if args.fwhmSeg is not None:
voxSize = pet.header.get_zooms()
sigmas = np.divide(args.fwhmSeg[0] / np.sqrt(8.0 * np.log(2.0)),
voxSize[0:3])
cbvData = filt.gaussian_filter(cbvData, sigmas).reshape(cbvData.shape)
cbvData[segData == 0] = 0.0
#Get mask where inputs are non_zero
maskData = np.where(
np.logical_and(
roiData != 0,
np.logical_and(np.logical_and(brainData != 0, cbfData != 0),
cbvData != 0)), 1, 0)
#Flatten the PET images and then mask
petMasked = nagini.reshape4d(petData)[maskData.flatten() > 0, :]
roiMasked = roiData[maskData > 0].flatten()
#Prep CBF and CBV data
if args.fwhm is None:
#Do not smooth,just mask
cbfMasked = cbfData[maskData > 0].flatten()
cbvMasked = cbvData[maskData > 0].flatten()
else:
#Prepare for smoothing
voxSize = pet.header.get_zooms()
sigmas = np.divide(args.fwhm[0] / np.sqrt(8.0 * np.log(2.0)), voxSize[0:3])
#Smooth and mask data
cbfMasked = filt.gaussian_filter(cbfData, sigmas)[maskData > 0].flatten()
cbvMasked = filt.gaussian_filter(cbvData, sigmas)[maskData > 0].flatten()
#!/usr/bin/python
#Parse arguments
import argparse, sys
argParse = argparse.ArgumentParser(description='Convert ECAT PET image to decay corrected well counts/mL/sec:')
argParse.add_argument('pet',help='Nifti ECAT count image',nargs=1,type=str)
argParse.add_argument('info',help='Yi Su style info file',nargs=1,type=str)
argParse.add_argument('pie',help='Pie calibration factor',nargs=1,type=float)
argParse.add_argument('out',help='Root for outputed files',nargs=1,type=str)
args = argParse.parse_args()
#Load in libraries we will need
import numpy as np, nibabel as nib, nagini
#Load in PET image
pet = nib.load(args.pet[0]); petData = nagini.reshape4d(pet.get_data())
#Load in info file
info = nagini.loadInfo(args.info[0])
#Loop through frames
for fIdx in range(petData.shape[1]):
#Convert to decay corrected well counts
petData[:,fIdx] *= 60.0 * args.pie[0] / info[fIdx,2] * info[fIdx,3]
#Write out result
well = nib.Nifti1Image(petData.reshape(pet.shape),pet.affine,header=pet.header)
well.to_filename(args.out[0]+'.nii.gz')
#Load in mask data
brainData = brain.get_data()
else:
#Make a fake mask
brainData = np.ones(pet.shape[0:3])
#Make flat mask
if args.ps is not None:
flatMask = np.logical_and(brainData.flatten() > 0, cbfData.flatten() > 0)
cbfMasked = cbfData.flatten()[flatMask]
else:
flatMask = brainData.flatten() > 0
#Flatten the PET images and then mask
petMasked = nagini.reshape4d(petData)[flatMask, :]
#Get pet start and end times. Assume start of first recorded frame is injection (has an offset which is subtracted out)
startTime = info[:, 0] - info[0, 0]
midTime = info[:, 1] - info[0, 0]
endTime = info[:, 2] + startTime
#Get aif time variable. Assume clock starts at injection.
