Abstract— Perturbations in the normal heart rate are generally represented by the presence of premature activation (PA) beats in the surface electrocardiogram (ECG). The presence of PA is one of the main reasons of instability in QT dynamics which could initiate arrhythmia. Analyzing Boundary-Input Boundary-Output (BIBO) stability of the short term linear autoregressive QT-RR model is a way of detecting instability in QT dynamics from the ECG. The aim of this paper is to investigate if PA is the only reason for instability in the ventricular repolarisation process, which is denoted by QT interval of surface ECG. Ten healthy subjects with normal sinus rhythm and seven patients with sustained ventricular tachycardia (VT) were analyzed in this study. 10 min long ECG data were collected from each subject of the healthy group and 10 min ECG before the start of VT were taken for each subject of the VT group. Autoregressive QT-RR model was derived for each non-overlapping 1 min long ECG segment of the 10 min long ECG data. Instability in QT dynamics was quantified by measuring the numbers of unstable segments in ECG data for each subject (  ). Results of this study revealed that like the VT group subjects, QT instability detected by QT-RR model is also found in healthy subjects whose ECG segments are mostly free from PA beats. This finding indicates that BIBO unstable QT characteristics might arise from other inherent factors of cardiovascular system in addition to PA. I. INTRODUCTION Mathematical modeling of the interaction between QT and RR intervals form surface ECG is a non-invasive technique to comprehend the relationship between heart rate variability and ventricular repolarisation variability [1]. Ventricular repolarisation (VR) phenomenon can be understood by monitoring the QT interval and any irregularity in this interval like duration change or the distortion in the shape in T wave could be an indication of the initiation of arrhythmia [3]. The linear parametric model developed by Porta et al. [2] explains the effect of Autonomic Nervous System (ANS) control on the VR process. The adaptation of QT interval with the alteration in heart rate was investigated using a transfer function based model derived by Halamek et al. [3] where the effect of QT hysteresis was also M. H. Imam, C. Karmakar, A. Khandoker, and M. Palaniswami are with the Electrical and Electronic Engineering Department, University of Melbourne, Melbourne, VIC 3010, Australia (phone: +61-(0) 3-8344-0377; fax: +61- (0)3-555-5555; e-mail: m.imam@student.unimelb.edu.au; ahsank@unimelb.edu.au;karmakar@unimelb.edu.au;palani@unimelb.edu.a u). A. Khandoker is also with the Dept of Biomedical Engineering, Khalifa University of Science, Technology and Research, Abu Dhabi, UAE (e-mail: ahsan.khandoker@kustar.ac.ae). considered. Besides the understanding of VR process, the QT-RR model can also be used to determine the QT dynamics stability. Some recent studies used this technique to detect instability in QT dynamics and used it as an indicator of the onset of ventricular tachycardia (VT) and ventricular fibrillation (VF) in patients having structural heart disease [4,5]. The instability in the QT dynamics could be a prognostic marker of the arrhythmia susceptibility for diseased human heart having prolonged QT interval, acute myocardial infarction (AMI) and dilated cardiomyopathy [4]. These studies showed that the presence of premature activations (PA) in the ECG which are directly related with the unstable action potential dynamics (APD), could alter the normal QT variability and this perturbed QT dynamics could start arrhythmias like sustained ventricular tachycardia in AMI patients [5]. Chen et al.[4] developed a method by calculating the BIBO (bounded-input bounded-output) stability index from the linear autoregressive QT-RR model to detect the unstable ventricular repolarisation characteristics and demonstrated how the amount of PA affects the QT dynamics stability by calculating the frequency of PA in the ECG [4]. They hypothesized that the occurrence and the frequency of PAs affect the stability in QT dynamics and before the onset of arrhythmia (i.e. VT) the QT dynamics instability increases. Their study analyzed 10 minute long ECG signal segments of the patients having sustained VT by forming two groups of ECG form the total data. The Two segments were termed as the Control segment (i.e. the 10 minute duration ECG segment extracted long before the beginning of VT) and the VT segment (i.e. the 10 minute long ECG segment collected just before the initiation of VT). They collected these two groups of ECGs from the same subject. The 10 minute ECG data was then divided into ten 1 min long segments for each subject for the purpose of modeling and the calculation of stability index. The number of unstable segments and the frequency of PA were found to increase before the occurrence of VT [4, 5]. An important conclusion of their study was that, ECG segments which have no PA did not show instability in the analysis [5]. The objective of this paper is to explore whether the instability characteristics in QT dynamics determined from the QT-RR model [4, 5] is due only to the presence of PA beats in the ECG. Another target is to examine how model complexity changes for the identification of the dynamic relation between heart rate and ventricular repolarisation in healthy subjects in comparison the subjects having heart disease. Effect of premature activation in analyzing QT dynamics instability using QT-RR model for ventricular fibrillation and healthy subjects Mohammad H. Imam, Student Member, IEEE, Chandan K. Karmakar, Member, IEEE, Ahsan H. Khandoker, Senior Member, IEEE, and Marimuthu Palaniswami, Fellow, IEEE 35th Annual International Conference of the IEEE EMBS Osaka, Japan, 3 - 7 July, 2013 978-1-4577-0216-7/13/$26.00 ©2013 IEEE 2559