def intensity(self, x):
transform = tio.OneOf([
tio.RandomMotion(translation=1),
tio.RandomBlur(),
tio.RandomGamma(),
tio.RandomSpike(intensity=[0.2, 0.5]),
tio.RandomBiasField()
])
x = transform(x.unsqueeze(0)).squeeze(0)
return x
def __init__(self, X, Y, mixup=0, aug=False):
self.X = X.astype('float32')
self.Y = Y.astype('float32')
self.aug = aug
if self.aug:
self.intensity = [
tio.RandomMotion(translation=1),
tio.RandomBlur(),
tio.RandomGamma(),
tio.RandomSpike(intensity=[0.2, 0.5]),
tio.RandomBiasField()
]
else:
self.intensity = None
self.mixup = mixup
def train_transform(self):
# tf_list = [iotf.CropOrPad(self.volume_size, padding_mode='edge')]
tf_list = []
if self.randomflip['enable']:
params = {
key: val
for key, val in self.randomflip.items() if key != 'enable'
}
tf_list.append(iotf.RandomFlip(**params))
if self.randomaffine['enable']:
params = {
key: val
for key, val in self.randomaffine.items() if key != 'enable'
}
tf_list.append(iotf.RandomAffine(**params))
if self.randomblur['enable']:
params = {
key: val
for key, val in self.randomblur.items() if key != 'enable'
}
tf_list.append(iotf.RandomBlur(**params))
if self.randomnoise['enable']:
params = {
key: val
for key, val in self.randomnoise.items() if key != 'enable'
}
tf_list.append(iotf.RandomNoise(**params))
if self.randomswap['enable']:
params = {
key: val
for key, val in self.randomswap.items() if key != 'enable'
}
tf_list.append(iotf.RandomSwap(**params))
if self.randomelasticdeformation['enable']:
params = {
key: val
for key, val in self.randomelasticdeformation.items()
if key != 'enable'
}
tf_list.append(iotf.RandomElasticDeformation(**params))
transform = iotf.Compose(tf_list)
return transform