Related Experiment Video
Updated: Sep 5, 2025
A Model to Simulate Clinically Relevant Hypoxia in Humans
Published on: December 22, 2016
A 2D convolutional neural network to detect sleep apnea in children using airflow and oximetry
Jorge Jiménez-García1, María García1, Gonzalo C Gutiérrez-Tobal1
1Biomedical Engineering Group, University of Valladolid, Valladolid, Spain; CIBER-BBN, Centro de Investigación Biomédica en Red en Bioingeniería, Biomateriales y Nanomedicina, Instituto de Salud Carlos III, Valladolid, Spain.
Insights
A new deep learning model using airflow and oximetry signals can accurately diagnose childhood obstructive sleep apnea (OSA). This approach simplifies diagnosis, improving accessibility and reducing waitlists for pediatric sleep studies.
Area of Science:
- Pediatric Pulmonology
- Sleep Medicine
- Artificial Intelligence in Healthcare
- Medical Diagnostics
Background:
- Obstructive sleep apnea (OSA) diagnosis in children typically relies on overnight polysomnography (PSG), which is resource-intensive.
- Simplifying PSG is crucial to increase accessibility, comfort, and reduce healthcare burdens.
- Deep learning analysis of airflow (AF) and oximetry (SpO2) signals for pediatric OSA has been underexplored.
Purpose of the Study:
- To evaluate a convolutional neural network (CNN) for estimating pediatric OSA severity using AF and SpO2 signals.
- To assess the diagnostic performance of the CNN in two independent pediatric datasets.
Main Methods:
- A 2D CNN architecture was developed to analyze AF and SpO2 signals from PSG recordings.
- The CNN was trained and validated on data from the Childhood Adenotonsillectomy Trial (CHAT) and a clinical database.
- The model estimated the apnea-hypopnea index (AHI) and classified OSA severity into four levels.
Main Results:
- The CNN achieved 4-class OSA severity classification accuracies of 72.55% (CHAT) and 61.79% (clinical).
- Binary classification accuracy for OSA detection ranged from 84.64% to 94.44% (CHAT) and 84.10% to 90.26% (clinical).
- The deep learning approach outperformed previous methods using SpO2 alone or traditional feature engineering.
Conclusions:
- Deep learning analysis of AF and SpO2 signals demonstrates high diagnostic ability for pediatric OSA.
- This CNN-based approach shows promise for developing reliable computer-aided diagnostic tools for childhood OSA.
- The findings suggest a potential for more accessible and efficient pediatric OSA screening and diagnosis.
Abstract:
The gold standard approach to diagnose obstructive sleep apnea (OSA) in children is overnight in-lab polysomnography (PSG), which is labor-intensive for clinicians and onerous to healthcare systems and families. Simplification of PSG should enhance availability and comfort, and reduce complexity and waitlists. Airflow (AF) and oximetry (SpO2) signals summarize most of the information needed to detect apneas and hypopneas, but automatic analysis of these signals using deep-learning algorithms has not been extensively investigated in the pediatric context. The aim of this study was to evaluate a convolutional neural network (CNN) architecture based on these two signals to estimate the severity of pediatric OSA. PSG-derived AF and SpO2 signals from the Childhood Adenotonsillectomy Trial (CHAT) database (1638 recordings), as well as from a clinical database (974 recordings), were analyzed. A 2D CNN fed with AF and SpO2 signals was implemented to estimate the number of apneic events, and the total apnea-hypopnea index (AHI) was estimated. A training-validation-test strategy was used to train the CNN, adjust the hyperparameters, and assess the diagnostic ability of the algorithm, respectively. Classification into four OSA severity levels (no OSA, mild, moderate, or severe) reached 4-class accuracy and Cohen's Kappa of 72.55% and 0.6011 in the CHAT test set, and 61.79% and 0.4469 in the clinical dataset, respectively. Binary classification accuracy using AHI cutoffs 1, 5 and 10 events/h ranged between 84.64% and 94.44% in CHAT, and 84.10%-90.26% in the clinical database. The proposed CNN-based architecture achieved high diagnostic ability in two independent databases, outperforming previous approaches that employed SpO2 signals alone, or other classical feature-engineering approaches. Therefore, analysis of AF and SpO2 signals using deep learning can be useful to deploy reliable computer-aided diagnostic tools for childhood OSA.
Related Concept Videos
01:21Sleep Apnea
The condition is more prevalent among...
01:24Pulse Oximetry
Purpose
Average SpO2 values are greater than 95%. If the readings fall below 90%, it indicates that...
01:26Oxygen Delivering System I: Nasal Cannula and Face Mask
Nasal Cannula
A nasal cannula is a lightweight tube split at one end into two prongs and placed in the nostrils. It is typically used to deliver low to medium levels of oxygen.
Suggested flow rate: The suggested flow rate for a nasal cannula typically ranges between 1 and 6 L/min.
Oxygen percentage setting:...
01:18Neural Control of Respiration
Respiratory Centers in the Brainstem
Two primary areas comprise the respiratory center: the medullary respiratory center in the medulla oblongata and the pontine respiratory group in the pons. The...
01:16Oxygen Delivering System II: Venturi Mask and Transtracheal Oxygen
Venturi Mask
The Venturi mask, named after the Venturi effect, is designed to deliver precise oxygen concentrations. It consists of a large tube with an oxygen inlet that narrows down, causing a pressure drop that pulls air in through adjustable side ports. The mask is a lightweight,...
01:29Assessment of Ventilation II: Respiratory Depth and Rhythm
Respiratory depth measures the volume of air inhaled or exhaled during a breath. It can vary from shallow to deep and typically remains consistent when a person is at rest or asleep. Occasionally, individuals will automatically inhale deeply, known as sighing, which inflates the lungs with more air than normal breathing.
To assess respiratory depth, observe the degree of chest excursion or movement: