No abstract

This paper presents a novel method for QRS detection in electrocardiograms (ECG). It is based on the S-Transform, a new time frequency representation (TFR). The S-Transform provides frequency-dependent resolution while maintaining a direct relationship with the Fourier spectrum. We exploit the advantages of the S-Transform to isolate the QRS complexes in the time-frequency domain. Shannon energy of each obtained local spectrum is then computed in order to localize the R waves in the time domain. Significant performance enhancement is confirmed when the proposed approach is tested with the MIT-BIH arrhythmia database (MITDB). The obtained results show a sensitivity of 99.84%, a positive predictivity of 99.91% and an error rate of 0.25%. Furthermore, to be more convincing, the authors illustrated the detection parameters in the case of certain ECG segments with complicated patterns.

In this paper, we introduce a new system for ECG beat classification using Support Vector Machines (SVMs) classifier with rejection. After ECG preprocessing, the QRS complexes are detected and segmented. A set of features including frequency information, RR intervals, QRS morphology and AC power of QRS detail coefficients is exploited to characterize each beat. An SVM follows to classify the feature vectors. Our decision rule uses dynamic reject thresholds following the cost of misclassifying a sample and the cost of rejecting a sample. Significant performance enhancement is observed when the proposed approach is tested with the MIT-BIH arrhythmia database. The achieved results are represented by the average accuracy of 97.2% with no rejection and 98.8% for the minimal classification cost.

In this paper, the use of a Compact Support Kernel (CSK) instead of the Gaussian window in the Stransform is proposed. The CSK is derived from the Gaussian but overcomes its practical drawbacks while preserving a large number of its useful properties. The width of the CSK is controlled by some parameters making it more flexible. These parameters are selected to optimize the energy concentration in the timefrequency domain. Compared to other versions of S-transform, other time-frequency representations and continuous wavelet transform, the achieved results obtained using synthetic and real data show a significant improvement in the time and frequency resolution, energy concentration and instantaneous frequency estimation.

In this paper, we introduce a new system for ECG beat classification using support vector machines classifier with a double hinge loss. The proposed classifier rejects samples that cannot be classified with enough confidence. Specifically in medical diagnoses, the consequence of a wrong classification can be so harmful that it is convenient to reject such sample. After ECG preprocessing, feature selection and extraction, our decision rule uses dynamic reject thresholds according to the cost of rejecting or misclassifying a sample. Significant performance enhancement is observed when the proposed approach is tested with the MIT-BIH arrythmia database. The achieved results are represented by the error reject tradeoff. We obtained 98.2% of sensitivity with no rejection and more than 99% of sensitivity for the optimal classification cost being competitive to other published studies.