niftynet.network.interventional_hybrid_net module

class INetHybridPreWarp(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')[source]

Bases: niftynet.network.base_net.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
__init__(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')[source]
Parameters:
  • decay – float, regularisation decay
  • affine_w_initializer – weight initialisation for affine registration network
  • affine_b_initializer – bias initialisation for affine registration network
  • disp_w_initializer – weight initialisation for dense registration network
  • disp_b_initializer – bias initialisation for dense registration network
  • acti_func – activation function to use
  • interp – string, type of interpolation for the resampling [default:linear]
  • boundary – string, padding mode to deal with image boundary
  • name – layer name
layer_op(fixed_image, moving_image, is_training=True, **unused_kwargs)[source]
Parameters:
  • fixed_image – tensor, fixed image for registration (defines reference space)
  • moving_image – tensor, moving image to be registered to fixed
  • is_training – boolean, True if network is in training mode
  • unused_kwargs – not in use
Returns:

estimated final dense and affine displacement fields

class INetHybridTwoStream(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')[source]

Bases: niftynet.network.base_net.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
__init__(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')[source]
Parameters:
  • decay – float, regularisation decay
  • affine_w_initializer – weight initialisation for affine registration network
  • affine_b_initializer – bias initialisation for affine registration network
  • disp_w_initializer – weight initialisation for dense registration network
  • disp_b_initializer – bias initialisation for dense registration network
  • acti_func – activation function to use
  • interp – string, type of interpolation for the resampling [default:linear] - not in use
  • boundary – string, padding mode to deal with image boundary [default: replicate] - not is use
  • name – layer name
layer_op(fixed_image, moving_image, is_training=True, **unused_kwargs)[source]
Parameters:
  • fixed_image – tensor, fixed image for registration (defines reference space)
  • moving_image – tensor, moving image to be registered to fixed
  • is_training – boolean, True if network is in training mode
  • unused_kwargs – not in use
Returns:

estimated total, dense and affine displacement fields