Biomedical Signal Processing and Control 8 (2013) 733–739 Contents lists available at ScienceDirect Biomedical Signal Processing and Control jou rn al h om epage: www.elsevier.com/locate/bspc Analysis of the robustness of spectral indices during ventricular fibrillation Jesús Requena-Carrión a,∗ , Felipe Alonso-Atienza a , Estrella Everss a , Juan José Sánchez-Mu ˜ noz b , Mercedes Ortiz c , Arcadi García-Alberola b , José Luis Rojo-Álvarez a a Department of Signal Theory and Communications, University Rey Juan Carlos, Fuenlabrada, Spain b Arrhythmia Unit, Hospital Universitario Virgen de la Arrixaca, Murcia, Spain c Cardiac Electrophysiology and Clinical Arrhythmology Unit, Madrid Hospital Group, CEU-San Pablo University, Madrid, Spain a r t i c l e i n f o Article history: Received 19 November 2012 Received in revised form 3 May 2013 Accepted 25 June 2013 Available online 1 August 2013 Keywords: Spectral analysis Dominant frequency Cardiac organization Implantable cardioverter-defibrillator Intracardiac electrogram a b s t r a c t The spatiotemporal characteristics of cardiac fibrillation are often investigated by using indices extracted from the spectrum of cardiac signals. However different signal acquisition systems may produce signals of different spectra and affect the estimation of some spectral indices. In this study, we investigate the robustness of four spectral indices previously proposed for describing fibrillation, namely the dominant frequency (DF), the peak frequency (PF), the median frequency (MF) and the organization index (OI). The effects of different lead configurations on the values of the spectral indices are statistically quan- tified and further analyzed in a database consisting of unipolar and bipolar intracardiac electrograms (EGM), recorded by implantable cardioverter-defibrillators during ventricular fibrillation. Our analysis shows that the lead configuration significantly affects the PF, the MF and the OI, whereas the DF remains unaffected. We further explore the nature of cardiac spectrum and show that unipolar EGM concentrate power at lower frequencies than bipolar EGM. We conclude that indices that depend on the envelope of the spectrum of cardiac signals are in general sensitive to the lead configuration. © 2013 Elsevier Ltd. All rights reserved. 1. Introduction Ventricular fibrillation (VF) has been traditionally described as a highly irregular and disorganized cardiac rhythm [1]. This descrip- tion is motivated by the observation of VF in the electrocardiogram (ECG), in which VF lacks any discernible pattern and is character- ized by irregular deflections whose frequency, amplitude and shape are continually changing. However, some studies based on spectral analysis of cardiac signals have suggested otherwise, that fibrilla- tion is not completely chaotic and could possess a high degree of spatiotemporal organization. Early applications of spectral analy- sis revealed that ECG spectrum during VF has a distinct dominant frequency (DF) between 3 and 7 Hz [2], and that median frequency (MF) reflects the evolution of an episode of VF [3]. More recently, two methods based on spectral analysis have been developed for quantifying cardiac spatiotemporal organization during fibrillation. In the first method an organization index (OI) extracted from the ∗ Corresponding author at: Universidad Rey Juan Carlos, Campus de Fuenlabrada, Camino del Molino s/n, Departamental III, D207, 28943 Fuenlabrada, Madrid, Spain. Tel.: +34 91 488 8463; fax: +34 91 488 7500. E-mail address: jesus.requena@urjc.es (J. Requena-Carrión). spectrum of intracardiac electrograms (EGM) has been used to quantify atrial fibrillation (AF) organization [4]. The second method consists of computing the DF in either optical or electrical recor- dings to estimate the local activation rate of cardiac tissue during fibrillation. The analysis of DF maps has revealed regional differ- ences in the atria during AF [5–7] and in the ventricles during VF [8,9], which have been interpreted as a manifestation of cardiac spatiotemporal organization. In other practical applications, spec- tral analysis has been used to detect ventricular tachyarrhythmias [10–13], to predict the success of the defibrillation shock [2,14–16], to characterize VF in implantable cardioverter-defibrillator (ICD) EGM [17–20] and to identify artifact events in ICD EGM [21]. The widespread use of spectral indices for describing fibrillation has motivated to further investigate their nature and their relation- ship to cardiac spatiotemporal characteristics. On the one hand, from a signal processing perspective it has been shown that EGM complexity and fractionation can hinder the estimation of the local activation rate during DF analysis [22]. Also, the DF and a spec- tral index used for quantifying EGM regularity have been analyzed in [23]. On the other hand, from a signal measurement perspec- tive there exists the possibility that lead configuration may have an effect on the estimation of some spectral indices. For instance, it has been previously shown that during spatially correlated rhythms 1746-8094/$ – see front matter © 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.bspc.2013.06.013