BIRNet: Brain Image Registration Using Dual-Supervised Fully Convolutional Networks
This work addresses brain image registration, a domain-specific problem in medical imaging, with incremental improvements in training strategies.
The paper tackles the challenge of obtaining ground-truth deformation fields for brain image registration by proposing BIRNet, a dual-supervised fully convolutional network that uses coarse guidance from existing methods and fine guidance from image similarity, achieving promising registration accuracy and efficiency compared to state-of-the-art methods.
In this paper, we propose a deep learning approach for image registration by predicting deformation from image appearance. Since obtaining ground-truth deformation fields for training can be challenging, we design a fully convolutional network that is subject to dual-guidance: (1) Coarse guidance using deformation fields obtained by an existing registration method; and (2) Fine guidance using image similarity. The latter guidance helps avoid overly relying on the supervision from the training deformation fields, which could be inaccurate. For effective training, we further improve the deep convolutional network with gap filling, hierarchical loss, and multi-source strategies. Experiments on a variety of datasets show promising registration accuracy and efficiency compared with state-of-the-art methods.