[2603.00124] OrthoAI: A Lightweight Deep Learning Framework for Automated Biomechanical Analysis in Clear Aligner Orthodontics -- A Methodological Proof-of-Concept

Computer Science > Computer Vision and Pattern Recognition arXiv:2603.00124 (cs) [Submitted on 23 Feb 2026] Title:OrthoAI: A Lightweight Deep Learning Framework for Automated Biomechanical Analysis in Clear Aligner Orthodontics -- A Methodological Proof-of-Concept Authors:Edouard Lansiaux, Margaux Leman, Mehdi Ammi View a PDF of the paper titled OrthoAI: A Lightweight Deep Learning Framework for Automated Biomechanical Analysis in Clear Aligner Orthodontics -- A Methodological Proof-of-Concept, by Edouard Lansiaux and Margaux Leman and Mehdi Ammi View PDF HTML (experimental) Abstract:Clear aligner therapy now dominates orthodontics, yet clinician review of digitally planned tooth movements-typically via ClinCheck (Align Technology)-remains slow and error-prone. We present OrthoAI, an open-source proof-of-concept decision-support system combining lightweight 3D dental segmentation with automated biomechanical analysis to assist treatment-plan evaluation. The framework uses a Dynamic Graph CNN trained on landmark-reconstructed point clouds from 3DTeethLand (MICCAI) and integrates a rule-based biomechanical engine grounded in orthodontic evidence (Kravitz et al 2009; Simon et al 2014). The system decomposes per-tooth motion across six degrees of freedom, computes movement-specific predictability, issues alerts when biomechanical limits are exceeded, and derives an exploratory composite index. With 60,705 trainable parameters, segmentation reaches a Tooth Identification Rate o...