Diagnosis of pediatric obstructive sleep apnea: Preliminary findingsusing automatic analysis of airflow and oximetry recordings obtainedat patients’ home

Abstract

The obstructive sleep apnea syndrome (OSAS) greatly affects both the health and the quality of life of chil-dren. Therefore, an early diagnosis is crucial to avoid their severe consequences. However, the standarddiagnostic test (polysomnography, PSG) is time-demanding, complex, and costly. We aim at assessinga new methodology for the pediatric OSAS diagnosis to reduce these drawbacks. Airflow (AF) and oxy-gen saturation (SpO2) at-home recordings from 50 children were automatically processed. Informationfrom the spectrum of AF was evaluated, as well as combined with 3% oxygen desaturation index (ODI3)through a logistic regression model. A bootstrap methodology was conducted to validate the results.OSAS significantly increased the spectral content of AF at two abnormal frequency bands below (BW1)and above (BW2) the normal respiratory range. These novel bands are consistent with the occurrenceof apneic events and the posterior respiratory overexertion, respectively. The spectral information fromBW1 and BW2 showed complementarity both between them and with ODI3. A logistic regression modelbuilt with 3 AF spectral features (2 from BW1 and 1 from BW2) and ODI3 achieved (mean and 95% confi-dence interval): 85.9% sensitivity [64.5–98.7]; 87.4% specificity [70.2–98.6]; 86.3% accuracy [74.9–95.4];0.947 area under the receiver-operating characteristics curve [0.826–1]; 88.4% positive predictive value[72.3–98.5]; and 85.8% negative predictive value [65.8–98.5]. The combination of the spectral informationfrom two novel AF bands with the ODI3 from SpO2is useful for the diagnosis of OSAS in children.