def iterativeBackPropagation(hrImage, lrImages, lrMasks, transforms, H,
itermax, interpOrder):
#Convert LR images to a list of vectors
y = []
for i in range(len(lrImages)):
y.append(
convert_image_to_vector(lrImages[i]) *
convert_image_to_vector(lrMasks[i]))
#Convert HR Image to vector
x = convert_image_to_vector(hrImage)
outputImage = nibabel.Nifti1Image(hrImage.get_data(), hrImage.affine)
#Compute HR mask
hrMaskSum = np.zeros(hrImage.get_data().shape, dtype=np.float32)
for i in range(len(lrImages)):
tmp1 = apply_affine_itk_transform_on_image(input_image=lrMasks[i],
transform=transforms[i][0],
center=transforms[i][1],
reference_image=hrImage,
order=0)
hrMaskSum += tmp1.get_data()
#index = np.nonzero(hrMaskSum)
for j in range(itermax):
error = ibpComputeError(x, H, y,
nibabel.Nifti1Image(hrMaskSum, hrImage.affine),
lrImages, transforms, interpOrder)
# #simulation and error computation
# hrError = np.zeros(hrImage.get_data().shape, dtype=np.float32)
#
# for i in range(len(lrImages)):
# lrError = convert_vector_to_image(H[i].dot(x)-y[i], lrImages[i])
# tmp2 = apply_affine_itk_transform_on_image(input_image = lrError, transform=transforms[i][0], center=transforms[i][1], reference_image=hrImage, order=interpOrder)
# hrError += tmp2.get_data()
#
# hrError2 = np.zeros(hrImage.get_data().shape, dtype=np.float32)
# hrError2[index] = hrError[index] / hrMaskSum[index]
#
#filter error map
from skimage.restoration import denoise_tv_chambolle
hrError2 = denoise_tv_chambolle(error, weight=5)
#update hr image and x
outputImage = nibabel.Nifti1Image(outputImage.get_data() - hrError2,
hrImage.affine)
nibabel.save(nibabel.Nifti1Image(hrError2, hrImage.affine),
'error_iter' + str(j) + '.nii.gz')
nibabel.save(outputImage, 'ibp_iter' + str(j) + '.nii.gz')
x = convert_image_to_vector(outputImage)
return outputImage
def iterativeBackPropagation(hrImage, lrImages, lrMasks, transforms, H, itermax, interpOrder):
# Convert LR images to a list of vectors
y = []
for i in range(len(lrImages)):
y.append(convert_image_to_vector(lrImages[i]) * convert_image_to_vector(lrMasks[i]))
# Convert HR Image to vector
x = convert_image_to_vector(hrImage)
outputImage = nibabel.Nifti1Image(hrImage.get_data(), hrImage.affine)
# Compute HR mask
hrMaskSum = np.zeros(hrImage.get_data().shape, dtype=np.float32)
for i in range(len(lrImages)):
tmp1 = apply_affine_itk_transform_on_image(
input_image=lrMasks[i],
transform=transforms[i][0],
center=transforms[i][1],
reference_image=hrImage,
order=0,
)
hrMaskSum += tmp1.get_data()
# index = np.nonzero(hrMaskSum)
for j in range(itermax):
error = ibpComputeError(
x, H, y, nibabel.Nifti1Image(hrMaskSum, hrImage.affine), lrImages, transforms, interpOrder
)
# #simulation and error computation
# hrError = np.zeros(hrImage.get_data().shape, dtype=np.float32)
#
# for i in range(len(lrImages)):
# lrError = convert_vector_to_image(H[i].dot(x)-y[i], lrImages[i])
# tmp2 = apply_affine_itk_transform_on_image(input_image = lrError, transform=transforms[i][0], center=transforms[i][1], reference_image=hrImage, order=interpOrder)
# hrError += tmp2.get_data()
#
# hrError2 = np.zeros(hrImage.get_data().shape, dtype=np.float32)
# hrError2[index] = hrError[index] / hrMaskSum[index]
#
# filter error map
from skimage.restoration import denoise_tv_chambolle
hrError2 = denoise_tv_chambolle(error, weight=5)
# update hr image and x
outputImage = nibabel.Nifti1Image(outputImage.get_data() - hrError2, hrImage.affine)
nibabel.save(nibabel.Nifti1Image(hrError2, hrImage.affine), "error_iter" + str(j) + ".nii.gz")
