# Applications of Artificial Intelligence in Oral and Maxillofacial Cosmetic Surgery: A Systematic Review of Diagnostic, Planning, and Outcome Assessment Tools

## TL;DR

This paper reviews how artificial intelligence is being used in facial and mouth cosmetic surgery to improve diagnosis, planning, and outcome predictions, but notes that more rigorous research is needed.

## Contribution

A systematic review and critical appraisal of AI applications in oral and maxillofacial cosmetic surgery, highlighting current capabilities and limitations.

## Key findings

- AI models achieved 78.9-90.2% accuracy in predicting third molar extraction difficulty.
- Neural networks predicted postoperative swelling with 98% accuracy in third molar surgery.
- AI predicted soft-tissue changes with sub-millimeter accuracy in surgical planning.

## Abstract

Artificial intelligence (AI) has rapidly expanded into oral and maxillofacial cosmetic surgery, offering potential improvements in diagnosis, surgical planning, perioperative risk assessment, and outcome prediction. Despite promising results, the extent of its clinical utility and methodological rigor across published studies remains unclear. This systematic review synthesized and critically appraised the applications of AI in oral and maxillofacial cosmetic surgery, focusing on diagnostic support, preoperative planning, and esthetic or functional outcome evaluation. A comprehensive search of PubMed, Cochrane, Scopus, and Web of Science from inception to November 2024 identified 11,031 records, of which 14 studies were included after screening. Data extraction captured study characteristics, AI techniques, input modalities, comparators, and outcomes, and the study quality was appraised using the PROBAST-AI tool. The included studies encompassed diverse AI applications. In third molar surgery, convolutional neural networks achieved 78.9-90.2% accuracy for extraction difficulty, while neural networks predicted postoperative swelling with 98% accuracy. In orthognathic diagnostics, models using cephalograms and facial photographs reported accuracies above 90%, with specificity up to 99%. For surgical planning, AI predicted soft-tissue changes with sub-millimeter error margins, outperforming conventional models. Perioperative risk models predicted blood loss with mean errors <10 mL, while aesthetic applications quantified age reduction post-rhinoplasty and generated simulations with high surgeon agreement. Despite these advances, most studies were retrospective, single-center, and limited by small or homogeneous datasets, with an overall high risk of bias. Overall, AI demonstrates strong potential for enhancing diagnostic accuracy, surgical planning, risk prediction, and esthetic evaluation in oral and maxillofacial cosmetic surgery. However, current evidence is constrained by methodological weaknesses, limited validation, and ethical concerns, including dataset bias and subjective outcome measures. Future research should prioritize prospective multicenter validation, standardized reporting frameworks, multimodal data integration, and transparent model design to enable safe and effective clinical translation.