Topo-Morphological Edge Logits Field for 3D Dental Proposal Generation and Instance Segmentation

Area of Science:

• Computer-aided dentistry
• Medical image analysis
• Computational geometry

Background:

• Automatic tooth segmentation is crucial for dental diagnosis and treatment planning.
• Existing methods struggle with accuracy due to offset bias, spatial information loss, and centroid shift.
• These limitations lead to missed detections or merged instances, especially with crowded or missing teeth.

Purpose of the Study:

• To develop a novel proposal generation framework for accurate automatic tooth segmentation.
• To overcome the limitations of existing Euclidean space-based methods.
• To improve the detection and segmentation of individual tooth instances.

Main Methods:

• Introduced Topological Morphological Clustering (TMC) integrating edge logits fields with topological morphology.
• Utilized morphological erosion on tooth and edge masks for instance separation.
• Developed an edge enhancement module fusing alpha-shape and KNN-based iterative contour approximation (KNN-ICA) for geometric edge logits field construction.

Main Results:

• The proposed TMC framework achieved superior detection performance compared to existing methods.
• The method outperformed prior instance segmentation techniques on two public benchmarks.
• Demonstrated enhanced accuracy in delineating tooth boundaries, particularly in challenging cases.

Conclusions:

• Topological Morphological Clustering (TMC) offers a robust solution for automatic tooth segmentation.
• The novel framework effectively addresses limitations of previous methods, improving dental image analysis.
• The approach shows significant potential for advancing computer-aided diagnosis and treatment planning in dentistry.