Reliability of Registration Methods for Local TN Visualization

Determine the reliability of ANTs, FireANTs, ConvexAdam, EasyReg, SynthMorph, and related registration methods for local trigeminal-neurovascular visualization under MRI–MRA field-of-view mismatch, variable MRA conspicuity, and partial vessel annotations.

Background

The paper explains that existing registration methods are commonly evaluated on whole-brain alignment tasks with adequate anatomical overlap. Trigeminal neuralgia MRI–MRA fusion instead requires reliable visualization within a small trigeminal region of interest, where MRI–MRA field-of-view mismatch, variable local TOF-MRA contrast, and incomplete clinically focused vessel annotations can substantially affect evaluation.

The reliability of these methods under the combined conditions relevant to TN-specific local visualization is not resolved in the paper. The benchmark evaluates representative pipelines and identifies important confounds, but the quoted passage frames their general reliability in these challenging settings as an unresolved issue requiring further determination.

References

These methods are typically evaluated on whole-brain alignment tasks with adequate anatomical overlap, and their reliability for local TN visualization under FOV mismatch, variable MRA conspicuity, and partial vessel annotations remains unclear.

— Evaluation Principles for MRI-MRA Registration in Trigeminal Neuralgia: An ROI-Centered Neurovascular Benchmark  (2609.08805 - Zhang et al., 8 Sep 2026) in Section 1, Related Work, subsection “Registration methodology”