# Copyright (c) MONAI Consortium # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # http://www.apache.org/licenses/LICENSE-2.0 # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. from __future__ import annotations from collections.abc import Sequence import torch from torch import nn from torch.nn import functional as F from monai.networks.blocks import Convolution from monai.networks.layers import same_padding from monai.networks.layers.factories import Conv, Norm, Pool def get_conv_block( spatial_dims: int, in_channels: int, out_channels: int, kernel_size: Sequence[int] | int = 3, act: tuple | str | None = "RELU", norm: tuple | str | None = "BATCH", ) -> nn.Module: padding = same_padding(kernel_size) mod: nn.Module = Convolution( spatial_dims, in_channels, out_channels, kernel_size=kernel_size, act=act, norm=norm, bias=False, conv_only=False, padding=padding, ) return mod def get_conv_layer( spatial_dims: int, in_channels: int, out_channels: int, kernel_size: Sequence[int] | int = 3 ) -> nn.Module: padding = same_padding(kernel_size) mod: nn.Module = Convolution( spatial_dims, in_channels, out_channels, kernel_size=kernel_size, bias=False, conv_only=True, padding=padding ) return mod def get_deconv_block(spatial_dims: int, in_channels: int, out_channels: int) -> nn.Module: mod: nn.Module = Convolution( spatial_dims=spatial_dims, in_channels=in_channels, out_channels=out_channels, strides=2, act="RELU", norm="BATCH", bias=False, is_transposed=True, padding=1, output_padding=1, ) return mod class ResidualBlock(nn.Module): def __init__( self, spatial_dims: int, in_channels: int, out_channels: int, kernel_size: Sequence[int] | int ) -> None: super().__init__() if in_channels != out_channels: raise ValueError( f"expecting in_channels == out_channels, " f"got in_channels={in_channels}, out_channels={out_channels}" ) self.conv_block = get_conv_block( spatial_dims=spatial_dims, in_channels=in_channels, out_channels=out_channels, kernel_size=kernel_size ) self.conv = get_conv_layer( spatial_dims=spatial_dims, in_channels=out_channels, out_channels=out_channels, kernel_size=kernel_size ) self.norm = Norm[Norm.BATCH, spatial_dims](out_channels) self.relu = nn.ReLU() def forward(self, x) -> torch.Tensor: out: torch.Tensor = self.relu(self.norm(self.conv(self.conv_block(x))) + x) return out class LocalNetResidualBlock(nn.Module): def __init__(self, spatial_dims: int, in_channels: int, out_channels: int) -> None: super().__init__() if in_channels != out_channels: raise ValueError( f"expecting in_channels == out_channels, " f"got in_channels={in_channels}, out_channels={out_channels}" ) self.conv_layer = get_conv_layer(spatial_dims=spatial_dims, in_channels=in_channels, out_channels=out_channels) self.norm = Norm[Norm.BATCH, spatial_dims](out_channels) self.relu = nn.ReLU() def forward(self, x, mid) -> torch.Tensor: out: torch.Tensor = self.relu(self.norm(self.conv_layer(x)) + mid) return out class LocalNetDownSampleBlock(nn.Module): """ A down-sample module that can be used for LocalNet, based on: `Weakly-supervised convolutional neural networks for multimodal image registration `_. `Label-driven weakly-supervised learning for multimodal deformable image registration `_. Adapted from: DeepReg (https://github.com/DeepRegNet/DeepReg) """ def __init__( self, spatial_dims: int, in_channels: int, out_channels: int, kernel_size: Sequence[int] | int ) -> None: """ Args: spatial_dims: number of spatial dimensions. in_channels: number of input channels. out_channels: number of output channels. kernel_size: convolution kernel size. Raises: NotImplementedError: when ``kernel_size`` is even """ super().__init__() self.conv_block = get_conv_block( spatial_dims=spatial_dims, in_channels=in_channels, out_channels=out_channels, kernel_size=kernel_size ) self.residual_block = ResidualBlock( spatial_dims=spatial_dims, in_channels=out_channels, out_channels=out_channels, kernel_size=kernel_size ) self.max_pool = Pool[Pool.MAX, spatial_dims](kernel_size=2) def forward(self, x) -> tuple[torch.Tensor, torch.Tensor]: """ Halves the spatial dimensions. A tuple of (x, mid) is returned: - x is the downsample result, in shape (batch, ``out_channels``, insize_1 / 2, insize_2 / 2, [insize_3 / 2]), - mid is the mid-level feature, in shape (batch, ``out_channels``, insize_1, insize_2, [insize_3]) Args: x: Tensor in shape (batch, ``in_channels``, insize_1, insize_2, [insize_3]) Raises: