forked from NVlabs/noise2noise
-
Notifications
You must be signed in to change notification settings - Fork 0
/
dataset_tool_mri.py
183 lines (150 loc) · 7.79 KB
/
dataset_tool_mri.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
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
# Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved.
#
# This work is licensed under the Creative Commons Attribution-NonCommercial
# 4.0 International License. To view a copy of this license, visit
# http://creativecommons.org/licenses/by-nc/4.0/ or send a letter to
# Creative Commons, PO Box 1866, Mountain View, CA 94042, USA.
# See README.md in this directory for instructions on how to use this script.
import re
import argparse
import glob
import os
import PIL.Image
import numpy as np
import sys
import util
import nibabel as nib
OUT_RESOLUTION = 256
# Select z-slices from [25,124]
slice_min = 25
slice_max = 125
# Select train and validation subsets from IXI-T1 (these two lists shouldn't overlap)
train_basenames=['IXI002-Guys-0828', 'IXI012-HH-1211', 'IXI013-HH-1212', 'IXI014-HH-1236', 'IXI015-HH-1258', 'IXI016-Guys-0697', 'IXI017-Guys-0698', 'IXI019-Guys-0702', 'IXI020-Guys-0700', 'IXI021-Guys-0703', 'IXI022-Guys-0701', 'IXI023-Guys-0699', 'IXI024-Guys-0705', 'IXI025-Guys-0852', 'IXI026-Guys-0696', 'IXI027-Guys-0710', 'IXI028-Guys-1038', 'IXI029-Guys-0829', 'IXI030-Guys-0708', 'IXI031-Guys-0797', 'IXI033-HH-1259', 'IXI034-HH-1260', 'IXI035-IOP-0873', 'IXI036-Guys-0736', 'IXI037-Guys-0704', 'IXI038-Guys-0729', 'IXI039-HH-1261', 'IXI040-Guys-0724', 'IXI041-Guys-0706', 'IXI042-Guys-0725', 'IXI043-Guys-0714', 'IXI044-Guys-0712', 'IXI045-Guys-0713', 'IXI046-Guys-0824', 'IXI048-HH-1326', 'IXI049-HH-1358', 'IXI050-Guys-0711', 'IXI051-HH-1328', 'IXI052-HH-1343', 'IXI053-Guys-0727', 'IXI054-Guys-0707', 'IXI055-Guys-0730', 'IXI056-HH-1327', 'IXI057-HH-1342', 'IXI058-Guys-0726', 'IXI059-HH-1284', 'IXI060-Guys-0709', 'IXI061-Guys-0715', 'IXI062-Guys-0740', 'IXI063-Guys-0742']
valid_basenames=['IXI064-Guys-0743', 'IXI065-Guys-0744', 'IXI066-Guys-0731', 'IXI067-HH-1356', 'IXI068-Guys-0756', 'IXI069-Guys-0769', 'IXI070-Guys-0767', 'IXI071-Guys-0770', 'IXI072-HH-2324', 'IXI073-Guys-0755']
def fftshift2d(x, ifft=False):
assert (len(x.shape) == 2) and all([(s % 2 == 1) for s in x.shape])
s0 = (x.shape[0] // 2) + (0 if ifft else 1)
s1 = (x.shape[1] // 2) + (0 if ifft else 1)
x = np.concatenate([x[s0:, :], x[:s0, :]], axis=0)
x = np.concatenate([x[:, s1:], x[:, :s1]], axis=1)
return x
def preprocess_mri(input_files,
output_file):
all_files = sorted(input_files)
num_images = len(all_files)
print('Input images: %d' % num_images)
assert num_images > 0
resolution = np.asarray(PIL.Image.open(all_files[0]), dtype=np.uint8).shape
assert len(resolution) == 2 # Expect monochromatic images
print('Image resolution: %s' % str(resolution))
crop_size = tuple([((r - 1) | 1) for r in resolution])
crop_slice = np.s_[:crop_size[0], :crop_size[1]]
print('Crop size: %s' % str(crop_size))
img_primal = np.zeros((num_images,) + resolution, dtype=np.uint8)
img_spectrum = np.zeros((num_images,) + crop_size, dtype=np.complex64)
print('Processing input files..')
