Development of Age-Specific Parametric Signal Models for Pediatric Sleep Apnea Integrating Respiratory Inductance Plethysmography with Deep Neural Networks

• Abiodun Okunola

  Ladoke Akintola University Technology

  Author

Pediatric sleep apnea remains significantly underdiagnosed due to the reliance on costly, labor-intensive polysomnography (PSG) and the absence of validated automated screening tools adapted to children's unique physiological characteristics. Current machine learning approaches for sleep apnea detection predominantly utilize adult-derived models that fail to account for age-dependent variations in respiratory mechanics and sleep architecture. This study addresses this critical gap by developing age-specific parametric signal models for pediatric sleep apnea detection, integrating respiratory inductance plethysmography (RIP) signals with deep neural network architectures. Using retrospective PSG data from 2,379 pediatric recordings sourced from the Nationwide Children's Hospital Sleep DataBank, we designed a hybrid CNN-LSTM framework incorporating an age-stratified parametric estimation module. The proposed model achieved a detection accuracy of 89.4% with a sensitivity of 91.2% and specificity of 87.6%, significantly outperforming conventional non-age-adjusted models (accuracy: 78.3%, p<0.001). Key findings demonstrate that age-specific parameterization of respiratory signal features critically enhances model performance, with the strongest improvements observed in children under 6 years and adolescents aged 12-17 years. This research provides a replicable, computationally efficient framework suitable for real-time screening applications, offering a practical pathway toward accessible, non-invasive pediatric sleep apnea detection outside traditional sleep laboratory settings.

• PDF

Copyright (c) 2026 Abiodun Okunola (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.