aifTime = aif[:, 0]
#Extract AIF counts
aifC = aif[:, 1]
#Logic for preparing blood sucker curves
if args.dcv != 1:
import numpy as np, nibabel as nib, matplotlib.pyplot as plt, nagini
#Load in pet and mask headers
pet = nib.load(args.pet[0])
mask = nib.load(args.mask[0])
#Load in info file
info = nagini.loadInfo(args.info[0])
#Mask sure mask and pet dimensions mask
if pet.shape[0:3] != mask.shape[0:3]:
print 'Error: Images do not have the same dimensions'
sys.exit()
#Get image data
petData = nagini.reshape4d(pet.get_data())
maskData = mask.get_data().flatten()
#Mask pet data
petMasked = petData[maskData == 1, :]
#Get PET time vetor
petTime = info[:, 1]
#Caculate whole-brain timecourse
petMean = np.mean(petMasked, axis=0)
#Get spline knots and basis
petKnots = nagini.knotLoc(petTime, args.nKnots[0])
petKnots[0] = 10
petBasis = nagini.rSplineBasis(petTime, petKnots)
pet = nagini.loadHeader(args.pet[0]) brain = nagini.loadHeader(args.brain[0]) roi = nagini.loadHeader(args.roi[0]) #Check to make sure dimensions match if pet.shape[0:3] != brain.shape[0:3] or pet.shape[3] != info.shape[0] or brain.shape[0:3] != roi.shape[0:3]: print 'ERROR: Data dimensions do not match. Please check...' sys.exit() #Get the image data petData = pet.get_data() brainData = brain.get_data() roiData = roi.get_data() #Flatten the PET images and then mask petFlat = nagini.reshape4d(petData) petMasked = petFlat[brainData.flatten()>0,:] #Get pet into rois petRoi = nagini.roiAvg(petFlat,roiData.flatten()) #If cbv image is given, load it up if args.cbv is not None: #Load in CBV image cbv = nagini.loadHeader(args.cbv[0]) if cbv.shape[0:3] != pet.shape[0:3]: print 'ERROR: CBV image does not match PET resolution...' sys.exit() cbvData = cbv.get_data()
#Remask CBV image from ROI averages cbvData = nagini.roiBack(cbvAvgs,segData.flatten()).reshape(cbvData.shape) #Smooth if args.fwhmSeg is not None: voxSize = pet.header.get_zooms() sigmas = np.divide(args.fwhmSeg[0]/np.sqrt(8.0*np.log(2.0)),voxSize[0:3]) cbvData = filt.gaussian_filter(cbvData,sigmas).reshape(cbvData.shape) cbvData[segData==0] = 0.0 #Get mask where inputs are non_zero maskData = np.where(np.logical_and(np.logical_and(brainData!=0,cbfData!=0),cbvData!=0),1,0) #Flatten the PET images and then mask petMasked = nagini.reshape4d(petData)[maskData.flatten()>0,:] #Prep CBF and CBV data if args.fwhm is None: #Do not smooth,just mask cbfMasked = cbfData[maskData>0].flatten() cbvMasked = cbvData[maskData>0].flatten() else: #Prepare for smoothing voxSize = pet.header.get_zooms()[0:3] sigmas = np.divide(args.fwhm[0]/np.sqrt(8.0*np.log(2.0)),voxSize[0:3]) #Smooth and mask data cbfMasked = filt.gaussian_filter(cbfData,sigmas)[maskData>0].flatten() cbvMasked = filt.gaussian_filter(cbvData,sigmas)[maskData>0].flatten()
if pet.shape[0:3] != brain.shape[0:3] or pet.shape[0:3] != cbf.shape[0:3] or \
pet.shape[0:3] != lmbda.shape[0:3] or pet.shape[0:3] != cbv.shape[0:3] or \
pet.shape[3] != idaif.shape[0] or pet.shape[3] != info.shape[0]:
print 'ERROR: Data dimensions do not match. Please check...'
sys.exit()
#Get the image data
petData = pet.get_data()
cbfData = cbf.get_data()
lmbdaData = lmbda.get_data()
cbvData = cbv.get_data()
brainData = brain.get_data()
#Flatten the PET images and then mask. Also convert parameteric images back to orginal PET units.
