Evaluation of Deformable Image Registration under Alignment-Regularity Trade-off
This work provides a nuanced evaluation framework for practitioners and researchers in medical imaging, though it is incremental as it builds on existing DIR methods by refining evaluation practices rather than introducing a new registration technique.
The paper tackles the challenge of evaluating deformable image registration methods by addressing the trade-off between alignment accuracy and deformation regularity, proposing an evaluation scheme that uses alignment regularity characteristic curves to reveal insights not captured by existing practices, with experiments on deep learning methods showing improved sample density and interpolation across regularization ranges.
Evaluating deformable image registration (DIR) is challenging due to the inherent trade-off between achieving high alignment accuracy and maintaining deformation regularity. However, most existing DIR works either address this trade-off inadequately or overlook it altogether. In this paper, we highlight the issues with existing practices and propose an evaluation scheme that captures the trade-off continuously to holistically evaluate DIR methods. We first introduce the alignment regularity characteristic (ARC) curves, which describe the performance of a given registration method as a spectrum under various degrees of regularity. We demonstrate that the ARC curves reveal unique insights that are not evident from existing evaluation practices, using experiments on representative deep learning DIR methods with various network architectures and transformation models. We further adopt a HyperNetwork based approach that learns to continuously interpolate across the full regularization range, accelerating the construction and improving the sample density of ARC curves. Finally, we provide general guidelines for a nuanced model evaluation and selection using our evaluation scheme for both practitioners and registration researchers.