-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathnnunet_retouch.py
176 lines (166 loc) · 6.42 KB
/
nnunet_retouch.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
import argparse
import os
import random
from concurrent.futures import ProcessPoolExecutor
from pathlib import Path
import numpy as np
import SimpleITK as sitk
from nnunetv2.dataset_conversion.generate_dataset_json import generate_dataset_json
from segmentation_failures.utils.io import load_json, save_json
def process_case(
case_dir, img_out_dir, label_out_dir, scanner, only_stats=False, merge_labels=False
):
# load with simpleITK
img = sitk.ReadImage(str(case_dir / "oct.mhd"))
seg = sitk.ReadImage(str(case_dir / "reference.mhd"))
# for the image, we need to normalize the intensities
# Cirrus: 0-255, Spectralis: 0-2**16, Topcon: 0-255
img_arr = sitk.GetArrayFromImage(img)
image_stats = {
"min": img_arr.min(),
"max": img_arr.max(),
"dtype": str(img_arr.dtype),
}
if img_arr.dtype == np.uint8:
img_arr = img_arr.astype(np.float32) / 255
elif img_arr.dtype == np.uint16:
img_arr = img_arr.astype(np.float32) / 2**16
if merge_labels:
seg_old = seg
seg_arr = sitk.GetArrayFromImage(seg)
seg_arr[seg_arr != 0] = 1
seg = sitk.GetImageFromArray(seg_arr)
seg.CopyInformation(seg_old)
# save in new location
if not only_stats:
img_out = sitk.GetImageFromArray(img_arr)
img_out.CopyInformation(img)
sitk.WriteImage(img_out, str(img_out_dir / f"{case_dir.name}_0000.nii.gz"))
sitk.WriteImage(seg, str(label_out_dir / f"{case_dir.name}.nii.gz"))
return image_stats
def main():
subsets = ["Cirrus", "Spectralis", "Topcon"]
random.seed(420000)
parser = argparse.ArgumentParser()
parser.add_argument("raw_data_dir", type=str, help="Path to raw data directory (MSD)")
parser.add_argument(
"--ood_subset", default="Cirrus", choices=subsets, help="OOD subset to use."
)
parser.add_argument(
"--use_default_splits",
action="store_true",
help="Use default train/test split.",
)
parser.add_argument(
"--merge_labels",
action="store_true",
help="Merge labels to a single fluid class.",
)
args = parser.parse_args()
source_dir = Path(args.raw_data_dir)
if not source_dir.exists():
raise FileNotFoundError("One of the specified directories does not exist")
TASK_NAME = f"Dataset5{4 + args.merge_labels}{subsets.index(args.ood_subset)}_RETOUCH_ood={args.ood_subset}"
print(f"Processing data for task {TASK_NAME}")
num_id_test_cases_per_scanner = 6
nnunet_split = None
if args.use_default_splits:
default_split_path = (
Path(__file__).resolve().parents[4]
/ "dataset_splits"
/ TASK_NAME
/ "splits_final.json"
)
nnunet_split = load_json(default_split_path)
target_root_dir = Path(os.environ["nnUNet_raw"]) / TASK_NAME
target_root_dir.mkdir(exist_ok=True)
images_train_dir = target_root_dir / "imagesTr"
images_test_dir = target_root_dir / "imagesTs"
labels_train_dir = target_root_dir / "labelsTr"
labels_test_dir = target_root_dir / "labelsTs"
images_train_dir.mkdir()
labels_train_dir.mkdir()
images_test_dir.mkdir()
labels_test_dir.mkdir()
# get a list of all patients for each scanner
# and do the train/test split
case_dict = {}
all_cases = []
for curr_scanner in subsets:
subset_dir = source_dir / f"RETOUCH-TrainingSet-{curr_scanner}"
case_dict[curr_scanner] = [x.name for x in subset_dir.iterdir() if x.is_dir()]
all_cases.extend(case_dict[curr_scanner])
if nnunet_split is not None:
train_cases = []
for fold in nnunet_split:
train_cases.extend(fold["train"])
test_cases = list(set(all_cases) - set(train_cases))
else:
train_cases = []
test_cases = []
for scanner, case_list in case_dict.items():
if scanner == args.ood_subset:
test_cases.extend(case_list)
else:
selected_cases = random.sample(case_list, num_id_test_cases_per_scanner)
train_cases.extend([x for x in case_list if x not in selected_cases])
test_cases.extend(selected_cases)
# process all cases
all_stats = {k: [] for k in subsets}
case_to_domain_map = {}
results = []
print(f"Processing {len(train_cases)} training cases and {len(test_cases)} test cases")
with ProcessPoolExecutor(max_workers=6) as executor:
for scanner, case_list in case_dict.items():
for case_id in case_list:
if case_id in test_cases:
img_out_dir = images_test_dir
label_out_dir = labels_test_dir
case_to_domain_map[case_id] = scanner
else:
img_out_dir = images_train_dir
label_out_dir = labels_train_dir
results.append(
(
scanner,
executor.submit(
process_case,
source_dir / f"RETOUCH-TrainingSet-{scanner}" / case_id,
img_out_dir,
label_out_dir,
scanner=scanner,
merge_labels=args.merge_labels,
),
)
)
for scanner, r in results:
stats = r.result()
all_stats[scanner].append(stats)
for scanner, stats in all_stats.items():
print(f"=== Stats for {scanner} ===")
overall_min = min([x["min"] for x in stats])
overall_max = max([x["max"] for x in stats])
all_dtypes = set([x["dtype"] for x in stats])
print(f"Overall min: {overall_min}, max: {overall_max}")
print(f"Overall dtypes: {all_dtypes}")
save_json(case_to_domain_map, target_root_dir / "domain_mapping_00.json")
label_dict = {
"background": 0,
"Intraretinal Fluid": 1,
"Subretinal Fluid": 2,
"Pigment Epithelium Detachments": 3,
}
if args.merge_labels:
label_dict = {"background": 0, "fluid": 1}
generate_dataset_json(
output_folder=str(target_root_dir),
channel_names={0: "OCT"},
labels=label_dict,
num_training_cases=len(train_cases),
file_ending=".nii.gz",
dataset_name=TASK_NAME,
dim=3,
)
if __name__ == "__main__":
main()
# no need for special train-val splits here