Artificial intelligence in joint face-skeletal prediction for orthognathic procedures: A systematic review.

OBJECTIVE

This systematic review aimed to produce an overview of the development of AI in joint skeletal-soft tissue prediction after orthognathic surgery, and to assess its accuracy comprehensively.

MATERIALS AND

METHODS

Following PRISMA guidelines, a systematic search was conducted in PubMed, Web of Science, IEEE Xplore, EMBASE, and Scopus (as of July 20, 2025). The initial search yielded 885 records, with eight studies meeting the inclusion criteria. The PICOS principle was applied to focus on AI models' joint predictive ability and their clinical validation. QUADAS-2 and JBI tools were used to assess bias, and GRADE was applied to evaluate outcome certainty.

RESULTS

The included AI models significantly enhanced prediction accuracy through innovations like CPSA, P2P-Conv, and DGCFP. Surface bias errors in critical areas were reduced, and clinical acceptance rates increased. However, challenges remain, including insufficient modeling of anatomical details, high sample homogeneity, and lack of standardized assessments.

CONCLUSION

AI-driven joint predictive modeling shows great potential for improving surgical accuracy and efficiency. Future work should focus on creating multicenter, cross-ethnic datasets, exploring hybrid architectures, and conducting prospective clinical validation to enhance generalizability. Establishing uniform evaluation standards, improving model interpretability, and enhancing ethical and privacy frameworks are also crucial for clinical translation.

CLINICAL RELEVANCE: Based on model evaluations, AI-based orthognathic prediction shows region-dependent accuracy: highest for the chin (genioplasty), followed by the mandibular body, and partial for the maxillary nasal base. Current models lack precision for lip aesthetics and 3D rotational displacement, but offer high value for rapid chin-mandible continuity predictions.