GeoReg: Direct biplanar DSA-to-CTA registration with geodesic consistency for acute ischemic stroke

Rudolf Leonardus Mirjam Van Herten, Robert Graf, Felix Bitzer, Jan Kirschke, Johannes C. Paetzold

Keywords: Pose estimation, image registration, digital subtraction angiography, computed tomography angiography

TL;DR: We perform direct, vessel-segmentation free registration of biplanar DSA to CTA

Abstract: The complementary nature of pre-procedural computed tomography angiography (CTA) and intraoperative digital subtraction angiography (DSA) has motivated significant interest in image registration to enhance therapeutic decision-making during stroke interventions. However, current methods typically depend on accurate vessel segmentation, creating a deployment bottleneck due to the requirement for extensive annotated training data. Here, we present an alternative approach that establishes the feasibility of registration without this dependency. Instead of extracting vascular features using pre-trained models, we optimize a direct registration framework using maximum intensity projections (MAP) of DSA sequences to align a silhouette of the subtracted X-ray image. We introduce a geodesic consistency formulation that jointly optimizes biplanar views, employing soft geometric constraints on SO(3) to maintain consistency while accommodating non-orthogonal scanner configurations. We demonstrate the effectiveness of this model on clinical stroke data and find that it outperforms baseline methods, proving particularly effective in escaping local minima where single-view optimization fails. These results indicate that reliable DSA-to-CTA registration is achievable without vessel-specific training data, simplifying the path toward clinical integration.

Primary Subject Area: Image Registration

Secondary Subject Area: Image Acquisition and Reconstruction

Registration Requirement: Yes

Reproducibility: https://github.com/RoelvH97/GeoReg

Visa & Travel: No

Read CFP & Author Instructions: Yes

Originality Policy: Yes

Single-blind & Not Under Review Elsewhere: Yes

LLM Policy: Yes

Midl Latex Submission Checklist: Ensure no LaTeX errors during compilation., Replace NNN with your OpenReview submission ID., Includes \documentclass{midl}, \jmlryear{2026}, \jmlrworkshop, \jmlrvolume, \editors, and correct \bibliography command., Did not override options of the hyperref package., Did not use the times package., Use the correct spelling and format, avoid Unicode characters, and use LaTeX equivalents instead., Any math in the title and abstract must be enclosed within $...$., Did not override the bibliography style defined in midl.cls and did not use \begin{thebibliography} directly to insert references., Avoid using \scalebox; use \resizebox when needed., Included all necessary figures and removed *unused* files in the zip archive., Removed special formatting, visual annotations, and highlights used during rebuttal., All special characters in the paper and .bib file use LaTeX commands (e.g., \'e for é)., No separate supplementary PDF uploads., Acknowledgements, references, and appendix must start after the main content.

Submission Number: 13