The paper seems to contribute in a significant way in proposing an alternative RNN-based approach for deformable image registration. Although the experimental setting is not extremely strong, the proposed approach seems to give significant computational advantages. Rebuttal clarified most of the reviewers concerns.