-
Notifications
You must be signed in to change notification settings - Fork 0
/
simulate_projections_for_train_and_test.py
158 lines (113 loc) · 6.28 KB
/
simulate_projections_for_train_and_test.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
Created on Mon Mar 29 12:39:02 2021
@author: subhadip
"""
import numpy as np
import torch
import odl
import os
import shutil
from torch.utils.data import DataLoader
from torchvision import transforms
import torch_wrapper, mayo_utils
from skimage.measure import compare_psnr
device = 'cuda' if torch.cuda.is_available() else 'cpu'
##############specify geometry parameters#################
img_size, space_range = 512, 128 #space discretization
num_angles, det_shape = 200, 400 #projection parameters
noise_std_dev = 2.0
geom = 'parallel_beam' # 'cone_beam' or 'parallel_beam': The network and optimizer hyper-parameters are optimized for 'parallel_beam'
######computing the projection#############
def compute_projection(phantom, num_angles=num_angles, det_shape=det_shape, space_range=space_range, geom=geom, noise_std_dev=noise_std_dev):
space = odl.uniform_discr([-space_range, -space_range], [space_range, space_range],\
(phantom.size(2), phantom.size(3)), dtype='float32', weighting=1.0)
if(geom=='parallel_beam'):
geometry = odl.tomo.geometry.parallel.parallel_beam_geometry(space, num_angles=num_angles, det_shape=det_shape)
else:
geometry = odl.tomo.geometry.conebeam.cone_beam_geometry(space, src_radius=1.6*space_range, \
det_radius=1.6*space_range, num_angles=num_angles, det_shape=det_shape)
fwd_op_odl = odl.tomo.RayTransform(space, geometry, impl='astra_cuda')
fbp_op_odl = odl.tomo.fbp_op(fwd_op_odl)
fwd_op = torch_wrapper.OperatorModule(fwd_op_odl).to(device)
fbp_op = torch_wrapper.OperatorModule(fbp_op_odl).to(device)
#print('op_norm = %.4f'%(1.1 * odl.power_method_opnorm(fwd_op_odl)))
sinogram = fwd_op(phantom)
sinogram_noisy = sinogram + noise_std_dev*torch.randn(sinogram.size()).to(device)
fbp = fbp_op(sinogram_noisy)
return sinogram_noisy, fbp
######computing the projection#############
def compute_adjoint(sinogram, img_size=img_size, num_angles=num_angles, det_shape=det_shape, space_range=space_range, geom=geom, noise_std_dev=noise_std_dev):
space = odl.uniform_discr([-space_range, -space_range], [space_range, space_range], (img_size, img_size), dtype='float32', weighting=1.0)
if(geom=='parallel_beam'):
geometry = odl.tomo.geometry.parallel.parallel_beam_geometry(space, num_angles=num_angles, det_shape=det_shape)
else:
geometry = odl.tomo.geometry.conebeam.cone_beam_geometry(space, src_radius=20.0, \
det_radius=20.0, num_angles=num_angles, det_shape=det_shape)
fwd_op_odl = odl.tomo.RayTransform(space, geometry, impl='astra_cuda')
adjoint_op_odl = fwd_op_odl.adjoint
adjoint_op = torch_wrapper.OperatorModule(adjoint_op_odl).to(device)
x_adj = adjoint_op(sinogram)
return x_adj
####################arrange the slices into training and test data ###############################
if __name__ == '__main__':
print('creating training and test data...')
datapath = './mayo_data/'
output_datapath = './mayo_data_arranged_patientwise/'
shutil.rmtree(output_datapath, ignore_errors=True) #delete pre-existing folders
files = sorted(os.listdir(datapath))
for idx in range(len(files)):
filename = datapath + files[idx]
image = np.load(filename)
image = (image - image.min())/(image.max() - image.min()) #normalize range to [0,1]
###compute projection and FBP #############
phantom = torch.from_numpy(image).view(1, 1, img_size, img_size).to(device)
sinogram, fbp = compute_projection(phantom)
######save the images as numpy files###############
sinogram_image = sinogram.cpu().numpy().squeeze()
fbp_image = fbp.cpu().numpy().squeeze()
psnr = compare_psnr(image, fbp_image, data_range=1.0)
nmse = torch.mean((phantom - fbp)**2)/torch.mean(phantom**2)
print('FBP: NSME = {:.6f}\t PSNR = {:.6f}'.format(nmse, psnr))
#####use patient L109 for testing, rest for training
if('L109' not in filename):
#####save phantom#####
path = output_datapath + 'train/' + 'Phantom/'
out_filename = path + 'phantom_%d'%idx + '.npy'
os.makedirs(path, exist_ok=True)
np.save(out_filename, image)
#####save FBP#####
path = output_datapath + 'train/' + 'FBP/'
out_filename = path + 'fbp_%d'%idx + '.npy'
os.makedirs(path, exist_ok=True)
np.save(out_filename, fbp_image)
#####save sinogram#####
path = output_datapath + 'train/' + 'Sinogram/'
out_filename = path + 'sinogram_%d'%idx + '.npy'
os.makedirs(path, exist_ok=True)
np.save(out_filename, sinogram_image)
else:
#####save phantom#####
path = output_datapath + 'test/' + 'Phantom/'
out_filename = path + 'phantom_%d'%idx + '.npy'
os.makedirs(path, exist_ok=True)
np.save(out_filename, image)
#####save FBP#####
path = output_datapath + 'test/' + 'FBP/'
out_filename = path + 'fbp_%d'%idx + '.npy'
os.makedirs(path, exist_ok=True)
np.save(out_filename, fbp_image)
#####save sinogram#####
path = output_datapath + 'test/' + 'Sinogram/'
out_filename = path + 'sinogram_%d'%idx + '.npy'
os.makedirs(path, exist_ok=True)
np.save(out_filename, sinogram_image)
####### verify training and testing dataloader
print('creating dataloaders...')
transform_to_tensor = [transforms.ToTensor()]
train_dataloader = DataLoader(mayo_utils.mayo_dataset(output_datapath, transforms_=transform_to_tensor, mode = 'train'),\
batch_size = 1, shuffle = True)
#testing dataloader
eval_dataloader = DataLoader(mayo_utils.mayo_dataset(output_datapath, transforms_=transform_to_tensor, mode = 'test'),\
batch_size = 1, shuffle = True)