ARTIFICIAL INTELLIGENCE IN ORTHOGNATHIC SURGERY: A REVIEW OF DIAGNOSTIC SUPPORT, VIRTUAL PLANNING, AND SOFT TISSUE OUTCOME PREDICTION

Objectives. Orthognathic treatment of dentofacial deformities is highly experience-dependent, and the prediction of post-operative facial soft tissue change has historically been the least reliable step of surgical planning. This review synthesises the current applications, reported performance, and limitations of artificial intelligence (AI) across the orthognathic care pathway, with particular emphasis on diagnostic support, virtual surgical planning, and soft tissue outcome prediction. Methods. A narrative review of peer-reviewed studies applying machine learning (ML), deep learning (DL), and related methods to diagnosis, patient qualification, cephalometric analysis, treatment planning, soft tissue prediction, and outcome evaluation in orthognathic surgery was conducted. Findings were organised thematically and appraised against commonly reported metrics. Results. AI is most mature for automated cephalometric landmarking and for screening the need for surgery, where convolutional neural networks and ensemble classifiers approach expert-level agreement. For soft tissue prediction, DL methods that learn nonlinear bone-to-soft-tissue mappings achieve clinically relevant accuracy that exceeds traditional proportional and software-based simulations, with region-specific surface errors typically in the sub-millimetre to low-millimetre range. Applications in planning, outcome verification, and patient communication are emerging but less validated. Conclusions. AI provides meaningful decision support across the orthognathic pathway and improves soft tissue simulation, yet evidence remains limited by small single-centre datasets, weak external validation, and restricted interpretability. Multicentre data, prospective validation, and explainable, clinician-supervised models are required before routine clinical adoption.