FiRework: Field Refinement Framework for Efficient Enhancement of Deformable Registration

TL;DR

This work redesigns the continuous deformation framework to mitigate the aforementioned errors, and proposes a novel approach, the field refinement framework (FiRework), tailored for unsupervised deformable registration, aiming to address these challenges.

Abstract

Deformable image registration remains a fundamental task in clinical practice, yet solving registration problems involving complex deformations remains challenging. Current deep learning-based registration methods employ continuous deformation to model large deformations, which often suffer from accumulated registration errors and interpolation inaccuracies. Moreover, achieving satisfactory results with these frameworks typically requires a large number of cascade stages, demanding substantial computational resources. Therefore, we propose a novel approach, the field refinement framework (FiRework), tailored for unsupervised deformable registration, aiming to address these challenges. In FiRework, we redesign the continuous deformation framework to mitigate the aforementioned errors. Notably, our FiRework requires only one level of recursion during training and supports continuous inference, offering improved efficacy compared to continuous deformation frameworks. We conducted experiments on two brain MRI datasets, enhancing two existing deformable registration networks with FiRework. The experimental results demonstrate the superior performance of our proposed framework in deformable registration. The code is publicly available at https://github.com/ZAX130/FiRework.

Figures (7)

• Figure 1: Illustration of a two-stage continuous registration framework. $I_f$ and $I_m$ are the fixed and moving images, respectively. The network at stage 1 predicts the initial deformation field $\phi_1$. Then the warped moving image $I_m \circ \phi_1$, together with $I_f$, are sent into the network at stage 2 to compute the residual deformation field $\varphi_2$. Finally, $\phi_1$ and $\varphi_2$ are fused together to generate the overall deformation field $\phi_2$ for registering.
• Figure 2: Illustration of the training process of the proposed field refinement framework (FiRework). The training process is primarily divided into the initial stage and refinement stage.
• Figure 3: Illustration of the inference process of the proposed FiRework. The figure illustrates the generation process from $\phi_1$ to $\phi_T$.
• Figure 4: Registration results from different methods on LPBA (top row) and Mindboggle (bottom row).
• Figure 5: Illustration of the DSC values obtained for each level of continuous deformation using the methods with proposed FiRework framework on two datasets.