Morphology in Oral Rare Syndromes & Artificial Intelligence for Clinical Diagnosis

Overview

MOSAIC aims to determine whether oro-dental morphological anomalies, particularly palatal morphology, associated with rare bone and cartilage diseases can be precisely characterized using 3D digital models analysed through geometric morphometrics. The study will also evaluate whether these morphological signatures can train an artificial intelligence (AI) algorithm to classify syndromes. A prospective monocentric case-control cohort will be constituted, including 3D intra-oral scans and associated clinical data. The final goal is to improve diagnostic accuracy and reduce diagnostic delay in rare bone disorders.

Rare bone and cartilage diseases are genetically heterogeneous conditions in which oro-dental anomalies are frequent yet insufficiently characterized, partly due to subjective clinical assessment and the absence of quantitative tools. Palatal morphology and tooth number/shape anomalies may represent key phenotypic markers but remain underused in diagnosis. Advances in 3D intra-oral scanning and geometric morphometrics now allow precise, reproducible shape analysis of complex anatomical structures. In parallel, artificial intelligence has shown promising results in classifying craniofacial phenotypes from 2D images. However, no study has yet combined 3D digital oral data, geometric morphometrics, and machine learning for rare bone disorders. MOSAIC addresses this gap by building the first structured 3D database dedicated to these conditions and developing a classification model capable of identifying syndrome-specific morphological patterns. Participants will undergo a single visit including an intra-oral 3D optical impression and collection of clinical/genetic data. Geometric morphometric analysis (Generalized Procrustes Analysis, Principal Component Analysis, ProcMANOVA/MANCOVA, Pairwise comparison) will be performed on palatal landmarks configuration. Morphometric outputs will feed supervised machine-learning models (Random Forest, SVM, XGBoost) trained and validated for syndrome classification. Each participant will take part in one single visit (T0) without longitudinal follow-up. Data will then be pseudonymized, processed, and analysed in successive workpackages: (1) database constitution, (2) geometric morphometric analysis, (3) AI model training and validation, (4) internal independent testing. Further external validation is expected through a dedicated follow-up protocol using an independent external dataset. No clinical intervention or therapeutic modification is involved.