brainMask = brainData.flatten()
petMasked = nagini.reshape4d(petData)[brainMask > 0, :]
cbfMasked = cbfData.flatten()[brainMask > 0] / 6000.0 * args.d
lmbdaMasked = lmbdaData.flatten()[brainMask > 0] * args.d
cbvMasked = cbvData.flatten()[brainMask > 0] / 100 * args.d
#Limit pet range
timeMask = np.logical_and(petTime >= petRange[0], petTime <= petRange[1])
petTime = petTime[timeMask]
idaif = idaif[timeMask]
petMasked = petMasked[:, timeMask]
#Interpolate the aif to minimum sampling time
minTime = np.min(np.diff(petTime))
interpTime = np.arange(petTime[0], np.ceil(petTime[-1] + minTime), minTime)
nTime = interpTime.shape[0]
aifLinear = interp.interp1d(petTime,
maskData = np.ones((seg.shape[0], seg.shape[1], seg.shape[2]))
#Load in image data
petData = pet.get_data()
segData = seg.get_data()
#Remove 4th dimension of segmentation
if len(segData.shape) == 4:
segData = segData[:, :, :, 0]
#Make a flattened version of the segmentation for use later
segFlat = segData.flatten()
#Reshape PET data as necessary
if len(petData.shape) == 4:
petData = nagini.reshape4d(petData)
nPet = petData.shape[1]
else:
petData = petData.reshape((pet.shape[0] * pet.shape[1] * pet.shape[2], 1))
nPet = 1
#Get ROI list and number of ROIs
roiList = np.unique(segData)
if args.noZero == 1:
roiList = roiList[roiList != 0]
nRoi = roiList.shape[0]
#Make weight matrices
if args.weight is not None:
#Load w matrix
if pet.shape[0:3] != brain.shape[0:3] or pet.shape[0:3] != cbf.shape[0:3] or \ pet.shape[0:3] != cbv.shape[0:3] or pet.shape[3] != idaif.shape[0] or \ pet.shape[3] != info.shape[0]: print 'ERROR: Data dimensions do not match. Please check...' sys.exit() #Get the image data petData = pet.get_data() cbfData = cbf.get_data() cbvData = cbv.get_data() brainData = brain.get_data() #Flatten the PET images and then mask. Also convert parameteric images back to original PET units. brainData = np.logical_and(np.logical_and(cbfData!=0,cbvData!=0),brainData!=0) brainMask = brainData.flatten() petMasked = nagini.reshape4d(petData)[brainMask,:] cbfMasked = cbfData.flatten()[brainMask] / 6000 * args.d cbvMasked = cbvData.flatten()[brainMask] / 100 * args.d sys.exit() #Interpolate the aif to minimum sampling time. Make sure we get the last time even if we have to go over range. minTime = np.min(np.diff(petTime[petTime<petRange[1]])) interpTime = np.arange(petRange[0],np.ceil(petRange[1]+minTime),minTime) nTime = interpTime.shape[0] aifLinear = interp.interp1d(petTime,idaif,kind="linear",fill_value="extrapolate") aifInterp = aifLinear(interpTime) #Get input function for h20 and oxygen seperately aifDelay = aifLinear(interpTime-args.delay); aifDelay[aifDelay<0] = 0 aifWater = args.decay*np.convolve(aifDelay,np.exp(-args.decay*interpTime))[0:nTime]*minTime aifOxy = aifInterp - aifWater
#Load libraries import numpy as np, nibabel as nib, matplotlib.pyplot as plt, nagini #Load in pet and mask headers pet = nib.load(args.pet[0]); mask = nib.load(args.mask[0]); #Load in info file info = nagini.loadInfo(args.info[0]) #Mask sure mask and pet dimensions mask if pet.shape[0:3] != mask.shape[0:3]: print 'Error: Images do not have the same dimensions' sys.exit() #Get image data petData = nagini.reshape4d(pet.get_data()) maskData = mask.get_data().flatten() #Mask pet data petMasked = petData[maskData==1,:] #Get PET time vetor petTime = info[:,1] #Caculate whole-brain timecourse petMean = np.mean(petMasked,axis=0) #Get spline knots and basis petKnots = nagini.knotLoc(petTime,args.nKnots[0]); petKnots[0] = 10 petBasis = nagini.rSplineBasis(petTime,petKnots)
args = argParse.parse_args()
#Load needed libraries
import numpy as np, nibabel as nib, nagini, sys
#Load images headers
val = nagini.loadHeader(args.val[0])
roi = nagini.loadHeader(args.roi[0])
#Load image data
valData = val.get_data()
roiData = roi.get_data()
#Reshape image data as necessary
if len(valData.shape) == 4:
valData = nagini.reshape4d(valData)
else:
valData = valData.reshape((val.shape[0],1))
roiData = roiData.flatten()
#Sample into ROIs
projData = nagini.roiBack(valData,roiData)
#Save projected ROI image
proj = nib.Nifti1Image(projData.reshape((roi.shape[0],roi.shape[1],roi.shape[2],projData.shape[-1])),roi.affine,header=roi.header)
proj.to_filename('%s_roiProj.nii.gz'%(args.out[0]))
#Load image headers
img = nagini.loadHeader(args.img[0])
roi = nagini.loadHeader(args.roi[0])
#Check to make sure images have same dimensions
if img.shape[0:3] != roi.shape[0:3]:
print 'ERROR: Images do not have same dimensions. Exiting...'
