SVoRT: Iterative Transformer for Slice-to-Volume Registration in Fetal Brain MRI
This addresses the problem of accurate 3D reconstruction for fetal brain MRI in clinical settings, where motion artifacts are common, representing an incremental improvement over existing methods.
The paper tackles the challenge of reconstructing 3D fetal brain volumes from MRI slices affected by severe motion by proposing a Transformer-based slice-to-volume registration method that uses attention to relate slices and iteratively updates transformations and volume estimates. Results show lower registration error and better reconstruction quality compared to state-of-the-art methods on synthetic data, with real-world experiments demonstrating improved 3D reconstruction under severe motion.
Volumetric reconstruction of fetal brains from multiple stacks of MR slices, acquired in the presence of almost unpredictable and often severe subject motion, is a challenging task that is highly sensitive to the initialization of slice-to-volume transformations. We propose a novel slice-to-volume registration method using Transformers trained on synthetically transformed data, which model multiple stacks of MR slices as a sequence. With the attention mechanism, our model automatically detects the relevance between slices and predicts the transformation of one slice using information from other slices. We also estimate the underlying 3D volume to assist slice-to-volume registration and update the volume and transformations alternately to improve accuracy. Results on synthetic data show that our method achieves lower registration error and better reconstruction quality compared with existing state-of-the-art methods. Experiments with real-world MRI data are also performed to demonstrate the ability of the proposed model to improve the quality of 3D reconstruction under severe fetal motion.