Artificial intelligence-driven soft tissue prediction in virtual surgical planning for orthognathic surgery: a scoping review

The objective of this study was to map the current evidence on artificial intelligence (AI)-based facial soft tissue prediction in virtual surgical planning (VSP) workflows for orthognathic surgery. A scoping review was conducted using PubMed/MEDLINE, Embase, Scopus, Web of Science and IEEE Xplore databases, in accordance with the Joanna Briggs Institute (JBI) methodology and reported following the PRISMA Extension for Scoping Reviews (PRISMA-ScR). Data on study characteristics, AI methodologies, imaging inputs, accuracy metrics, validation strategies, and levels of integration into VSP were extracted. Six studies met the eligibility criteria. All used deep learning-based models, predominantly using point cloud or surface mesh representations derived from computed tomography (CT), cone-beam computed tomography (CBCT), and/or three-dimensional (3D) surface scans. The reported geometric prediction accuracy was frequently within clinically accepted thresholds (≤ 2-3 mm), although reduced accuracy was consistently observed in dynamic regions. None of the studies adequately evaluated patient-reported outcomes, clinician usability, or full workflow integration with VSP. AI-based soft tissue prediction in VSP workflows demonstrates promising results but remains in an early translational phase. Further research is required to establish its generalisability, workflow integration, and clinical impact.