SOFTooth: Semantics-Enhanced Order-Aware Fusion for Tooth Instance Segmentation

2025-12-29

AI Summary

3D tooth instance segmentation faces clinical challenges including dental arch crowding, ambiguous gingival boundaries, edentulism, and difficulty identifying third molars. Existing 3D methods rely heavily on geometric features, leading to boundary leakage, centroid drift, and identity inconsistency; direct adaptation of 2D foundation models (e.g., SAM) to 3D clinical workflows is infeasible. This paper proposes the first framework integrating frozen SAM’s semantic priors with 3D point cloud geometry. It introduces a point-wise residual gating mechanism to inject fine-grained boundary semantics, a center-guided mask optimization module to enhance localization accuracy, and an anatomy-order-aware Hungarian matching strategy to ensure label consistency under edentulism or dense dentition. Evaluated on 3DTeethSeg’22, our method achieves state-of-the-art performance, with significant mIoU improvement and a 12.6% gain in third molar segmentation accuracy.

Technology Category

• Computer Vision: 3D Computer Vision
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories
• Machine Learning: Deep Neural Architectures and Foundation Models

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Security and Privacy: Data transparency and provenance
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Three-dimensional (3D) tooth instance segmentation remains challenging due to crowded arches, ambiguous tooth-gingiva boundaries, missing teeth, and rare yet clinically important third molars. Native 3D methods relying on geometric cues often suffer from boundary leakage, center drift, and inconsistent tooth identities, especially for minority classes and complex anatomies. Meanwhile, 2D foundation models such as the Segment Anything Model (SAM) provide strong boundary-aware semantics, but directly applying them in 3D is impractical in clinical workflows. To address these issues, we propose SOFTooth, a semantics-enhanced, order-aware 2D-3D fusion framework that leverages frozen 2D semantics without explicit 2D mask supervision. First, a point-wise residual gating module injects occlusal-view SAM embeddings into 3D point features to refine tooth-gingiva and inter-tooth boundaries. Second, a center-guided mask refinement regularizes consistency between instance masks and geometric centroids, reducing center drift. Furthermore, an order-aware Hungarian matching strategy integrates anatomical tooth order and center distance into similarity-based assignment, ensuring coherent labeling even under missing or crowded dentitions. On 3DTeethSeg'22, SOFTooth achieves state-of-the-art overall accuracy and mean IoU, with clear gains on cases involving third molars, demonstrating that rich 2D semantics can be effectively transferred to 3D tooth instance segmentation without 2D fine-tuning.

Problem

• Addresses 3D tooth segmentation challenges like crowded arches and ambiguous boundaries
• Integrates 2D semantics into 3D without requiring 2D mask supervision
• Ensures consistent tooth labeling in cases of missing or crowded teeth

Innovation

• Fuses 2D SAM embeddings into 3D point features
• Refines masks using geometric centroids to reduce drift
• Integrates anatomical tooth order into instance matching