Motivation: Developing new MR acquisition and reconstruction techniques requires data for testing, benchmarking, and, for deep-learning approaches, training. Digital twins help meet this need. While some frameworks include phantom generation routines, no dedicated, comprehensive package exists, especially for quantitative MR (qMRI). Goal(s): To provide a lightweight, user-friendly tool for generating virtual phantoms for benchmarking and neural network training. Approach: Our framework includes routines for generating virtual objects (e.g., numerical and brain-like phantoms), field inhomogeneities, coil sensitivities, gradient imperfections, and motion patterns. Results: Our package produces realistic MR parameter distributions and field maps, validated by simulating a multi-echo SPGR dataset. Impact: MRTwin represents a useful tool for sequence design, reconstruction optimization and benchmarking, by providing a framework for the generation of digital twins for quantitative imaging.
Cencini et al. (Tue,) studied this question.