🤖 AI Summary
To address the dual challenges of weak feature representation—specifically, the lack of global anatomical semantics in local descriptors—and low optimization efficiency—namely, high computational cost and poor handling of large deformations—in CT image registration, this paper proposes a coarse-to-fine discrete optimization framework. Methodologically: (i) we design a decoupled self-supervised anatomical embedding (SAM) that jointly models local texture and global organ structure; (ii) we construct a 6D correlation pyramid for efficient deformation field estimation; and (iii) we introduce a novel convex-relaxed discrete optimization pipeline. Evaluated on abdominal, head-and-neck, and pulmonary CT datasets, our method consistently outperforms state-of-the-art learning-based and optimization-based approaches. It achieves high accuracy with real-time efficiency: average registration time is ~2 seconds per case (~5 seconds including instance-specific optimization).
📝 Abstract
Estimating displacement vector field via a cost volume computed in the feature space has shown great success in image registration, but it suffers excessive computation burdens. Moreover, existing feature descriptors only extract local features incapable of representing the global semantic information, which is especially important for solving large transformations. To address the discussed issues, we propose SAMConvex, a fast coarse-to-fine discrete optimization method for CT registration that includes a decoupled convex optimization procedure to obtain deformation fields based on a self-supervised anatomical embedding (SAM) feature extractor that captures both local and global information. To be specific, SAMConvex extracts per-voxel features and builds 6D correlation volumes based on SAM features, and iteratively updates a flow field by performing lookups on the correlation volumes with a coarse-to-fine scheme. SAMConvex outperforms the state-of-the-art learning-based methods and optimization-based methods over two inter-patient registration datasets (Abdomen CT and HeadNeck CT) and one intra-patient registration dataset (Lung CT). Moreover, as an optimization-based method, SAMConvex only takes $sim2$s ($sim5s$ with instance optimization) for one paired images.
