English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

We present KeyMorph, a deep learning-based image registration framework thatrelies on automatically detecting corresponding keypoints. State-of-the-artdeep learning methods for registration often are not robust to largemisalignments, are not interpretable, and do not incorporate the symmetries ofthe problem. In addition, most models produce only a single prediction attest-time. Our core insight which addresses these shortcomings is thatcorresponding keypoints between images can be used to obtain the optimaltransformation via a differentiable closed-form expression. We use thisobservation to drive the end-to-end learning of keypoints tailored for theregistration task, and without knowledge of ground-truth keypoints. Thisframework not only leads to substantially more robust registration but alsoyields better interpretability, since the keypoints reveal which parts of theimage are driving the final alignment. Moreover, KeyMorph can be designed to beequivariant under image translations and/or symmetric with respect to the inputimage ordering. Finally, we show how multiple deformation fields can becomputed efficiently and in closed-form at test time corresponding to differenttransformation variants. We demonstrate the proposed framework in solving 3Daffine and spline-based registration of multi-modal brain MRI scans. Inparticular, we show registration accuracy that surpasses currentstate-of-the-art methods, especially in the context of large displacements. Ourcode is available at https://github.com/alanqrwang/keymorph.

A Robust and Interpretable Deep Learning Framework for Multi-modal Registration via Keypoints