = nb.load(measImg)
if not list(mask_shape[0]) == list(np.shape(t_img)):
error_log("The diffusion tensor measure volume %s (%s) does not have a dimension that matches those of the mask files" \
% (measImg, measName))
t_img_dat = np.ndarray.flatten(t_img.get_data())
tensMeas[measName][i2, i1] = \
np.mean(t_img_dat[nzIdx[i1][i2]])
#=== Write data to file ===#
from scipy.io import savemat
res = {"roiNames": roiNames,
"bSpeech": bSpeech,
"parcName": parcName,
"maskShape": mask_shape,
"wmDepths": wmDepths,
"tensMeas": tensMeas}
savemat(tensMeasMatFN, res)
check_file(tensMeasMatFN, logFN=logFileName)
info_log("Tensor measures (%d types) and associated data were saved at: %s"
% (len(TENSOR_MEASURES), tensMeasMatFN),
logFN=logFileName)
def generate_conn_mat(roiList, sc_roiList,
parcTypeDir, parcTracksDir, hemi,
arg_bSpeech, maskType, connFN, logFN=None):
import os
import sys
import numpy as np
import nibabel as nb
from scai_utils import check_file, check_dir, info_log, error_log
bSC = sc_roiList != None # Subcortical mode flag
# Process cortical ROIs (this is needed for both SC and C matrix types)
mask_shapes = []
roiNames = []
nzIdx = []
bSpeech = []
for (i0, troi) in enumerate(roiList):
targROI = troi[0]
maskFN = os.path.join(parcTypeDir, \
"%s_%s.diff.nii.gz" % (hemi, targROI))
check_file(maskFN, logFN=logFN)
t_img = nb.load(maskFN)
t_img_dat = t_img.get_data()
mask_shapes.append(np.shape(t_img_dat))
t_img_dat = np.ndarray.flatten(t_img_dat)
nzIdx.append(np.nonzero(t_img_dat)[0])
roiNames.append(troi[0])
if troi[2] == 'N':
bSpeech.append(0)
else:
bSpeech.append(1)
roiNames = np.array(roiNames)
bSpeech = np.array(bSpeech)
nzIdx = np.array(nzIdx)
if arg_bSpeech:
roiNames = roiNames[np.nonzero(bSpeech)[0]]
nzIdx = nzIdx[np.nonzero(bSpeech)[0]]
#print(roiNames) # DEBUG
#print(bSpeech) # DEBUG
# Process subcortical ROIs
if bSC:
parcSCDir = os.path.join(os.path.split(parcTypeDir)[0], "subcort")
check_dir(parcSCDir)
sc_roiNames = []
sc_nzIdx = []
for (i0, troi) in enumerate(sc_roiList):
if (hemi == "lh" and troi.startswith("Left-")) or \
(hemi == "rh" and troi.startswith("Right")):
sc_roiNames.append(troi)
maskFN = os.path.join(parcSCDir, \
"%s.diff.nii.gz" % (troi))
check_file(maskFN, logFN=logFN)
t_img = nb.load(maskFN)
t_img_dat = t_img.get_data()
mask_shapes.append(np.shape(t_img_dat))
t_img_dat = np.ndarray.flatten(t_img_dat)
sc_nzIdx.append(np.nonzero(t_img_dat)[0])
#print(sc_nzIdx[-1]) # DEBUG
#print(maskFN) # DEBUG
sc_roiNames = np.array(sc_roiNames)
sc_nzIdx = np.array(sc_nzIdx)
#print(sc_roiNames) # DEBUG
nROIs = len(roiNames)
assert(len(nzIdx) == nROIs)
if len(np.unique(mask_shapes)) != 1:
error_log("Non-unique matrix size among the mask files", logFN=logFN)
imgShape = np.unique(mask_shapes)[0]
if bSC:
nROIs_sc = len(sc_roiNames)
#=== Check the completion of seed-only probtrackx ===#
#=== and calculate the conn matrix ===#
if not bSC:
d1_roiNames = roiNames
d2_roiNames = roiNames
else:
d1_roiNames = sc_roiNames
d2_roiNames = np.array(list(sc_roiNames) + list(roiNames))
connMat = np.zeros([len(d1_roiNames), len(d2_roiNames)])
#print(d2_roiNames) # DEBUG
#print(len(connMat)) # DEBUG
#print(len(connMat[0])) # DEBUG
#print(parcTracksDir) # DEBUG
if bSC:
tmp_dir = os.path.split(parcTracksDir)[1]
parcTracksSCDir = os.path.split(os.path.split(parcTracksDir)[0])[0]
parcTracksSCDir = os.path.join(parcTracksSCDir, "tracks_sc", tmp_dir)
#print(parcTracksSCDir) # DEBUG
check_dir(parcTracksSCDir)
for (i0, troi) in enumerate(d1_roiNames):
seedROI = troi
if not bSC:
trackResDir = os.path.join(parcTracksDir,
"%s_%s_%s" % \
(hemi, seedROI, maskType))
else:
trackResDir = os.path.join(parcTracksSCDir, seedROI)
check_probtrackx_complete(trackResDir, "seedOnly",
doSeedNorm=True, doSize=True,
logFN=logFN)
fdt_norm = os.path.join(trackResDir, "fdt_paths_norm.nii.gz")
t_img = nb.load(fdt_norm)
t_img_dat = t_img.get_data()
assert(list(np.shape(t_img_dat)) == list(imgShape))
t_img_dat = np.ndarray.flatten(t_img_dat)
for (i1, troi1) in enumerate(d2_roiNames):
if not bSC:
connMat[i0, i1] = np.mean(t_img_dat[nzIdx[i1]])
else:
if i1 < nROIs_sc:
connMat[i0, i1] = np.mean(t_img_dat[sc_nzIdx[i1]])
else:
connMat[i0, i1] = np.mean(t_img_dat[nzIdx[i1 - nROIs_sc]])
