# -*- coding: utf-8 -*-
from __future__ import absolute_import, print_function
from niftynet.layer.resampler import ResamplerLayer as resampler
from niftynet.network.base_net import BaseNet
from niftynet.network.interventional_affine_net import INetAffine
from niftynet.network.interventional_dense_net import INetDense
[docs]class INetHybridPreWarp(BaseNet):
"""
### Description
Re-implementation of the registration network proposed in:
Hu et al., Label-driven weakly-supervised learning for
multimodal deformable image registration, arXiv:1711.01666
https://arxiv.org/abs/1711.01666
Hu et al., Weakly-Supervised Convolutional Neural Networks for
Multimodal Image Registration, Medical Image Analysis (2018)
https://doi.org/10.1016/j.media.2018.07.002
see also:
https://github.com/YipengHu/label-reg
### Building blocks
[GLOBAL] - INetAffine from interventional_affine_net.py
[RESAMPLER] - Layer to resample the moving image with estimated affine
[DENSE] - INetDense from intervetional_dense_net.py
### Diagram
INPUT PAIR --> [GLOBAL] --> [RESAMPLER] --> [DENSE] --> DENSE FIELD, AFFINE FIELD
### Constraints
- input spatial rank should be either 2 or 3 (2D or 3D images only)
- fixed image size should be divisible by 16
"""
[docs] def __init__(self,
decay,
affine_w_initializer=None,
affine_b_initializer=None,
disp_w_initializer=None,
disp_b_initializer=None,
acti_func='relu',
interp='linear',
boundary='replicate',
name='inet-hybrid-pre-warp'):
"""
:param decay: float, regularisation decay
:param affine_w_initializer: weight initialisation for affine registration network
:param affine_b_initializer: bias initialisation for affine registration network
:param disp_w_initializer: weight initialisation for dense registration network
:param disp_b_initializer: bias initialisation for dense registration network
:param acti_func: activation function to use
:param interp: string, type of interpolation for the resampling [default:linear]
:param boundary: string, padding mode to deal with image boundary
:param name: layer name
"""
BaseNet.__init__(self, name=name)
self.global_net = INetAffine(decay=decay,
affine_w_initializer=affine_w_initializer,
affine_b_initializer=affine_b_initializer,
acti_func=acti_func,
name='inet-global')
self.local_net = INetDense(decay=decay,
disp_w_initializer=disp_w_initializer,
disp_b_initializer=disp_b_initializer,
acti_func=acti_func,
name='inet-local')
self.interp = interp
self.boundary = boundary
[docs] def layer_op(self,
fixed_image,
moving_image,
is_training=True,
**unused_kwargs):
"""
:param fixed_image: tensor, fixed image for registration (defines reference space)
:param moving_image: tensor, moving image to be registered to fixed
:param is_training: boolean, True if network is in training mode
:param unused_kwargs: not in use
:return: estimated final dense and affine displacement fields
"""
affine_field = self.global_net(fixed_image, moving_image, is_training)
moving_image = resampler(
interpolation=self.interp,
boundary=self.boundary)(moving_image, affine_field)
dense_field = self.local_net(
fixed_image, moving_image, affine_field, is_training)
return dense_field, affine_field
[docs]class INetHybridTwoStream(BaseNet):
"""
### Description
Re-implementation of the registration network proposed in:
Hu et al., Label-driven weakly-supervised learning for
multimodal deformable image registration, arXiv:1711.01666
https://arxiv.org/abs/1711.01666
Hu et al., Weakly-Supervised Convolutional Neural Networks for
Multimodal Image Registration, Medical Image Analysis (2018)
https://doi.org/10.1016/j.media.2018.07.002
see also:
https://github.com/YipengHu/label-reg
### Building blocks
[GLOBAL] - INetAffine from interventional_affine_net.py
[DENSE] - INetDense from intervetional_dense_net.py
### Diagram
INPUT PAIR --> [GLOBAL] --> AFFINE FIELD --- DENSE + AFFINE FIELD
| |
-------> [DENSE] --> DENSE FIELD ------
### Constraints
- input spatial rank should be either 2 or 3 (2D or 3D images only)
- fixed image size should be divisible by 16
"""
[docs] def __init__(self,
decay,
affine_w_initializer=None,
affine_b_initializer=None,
disp_w_initializer=None,
disp_b_initializer=None,
acti_func='relu',
interp='linear',
boundary='replicate',
name='inet-hybrid-two-stream'):
"""
:param decay: float, regularisation decay
:param affine_w_initializer: weight initialisation for affine registration network
:param affine_b_initializer: bias initialisation for affine registration network
:param disp_w_initializer: weight initialisation for dense registration network
:param disp_b_initializer: bias initialisation for dense registration network
:param acti_func: activation function to use
:param interp: string, type of interpolation for the resampling [default:linear] - not in use
:param boundary: string, padding mode to deal with image boundary [default: replicate] - not is use
:param name: layer name
"""
BaseNet.__init__(self, name=name)
self.global_net = INetAffine(decay=decay,
affine_w_initializer=affine_w_initializer,
affine_b_initializer=affine_b_initializer,
acti_func=acti_func,
name='inet-global')
self.local_net = INetDense(decay=decay,
disp_w_initializer=disp_w_initializer,
disp_b_initializer=disp_b_initializer,
acti_func=acti_func,
name='inet-local')
self.interp = interp
self.boundary = boundary
[docs] def layer_op(self,
fixed_image,
moving_image,
is_training=True,
**unused_kwargs):
"""
:param fixed_image: tensor, fixed image for registration (defines reference space)
:param moving_image: tensor, moving image to be registered to fixed
:param is_training: boolean, True if network is in training mode
:param unused_kwargs: not in use
:return: estimated total, dense and affine displacement fields
"""
affine_field = self.global_net(fixed_image, moving_image, is_training)
dense_field = self.local_net(fixed_image, moving_image, is_training)
return dense_field + affine_field, dense_field, affine_field