ValueError: when input spatial dimensions are not even. """ for i in x.shape[2:]: if i % 2 != 0: raise ValueError("expecting x spatial dimensions be even, " f"got x of shape {x.shape}") x = self.conv_block(x) mid = self.residual_block(x) x = self.max_pool(mid) return x, mid class LocalNetUpSampleBlock(nn.Module): """ An up-sample module that can be used for LocalNet, based on: `Weakly-supervised convolutional neural networks for multimodal image registration `_. `Label-driven weakly-supervised learning for multimodal deformable image registration `_. Adapted from: DeepReg (https://github.com/DeepRegNet/DeepReg) """ def __init__( self, spatial_dims: int, in_channels: int, out_channels: int, mode: str = "nearest", align_corners: bool | None = None, ) -> None: """ Args: spatial_dims: number of spatial dimensions. in_channels: number of input channels. out_channels: number of output channels. mode: interpolation mode of the additive upsampling, default to 'nearest'. align_corners: whether to align corners for the additive upsampling, default to None. Raises: ValueError: when ``in_channels != 2 * out_channels`` """ super().__init__() self.deconv_block = get_deconv_block( spatial_dims=spatial_dims, in_channels=in_channels, out_channels=out_channels ) self.conv_block = get_conv_block(spatial_dims=spatial_dims, in_channels=out_channels, out_channels=out_channels) self.residual_block = LocalNetResidualBlock( spatial_dims=spatial_dims, in_channels=out_channels, out_channels=out_channels ) if in_channels / out_channels != 2: raise ValueError( f"expecting in_channels == 2 * out_channels, " f"got in_channels={in_channels}, out_channels={out_channels}" ) self.out_channels = out_channels self.mode = mode self.align_corners = align_corners def additive_upsampling(self, x, mid) -> torch.Tensor: x = F.interpolate(x, mid.shape[2:], mode=self.mode, align_corners=self.align_corners) # [(batch, out_channels, ...), (batch, out_channels, ...)] x = x.split(split_size=int(self.out_channels), dim=1) # (batch, out_channels, ...) out: torch.Tensor = torch.sum(torch.stack(x, dim=-1), dim=-1) return out def forward(self, x, mid) -> torch.Tensor: """ Halves the channel and doubles the spatial dimensions. Args: x: feature to be up-sampled, in shape (batch, ``in_channels``, insize_1, insize_2, [insize_3]) mid: mid-level feature saved during down-sampling, in shape (batch, ``out_channels``, midsize_1, midsize_2, [midsize_3]) Raises: ValueError: when ``midsize != insize * 2`` """ for i, j in zip(x.shape[2:], mid.shape[2:]): if j != 2 * i: raise ValueError( "expecting mid spatial dimensions be exactly the double of x spatial dimensions, " f"got x of shape {x.shape}, mid of shape {mid.shape}" ) h0 = self.deconv_block(x) + self.additive_upsampling(x, mid) r1 = h0 + mid r2 = self.conv_block(h0) out: torch.Tensor = self.residual_block(r2, r1) return out class LocalNetFeatureExtractorBlock(nn.Module): """ A feature-extraction module that can be used for LocalNet, based on: `Weakly-supervised convolutional neural networks for multimodal image registration `_. `Label-driven weakly-supervised learning for multimodal deformable image registration `_. Adapted from: DeepReg (https://github.com/DeepRegNet/DeepReg) """ def __init__( self, spatial_dims: int, in_channels: int, out_channels: int, act: tuple | str | None = "RELU", initializer: str = "kaiming_uniform", ) -> None: """ Args: spatial_dims: number of spatial dimensions. in_channels: number of input channels. out_channels: number of output channels. act: activation type and arguments. Defaults to ReLU. kernel_initializer: kernel initializer. Defaults to None. """ super().__init__() self.conv_block = get_conv_block( spatial_dims=spatial_dims, in_channels=in_channels, out_channels=out_channels, act=act, norm=None ) conv_type: type[nn.Conv1d | nn.Conv2d | nn.Conv3d] = Conv[Conv.CONV, spatial_dims] for m in self.conv_block.modules(): if isinstance(m, conv_type): if initializer == "kaiming_uniform": nn.init.kaiming_normal_(torch.as_tensor(m.weight)) elif initializer == "zeros": nn.init.zeros_(torch.as_tensor(m.weight)) else: raise ValueError( f"initializer {initializer} is not supported, " "currently supporting kaiming_uniform and zeros" ) def forward(self, x) -> torch.Tensor: """ Args: x: Tensor in shape (batch, ``in_channels``, insize_1, insize_2, [insize_3]) """ out: torch.Tensor = self.conv_block(x) return out