nibabel.save(outputImage, "ibp_iter" + str(j) + ".nii.gz")
x = convert_image_to_vector(outputImage)
return outputImage
def iterativeBackPropagation(hrImage, lrImages, lrMasks, transforms, H,
itermax):
y = []
for i in range(len(lrImages)):
y.append(
convert_image_to_vector(lrImages[i]) *
convert_image_to_vector(lrMasks[i]))
x = convert_image_to_vector(hrImage)
outputImage = nibabel.Nifti1Image(hrImage.get_data(), hrImage.affine)
hrMaskSum = np.zeros(hrImage.get_data().shape, dtype=np.float32)
for i in range(len(lrImages)):
tmp1 = apply_affine_itk_transform_on_image(input_image=lrMasks[i],
transform=transforms[i][0],
center=transforms[i][1],
reference_image=hrImage,
order=0)
hrMaskSum += tmp1.get_data()
index = np.nonzero(hrMaskSum)
for j in range(itermax):
#simulation and error computation
hrError = np.zeros(hrImage.get_data().shape, dtype=np.float32)
for i in range(len(lrImages)):
lrError = convert_vector_to_image(H[i].dot(x) - y[i], lrImages[i])
tmp2 = apply_affine_itk_transform_on_image(
input_image=lrError,
transform=transforms[i][0],
center=transforms[i][1],
reference_image=hrImage,
order=1)
hrError += tmp2.get_data()
hrError2 = np.zeros(hrImage.get_data().shape, dtype=np.float32)
hrError2[index] = hrError[index] / hrMaskSum[index]
#update hr image and x
outputImage = nibabel.Nifti1Image(outputImage.get_data() - hrError2,
hrImage.affine)
nibabel.save(outputImage, 'ibp_iter' + str(j) + '.nii.gz')
x = convert_image_to_vector(outputImage)
return outputImage
def optimization(HList,x,yList,maxiter,initHRImage):
#prepare the data for optimization
#compress x and H : to perform the optimizaton only on nonzero points (i.e reduce the dimension of the parameter vector)
index = np.nonzero(x)[0]
xc = x[index]
HListc = []
for i in range(len(HList)):
HListc.append(HList[i][:,index])
from skimage.restoration import denoise_tv_chambolle
HRDenoised = denoise_tv_chambolle(initHRImage.get_data(),weight=10)
nibabel.save(nibabel.Nifti1Image(HRDenoised,initHRImage.affine),'cham.nii.gz')
xRef = convert_image_to_vector(nibabel.Nifti1Image(HRDenoised, initHRImage.affine))[index]
res = optimize(HListc,xc,yList,maxiter,index,xRef,lambdaL2=0.6)
#decompress res.x
x[index] = res
return x.reshape(initHRImage.get_data().shape)
def optimization(HList, x, yList, maxiter, initHRImage):
#prepare the data for optimization
#compress x and H : to perform the optimizaton only on nonzero points (i.e reduce the dimension of the parameter vector)
index = np.nonzero(x)[0]
xc = x[index]
HListc = []
for i in range(len(HList)):
HListc.append(HList[i][:, index])
from skimage.restoration import denoise_tv_chambolle
HRDenoised = denoise_tv_chambolle(initHRImage.get_data(), weight=10)
nibabel.save(nibabel.Nifti1Image(HRDenoised, initHRImage.affine),
'cham.nii.gz')
xRef = convert_image_to_vector(
nibabel.Nifti1Image(HRDenoised, initHRImage.affine))[index]
res = optimize(HListc, xc, yList, maxiter, index, xRef, lambdaL2=0.6)
#decompress res.x
x[index] = res
return x.reshape(initHRImage.get_data().shape)
#Compute H
HList = []
for i in range(len(inputImages)):
HList.append(
compute_H(inputImages[i], initHRImage, inputTransforms[i],
psfList[i], maskImages[i]))
#Intensity correction To do
#N4 on initHR
#local correction
#New init HR
if args.bias == True:
initHRImage_N4 = apply_N4_on_image(initHRImage, shrink_factor=1)
xN4 = convert_image_to_vector(initHRImage_N4)
hrN4Data = np.zeros(initHRImage.get_data().shape)
for i in range(len(inputImages)):
simu = convert_vector_to_image(HList[i].dot(xN4), inputImages[i])
im = gaussian_biais_correction(inputImages[i], simu, 5)
