Craniofacial and airway changes after skeletal Class III orthognathic surgery: A machine learning study

Abstract

Mandibular setback surgery is a common orthognathic procedure for correcting Class III malocclusion. This study aimed to evaluate posterior airway space (PAS) changes in Class III patients following mandibular setback surgery, focusing on skeletal, dental, and soft tissue alterations and their association with airway changes using conventional statistics and machine learning. Fifty patients who underwent mandibular setback surgery were assessed using cephalometric radiographs taken preoperatively (T1) and postoperatively (T2). Measurements included PAS area and anteroposterior width, alongside angular and linear cephalometric parameters. Paired t -tests evaluated pre- and post-treatment differences. Multivariate regression, Pearson correlation and Random Forest regression were used to analyse associations between craniofacial changes and PAS alterations. Significant postoperative changes were observed in most craniofacial and airway variables, except for SNA, overbite, lower lip to E-plane, and hypopharynx area. Multivariate regression showed that age, sex and several other variables influenced PAS measurements, but their effects were not uniform across all airway levels. Pearson correlation showed the strongest associations between PAS changes and SNB, ANB, mandibular body length, overjet, SN-GoGn, and Z-angle. Random forest regression demonstrated low overall predictive performance; however, the model identified SNB, nasolabial angle, overjet as the most influential predictors of airway changes. Mandibular setback surgery significantly reduces PAS. Skeletal changes exert the greatest influence on airway, followed by soft tissue and dental variables. Horizontal craniofacial landmarks movements are more predictive of airway changes than vertical shifts. SNB emerged as the most influential variable in both correlation and machine learning analyses.

Access to Document

Other files and links

• Link to publication in Scopus
• Open Access Version Available

Fingerprint