Kernel-Based Feature Relevance Analysis for ECG Beat Classification
The analysis of Electrocardiogram (ECG) records for arrhythmia classification favors the developing of aid diagnosis systems. However, current devices provide large amounts of data being necessary the development of signal processing methodologies to reveal relevant information. Here, a kernel-based feature relevance analysis approach is introduced to highlight discriminative attributes in ECG-based arrhythmia classification tasks. For such purpose, morphological and spectral-based features are extracted from each provided heartbeat. Then, a linear mapping is learned by using a Kernel Centered Alignment-based scheme to highlight the most relevant features when estimating nonlinear dependencies among samples. The proposed approach is performed as a cascade classification scheme to avoid biased results due to unbalance issue of the studied phenomenon. The results yield a performance rate of \(86.52\,\%\) (sensitivity), \(97.57\,\%\) (specificity), and \(92.57\,\%\) (accuracy) in a well-known database, which validate the reliability of the proposed algorithm in comparison to the state-of-art.
KeywordsKernel functions ECG feature extraction Relevance analysis
This work is supported by Programa Nacional de Formación de Investigadores “Generación del Bicentenario”, 2011/2012 funded by COLCIENCIAS and the project “Plataforma tecnológica para los servicios de teleasistencia, emergencias médicas, seguimiento y monitoreo permanente de pacientes y apoyo a los programas de prevención” Eje 3 - ARTICA.
- 2.Karpagachelvi, S., Arthanari, M., Sivakumar, M.: Ecg feature extraction techniques-a survey approach. arXiv preprint arXiv:1005.0957 (2010)
- 7.Zhang, L., Peng, H., Yu, C.: An approach for ecg classification based on wavelet feature extraction and decision tree. In: 2010 International Conference on Wireless Communications and Signal Processing (WCSP), pp. 1–4. IEEE (2010)Google Scholar
- 10.Castro Hoyos, C., Peluffo Ordonez, D.H., Rodríguez-Sotelo, J.L., Dominguez, G.C.: Effectiveness of morphological and spectral heartbeat characterization on arrhythmia clustering for holter recordings. In: SIPAIM (2014)Google Scholar