A collection of some of MRI/Pytorch-related code fragments.

AdjointGridSample

The adjoint of torch.nn.functional.grid_sample as a differentiable pytorch function

ChannelSeparateConv

Apply the same convolutional filers to all input channels

SeparateConv3d

Flexible Building block for xy+t convolutions. Describe separable convolution by a format string. Allows, for example

• yx, C -> Depthwise Separable Convolution, i.e. grouped 2D convolution in x and y with groups=channels_in, followed by kernel_size=1 convolution
• zyC, zxC -> 2D convolution in zy followed by a 2D convolution in zx.
• yx,z,C -> 2D convolution in yx, followed by a 1D convoution in z, both grouped. Followed by a kernel_size=1 convolution to out_channels
• c, zyx, C - > 1x1x1 channel-mixing convoltion of the input channels. 3d depthwise convolution. 1x1x1 convolution to the number of out_channnels.

Useful to implement networks of PINQI: An End-to-End Physics-Informed Approach to Learned Quantitative MRI Reconstruction (arxiv)

Inati Coil Sensitivity estimation

Implementation of [A Solution to the Phase Problem in Adaptive Coil Combination] (https://archive.ismrm.org/2013/2672.html) as used in (PINQI)

Sliding Window View

Zero-Copy sliding window