Single Lead ECG Discrimination Between Normal Sinus Rhythm and Sleep Apnea with Intrinsic Mode Function Complexity Index Using Empirical Mode Decomposition

Diagnosis and treatment of sleep apnea in its various forms such as obstructive, central and complex syndrome is extremely important to prevent various diseases such as hypertension, diabetes, coronary artery disease, metabolic syndrome, and cerebrovascular diseases. Current methods to detect sleep apnea interfere with sleep and also require long hours of data recording and therefore, single lead ECG based sleep apnea detection is gaining popularity due to its simplicity and practicality for real-time sleep apnea monitoring. The purpose of this research was to test the feasibility of discriminating single lead ECG's with normal sinus rhythm (NSR) and sleep apnea with intrinsic mode function (IMF) complexity index using empirical mode decomposition for real-time detection of sleep apnea. Ten sets of ECG's with NSR and ECG's with sleep apnea were obtained from Physionet database. Custom MATLAB® software was written to compute IMF complexity index for each of the data set and compared for statistical significance test $(\mathbf{p} < 0.01)$. The mean IMF complexity for NSR across 10 data sets was $0.41\pm 0.06$ and the mean MSF for ECG with sleep apnea was $0.32 \pm 0.05$ showing robust discrimination with statistical significance $(\mathbf{p} < 0.01)$. IMF complexity robustly discriminates single lead ECG with normal sinus rhythm and sleep apnea. Further validation of this result is required on a larger dataset.