for i, fn in enumerate(all_files):
if i % 100 == 0:
print('%d / %d ..' % (i, num_images))
img = np.asarray(PIL.Image.open(fn), dtype=np.uint8)
img_primal[i] = img
img = img.astype(np.float32) / 255.0 - 0.5
img = img[crop_slice]
spec = np.fft.fft2(img).astype(np.complex64)
spec = fftshift2d(spec)
img_spectrum[i] = spec
print('Saving: %s' % output_file)
util.save_pkl((img_primal, img_spectrum), output_file)
def genpng(args):
if args.outdir is None:
print ('Must specify output directory with --outdir')
sys.exit(1)
if args.ixi_dir is None:
print ('Must specify input IXI-T1 directory with --ixi-dir')
sys.exit(1)
mri_directory = args.ixi_dir
out_directory = args.outdir
os.makedirs(out_directory, exist_ok=True)
nii_files = glob.glob(os.path.join(mri_directory, "*.nii.gz"))
for nii_file in nii_files:
print('Processing', nii_file)
nii_img = nib.load(nii_file)
name = os.path.basename(nii_file).split(".")[0]
print("name", name)
hborder = (np.asarray([OUT_RESOLUTION, OUT_RESOLUTION]) - nii_img.shape[0:2]) // 2
print("Img: ", nii_img.shape, " border: ", hborder)
# Normalize image to [0,1]
img = nii_img.get_data().astype(np.float32)
img = img / np.max(img)
print('Max value', np.max(img))
# # Slice along z dimension
#for s in range(70, nii_img.shape[2]-25):
for s in range(slice_min, slice_max):
slice = img[:, :, s]
# Pad to output resolution by inserting zeros
output = np.zeros([OUT_RESOLUTION, OUT_RESOLUTION])
output[hborder[0] : hborder[0] + nii_img.shape[0], hborder[1] : hborder[1] + nii_img.shape[1]] = slice
output = np.minimum(output, 1.0)
output = np.maximum(output, 0.0)
output = output * 255
# Save to png
if np.max(output) > 1.0:
outname = os.path.join(out_directory, "%s_%03d.png" % (name, s))
PIL.Image.fromarray(output).convert('L').save(outname)
def make_slice_name(basename, slice_idx):
return basename + ('-T1_%03d.png' % slice_idx)
def genpkl(args):
if args.png_dir is None:
print ('Must specify PNG directory directory with --png-dir')
sys.exit(1)
if args.pkl_dir is None:
print ('Must specify PKL output directory directory with --pkl-dir')
sys.exit(1)
input_train_files = []
input_valid_files = []
for base in train_basenames:
for sidx in range(slice_min, slice_max):
input_train_files.append(os.path.join(args.png_dir, make_slice_name(base, sidx)))
for base in valid_basenames:
for sidx in range(slice_min, slice_max):
input_valid_files.append(os.path.join(args.png_dir, make_slice_name(base, sidx)))
print ('Num train samples', len(input_train_files))
print ('Num valid samples', len(input_valid_files))
preprocess_mri(input_files=input_train_files, output_file=os.path.join(args.pkl_dir, 'ixi_train.pkl'))
preprocess_mri(input_files=input_valid_files, output_file=os.path.join(args.pkl_dir, 'ixi_valid.pkl'))
def extract_basenames(lst):
s = set()
name_re = re.compile('^(.*)-T1_[0-9]+.png')
for fname in lst:
m = name_re.match(os.path.basename(fname))
if m:
s.add(m[1])
return sorted(list(s))
examples='''examples:
# Convert the IXI-T1 dataset into a set of PNG image files:
python %(prog)s genpng --ixi-dir=~/Downloads/IXI-T1 --outdir=datasets/ixi-png
# Convert the PNG image files into a Python pickle for use in training:
python %(prog)s genpkl --png-dir=datasets/ixi-png --pkl-dir=datasets
'''
def main():
parser = argparse.ArgumentParser(
description='Convert the IXI-T1 dataset into a format suitable for network training',
epilog=examples,
formatter_class=argparse.RawDescriptionHelpFormatter
)
subparsers = parser.add_subparsers(help='Sub-commands')
parser_genpng = subparsers.add_parser('genpng', help='IXI nifti to PNG converter (intermediate step)')
parser_genpng.add_argument('--ixi-dir', help='Directory pointing to unpacked IXI-T1.tar')
parser_genpng.add_argument('--outdir', help='Directory where to save .PNG files')
parser_genpng.set_defaults(func=genpng)
parser_genpkl = subparsers.add_parser('genpkl', help='PNG to PKL converter (used in training)')
parser_genpkl.add_argument('--png-dir', help='Directory containing .PNGs saved by with the genpng command')
parser_genpkl.add_argument('--pkl-dir', help='Where to save the .pkl files for train and valid sets')
parser_genpkl.set_defaults(func=genpkl)
args = parser.parse_args()
if 'func' not in args:
print ('No command given. Try --help.')
sys.exit(1)
args.func(args)
if __name__ == "__main__":
main()