Oxygen saturation in children with and without obstructive sleep apnea using the phone-oximeter

Obstructive sleep apnea (OSA) in children can lead to daytime sleepiness, growth failure and developmental delay. Polysomnography (PSG), the gold standard to diagnose OSA is highly resource intensive and is confined to the sleep laboratory. In this study we propose to identify children with OSA using blood oxygen saturation (SpO₂) obtained from the Phone Oximeter. This portable, in-home device is able to monitor patients over multiple nights, causes less sleep disturbance and facilitates a more natural sleep pattern. The proposed algorithm analyzes the SpO₂ signal in the time and frequency domain using a 90-s sliding window. Three spectral parameters are calculated from the power spectral density (PSD) to evaluate the modulation in the SpO₂ due to the oxyhemoblobin desaturations. The power P, slope S in the discriminant band (DB), and ratio R between P and total power are calculated for each window. Tendency and variability indices, number of SpO₂ desaturations and time spent under 2% or 3% of baseline saturation level are computed for each time window. The statistical distribution of the temporal evolution of all parameters is analyzed to identify 68 children, 30 with OSA and 38 without OSA (nonOSA). This characterization was evaluated by a feature selection based on a linear discriminant. The combination of temporal and spectral parameters provided the best leave one out crossvalidation results with an accuracy of 86.8%, a sensitivity of 80.0%, and a specificity of 92.1% using only 5 parameters. The median of R, mean of P and S and mean and standard deviation of the number of desaturations below 3% of baseline saturation level, were the most representative parameters. Hence, a better knowledge of SpO₂ dynamics could help identifying children with OSA with the Phone Oximeter.