Efficient Point-Based Neural Network For Finish Line Extraction

Accurate extraction of finish lines from prepared teeth is critical for optimizing CAD/CAM workflows and is essential for enabling controlled automatic crown design. This study proposes a lightweight point-based neural network for interactive finish line extraction from 3D dental meshes, balancing efficiency, accuracy, and clinical usability. A dataset of 600 full-crown preparations was used to train the model, the training process incorporates various data augmentation techniques to enhance model generalizability. The proposed network comprises only 1.7 million parameters, making it suitable for client-side clinical deployment. The segmentation process is semi-automatic, allowing users to click on any point of a prepared tooth to initiate finish line extraction. The model was evaluated on full-resolution dental meshes and tested on low-end GPUs to assess its practical feasibility. The proposed model achieved performance comparable to the state-of-the-art deep learning approaches while achieving significant computational efficiency. It successfully processed full-resolution meshes with reasonable computational resources. The semi-automatic interaction improved adaptability, making the system robust across diverse clinical cases. The proposed neural network provides an efficient and accurate solution for finish line extraction from 3D dental meshes. Its compact design, fast processing capability, and user-guided segmentation approach make it a promising tool for enhancing clinical workflows in digital dentistry.