Applying neural network classifiers in the diagnosis of the obstructive sleep apnea syndrome from nocturnal pulse oximetric recordings

Artificial neural network analysis of the oxygen saturation signal enables accurate diagnostics of sleep apnea

The severity of obstructive sleep apnea (OSA) is classified using apnea-hypopnea index (AHI). Accurate determination of AHI currently requires manual analysis and complicated registration setup making it expensive and labor intensive. Partially for these reasons, OSA is a heavily underdiagnosed disease as only 7% of women and 18% of men suffering from OSA have diagnosis. To resolve these issues, we introduce an artificial neural network (ANN) that estimates AHI and oxygen desaturation index (ODI) using only the blood oxygen saturation signal (SpO2), recorded during ambulatory polygraphy, as an input. Therefore, hypopneas associated only with an arousal were not considered in this study. SpO2 signals from 1692 patients were used for training and 99 for validation. Two test sets were used consisting of 198 and 1959 patients. In the primary test set, the median absolute errors of ANN estimated AHI and ODI were 0.78 events/hour and 0.68 events/hour respectively. Based on the ANN estimated AHI and ODI, 90.9% and 94.4% of the test patients were classified into the correct OSA severity category. In conclusion, AHI and ODI can be reliably determined using neural network analysis of SpO2 signal. The developed method may enable a more affordable screening of OSA.

Single layer network classifiers to assist in the detection of obstructive sleep apnea syndrome from oximetry data

The aim of this study is to assess the utility of single layer network classifiers to help in the diagnosis of the obstructive sleep apnea syndrome (SAOS). Oxygen saturation (SaO(2)) recordings from a total of 157 subjects suspected of suffering from OSAS were used. These were divided into a training set and a test set with 74 and 83 subjects, respectively. Four classification schemes were developed by using generalized linear models (GLM). Two GLM classifiers were built with spectral (GLM-SP) and non-linear (GLM-NL) features from SaO(2) signals, respectively. In addition, both algorithms were combined in order to improve their classification capability. The performance of two different ensemble classifiers was analyzed. The highest diagnostic accuracy was reached by the GLM-SP classifier (88%). The ensemble built from the combination of GLM-SP and GLM-NL by means of an additional GLM structure provided the best sensitivity value (87.8%). Applying spectral and non-linear features from SaO(2) data simultaneously could be useful in OSAS diagnosis.