def Matching(cf):
if cf.compute.useCUDA and cf.compute.gpuID is not None:
ca.SetCUDADevice(cf.compute.gpuID)
# prepare output directory
common.Mkdir_p(os.path.dirname(cf.io.outputPrefix))
# Output loaded config
if cf.io.outputPrefix is not None:
cfstr = Config.ConfigToYAML(MatchingConfigSpec, cf)
with open(cf.io.outputPrefix + "parsedconfig.yaml", "w") as f:
f.write(cfstr)
mType = ca.MEM_DEVICE if cf.compute.useCUDA else ca.MEM_HOST
I0 = common.LoadITKImage(cf.study.I0, mType)
I1 = common.LoadITKImage(cf.study.I1, mType)
#ca.DivC_I(I0,255.0)
#ca.DivC_I(I1,255.0)
grid = I0.grid()
ca.ThreadMemoryManager.init(grid, mType, 1)
#common.DebugHere()
# TODO: need to work on these
t = [x*1./cf.optim.nTimeSteps for x in range(cf.optim.nTimeSteps+1)]
checkpointinds = range(1,len(t))
checkpointstates = [(ca.Field3D(grid,mType),ca.Field3D(grid,mType)) for idx in checkpointinds]
p = MatchingVariables(I0,I1, cf.vectormomentum.sigma, t,checkpointinds, checkpointstates, cf.vectormomentum.diffOpParams[0], cf.vectormomentum.diffOpParams[1], cf.vectormomentum.diffOpParams[2], cf.optim.Niter, cf.optim.stepSize, cf.optim.maxPert, cf.optim.nTimeSteps, integMethod = cf.optim.integMethod, optMethod=cf.optim.method, nInv=cf.optim.NIterForInverse,plotEvery=cf.io.plotEvery, plotSlice = cf.io.plotSlice, quiverEvery = cf.io.quiverEvery, outputPrefix = cf.io.outputPrefix)
RunMatching(p)
# write output
if cf.io.outputPrefix is not None:
# reset all variables by shooting once, may have been overwritten
CAvmCommon.IntegrateGeodesic(p.m0,p.t,p.diffOp,\
p.m, p.g, p.ginv,\
p.scratchV1, p.scratchV2,p. scratchV3,\
p.checkpointstates, p.checkpointinds,\
Ninv=p.nInv, integMethod = p.integMethod)
common.SaveITKField(p.m0, cf.io.outputPrefix+"m0.mhd")
common.SaveITKField(p.ginv, cf.io.outputPrefix+"phiinv.mhd")
common.SaveITKField(p.g, cf.io.outputPrefix+"phi.mhd")
def AtlasWriteOutput(cf, atlas, m_array, nodeSubjectsIds, isReporter):
# save initial momenta for all individuals
for itsub in range(len(nodeSubjectsIds)):
common.SaveITKField(
m_array[itsub], cf.io.outputPrefix +
str(nodeSubjectsIds[itsub]).replace('.', '_') + "_m0.mhd")
# save the atlas
if isReporter:
common.SaveITKImage(atlas, cf.io.outputPrefix + "atlas.mhd")
def HGMWriteOutput(cf, groupState, tDiscGroup, isReporter):
# save initial momenta for residual geodesics, p, for all individuals
for i in range(len(groupState.t)):
if tDiscGroup[i].J is not None:
common.SaveITKField(
tDiscGroup[i].p0, cf.io.outputPrefix +
str(tDiscGroup[i].subjectId).replace('.', '_') + "_p0.mhd")
# write individual's energy history
energyFilename = cf.io.outputPrefix + str(
tDiscGroup[i].subjectId).replace('.',
'_') + "ResidualEnergy.csv"
HGMWriteEnergyHistoryToFile(tDiscGroup[i].Energy, energyFilename)
# save initial image and momenta for group gedoesic
if isReporter:
common.SaveITKImage(groupState.I0, cf.io.outputPrefix + "I0.mhd")
common.SaveITKField(groupState.m0, cf.io.outputPrefix + "m0.mhd")
# write energy history
energyFilename = cf.io.outputPrefix + "TotalEnergyHistory.csv"
HGMWriteEnergyHistoryToFile(groupState.EnergyHistory, energyFilename)
def GeoRegWriteOuput(subjectId, cf, p, t, Imsmts, cpinds, cpstates, msmtinds,
gradAtMsmts, EnergyHistory):
# save initial image and momenta for regression geodesic
common.SaveITKImage(p.I0, cf.io.outputPrefix + subjectId + "I0.mhd")
common.SaveITKField(p.m0, cf.io.outputPrefix + subjectId + "m0.mhd")
# save residual images for regression geodesic
# TODO:
# write Energy details
energyFilename = cf.io.outputPrefix + subjectId + "Energy.csv"
with open(energyFilename, 'w') as f:
csv_writer = csv.writer(f, delimiter='\t')
csv_writer.writerows(EnergyHistory)
def MatchingImageMomentaWriteOuput(cf, geodesicState, EnergyHistory, m0, n1):
grid = geodesicState.J0.grid()
mType = geodesicState.J0.memType()
# save momenta for the gedoesic
common.SaveITKField(geodesicState.p0, cf.io.outputPrefix + "p0.mhd")
# save matched momenta for the geodesic
if cf.vectormomentum.matchImOnly:
m0 = common.LoadITKField(cf.study.m, mType)
ca.CoAd(geodesicState.p, geodesicState.rhoinv, m0)
