Artificial intelligence prediction model for the relationship between obstructive sleep Apnea severity and maxillofacial developmental disorders in children: a prospective cohort study

This study aimed to develop an artificial intelligence-based prediction model for evaluating the relationship between obstructive sleep apnea (OSA) severity and maxillofacial developmental disorders in children. A prospective cohort design was employed, monitoring 50 children (mean age 8.4±2.3 years, 58% male) with varying degrees of maxillofacial abnormalities over a 12-month period. Participants were stratified into four groups: maxillary constriction (n=15), mandibular retrognathia (n=15), mixed phenotype (n=10), and control (n=10). Comprehensive assessments included cephalometric measurements, intraoral scans, and polysomnography performed at baseline, 6-month, and 12-month intervals. A hybrid artificial intelligence architecture integrating gradient boosting algorithms and deep neural networks was developed using multimodal data. Results demonstrated significant correlations between specific maxillofacial parameters and OSA severity, with SNB angle (r=-0.68, p<0.001) and maxillary width (r=-0.61, p<0.001) showing the strongest associations. Multiple regression analysis identified SNB angle (β=-0.46, p<0.001), maxillary width (β=-0.39, p<0.001), and BMI (β=0.28, p=0.012) as significant independent predictors of AHI, collectively explaining 72% of OSA severity variance. The AI model achieved an overall accuracy of 89.6% in classifying OSA severity, with differential performance across phenotype groups (mandibular retrognathia: 93.1%, maxillary constriction: 88.5%, mixed phenotype: 85.2%). Longitudinal follow-up revealed significant correlations between improvements in maxillofacial parameters and reductions in AHI, with stronger associations in younger children (5-8 years) compared to older children (9-12 years). This research provides an effective tool for assessing the relationship between OSA severity and maxillofacial developmental abnormalities in children, offering valuable insights for early risk stratification and personalized treatment strategies in pediatric sleep medicine.