Analysis of electrocardiogram pre-shock waveforms during ventricular fibrillation. (August 2015)

• Record Type: Journal Article
• Authors: Rasooli, M.; Foomany, F.H.; Balasundaram, K.; Masse, S.; Zamiri, N.; Ramadeen, A.; Hu, X.; Dorian, P.; Nanthakumar, K.; Krishnan, S.; Beheshti, S.; Umapathy, K.
• Is Part Of: Biomedical signal processing and control. Volume 21(2015)
• Page Start: 26
• Page End: 33
• Publication Date: 2015-08
• Subjects: Blind source separation -- Ventricular fibrillation -- Independent component analysis -- Shock prediction -- Feature extraction -- Pattern classification; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2015.05.003
• Languages: English
• ISSNs: 1746-8094

Abstract

Abstract : Highlights: ECG waveform markers during fibrillation could be associated to underlying sources. Blind source separation approach can resolve ECG into independent sources (ISs). Proposed approach performed comparatively well with balanced prediction accuracies. ISs could reveal hidden signal characteristics relevant for cardiac resuscitation. Abstract: Pre-shock waveform analysis for optimizing the timing of shock delivery could be immensely helpful to emergency medical personnel in treating ventricular fibrillation. For this purpose, our proposed method resolves the pre-shock surface electrocardiogram into independent sources using a blind source separation approach. The electrocardiogram pre-shock waveforms were transformed into the wavelet domain and the independent sources were extracted using component analysis. A database consisting of 50 pre-shock waveforms from 50 pigs was used in this study. The pre-shock waveforms were obtained using a controlled protocol. After ventricular fibrillation was induced and left untreated for 2–5 min, cardio pulmonary resuscitation was administered for 3 min, followed by defibrillation. Energy-based features were extracted from the independent sources and a linear discriminant analysis based pattern classifier was used to evaluate the features for their ability to discriminate between successful and unsuccessful shock outcomes. The proposed method achieved a classification accuracy of 68% ( P < 0.02), and the classification results were cross-validated using the leave-one-out method. A comparative study demonstrated that the proposed approach performed relatively well compared to existing methods for the given database.