APCReg: Anatomical-Prior-Guided Coarse-to-Fine CBCT--IOS Registration via Multi-View Projection and Reliability-Controlled Residual Correction

2026-08-07

AI Summary

This study addresses the challenge of automatic registration between cone-beam computed tomography (CBCT) and intraoral scans (IOS), which is hindered by modality discrepancies, limited overlap, and large initial pose offsets. To overcome these issues, the authors propose an anatomy-prior-guided coarse-to-fine registration framework. The method innovatively decomposes the six-degree-of-freedom search space into sequential orthogonal projections to achieve efficient global alignment. It further introduces an overlap-aware residual registration module and a ground-truth-free coarse registration preservation mechanism, integrating KPConv features, dental arch length priors, overlap-gated cross-attention, Sinkhorn matching, and multi-hypothesis pose selection to significantly enhance robustness and accuracy. Evaluated on 60 mandibular datasets, the approach achieves a mean Chamfer distance of 0.87 mm and Hausdorff distance of 2.92 mm, outperforming existing open-source baselines across all six evaluation metrics.

Abstract

Registration between cone-beam computed tomography (CBCT) and intraoral scans (IOS) is essential for patient-specific surgical planning. However, disparate imaging modalities, limited overlap, and large pose offsets make automated registration unreliable. Consequently, clinical registration remains dependent on conventional geometry pipelines and manual clinician adjustment. To address these challenges, we propose APCReg, an anatomical-prior-guided coarse-to-fine framework for global registration and reliability-controlled residual correction. Specifically, multi-view anatomical coarse registration (MACR) performs ordered orthogonal projection alignment (buccal, proximal, and occlusal) to decompose the six-degree-of-freedom search before three-dimensional refinement. Overlap-aware residual registration (OARR) combines shared KPConv features, a folded arch-length cue, overlap-gated cross-attention, and Sinkhorn matching. Finally, dental-arch-structured hypothesis selection evaluates diverse poses on held-out reliable correspondences, while a ground-truth-free coarse-retention guard conditionally retains a geometrically reliable coarse pose. On 60 held-out jaw pairs, APCReg achieves a submillimeter mean Chamfer distance of 0.87 mm and a Hausdorff distance of 2.92 mm under this evaluation protocol, and ranks first across the six reported metrics among the evaluated open-source baselines.

Problem

• CBCT- IOS registration
• multi-modality registration
• limited overlap
• large pose offset
• automated registration

Innovation

• coarse-to-fine registration
• anatomical prior
• multi-view projection
• overlap-aware residual correction
• reliability-controlled matching