label='Sphere parameter')
fig = ax.get_figure()
fig.set_size_inches([10, 5])
ax.set_xlim([0, 1])
ax.set_xlabel('Parameter value')
ax.set_ylabel('# voxels')
ax.legend()
fig.savefig(
'/Users/arokem/Dropbox/DWI_figures/MultiCanonicalTensor_params_%s.png'
% b)
idx = np.logical_and(to_hist1 < 1, to_hist2 < 1)
T1 = ni.load(data_path + 'FP_t1_resampled_to_dwi.nii.gz').get_data()
fig = viz.mosaic(T1.T, cmap=matplotlib.cm.bone, cbar=False)
fig = viz.mosaic(Model1.model_params[:, :, :, 1].T,
fig=fig,
cmap=matplotlib.cm.hot,
vmax=1)
fig.set_size_inches([12, 8])
fig.savefig(
'/Users/arokem/Dropbox/DWI_figures/MultiCanonicalTensor_tensor_param%s.png'
% b)
fig = viz.mosaic(T1.T, cmap=matplotlib.cm.bone, cbar=False)
fig = viz.mosaic(Model1.model_params[:, :, :, 2].T,
fig=fig,
cmap=matplotlib.cm.hot,
vmax=1)
fig.set_size_inches([12, 8])
# Set the background to black:
t1_d[~np.isfinite(t1_d)] = 0
fig_hist_rmse, ax_hist_rmse = plt.subplots(1)
fig_hist_snr, ax_hist_snr = plt.subplots(1)
for bval_idx, bval in enumerate([1000, 2000, 4000]):
snr_file_name = "%ssnr_b%s.nii.gz" % (data_path, bval)
rmse_file_name = "%ssignal_rmse_b%s.nii.gz" % (data_path, bval)
signal_rmse = ni.load(rmse_file_name).get_data()
snr = ni.load(snr_file_name).get_data()
vol = ozu.nans(signal_rmse.shape)
vol[mask_idx] = signal_rmse[mask_idx]
fig = viz.mosaic(t1_d.T[23:-12], cmap=matplotlib.cm.bone, cbar=False)
fig = viz.mosaic(vol.T[23:-12], fig=fig, cmap=matplotlib.cm.hot)
fig.set_size_inches([15, 10])
fig.savefig("%ssignal_rmse_b%s.png" % (figure_path, bval))
vol = ozu.nans(snr.shape)
vol[mask_idx] = snr[mask_idx]
fig = viz.mosaic(t1_d.T[23:-12], cmap=matplotlib.cm.bone, cbar=False)
fig = viz.mosaic(vol.T[23:-12], fig=fig, cmap=matplotlib.cm.hot)
fig.set_size_inches([15, 10])
fig.savefig("%ssignal_snr_b%s.png" % (figure_path, bval))
rmse_mask = signal_rmse[mask_idx]
fig_hist_rmse = viz.probability_hist(
rmse_mask[np.isfinite(rmse_mask)], bins=100, fig=fig_hist_rmse, label="b=%s" % bval
)
# Set the background to black:
t1_d[~np.isfinite(t1_d)] = 0
fig_hist_rmse, ax_hist_rmse = plt.subplots(1)
fig_hist_snr, ax_hist_snr = plt.subplots(1)
for bval_idx, bval in enumerate([1000, 2000, 4000]):
snr_file_name = '%ssnr_b%s.nii.gz' % (data_path, bval)
rmse_file_name = '%ssignal_rmse_b%s.nii.gz' % (data_path, bval)
signal_rmse = ni.load(rmse_file_name).get_data()
snr = ni.load(snr_file_name).get_data()
vol = ozu.nans(signal_rmse.shape)
vol[mask_idx] = signal_rmse[mask_idx]
fig = viz.mosaic(t1_d.T[23:-12], cmap=matplotlib.cm.bone, cbar=False)
fig = viz.mosaic(vol.T[23:-12], fig=fig, cmap=matplotlib.cm.hot)
fig.set_size_inches([15, 10])
fig.savefig('%ssignal_rmse_b%s.png' % (figure_path, bval))
vol = ozu.nans(snr.shape)
vol[mask_idx] = snr[mask_idx]
fig = viz.mosaic(t1_d.T[23:-12], cmap=matplotlib.cm.bone, cbar=False)
fig = viz.mosaic(vol.T[23:-12], fig=fig, cmap=matplotlib.cm.hot)
fig.set_size_inches([15, 10])
fig.savefig('%ssignal_snr_b%s.png' % (figure_path, bval))
rmse_mask = signal_rmse[mask_idx]