common.SaveITKField(geodesicState.p, cf.io.outputPrefix + "m1.mhd")
# momenta match energy
if cf.vectormomentum.matchImOnly:
vecdiff = ca.ManagedField3D(grid, mType)
ca.Sub_I(geodesicState.p, n1)
ca.Copy(vecdiff, geodesicState.p)
geodesicState.diffOp.applyInverseOperator(geodesicState.p)
momentaMatchEnergy = ca.Dot(vecdiff, geodesicState.p) / (
float(geodesicState.p0.nVox()) * geodesicState.SigmaSlope *
geodesicState.SigmaSlope)
# save energy
energyFilename = cf.io.outputPrefix + "testMomentaMatchEnergy.csv"
with open(energyFilename, 'w') as f:
print >> f, momentaMatchEnergy
# save matched image for the geodesic
tempim = ca.ManagedImage3D(grid, mType)
ca.ApplyH(tempim, geodesicState.J0, geodesicState.rhoinv)
common.SaveITKImage(tempim, cf.io.outputPrefix + "I1.mhd")
# save energy
energyFilename = cf.io.outputPrefix + "energy.csv"
MatchingImageMomentaWriteEnergyHistoryToFile(EnergyHistory, energyFilename)
def MatchingImageMomentaWriteOuput(cf, geodesicState):
# save momenta for the gedoesic
common.SaveITKField(geodesicState.p0, cf.io.outputPrefix + "p0.mhd")
def write_result(result, output_prefix):
common.SaveITKImage(result['I1'], output_prefix+"I1.mhd")
common.SaveITKField(result['phiinv'], output_prefix+"phiinv.mhd")
def GeodesicShooting(cf):
# prepare output directory
common.Mkdir_p(os.path.dirname(cf.io.outputPrefix))
# Output loaded config
if cf.io.outputPrefix is not None:
cfstr = Config.ConfigToYAML(GeodesicShootingConfigSpec, cf)
with open(cf.io.outputPrefix + "parsedconfig.yaml", "w") as f:
f.write(cfstr)
mType = ca.MEM_DEVICE if cf.useCUDA else ca.MEM_HOST
#common.DebugHere()
I0 = common.LoadITKImage(cf.study.I0, mType)
m0 = common.LoadITKField(cf.study.m0, mType)
grid = I0.grid()
ca.ThreadMemoryManager.init(grid, mType, 1)
# set up diffOp
if mType == ca.MEM_HOST:
diffOp = ca.FluidKernelFFTCPU()
else:
diffOp = ca.FluidKernelFFTGPU()
diffOp.setAlpha(cf.diffOpParams[0])
diffOp.setBeta(cf.diffOpParams[1])
diffOp.setGamma(cf.diffOpParams[2])
diffOp.setGrid(grid)
g = ca.Field3D(grid, mType)
ginv = ca.Field3D(grid, mType)
mt = ca.Field3D(grid, mType)
It = ca.Image3D(grid, mType)
t = [
x * 1. / cf.integration.nTimeSteps
for x in range(cf.integration.nTimeSteps + 1)
]
checkpointinds = range(1, len(t))
checkpointstates = [(ca.Field3D(grid, mType), ca.Field3D(grid, mType))
for idx in checkpointinds]
scratchV1 = ca.Field3D(grid, mType)
scratchV2 = ca.Field3D(grid, mType)
scratchV3 = ca.Field3D(grid, mType)
# scale momenta to shoot
cf.study.scaleMomenta = float(cf.study.scaleMomenta)
if abs(cf.study.scaleMomenta) > 0.000000:
ca.MulC_I(m0, float(cf.study.scaleMomenta))
CAvmCommon.IntegrateGeodesic(m0,t,diffOp, mt, g, ginv,\
scratchV1,scratchV2,scratchV3,\
keepstates=checkpointstates,keepinds=checkpointinds,
Ninv=cf.integration.NIterForInverse, integMethod = cf.integration.integMethod)
else:
ca.Copy(It, I0)
ca.Copy(mt, m0)
ca.SetToIdentity(ginv)
ca.SetToIdentity(g)
# write output
if cf.io.outputPrefix is not None:
# scale back shotmomenta before writing
if abs(cf.study.scaleMomenta) > 0.000000:
ca.ApplyH(It, I0, ginv)
ca.CoAd(mt, ginv, m0)
ca.DivC_I(mt, float(cf.study.scaleMomenta))
common.SaveITKImage(It, cf.io.outputPrefix + "I1.mhd")
common.SaveITKField(mt, cf.io.outputPrefix + "m1.mhd")
common.SaveITKField(ginv, cf.io.outputPrefix + "phiinv.mhd")
common.SaveITKField(g, cf.io.outputPrefix + "phi.mhd")
GeodesicShootingPlots(g, ginv, I0, It, cf)
if cf.io.saveFrames:
SaveFrames(checkpointstates, checkpointinds, I0, It, m0, mt, cf)
def predict_image(args):
if (args.use_CPU_for_shooting):
mType = ca.MEM_HOST
else:
mType = ca.MEM_DEVICE
# load the prediction network
predict_network_config = torch.load(args.prediction_parameter)
prediction_net = create_net(args, predict_network_config)
batch_size = args.batch_size
patch_size = predict_network_config['patch_size']
input_batch = torch.zeros(batch_size, 2, patch_size, patch_size,
patch_size).cuda()
# start prediction
for i in range(0, len(args.moving_image)):
common.Mkdir_p(os.path.dirname(args.output_prefix[i]))
if (args.affine_align):
