Reconstructions Through TomoPy
Basic Reconstructions
This section will show a User how to use the raw hdf5 projections to reconstruct a CT scan. The basics are to import the correct modules, define a reconstruction function, and pass that function to the reconstruct_data function which handels everything else. The reconstruct_data function is a wrapper which has the ability to pull data from many different directories. Most use cases will be explained somewhere on this page.
import sys
sys.path.append('/path/to/tomosuite/github/clone/tomosuite/')
import tomosuite
import tomopy
# Import TomoSuite helper functions
from tomosuite.base.reconstruct import reconstruct_data, plot_reconstruction
# Define your own tomography reconstruction function. This is the TomoSuite's default
def tomo_recon(prj, theta, rot_center, user_extra=None):
recon = tomopy.recon(prj, theta,
center=rot_center,
algorithm='gridrec',
ncore=30)
return recon, user_extra
# Reconstruct the raw projection data
basedir = '/local/data/project_01/'
slcs, user_extra = reconstruct_data(basedir,
rot_center,
start_row=None, # If the User doesnt want to reconstruct all rows
end_row=None, # If the User doesnt want to reconstruct all rows
med_filter=False,
all_data_med_filter=False,
med_filter_kernel=(1, 3, 3),
reconstruct_func=tomo_recon, # Allows the user to define their own recon function
network=None, # None loads in the raw projection data
wedge_removal=0, # zero out the first and last wedge_removal number of projections
sparse_angle_removal=1, # only use every sparse_anlge_removal image for the recon
types='denoise', # used when network='TomoGAN'
second_basedir=None, # used to pull a second datasets data for the reconstruction
checkpoint_num=None, # used when network='Deepfillv2'
double_sparse=None,
power2pad=False, # Force the sinogram to be of shape power2
edge_transition=None # eliminate columns of sinogram which cause a harsh ring effect in the recon)
# Plot the reconstruction
plot_reconstruction(slcs[0:1])
DeNoise Type 1 or Type 2
Once the User predicts through tomogan they now have the ability to reconstruct that predicted data. In this case we are looking at DeNoise Type 1 or Type 2. Type 1 is where the User has imput fake noise into their projections, and used tomogan to denoise the original projections. While Type 2 is where the User has a sacraficial sample, which contains multiple projections of the save FOV.
The main concept is similar to that of the basic reconstruction. The main difference is now the User has to define the network=’tomogan’ and the types=’denoise_fake’ for Type 1 or types=’denoise_exp’ for Type 2. This tells the reconstruct_data function to import the data related to tomogan and make sure you import the denoised data based on the fake noise training or the sacraficial sample training.
import sys
sys.path.append('/path/to/tomosuite/github/clone/tomosuite/')
import tomosuite
import tomopy
# Import TomoSuite helper functions
from tomosuite.base.reconstruct import reconstruct_data, plot_reconstruction
# Define your own tomography reconstruction function. This is the TomoSuite's default
def tomo_recon(prj, theta, rot_center, user_extra=None):
recon = tomopy.recon(prj, theta,
center=rot_center,
algorithm='gridrec',
ncore=30)
recon = tomopy.circ_mask(recon, axis=0, ratio=0.95)
return recon, user_extra
# Reconstruct the raw projection data
basedir = '/local/data/project_01/'
slcs, user_extra = reconstruct_data(basedir,
rot_center=600,
reconstruct_func=tomo_recon,
network='tomogan',
types='denoise_fake', # or denoise_exp if you are reconstructing Type 2
power2pad=True, # forces the sinogram to be in a power of 2 shape
edge_transition=5 # removes harsh edge on sinogram
)
# Plot the reconstruction
plot_reconstruction(slcs[0:10])
DeWedge - DeepFillV2 Type 1
Once the User predicts the missing wedge data based upon sinogram inpainting, the User can then use TomoSuite to reconstruct the sinograms for them. To do this please follow the below reconstruction scripts.
import sys
sys.path.append('/path/to/tomosuite/github/clone/tomosuite/')
import tomosuite
import tomopy
# Import TomoSuite helper functions
from tomosuite.base.reconstruct import reconstruct_data, plot_reconstruction
# Define your own tomography reconstruction function. This is the TomoSuite's default
def tomo_recon(prj, theta, rot_center, user_extra=None):
recon = tomopy.recon(prj, theta,
center=rot_center,
algorithm='gridrec',
ncore=30)
recon = tomopy.circ_mask(recon, axis=0, ratio=0.95)
return recon, user_extra
# Reconstruct the deepfillv2 projection data
basedir = '/local/data/project_01/'
slcs, user_extra = reconstruct_data(basedir, rot_center=181,
reconstruct_func=tomo_recon, network='deepfillv2',
checkpoint_num='120000', power2pad=True)
# Plot the reconstruction
plot_reconstruction(slcs[0:10])
Defining Your Own Recon Function
In this seciton the User will learn how to define their own reconstruction function. to do this one must follow the template below of a function defined as tomo_recon(prj, theta, rot_center, user_extra=None). The user_extra parameter allows the user to pass data out of the recon function. This is mainly for debugging purposes. Next the User has to make sure that this tomo_recon function returns recon, user_extra. Everything in between can be set by the User. make sure you are using tomopy.recon() to reconstruct the slices.
# The inputs have to be prj, theta, and rot_center
# You can define this function however you like and pass your
#new function into the tomosuite recon parameters
# This is the standard defined tomo_recon function through TomoSuite
def tomo_recon(prj, theta, rot_center, user_extra=None):
recon_type='gridrec'
# Add preprocessing steps here
#prj = tomopy.remove_stripe_ti(prj, 2)
if recon_type == 'gridrec':
recon = tomopy.recon(prj, theta,
center=rot_center,
algorithm='gridrec',
ncore=16)
recon = tomopy.circ_mask(recon, axis=0, ratio=0.95)
elif recon_type == 'gridrec_parzen':
recon = tomopy.recon(prj, theta,
center=rot_center,
algorithm='gridrec',
ncore=16,
filter_name='parzen')
recon = tomopy.circ_mask(recon, axis=0, ratio=0.95)
elif recon_type == 'sirt':
extra_options ={'MinConstraint':0}
options = {'proj_type':'cuda', 'method':'SIRT_CUDA',
'num_iter':200, 'extra_options': extra_options}
recon = tomopy.recon(prj, theta,
center=rot_center,
algorithm=tomopy.astra,
ncore=1, options=options)
#Remove ring artifacts, this comes with a slight resolution cost
#recon = tomopy.remove_ring(recon, center_x=None, center_y=None, thresh=300.0)
return recon, user_extra
from tomosuite.base.reconstruct import reconstruct_data, plot_reconstruction
slcs, user_extra = reconstruct_data(basedir,
rot_center=600,
reconstruct_func=tomo_recon,
network='tomogan',
types='denoise_fake', # or denoise_exp
)
plot_reconstruction(slc[0:10], clim=(0, 1))