def inference(args, loader, model, transforms):
src = args.inference
dst = args.save
model.eval()
nvols = reduce(operator.mul, target_split, 1)
# assume single GPU / batch size 1
for data in loader:
data, series, origin, spacing = data[0]
shape = data.size()
# convert names to batch tensor
if args.cuda:
data.pin_memory()
data = data.cuda()
data = Variable(data, volatile=True)
output = model(data)
_, output = output.max(1)
output = output.view(shape)
output = output.cpu()
# merge subvolumes and save
results = output.chunk(nvols)
results = map(
lambda var: torch.squeeze(var.data).numpy().astype(np.int16),
results)
volume = utils.merge_image([*results], target_split)
print("save {}".format(series))
utils.save_updated_image(volume, os.path.join(dst, series + ".mhd"),
origin, spacing)
def normalize_lung_CT(**kwargs):
mean_values = []
var_values = []
MIN_BOUND = -1000
MAX_BOUND = 400
Z_MAX, Y_MAX, X_MAX = kwargs['Z_MAX'], kwargs['Y_MAX'], kwargs['X_MAX']
vox_spacing = kwargs['vox_spacing']
utils.init_dims3D(Z_MAX, Y_MAX, X_MAX, vox_spacing)
luna_subset_path = kwargs['src']
luna_save_path = kwargs['dst']
file_list=glob(luna_subset_path + "/" + "*.mhd")
img_spacing = (vox_spacing, vox_spacing, vox_spacing)
for img_file in file_list:
itk_img = sitk.ReadImage(img_file)
(x_space, y_space, z_space) = itk_img.GetSpacing()
spacing_old = (z_space, y_space, x_space)
img_array = sitk.GetArrayFromImage(itk_img) # indexes are z,y,x (notice the ordering)
img, mu, var = utils.resample_volume(img_array, spacing_old, img_spacing, bounds=(MIN_BOUND, MAX_BOUND))
utils.save_updated_image(img, luna_save_path+os.path.basename(img_file), itk_img.GetOrigin(), img_spacing)
mean_values.append(mu)
var_values.append(var)
dataset_mean = np.mean(mean_values)
dataset_stddev = np.sqrt(np.mean(var_values))
return (dataset_mean, dataset_stddev)
def inference(params, args, loader, model):
src = params['ModelParams']['dirInfer']
dst = params['ModelParams']['dirResult']
model.eval()
# assume single GPU / batch size 1
for batch_idx, data in enumerate(loader):
data, id = data
id = id[0]
itk_img = sitk.ReadImage(os.path.join(src, id))
origin = np.array(list(reversed(itk_img.GetOrigin())))
spacing = np.array(list(reversed(itk_img.GetSpacing())))
# pdb.set_trace()
_, _, z, y, x = data.shape # need to subset shape of 3-d. by Chao.
# convert names to batch tensor
if args.cuda:
data.pin_memory()
data = data.cuda()
with torch.no_grad():
data = Variable(data)
output = model(data)
_, output = output.max(1)
output = output.view((x, y, z))
# pdb.set_trace()
output = output.cpu()
print("save {}".format(id))
utils.save_updated_image(output,
os.path.join(dst, id + "_predicted.mhd"),
origin, spacing)
def normalize_lung_mask(**kwargs):
Z_MAX, Y_MAX, X_MAX = kwargs['Z_MAX'], kwargs['Y_MAX'], kwargs['X_MAX']
vox_spacing = kwargs['vox_spacing']
utils.init_dims3D(Z_MAX, Y_MAX, X_MAX, vox_spacing)
luna_seg_lungs_path = kwargs['src']
luna_seg_lungs_save_path = kwargs['dst']
file_list=glob(os.path.join(luna_seg_lungs_path, "*.mhd"))
img_spacing = (vox_spacing, vox_spacing, vox_spacing)
for img_file in file_list:
itk_img = sitk.ReadImage(img_file)
(x_space, y_space, z_space) = itk_img.GetSpacing()
spacing_old = (z_space, y_space, x_space)
img_array = sitk.GetArrayFromImage(itk_img) # indexes are z,y,x (notice the ordering)
img, _, _ = utils.resample_volume(img_array, spacing_old, img_spacing)
img[img < 1] = 0
utils.save_updated_image(img, os.path.join(luna_seg_lungs_save_path, os.path.basename(img_file)),
itk_img.GetOrigin(), img_spacing)