warped = apply_affine_itk_transform_on_image(
input_image=im,
transform=inputTransforms[i][0],
center=inputTransforms[i][1],
reference_image=initHRImage,
order=3)
hrN4Data += (warped.get_data() / np.float32(len(inputImages)))
initHRImage = nibabel.Nifti1Image(hrN4Data, initHRImage.affine)
#Compute x
m = np.identity(4)
c = np.array([0, 0, 0, 1])
inputTransform = (m, c)
print('Creating mask image using the following padding value:' +
str(args.padding))
data = np.zeros(HRimage.get_data().shape)
data[HRimage.get_data() > args.padding] = 1
maskHRImage = nibabel.Nifti1Image(data, HRimage.affine)
print('Percentage of HR masked values : %.2f ' %
(np.size(np.nonzero(
(data))) / (1.0 * np.size(data.shape)) * 100.0 / np.size(data)))
data = np.zeros(LRimage.get_data().shape)
data[LRimage.get_data() > args.padding] = 1
maskLRImage = nibabel.Nifti1Image(data, LRimage.affine)
print('Percentage of LR masked values : %.2f ' %
(np.size(np.nonzero(
(data))) / (1.0 * np.size(data.shape)) * 100.0 / np.size(data)))
HRSpacing = np.float32(np.array(HRimage.header['pixdim'][1:4]))
LRSpacing = np.float32(np.array(LRimage.header['pixdim'][1:4]))
psf = compute_psf(LRSpacing, HRSpacing, args.psf)
H = compute_H(LRimage, HRimage, inputTransform, psf, maskLRImage)
x = convert_image_to_vector(HRimage)
maskX = convert_image_to_vector(maskHRImage)
#Let mask the HR image
x = x * maskX
nibabel.save(convert_vector_to_image(H.dot(x), LRimage), args.output)
else:
#no transform provided : use identity as transform and zero as center
m = np.identity(4)
c = np.array([0, 0, 0, 1])
inputTransform = (m,c)
print('Creating mask image using the following padding value:'+str(args.padding))
data = np.zeros(HRimage.get_data().shape)
data[HRimage.get_data() > args.padding] = 1
maskHRImage = nibabel.Nifti1Image(data, HRimage.affine)
print('Percentage of HR masked values : %.2f '%( np.size(np.nonzero((data))) / (1.0*np.size(data.shape)) * 100.0 / np.size(data) ) )
data = np.zeros(LRimage.get_data().shape)
data[LRimage.get_data() > args.padding] = 1
maskLRImage = nibabel.Nifti1Image(data, LRimage.affine)
print('Percentage of LR masked values : %.2f '%( np.size(np.nonzero((data))) / (1.0*np.size(data.shape)) * 100.0 / np.size(data) ) )
HRSpacing = np.float32(np.array(HRimage.header['pixdim'][1:4]))
LRSpacing = np.float32(np.array(LRimage.header['pixdim'][1:4]))
psf = compute_psf(LRSpacing, HRSpacing, args.psf)
H = compute_H(LRimage, HRimage, inputTransform, psf, maskLRImage)
x = convert_image_to_vector(HRimage)
maskX = convert_image_to_vector(maskHRImage)
#Let mask the HR image
x = x*maskX
nibabel.save(convert_vector_to_image(H.dot(x),LRimage),args.output)
HRpsf = compute_psf(LRSpacing, HRSpacing, args.psf)
psfList.append(HRpsf)
# Compute H
HList = []
for i in range(len(inputImages)):
HList.append(compute_H(inputImages[i], initHRImage, inputTransforms[i], psfList[i], maskImages[i]))
# Intensity correction To do
# N4 on initHR
# local correction
# New init HR
if args.bias == True:
initHRImage_N4 = apply_N4_on_image(initHRImage, shrink_factor=1)
xN4 = convert_image_to_vector(initHRImage_N4)
hrN4Data = np.zeros(initHRImage.get_data().shape)
for i in range(len(inputImages)):
simu = convert_vector_to_image(HList[i].dot(xN4), inputImages[i])
im = gaussian_biais_correction(inputImages[i], simu, 5)
warped = apply_affine_itk_transform_on_image(
input_image=im,
transform=inputTransforms[i][0],
center=inputTransforms[i][1],
reference_image=initHRImage,
order=3,
)
hrN4Data += warped.get_data() / np.float32(len(inputImages))
initHRImage = nibabel.Nifti1Image(hrN4Data, initHRImage.affine)
