This work presents Recurrence with Correlation Network(RWCNet), a novel multi-scale recurrent neural network architecture for medical image registration that integrates core principles from optical flow, including correlation volume computation and inference-time instance optimization. In evaluations on the large-displacement National Lung Screening Test (NLST) dataset, RWCNet exhibited superior performance (total registration error (TRE) of 2.11 mm) compared to other deep learning alternatives, and achieved results on par with variational optimization techniques. In contrast, on the OASIS dataset, which is characterized by smaller displacements, RWCNet achieved an average Dice similarity of 81.7%, representing only a modest improvement over other multi-scale deep learning models. Ablation experiments showed that multi-scale features consistently improved performance, whereas the correlation volume, number of recurrent steps, and inference-time instance optimization had large impacts on performance within the large-displacement NLST dataset. The performance of RWCNet compared to approaches that use instance optimization show that deep learning-based methods can find local minima that escape instance optimization methods. The results highlight the need for algorithm hyperparameter selection that adjusts with the dataset characteristics. RWCNet’s promising results may improve registration accuracy and computation efficiency, enabling many potential applications such as treatment planning, intra-procedural guidance, and longitudinal monitoring.
Sivan et al. (Fri,) studied this question.