DDR-Net: Dividing and Downsampling Mixed Network for Diffeomorphic Image Registration

TL;DR

DDR-Net introduces a multi-scale image registration architecture that effectively balances global context and local details, reducing memory usage and improving accuracy for high-dimensional images like 3D medical volumes.

Contribution

The paper presents DDR-Net, a novel dividing and downsampling approach that preserves image information at multiple scales for improved diffeomorphic registration.

Findings

Outperforms existing registration methods on public datasets.

Reduces memory cost while maintaining registration accuracy.

Effectively captures both coarse and fine image details.

Abstract

Deep diffeomorphic registration faces significant challenges for high-dimensional images, especially in terms of memory limits. Existing approaches either downsample original images, or approximate underlying transformations, or reduce model size. The information loss during the approximation or insufficient model capacity is a hindrance to the registration accuracy for high-dimensional images, e.g., 3D medical volumes. In this paper, we propose a Dividing and Downsampling mixed Registration network (DDR-Net), a general architecture that preserves most of the image information at multiple scales. DDR-Net leverages the global context via downsampling the input and utilizes the local details from divided chunks of the input images. This design reduces the network input size and its memory cost; meanwhile, by fusing global and local information, DDR-Net obtains both coarse-level and fine-level alignments in the final deformation fields. We evaluate DDR-Net on three public datasets, i.e., OASIS, IBSR18, and 3DIRCADB-01, and the experimental results demonstrate our approach outperforms existing approaches.