Commit
·
7968536
1
Parent(s):
15c9383
Added --output, --erase, and --save-nifti flags, as with the other evaluation scripts
Browse files- SoA_methods/eval_ants.py +46 -32
SoA_methods/eval_ants.py
CHANGED
@@ -18,7 +18,7 @@ from DeepDeformationMapRegistration.utils.misc import DisplacementMapInterpolato
|
|
18 |
from DeepDeformationMapRegistration.utils.nifti_utils import save_nifti
|
19 |
from DeepDeformationMapRegistration.utils.visualization import save_disp_map_img, plot_predictions
|
20 |
import DeepDeformationMapRegistration.utils.constants as C
|
21 |
-
|
22 |
import medpy.metric as medpy_metrics
|
23 |
|
24 |
import voxelmorph as vxm
|
@@ -41,16 +41,19 @@ INV_TRFS = 'invtransforms'
|
|
41 |
if __name__ == '__main__':
|
42 |
parser = ArgumentParser()
|
43 |
parser.add_argument('--dataset', type=str, help='Directory with the images')
|
44 |
-
parser.add_argument('--
|
45 |
parser.add_argument('--gpu', type=int, help='GPU')
|
|
|
|
|
46 |
args = parser.parse_args()
|
47 |
|
48 |
os.environ['CUDA_DEVICE_ORDER'] = 'PCI_BUS_ID'
|
49 |
os.environ['CUDA_VISIBLE_DEVICES'] = str(args.gpu)
|
50 |
-
|
51 |
-
|
52 |
-
os.makedirs(
|
53 |
-
os.makedirs(os.path.join(args.
|
|
|
54 |
dataset_files = os.listdir(args.dataset)
|
55 |
dataset_files.sort()
|
56 |
dataset_files = [os.path.join(args.dataset, f) for f in dataset_files if re.match(DATASET_NAMES, f)]
|
@@ -59,7 +62,7 @@ if __name__ == '__main__':
|
|
59 |
|
60 |
f = h5py.File(dataset_files[0], 'r')
|
61 |
image_shape = list(f['fix_image'][:].shape[:-1])
|
62 |
-
nb_labels = f['fix_segmentations'][:].shape[-1]
|
63 |
f.close()
|
64 |
|
65 |
#### TF prep
|
@@ -96,10 +99,10 @@ if __name__ == '__main__':
|
|
96 |
print("Running ANTs using {} threads".format(os.environ.get("ITK_GLOBAL_DEFAULT_NUMBER_OF_THREADS")))
|
97 |
# dm_interp = DisplacementMapInterpolator(image_shape, 'griddata')
|
98 |
# Header of the metrics csv file
|
99 |
-
csv_header = ['File', 'SSIM', 'MS-SSIM', 'NCC', 'MSE', 'DICE', 'DICE_MACRO', 'HD', 'Time', 'TRE']
|
100 |
|
101 |
-
metrics_file = {'SyN': os.path.join(args.
|
102 |
-
'SyNCC': os.path.join(args.