#=== Make symmetric ===#
if not bSC:
connMat = 0.5 * (connMat + connMat.T)
#print(connMat) ## DEBUG
#=== Write result .mat file ===#
from scipy.io import savemat
if not bSC:
res = {"roiNames": roiNames,
"connMat": connMat}
else:
res = {"d1_roiNames": d1_roiNames,
"d2_roiNames": d2_roiNames,
"connMat": connMat}
savemat(connFN, res)
print("connFN = " + connFN)
check_file(connFN, logFN=logFN)
info_log("Connectivity matrix and associated data were saved at: %s" \
% (connFN),
logFN=logFN)
def run_probtrackx(seedMask, targMask, bedpBase, brainMask, outDir,
doSeedNorm=True, doSize=True,
doTargMaskedFDT=True,
ccStop=False,
bRedo=False,
logFN=None):
#=========================================================#
# Mode 1: from seed to targ
# Specify both seedMask and targMask
#
# Mode 2: from seed to all
# Specify only seedMask; set targMask=None
#
# Options:
# ccStop: Use corpus callosum stop mask
#
#=========================================================#
import os
from scai_utils import check_file, check_dir, check_bin_path, \
saydo, cmd_stdout, info_log, error_log
from mri_utils import nz_voxels
#== Get seed and targ nvox ==#
check_file(seedMask, logFN=logFN)
(seed_nVoxels, seed_mm3) = nz_voxels(seedMask)
seed_nVoxels = float(seed_nVoxels)
#assert(seed_nVoxels > 0)
if targMask != None:
check_file(targMask, logFN=logFN)
(targ_nVoxels, targ_mm3) = nz_voxels(targMask)
targ_nVoxels = float(targ_nVoxels)
assert(targ_nVoxels > 0)
check_bin_path("probtrackx", logFN=logFN)
check_dir(outDir, logFN=logFN)
if targMask != None:
#= Prepare waypoint file =#
wpfn = os.path.join(outDir, "waypoints.txt")
wptext = os.path.abspath(targMask) + "\n"
wpf = open(wpfn, "w")
wpf.write(wptext)
wpf.close()
check_file(wpfn, logFN=logFN)
cmd = 'probtrackx --mode=seedmask -x %s ' % seedMask + \
'-s %s ' % bedpBase + \
'-m %s ' % brainMask + \
'-l -c 0.2 -S 2000 --steplength=0.5 ' + \
'-P 5000 ' + \
'--forcedir --opd --pd --dir=%s ' % outDir
if targMask != None:
cmd += "--stop=%s --waypoints=%s " % (targMask, wpfn)
fdt_paths_fn = os.path.join(outDir, "fdt_paths.nii.gz")
#== Get the size of fdt_paths.nii.gz. If the size is zero, start over. ==#
if not os.path.isfile(fdt_paths_fn) \
or os.path.getsize(fdt_paths_fn) <= 1 \
or bRedo:
saydo(cmd, logFN=logFN)
#== Check for probtrackx completion ==#
check_file(fdt_paths_fn, logFN=logFN)
#== Save probtrackx command ==#
cmd_fn = os.path.join(outDir, "command.txt")
cmd_f = open(cmd_fn, "wt")
cmd_f.write("%s\n" % cmd)
cmd_f.close()
check_file(cmd_fn, logFN=logFN)
#== Generate seed size-normalized fdt_paths ==#
fdt_paths_norm_fn = os.path.join(outDir, "fdt_paths_norm.nii.gz")
check_bin_path("fslmaths", logFN=logFN)
norm_cmd = "fslmaths -dt float %s -div %d %s -odt float" % \
(fdt_paths_fn, seed_nVoxels, fdt_paths_norm_fn)
if not os.path.isfile(fdt_paths_norm_fn) or bRedo:
saydo(norm_cmd, logFN=logFN)
check_file(fdt_paths_norm_fn, logFN=logFN)
if doSize:
#== Write to seed size file ==#
seed_size_fn = os.path.join(outDir, 'seed_size.txt')
seed_size_f = open(seed_size_fn, 'w')
seed_size_f.write("%d %f" % (int(seed_nVoxels), seed_mm3))
seed_size_f.close()
check_file(seed_size_fn, logFN=logFN)
info_log("INFO: Saved seed size data to file: %s" % seed_size_fn,
logFN=logFN)
if targMask != None:
#== Write to targ size file ==#
targ_size_fn = os.path.join(outDir, 'targ_size.txt')
targ_size_f = open(targ_size_fn, 'w')
targ_size_f.write("%d %f" % (int(targ_nVoxels), targ_mm3))
targ_size_f.close()
check_file(targ_size_fn, logFN=logFN)
info_log("INFO: Saved targ size data to file: %s" % targ_size_fn,
logFN=logFN)
if (targMask != None) and doTargMaskedFDT:
#== Get target masked tract density ==#
check_bin_path("fslstats", logFN=logFN)
(so, se) = cmd_stdout("fslstats %s -k %s -m" \
% (fdt_paths_norm_fn, targMask))
assert(len(se) == 0)
so = so.split()
assert(len(so) >= 1)
targ_masked_norm_fdt = float(so[0])
targ_masked_norm_fdt_fn = \
os.path.join(outDir, "targ_masked_norm_fdt.txt")
tmnff = open(targ_masked_norm_fdt_fn, "wt")
tmnff.write("%f" % targ_masked_norm_fdt)
tmnff.close()
check_file(targ_masked_norm_fdt_fn, logFN=logFN)
info_log("INFO: Saved target-masked normalized FDT value tofile: %s" \
% targ_masked_norm_fdt_fn,
logFN=logFN)