# Perform affine registration to both moving and target image to the ICBM152 atlas space.
# Registration is done using Niftireg.
call([
"reg_aladin", "-noSym", "-speeeeed", "-ref", args.atlas,
"-flo", args.moving_image[i], "-res",
args.output_prefix[i] + "moving_affine.nii", "-aff",
args.output_prefix[i] + 'moving_affine_transform.txt'
])
call([
"reg_aladin", "-noSym", "-speeeeed", "-ref", args.atlas,
"-flo", args.target_image[i], "-res",
args.output_prefix[i] + "target_affine.nii", "-aff",
args.output_prefix[i] + 'target_affine_transform.txt'
])
moving_image = common.LoadITKImage(
args.output_prefix[i] + "moving_affine.nii", mType)
target_image = common.LoadITKImage(
args.output_prefix[i] + "target_affine.nii", mType)
else:
moving_image = common.LoadITKImage(args.moving_image[i], mType)
target_image = common.LoadITKImage(args.target_image[i], mType)
#preprocessing of the image
moving_image_np = preprocess_image(moving_image, args.histeq)
target_image_np = preprocess_image(target_image, args.histeq)
grid = moving_image.grid()
moving_image_processed = common.ImFromNPArr(moving_image_np, mType)
target_image_processed = common.ImFromNPArr(target_image_np, mType)
moving_image.setGrid(grid)
target_image.setGrid(grid)
predict_transform_space = False
if 'matlab_t7' in predict_network_config:
predict_transform_space = True
# run actual prediction
prediction_result = util.predict_momentum(moving_image_np,
target_image_np, input_batch,
batch_size, patch_size,
prediction_net,
predict_transform_space)
m0 = prediction_result['image_space']
m0_reg = common.FieldFromNPArr(prediction_result['image_space'], mType)
registration_result = registration_methods.geodesic_shooting(
moving_image_processed, target_image_processed, m0_reg,
args.shoot_steps, mType, predict_network_config)
phi = common.AsNPCopy(registration_result['phiinv'])
phi_square = np.power(phi, 2)
for sample_iter in range(1, args.samples):
print(sample_iter)
prediction_result = util.predict_momentum(
moving_image_np, target_image_np, input_batch, batch_size,
patch_size, prediction_net, predict_transform_space)
m0 += prediction_result['image_space']
m0_reg = common.FieldFromNPArr(prediction_result['image_space'],
mType)
registration_result = registration_methods.geodesic_shooting(
moving_image_processed, target_image_processed, m0_reg,
args.shoot_steps, mType, predict_network_config)
phi += common.AsNPCopy(registration_result['phiinv'])
phi_square += np.power(
common.AsNPCopy(registration_result['phiinv']), 2)
m0_mean = np.divide(m0, args.samples)
m0_reg = common.FieldFromNPArr(m0_mean, mType)
registration_result = registration_methods.geodesic_shooting(
moving_image_processed, target_image_processed, m0_reg,
args.shoot_steps, mType, predict_network_config)
phi_mean = registration_result['phiinv']
phi_var = np.divide(phi_square, args.samples) - np.power(
np.divide(phi, args.samples), 2)
#save result
common.SaveITKImage(registration_result['I1'],
args.output_prefix[i] + "I1.mhd")
common.SaveITKField(phi_mean,
args.output_prefix[i] + "phiinv_mean.mhd")
common.SaveITKField(common.FieldFromNPArr(phi_var, mType),
args.output_prefix[i] + "phiinv_var.mhd")