Spectral analysis of single-channel airflow and oxygen saturation recordings in obstructive sleep apnea detection

This study investigated the usefulness of the very low spectral content of single-channel airflow recordings to help in the diagnosis of the obstructive sleep apnea (OSA) syndrome. Additionally, we evaluated whether airflow frequency components in the 0.01 - 0.10 Hz band are linked with desaturations in blood oxygen saturation (SaO(2)) recordings due to apnea events. The relationship between changes in airflow and SaO(2) was analyzed by means of the magnitude squared coherence (MSC) function. Power spectral density (PSD) was used to obtain the power spectrum of single airflow and SaO(2) signals. Peak amplitude (PA) and relative power (P(R)) were used to parameterize the power spectrum in the very low frequency band. 148 subjects suspected of suffering from OSA were studied. Significant differences (p-value ≪ 0.01) between OSA positive and OSA negative subjects were obtained from PSD and MSC features. We found a power increase in the very low frequency band of single-channel airflow linked with the periodic desaturations of OSA. Diagnostic sensitivity, specificity and accuracy of 84.0%, 85.4% and 84.5%, respectively, were reached with the peak amplitude of the airflow PSD. Thus, spectral features from the very low frequency components of single-channel airflow recordings could provide useful information to help in OSA diagnosis.