sys.exit()
#Load image data
imgData = img.get_data()
roiData = roi.get_data()
#Reshape image data as necessary
if len(imgData.shape) == 4:
imgData = nagini.reshape4d(imgData)
else:
imgData = imgData.reshape((img.shape[0]*img.shape[1]*img.shape[2],1))
roiData = roiData.flatten()
#Sample into ROIs
avgData = nagini.roiAvg(imgData,roiData,stat=args.stat)
#Save roi averages in the format user wants.
if args.nii == 1:
avg = nib.Nifti1Image(avgData.reshape((avgData.shape[0],1,1,avgData.shape[1])),np.identity(4))
avg.to_filename('%s_roi_%s.nii.gz'%(args.out[0],args.stat))
else:
nagini.writeText('%s_roi_%s.txt'%(args.out[0],args.stat),avgData)
argParse.add_argument('out', help='Root for output file', nargs=1)
args = argParse.parse_args()
#Load needed libraries
import numpy as np, nibabel as nib, nagini, sys
#Load images headers
val = nagini.loadHeader(args.val[0])
roi = nagini.loadHeader(args.roi[0])
#Load image data
valData = val.get_data()
roiData = roi.get_data()
#Reshape image data as necessary
if len(valData.shape) == 4:
valData = nagini.reshape4d(valData)
else:
valData = valData.reshape((val.shape[0], 1))
roiData = roiData.flatten()
#Sample into ROIs
projData = nagini.roiBack(valData, roiData)
#Save projected ROI image
proj = nib.Nifti1Image(projData.reshape(
(roi.shape[0], roi.shape[1], roi.shape[2], projData.shape[-1])),
roi.affine,
header=roi.header)
proj.to_filename('%s_roiProj.nii.gz' % (args.out[0]))
#Load image headers
img = nagini.loadHeader(args.img[0])
roi = nagini.loadHeader(args.roi[0])
#Check to make sure images have same dimensions
if img.shape[0:3] != roi.shape[0:3]:
print 'ERROR: Images do not have same dimensions. Exiting...'
sys.exit()
#Load image data
imgData = img.get_data()
roiData = roi.get_data()
#Reshape image data as necessary
if len(imgData.shape) == 4:
imgData = nagini.reshape4d(imgData)
else:
imgData = imgData.reshape((img.shape[0] * img.shape[1] * img.shape[2], 1))
roiData = roiData.flatten()
#Sample into ROIs
avgData = nagini.roiAvg(imgData, roiData, stat=args.stat)
#Save roi averages in the format user wants.
if args.nii == 1:
avg = nib.Nifti1Image(
avgData.reshape((avgData.shape[0], 1, 1, avgData.shape[1])),
np.identity(4))
avg.to_filename('%s_roi_%s.nii.gz' % (args.out[0], args.stat))
else:
nagini.writeText('%s_roi_%s.txt' % (args.out[0], args.stat), avgData)
maskData = np.ones((seg.shape[0],seg.shape[1],seg.shape[2])) #Load in image data petData = pet.get_data() segData = seg.get_data() #Remove 4th dimension of segmentation if len(segData.shape) == 4: segData = segData[:,:,:,0] #Make a flattened version of the segmentation for use later segFlat = segData.flatten() #Reshape PET data as necessary if len(petData.shape) == 4: petData = nagini.reshape4d(petData) nPet = petData.shape[1] else: petData = petData.reshape((pet.shape[0]*pet.shape[1]*pet.shape[2],1)) nPet = 1 #Get ROI list and number of ROIs roiList = np.unique(segData) if args.noZero == 1: roiList = roiList[roiList!=0] nRoi = roiList.shape[0] #Make weight matrices if args.weight is not None: #Load w matrix
