Training XPos3R for cerebral-vessel registration

Develop an XPos3R model for intraoperative 2D/3D registration of cerebral vessels, addressing the current inability to train such a model because of limited public vessel datasets.

Background

XPos3R is demonstrated using the pelvis as a representative surgical region, but intraoperative registration also targets other anatomies, including cerebral vessels. The paper states that publicly available vessel datasets are insufficient for training an XPos3R model in this setting, leaving the extension to cerebral-vessel registration unresolved.

References

Other anatomies, such as cerebral vessels, are also common targets for intraoperative registration. However, due to the limited availability of public vessel datasets, we are currently unable to train a model in this setting.

— XPos3R: Cross-Modal Transformer for Intraoperative 2D/3D Registration  (2609.10733 - Su et al., 9 Sep 2026) in Section Discussion, subsection Limitations