|
103 |
for k in metrics_file.keys():
|
104 |
with open(metrics_file[k], 'w') as f:
|
105 |
f.write(';'.join(csv_header)+'\n')
|
@@ -113,11 +116,14 @@ if __name__ == '__main__':
|
|
113 |
fix_img = vol_file['fix_image'][:]
|
114 |
mov_img = vol_file['mov_image'][:]
|
115 |
|
116 |
-
fix_seg = vol_file['fix_segmentations'][:]
|
117 |
-
mov_seg = vol_file['mov_segmentations'][:]
|
|
|
|
|
|
|
118 |
|
119 |
-
|
120 |
-
|
121 |
|
122 |
# ndarray to ANTsImage
|
123 |
fix_img_ants = ants.make_image(fix_img.shape[:-1], np.squeeze(fix_img)) # SoA doesn't work fine with 1-ch images
|
@@ -155,6 +161,8 @@ if __name__ == '__main__':
|
|
155 |
fix_seg[np.newaxis,..., l]) for l in range(nb_labels)])
|
156 |
hd = np.mean(
|
157 |
[medpy_metrics.hd(pred_seg[np.newaxis,..., l], fix_seg[np.newaxis,..., l]) for l in range(nb_labels)])
|
|
|
|
|
158 |
dice_macro = np.mean(
|
159 |
[medpy_metrics.dc(pred_seg[np.newaxis,..., l], fix_seg[np.newaxis,..., l]) for l in range(nb_labels)])
|
160 |
|
@@ -178,32 +186,38 @@ if __name__ == '__main__':
|
|
178 |
disp_map = np.squeeze(np.asarray(nb.load(mov_to_fix_trf_list[0]).dataobj))
|
179 |
dm_interp = DisplacementMapInterpolator(fix_img.shape[:-1], 'griddata', step=2)
|
180 |
pred_centroids = dm_interp(disp_map, mov_centroids, backwards=True) + mov_centroids
|
|
|
181 |
# upsample_scale = 128 / 64
|
182 |
# fix_centroids_isotropic = fix_centroids * upsample_scale
|
183 |
# pred_centroids_isotropic = pred_centroids * upsample_scale
|
184 |
|
|
|
|
|
|
|
185 |
# fix_centroids_isotropic = np.divide(fix_centroids_isotropic, C.COMET_DATASET_iso_to_cubic_scales)
|
186 |
# pred_centroids_isotropic = np.divide(pred_centroids_isotropic, C.COMET_DATASET_iso_to_cubic_scales)
|
187 |
tre_array = target_registration_error(fix_centroids, pred_centroids, False).eval()
|
188 |
tre = np.mean([v for v in tre_array if not np.isnan(v)])
|
189 |
-
|
190 |
-
|
191 |
-
|
192 |
-
|
193 |
-
|
194 |
-
|
195 |
-
|
196 |
-
|
197 |
-
|
198 |
-
|
199 |
-
|
200 |
-
|
201 |
-
|
202 |
-
|
203 |
-
|
204 |
-
|
205 |
-
|
206 |
-
|
|
|
|
|
207 |
|
208 |
for k in metrics_file.keys():
|
209 |
print('Summary {}\n=======\n'.format(k))
|
|
|
18 |
from DeepDeformationMapRegistration.utils.nifti_utils import save_nifti
|
19 |
from DeepDeformationMapRegistration.utils.visualization import save_disp_map_img, plot_predictions
|
20 |
import DeepDeformationMapRegistration.utils.constants as C
|
21 |
+
import shutil
|
22 |
import medpy.metric as medpy_metrics
|
23 |
|
24 |
import voxelmorph as vxm
|
|
|
41 |
if __name__ == '__main__':
|
42 |
parser = ArgumentParser()
|
43 |
parser.add_argument('--dataset', type=str, help='Directory with the images')
|
44 |
+
parser.add_argument('--outdirname', type=str, help='Output directory')
|
45 |
parser.add_argument('--gpu', type=int, help='GPU')
|
46 |
+
parser.add_argument('--savenifti', type=bool, default=True)
|
47 |
+
parser.add_argument('--erase', type=bool, help='Erase the content of the output folder', default=False)
|
48 |
args = parser.parse_args()
|
49 |
|
50 |
