Flow Matching-enabled Test-Time Refinement for Unsupervised Cardiac MR Registration
This work addresses the practical use of cardiac MR registration for medical imaging by enabling faster and more accurate unsupervised alignment, though it is incremental as it builds on existing flow-matching and refinement techniques.
The paper tackled the problem of expensive multi-step inference in diffusion-based unsupervised cardiac MR registration by proposing FlowReg, a flow-matching framework that achieves strong registration in as few as two steps, outperforming state-of-the-art methods on five out of six tasks with a +0.6% mean Dice score improvement and reducing LVEF estimation error by -2.58 percentage points.
Diffusion-based unsupervised image registration has been explored for cardiac cine MR, but expensive multi-step inference limits practical use. We propose FlowReg, a flow-matching framework in displacement field space that achieves strong registration in as few as two steps and supports further refinement with more steps. FlowReg uses warmup-reflow training: a single-step network first acts as a teacher, then a student learns to refine from arbitrary intermediate states, removing the need for a pre-trained model as in existing methods. An Initial Guess strategy feeds back the model prediction as the next starting point, improving refinement from step two onward. On ACDC and MM2 across six tasks (including cross-dataset generalization), FlowReg outperforms the state of the art on five tasks (+0.6% mean Dice score on average), with the largest gain in the left ventricle (+1.09%), and reduces LVEF estimation error on all six tasks (-2.58 percentage points), using only 0.7% extra parameters and no segmentation labels. Code is available at https://github.com/mathpluscode/FlowReg.