DETECTION OF SLEEP APNEA IN ECG SIGNAL USING PAN-TOMPKINS ALGORITHM AND ANN CLASSIFIERS

In this paper, a novel methodology for Sleep apnea detection is proposed using ECG signal analysis. It involves the following sequential procedure: Pre-processing using digital filters, Peak or QRS complex detection using Pan-Tompkins algorithm, Feature extraction from detected QRS complex, Reduction of features using Principal Component Analysis (PCA) and finally the Classification using Artificial Neural Networks (ANNs). The result of classification of the input ECG signal record is as either belonging to apnea or normal category. For experimentation, the ECG-Apnea database from MIT's Physionet.org is used. The performance measures of Peak or QRS complex detection are Accuracy(Acc)=94%, Sensitivity(Se)=95%, Specificity(Sp)= 93% and Precision ( Pr) = 92%. The PCA is applied on the set of time and frequency features of ECG signal to achieve dimensionality reduction and thus reduce the computational time cost, both in training and testing phase of classification by 43% and 33% respectively. The performance of ANN clasifier trained using Scaled Conjugate Gradient (ANN_SCG) has marginally improved values of Acc, Se, Sp, Pr and F-measure , where as the execution time is significantly reduced by 66% as compared to that of ANN classifier trained with Levenberg-Marquardt algorithm (ANN_LM). The experimental results demonstrate the effectiveness of the proposed method in terms of significantly reduced time cost even as compared with two of the published results.