os.environ['CUDA_DEVICE_ORDER'] = 'PCI_BUS_ID'
|
51 |
os.environ['CUDA_VISIBLE_DEVICES'] = str(args.gpu)
|
52 |
+
if args.erase:
|
53 |
+
shutil.rmtree(args.outdirname, ignore_errors=True)
|
54 |
+
os.makedirs(args.outdirname, exist_ok=True)
|
55 |
+
os.makedirs(os.path.join(args.outdirname, 'SyN'), exist_ok=True)
|
56 |
+
os.makedirs(os.path.join(args.outdirname, 'SyNCC'), exist_ok=True)
|
57 |
dataset_files = os.listdir(args.dataset)
|
58 |
dataset_files.sort()
|
59 |
dataset_files = [os.path.join(args.dataset, f) for f in dataset_files if re.match(DATASET_NAMES, f)]
|
|
|
62 |
|
63 |
f = h5py.File(dataset_files[0], 'r')
|
64 |
image_shape = list(f['fix_image'][:].shape[:-1])
|
65 |
+
nb_labels = f['fix_segmentations'][:].shape[-1] - 1
|
66 |
f.close()
|
67 |
|
68 |
#### TF prep
|
|
|
99 |
print("Running ANTs using {} threads".format(os.environ.get("ITK_GLOBAL_DEFAULT_NUMBER_OF_THREADS")))
|
100 |
# dm_interp = DisplacementMapInterpolator(image_shape, 'griddata')
|
101 |
# Header of the metrics csv file
|
102 |
+
csv_header = ['File', 'SSIM', 'MS-SSIM', 'NCC', 'MSE', 'DICE', 'DICE_MACRO', 'HD', 'HD95', 'Time', 'TRE']
|
103 |
|
104 |
+
metrics_file = {'SyN': os.path.join(args.outdirname, 'SyN', 'metrics.csv'),
|
105 |
+
'SyNCC': os.path.join(args.outdirname, 'SyNCC', 'metrics.csv')}
|
106 |
for k in metrics_file.keys():
|
107 |
with open(metrics_file[k], 'w') as f:
|
108 |
f.write(';'.join(csv_header)+'\n')
|
|
|
116 |
fix_img = vol_file['fix_image'][:]
|
117 |
mov_img = vol_file['mov_image'][:]
|
118 |
|
119 |
+
fix_seg = vol_file['fix_segmentations'][..., 1:].astype(np.float32)
|
120 |
+
mov_seg = vol_file['mov_segmentations'][..., 1:].astype(np.float32)
|
121 |
+
|
122 |
+
fix_centroids = vol_file['fix_centroids'][1:, ...]
|
123 |
+
mov_centroids = vol_file['mov_centroids'][1:, ...]
|
124 |
|
125 |
+
isotropic_shape = vol_file['isotropic_shape'][:]
|
126 |
+
voxel_size = np.divide(fix_img.shape[:-1], isotropic_shape)
|
127 |
|
128 |
# ndarray to ANTsImage
|
129 |
fix_img_ants = ants.make_image(fix_img.shape[:-1], np.squeeze(fix_img)) # SoA doesn't work fine with 1-ch images
|
|
|
161 |
fix_seg[np.newaxis,..., l]) for l in range(nb_labels)])
|
162 |
hd = np.mean(
|
163 |
[medpy_metrics.hd(pred_seg[np.newaxis,..., l], fix_seg[np.newaxis,..., l]) for l in range(nb_labels)])
|
164 |
+
hd95 = np.mean(
|
165 |
+
[medpy_metrics.hd95(pred_seg[np.newaxis,..., l], fix_seg[np.newaxis,..., l]) for l in range(nb_labels)])
|
166 |
dice_macro = np.mean(
|
167 |
[medpy_metrics.dc(pred_seg[np.newaxis,..., l], fix_seg[np.newaxis,..., l]) for l in range(nb_labels)])
|
168 |
|
|
|
186 |
disp_map = np.squeeze(np.asarray(nb.load(mov_to_fix_trf_list[0]).dataobj))
|
187 |
dm_interp = DisplacementMapInterpolator(fix_img.shape[:-1], 'griddata', step=2)
|
188 |
pred_centroids = dm_interp(disp_map, mov_centroids, backwards=True) + mov_centroids
|
189 |
+
# Rescale the points back to isotropic space, where we have a correspondence voxel <-> mm
|
190 |
# upsample_scale = 128 / 64
|
191 |
# fix_centroids_isotropic = fix_centroids * upsample_scale
|
