MIRTorch: A PyTorch-powered Differentiable Toolbox for Fast Image Reconstruction and Scan Protocol Optimization

MIRTorch (Michigan Image Reconstruction Toolbox for PyTorch) is an image reconstruction toolbox implemented with native Python/PyTorch. It provides fast iterative and data-driven image reconstruction on both CPUs and GPUs. Researchers can rapidly develop new model-based and learning-based methods (i.e., unrolled neural networks) with its convenient abstraction layers. With the full support of auto-differentiation, one may optimize imaging protocols, such as sampling patterns and image reconstruction parameters with gradient-based methods.