文摘
Automatic sleep apnea detection using single lead ECG is a precondition for the implementation of a sleep apnea monitoring device. Computerized sleep apnea screening is also essential for expediting sleep apnea research and alleviating the onus of physicians of analyzing a large volume of data by visual inspection. However, most of the state-of-the-art works on automated sleep apnea identification are either based on multiple leads and multiple physiological signals or yield poor performance. In this article, normal inverse Gaussian (NIG) pdf modeling in the recently proposed tunable-Q factor wavelet transform (TQWT) domain is introduced for computer-assisted sleep apnea diagnosis from single-lead ECG signals. First, ECG signal segments are decomposed into sub-bands using TQWT. Afterwards, the corresponding NIG parameters are computed from each of the sub-bands. These parameters are used as features in the proposed apnea detection algorithm. Adaptive boosting (AdaBoost), an eminent ensemble learning based classification scheme is employed to perform classification. The suitability of TQWT is analyzed. The effectiveness of the selected features is validated by intuitive, statistical, and graphical analyses. The performance of the proposed feature extraction scheme is evaluated for various choices of classifiers. Optimal choices of TQWT and AdaBoost parameters are also determined. The performance of the proposed method, as compared to the state-of-the-art algorithms, is comparable or superior in terms of various performance metrics.