192 |
# pred_centroids_isotropic = pred_centroids * upsample_scale
|
193 |
|
194 |
+
fix_centroids_isotropic = fix_centroids * voxel_size
|
195 |
+
pred_centroids_isotropic = pred_centroids * voxel_size
|
196 |
+
|
197 |
# fix_centroids_isotropic = np.divide(fix_centroids_isotropic, C.COMET_DATASET_iso_to_cubic_scales)
|
198 |
# pred_centroids_isotropic = np.divide(pred_centroids_isotropic, C.COMET_DATASET_iso_to_cubic_scales)
|
199 |
tre_array = target_registration_error(fix_centroids, pred_centroids, False).eval()
|
200 |
tre = np.mean([v for v in tre_array if not np.isnan(v)])
|
201 |
+
if np.isnan(tre):
|
202 |
+
print('TRE is NaN for {} and file {}'.format(reg_method, step))
|
203 |
+
|
204 |
+
dataset_iterator.set_description('{} ({}): Saving data {}'.format(file_num, file_path, reg_method))
|
205 |
+
new_line = [step, ssim, ms_ssim, ncc, mse, dice, dice_macro, hd, hd95,
|
206 |
+
t1_syn-t0_syn if reg_method == 'SyN' else t1_syncc-t0_syncc,
|
207 |
+
tre]
|
208 |
+
with open(metrics_file[reg_method], 'a') as f:
|
209 |
+
f.write(';'.join(map(str, new_line))+'\n')
|
210 |
+
if args.savenifti:
|
211 |
+
save_nifti(fix_img, os.path.join(args.outdirname, reg_method, '{:03d}_fix_img_ssim_{:.03f}_dice_{:.03f}.nii.gz'.format(step, ssim, dice)), verbose=False)
|
212 |
+
save_nifti(mov_img, os.path.join(args.outdirname, reg_method, '{:03d}_mov_img_ssim_{:.03f}_dice_{:.03f}.nii.gz'.format(step, ssim, dice)), verbose=False)
|
213 |
+
save_nifti(pred_img, os.path.join(args.outdirname, reg_method, '{:03d}_pred_img_ssim_{:.03f}_dice_{:.03f}.nii.gz'.format(step, ssim, dice)), verbose=False)
|
214 |
+
save_nifti(fix_seg_card, os.path.join(args.outdirname, reg_method, '{:03d}_fix_seg_ssim_{:.03f}_dice_{:.03f}.nii.gz'.format(step, ssim, dice)), verbose=False)
|
215 |
+
save_nifti(mov_seg_card, os.path.join(args.outdirname, reg_method, '{:03d}_mov_seg_ssim_{:.03f}_dice_{:.03f}.nii.gz'.format(step, ssim, dice)), verbose=False)
|
216 |
+
save_nifti(pred_seg_card, os.path.join(args.outdirname, reg_method, '{:03d}_pred_seg_ssim_{:.03f}_dice_{:.03f}.nii.gz'.format(step, ssim, dice)), verbose=False)
|
217 |
+
|
218 |
+
plot_predictions(img_batches=[fix_img[np.newaxis, ...], mov_img[np.newaxis, ...], pred_img[np.newaxis, ...]], disp_map_batch=disp_map[np.newaxis, ...], seg_batches=[fix_seg_card[np.newaxis, ...], mov_seg_card[np.newaxis, ...], pred_seg_card[np.newaxis, ...]], filename=os.path.join(args.outdirname, reg_method, '{:03d}_figures_seg.png'.format(step)), show=False)
|
219 |
+
plot_predictions(img_batches=[fix_img[np.newaxis, ...], mov_img[np.newaxis, ...], pred_img[np.newaxis, ...]], disp_map_batch=disp_map[np.newaxis, ...], filename=os.path.join(args.outdirname, reg_method, '{:03d}_figures_img.png'.format(step)), show=False)
|
220 |
+
save_disp_map_img(disp_map[np.newaxis, ...], 'Displacement map', os.path.join(args.outdirname, reg_method, '{:03d}_disp_map_fig.png'.format(step)), show=False)
|
221 |
|
222 |
for k in metrics_file.keys():
|
223 |
print('Summary {}\n=======\n'.format(k))
|