HHT based cardiopulmonary coupling analysis for sleep apnea detection.

STUDY OBJECTIVES To validate the feasibility of the Hilbert-Huang transform (HHT) based cardiopulmonary coupling (CPC) technique in respiratory events detection and estimation of the severity of apnea/hypopnea. METHODS The HHT-CPC sleep spectrogram technique was applied to a total of 69 single-lead ECG signals downloaded from the Physionet Sleep Apnea Database. Sleep spectrograms generated by both the original and the improved CPC method were compared on the structure distribution and time-frequency resolution. The performance of respiratory events detection by using the power of low frequency coupling (pLFC) in the new method was estimated by receiver operating characteristic analysis. Furthermore, correlation between HHT-CPC index (temporal Variability of Dominant Frequency, TVDF) and conventional OSAHS scoring was computed. RESULTS The HHT-CPC spectrum provides much finer temporal resolution and frequency resolution (8 s and 0.001 Hz) compared with the original CPC (8.5 min and 0.004 Hz). The area under the ROC curve of pLFC was 0.79 in distinguishing respiratory events from normal breathing. Significant differences were found in TVDF among groups with different severities of OSAHS (normal, mild, moderate, and severe, p<0.001). TVDF has a strong negative correlation with the apnea/hypopnea index (AHI, correlation coefficient -0.71). CONCLUSIONS The HHT-CPC spectrum could exhibit more detailed temporal-frequency information about cardiopulmonary coupling during sleep. As two spectrographic markers, pLFC and TVDF can be used to identify respiratory events and represent the disruption extent of sleep architecture in patients with sleep apnea/hypopnea, respectively. The proposed technique might serve as a complementary approach to enhance diagnostic efforts.