fig_hist_rmse = viz.probability_hist(rmse_mask[np.isfinite(rmse_mask)],
bins=100,
fig=fig_hist_rmse,
for b_idx, file_root in enumerate(file_roots):
vol = np.ones((81,106,76,4)) * np.nan
for chunk in range(n_chunks):
if vox_per_chunk*chunk+1 > brain_idx.shape[-1]:
this_idx = brain_idx[:,vox_per_chunk * chunk:]
else:
this_idx = brain_idx[:,vox_per_chunk * chunk:vox_per_chunk*(chunk+1)]
mask_array = np.zeros((81, 106, 76))
mask_array[(this_idx[0], this_idx[1], this_idx[2])] = 1
params_file = '/home/arokem/data/ModelFits/MultiCanonicalTensor2/MultiCanonicalTensorb%s_%03d.nii.gz'%(all_bvals[b_idx], chunk+1)
this_params = ni.load(params_file).get_data()
mask_idx = np.where(mask_array)
vol[mask_idx] = this_params[mask_idx]
fig = viz.mosaic(vol[:,:,:,-1])
fig.set_size_inches([20,18])
ni.Nifti1Image(vol, ni.load(params_file).get_affine()).to_filename(
'/home/arokem/data/ModelFits/MultiCanonicalTensor/MultiCanonicalTensorCombinedb%s.nii.gz'%all_bvals[b_idx])
# Force a save of the params to file:
Model.model_params
for b_idx, file_root in enumerate(file_roots):
vol = np.ones((81, 106, 76, 4)) * np.nan
for chunk in range(n_chunks):
if vox_per_chunk * chunk + 1 > brain_idx.shape[-1]:
this_idx = brain_idx[:, vox_per_chunk * chunk:]
else:
this_idx = brain_idx[:, vox_per_chunk * chunk:vox_per_chunk *
(chunk + 1)]
mask_array = np.zeros((81, 106, 76))
mask_array[(this_idx[0], this_idx[1], this_idx[2])] = 1
params_file = '/home/arokem/data/ModelFits/MultiCanonicalTensor2/MultiCanonicalTensorb%s_%03d.nii.gz' % (
all_bvals[b_idx], chunk + 1)
this_params = ni.load(params_file).get_data()
mask_idx = np.where(mask_array)
vol[mask_idx] = this_params[mask_idx]
fig = viz.mosaic(vol[:, :, :, -1])
fig.set_size_inches([20, 18])
ni.Nifti1Image(
vol,
ni.load(params_file).get_affine()
).to_filename(
'/home/arokem/data/ModelFits/MultiCanonicalTensor/MultiCanonicalTensorCombinedb%s.nii.gz'
% all_bvals[b_idx])
# equivalent TensorModel, using the
# method outlined in Stikov et al. 2011):
# Add [0,1] to force the axes to be [0,1]
fig = viz.scatter_density(np.hstack([to_hist1[idx],[0,1]]),
np.hstack([TM1.fiber_volume_fraction[Model1.mask][idx],
[0,1]]))
fig.axes[0].set_xlabel('Tensor parameter')
fig.axes[0].set_ylabel('FVF')
fig.set_size_inches([10,10])
fig.axes[0].plot([100,0],[0,100],'--k')
fig.savefig('/home/arokem/Dropbox/DWI_figures/CanonicalTensor_fvf%s.png'%b)
T1 = ni.load(data_path + 'FP_t1_resampled_to_dwi.nii.gz').get_data()
fig = viz.mosaic(T1.T, cmap=matplotlib.cm.bone, cbar=False)
fig = viz.mosaic(Model1.model_params[:,:,:,1].T,
fig=fig, cmap=matplotlib.cm.hot, vmax=1)
fig.set_size_inches([12,8])
fig.savefig('/home/arokem/Dropbox/DWI_figures/CanonicalTensor_tensor_param%s.png'%b)
fig = viz.mosaic(T1.T, cmap=matplotlib.cm.bone, cbar=False)
fig = viz.mosaic(Model1.model_params[:,:,:,2].T,
fig=fig,
cmap=matplotlib.cm.hot,
vmax=1)
fig.set_size_inches([12,8])
fig.savefig('/home/arokem/Dropbox/DWI_figures/CanonicalTensor_sphere_param%s.png'%b)
fractional_anisotropy = (