def pointtrans(Fish, F):
#===============================================================================
import os
import ants
import pandas as pd
import numpy as np
# Apply registration to the CMN identified cells
#---------------------------------------------------------------------------
for c in range(len(Fish["Cond"])):
print('Transforming points to standard space for condition ' +
str(c + 1))
Freg = F["Freg"] + os.sep + Fish["Name"] + os.sep + Fish["Cond"][c][
"Name"]
Ff2cf = Freg + os.sep + 'FUN2CF'
os.listdir(Ff2cf)
cs = pd.DataFrame(Fish["Cond"][c]["Pixels"])
cs.columns = ['x', 'y', 'z']
tcs = np.multiply(cs, (-.3, .3, -6))
ncs = ants.apply_transforms_to_points(3, tcs, \
[ Ff2cf +os.sep+ 'cf2fun_R.mat',
Ff2cf +os.sep+ 'cf2fun_S.mat',
Ff2cf +os.sep+ 'cf2fun_S.nii.gz'], \
whichtoinvert = [True, True, False])
tcs = np.multiply(ncs, (1, 1, 1))
nncs = ants.apply_transforms_to_points(3, tcs, \
[ F["Ftrans"] +os.sep+ 'ref2cf_R.mat',
F["Ftrans"] +os.sep+ 'ref2cf_S.mat',
F["Ftrans"] +os.sep+ 'ref2cf_S.nii.gz'], \
whichtoinvert = [True,True,False])
Fish["Cond"][c]["ZBBCoord"] = nncs.values
return Fish
boldseg = ants.apply_transforms( und, t1seg['segmentation'], t1reg['fwdtransforms'], interpolator = 'nearestNeighbor' ) ## Template mapping # include prior information e.g. from meta-analysis or anatomy myvoxes = range(powers_areal_mni_itk.shape[0]) anat = powers_areal_mni_itk['Anatomy'] syst = powers_areal_mni_itk['SystemName'] Brod = powers_areal_mni_itk['Brodmann'] xAAL = powers_areal_mni_itk['AAL'] ch2 = ants.image_read( ants.get_ants_data( "ch2" ) ) treg = ants.registration( t1 * t1mask, ch2, 'SyN' ) concatx2 = treg['invtransforms'] + t1reg['invtransforms'] pts2bold = ants.apply_transforms_to_points( 3, powers_areal_mni_itk, concatx2,whichtoinvert = ( True, False, True, False ) ) locations = pts2bold.iloc[:,:3].values ptImg = ants.make_points_image( locations, bmask, radius = 3 ) networks = powers_areal_mni_itk['SystemName'].unique() dfnpts = np.where( powers_areal_mni_itk['SystemName'] == networks[5] ) dfnImg = ants.mask_image( ptImg, ptImg, level = dfnpts[0].tolist(), binarize=False ) # plot( und, ptImg, axis=3, window.overlay = range( ptImg ) ) bold2ch2 = ants.apply_transforms( ch2, und, concatx2, whichtoinvert = ( True, False, True, False ) ) # Extracting canonical functional network maps ## preprocessing csfAndWM = ( ants.threshold_image( boldseg, 1, 1 ) +
initial_transform='AffineFast',
type_of_transform='ElasticSyN',
grad_step=5,
aff_metric='mattes',
aff_sampling=32,
syn_metric='mattes',
syn_sampling=32,
reg_iterations=(40, 20, 10),
)
print('Transform: %r' % mytx)
t_elapsed = time.time() - t_start
print('Time: %r seconds' % t_elapsed)
warped_moving = ants.apply_transforms(fixed=fixed,
moving=moving,
transformlist=mytx['fwdtransforms'])
warped_points = ants.apply_transforms_to_points(
dim=2, points=lnds, transformlist=mytx['invtransforms'])
# # Visualisation
# import matplotlib.pyplot as plt
# plt.imshow(warped_moving.numpy(), cmap=plt.cm.Greys_r)
# plt.plot(warped_points['y'], warped_points['x'], '.')
# plt.show()
# Exporting results
with open(paths['time'], 'w') as fp:
fp.write(str(t_elapsed))
imsave(paths['warped'], warped_moving.numpy())
lnds = warped_points[['y', 'x']]
# transform landmarks coordinates back
lnds.columns = ['X', 'Y']
lnds.to_csv(paths['pts'])