Key result
The proposed combined ICA and LDA technique for ECG signal pre-processing and classification achieved a sensitivity of 99.92%, detection error rate of 0.122%, and output SNR of 37.77dB.
Why the study?
Automated ECG pattern recognition requires effective noise removal during pre-processing, prompting the combined use of ICA for pre-processing and LDA for classification.
A combined ICA and LDA technique for ECG signal processing demonstrates high sensitivity and signal-to-noise ratio for automated arrhythmia detection.
May enhance ECG pre-processing for automated analysis; leaves open clinical validation in prospective datasets.
Automated recognition of patterns in an ECG signal quintessentially requires removal of noises, such as Baseline wander (BLW), breathing activity, poor quality of electrodes and current flowing in the cables of the acquisition system, during its pre-processing for improving the signal quality to enable the identification of various physiological and pathological phenomena from it. In the literature, it has been well established that statistical domain technique viz. ICA (independent component analysis) surpasses the performance of even higher order filters in removing interferences by calculating independent components with much less computational/mathematical complexity and loss of information. Thus, it will result in higher signal-to-noise-ratios (SNRs) mitigating masking effects of various interferences. On the other hand, LDA (linear discriminant analysis) minimizes the variance and maximizes the distance between any two data-classes while detecting/classifying them resulting in very less false detections. Therefore in this paper, ICA and LDA are proposed to be used combinedly for pre-processing and classification of an ECG signal, respectively. Hence, important latent attributes of the ECG signal are retained with maximally statistically independent criteria using ICA and effective classification/detection is accomplished by cutting down the dimensional costs using LDA. The performance of ICA in ECG signal pre-processing is further compared with that obtained using ANF (adaptive notch filter) to further demonstrate its superiority. The proposed technique is able to achieve sensitivity (Se), detection error rate (DER) and output SNR of 99.92%, 0.122% and 37.77dB, respectively on Massachusetts Institute of Technology- Beth Israel Hospital Arrhythmia database (MB ARR DB).Abbreviations: SNR: Signal-to-Noise Ratio; CAMD: Computer-aided medical diagnosis; Se: Sensitivity; DR: Detection Rate; KNN: K-Nearest Neighbour; SVM: Support vector machines; MB ARR DB: MIT-BIH arrhythmia database; FoM: Figure-of-Merit; WT: Wavelet Transform; MIT-BIH: Massachusetts Institute of Technology- Beth Israel Hospital; PP: Positive Predictivity; SP: Specificity; STFT: Short Time Fourier Transform; ECG: Electrocardiogram; BLW: Baseline Wander; PLI: Power Line Interference; HoFs: Higher Order Filters.
No takes yet. Share an insight, caveat, or question.
Gupta et al. (2022) studied ECG signal processing. Independent Component Analysis (ICA) and Linear Discriminant Analysis (LDA) vs. Adaptive Notch Filter (ANF) was evaluated on Sensitivity, detection error rate, and output SNR. The proposed combined ICA and LDA technique for ECG signal pre-processing and classification achieved a sensitivity of 99.92%, detection error rate of 0.122%, and output SNR of 37.77dB.
Synapse has enriched 3 closely related papers on similar clinical questions. Consider them for comparative context: