Analysis of unpredictable intra-QRS potentials in signal-averaged electrocardiograms using an autoregressive moving average prediction model

https://doi.org/10.1016/j.medengphy.2009.11.001Get rights and content

Abstract

Instead of extracting the abnormal intra-QRS potentials (AIQP) waveform, this study proposes the analysis of the unpredictable intra-QRS potentials (UIQP) based on an autoregressive moving average (ARMA) prediction model to detect the signals with sudden slope change within the QRS complex for the diagnosis of high-risk patients with ventricular tachycardia (VT). The UIQP is detected as the slope changes at slope discontinuities by the prediction error of the ARMA prediction model. Because of the linearity of the ARMA prediction model, the UIQP is also proportional to the amplitude of the QRS complex if the input QRS waves have the same shapes. Hence this study further defines the UIQP-to-QRS ratio to normalize the UIQP by the root-mean-square (RMS) value of the QRS complex. The study subjects were composed of 42 normal Taiwanese and 30 patients with sustained VT. The clinical results show that the UIQP-to-QRS ratios of the VT patients in leads X, Y and Z were significantly higher than those of the normal subjects. The logical combination of any 4 of the UIQP-to-QRS ratios and conventional time-domain parameters can increase the diagnosis performance of VT patients to 92.9% specificity, 93.3% sensitivity and 93.1% total prediction accuracy.

Introduction

Ventricular late potentials (VLP) in signal-averaged electrocardiograms (SAECG) which outlast the normal QRS interval have been an important and non-invasive marker for the risk stratification of ventricular arrhythmias to prevent sudden cardiac death [1], [2], [3], [4]. According to the recommendations of an ACC Expert Consensus Document [4] for the use of SAECG, the established clinical values are for the stratification of the risk of development of sustained ventricular arrhythmias in patients who are recovering from myocardial infarction, and for the identification of patients with ischemic heart disease and unexplained syncope. Several recent studies further applied time-domain VLP analysis to evaluate the risk of ventricular arrhythmias for symptomatic and asymptomatic patients with Brugada syndrome [5], Chagas disease patients [6], thalassemia patients [7] and patients with arrhythmogenic right ventricular cardiomyopathy [8].

However, the weakness of VLP analysis is the low positive predictive accuracy [4] and the incomplete characterization of reentrant activity [9]. The VLP may be completely contained within the normal QRS complex and not prolong the normal QRS interval [10], [11]. In recent years, several studies [12], [13], [14], [15] have also been focused on the analysis of the abnormal intra-QRS potentials (AIQP) which are considered as low-amplitude notches and slurs with sudden changes in slope, to enhance the diagnostic performance of SAECG. The AIQP analysis has also been applied to noninvasively identify the mechanisms of premature ventricular beats (PVBs) [16] and to detect acute transmural myocardial ischemia [17].

The current autoregressive moving average (ARMA) modeling technique in the discrete cosine transform (DCT) domain [12] for extracting the AIQP can use a low model order to estimate the normal QRS complex and then separate out the AIQP. However, the true model order is unknown. Although our previous study [14] proposed a cross correlation method to automatically determine the model order, it cannot verify if the determined order is correct. The estimated AIQP may also contain a high estimation error due to the overlap between the normal QRS complex and the AIQP in the DCT domain [15]. Hence it is not an easy job to accurately extract the AIQP waveform, particularly with an extremely poor signal–noise-ratio (low-amplitude AIQP compared with a large QRS wave).

Instead of extracting the AIQP waveform, this study proposes the analysis of unpredictable intra-QRS potentials (UIQP) based on an autoregressive moving average prediction model to detect the signals with sudden slope change, which originate from the sharp QRS wave and the AIQP, for the diagnosis of high-risk patients with ventricular tachycardia (VT). The VT patients are expected to have higher UIQP because the presence of AIQP would bring more components with sudden changes in slope within the QRS complex. The aim of this study is to determine whether the UIQP detected as the slope changes at slope discontinuities by the prediction error of an ARMA prediction model can be applied to diagnose the VT patients and to improve the diagnostic performance of SAECG.

Section snippets

Data acquisition

This work followed the principles that (1) informed consent was obtained from each patient and (2) the Ethics Committee of Taipei Jen-Chi General Hospital had approved the study. The study subjects recruited were 42 normal Taiwanese (20 men and 22 women, aged 58 ± 14 years) and 30 patients with sustained VT (15 men and 15 women, aged 63 ± 16 years). The VT patients were suffering from chronic ischemic heart disease after surviving clinically documented myocardial infarction (MI). These study

Simulation analyses

To demonstrate the signals that can and cannot be predicted by an ARMA prediction model, this study used a positive sinusoidal wave of 900 μV peak value and 100 ms duration to simulate a smoothed R wave, and a low-amplitude, transient positive triangular wave of 30 μV peak value and 2.5 ms duration to simulate the AIQP. The simulated AIQP was located at time 150 ms. Fig. 2(a) shows the synthesized input signal (solid line) and the prediction output (dashed line) of an ARMA prediction model of order

Discussion

This study has demonstrated that the proposed ARMA prediction model can estimate the smoothed part of the input QRS complex, and the prediction error can be used to analyze the UIQP for the diagnosis of high-risk patients with VT. The simulation study in Fig. 2 showed that the prediction error can detect the signals with sudden slope changes as the slope changes at slope discontinuities. Although both the AIQP and the sharp QRS wave can produce sudden slope changes and cannot be further

Conclusions

This study has successfully demonstrated that the UIQP, defined as the signals with sudden slope changes, can be detected as the slope changes at the slope discontinuities using the ARMA prediction modeling technique. The clinical results further show that the UIQP-to-QRS ratios of VT patients were significantly higher than those of the normal group in leads X, Y and Z, and the logical combination of the UIQP-to-QRS ratios and the time-domain VLP parameters can enhance the diagnostic

Conflict of interest statement

No author had a financial or personal conflict of interest related to this research or its source of funding.

Acknowledgments

The author thanks the staff of the Hemodialysis Unit and patients of the Cardiology Department at Jen-Chi General Hospital for their kind assistance and cooperation in this investigation.

Financial support: This research was supported by Taiwanese National Science Council research grant NSC97-2320-B-167-001.

References (20)

There are more references available in the full text version of this article.

Cited by (17)

  • Facile immobilization of enzyme by entrapment using a plasma-deposited organosilicon thin film

    2014, Journal of Molecular Catalysis B: Enzymatic
    Citation Excerpt :

    Because multiple copies of the same functional group usually exist on protein surface, covalent immobilization is usually not selective [10]. When the protein contains just one copy of the reactive functional group, oriented immobilization can be achieved, which is usually realized through selective chemical reactions or by protein engineering [11]. In other words, covalent immobilization requires a good control over the orientation.

  • Principal component analysis in high resolution electrocardiogram for risk stratification of sustained monomorphic ventricular tachycardia

    2014, Biomedical Signal Processing and Control
    Citation Excerpt :

    Thus, in this setting, the term ‘VLP’ has been employed here to describe the occurrence of potentially arrhythmogenic high frequency electrical transients, indistinctively to its time of occurrence during ventricular activation. Those transients represent electrical activity originated from areas where propagation of the ventricular activation is delayed and fragmented [23–25]. An increased risk for ventricular tachyarrhythmia post-MI is usually related to extensive myocardial injury and substantial QRS complex waveform changes, when compared to healthy control, thus affecting both high and low frequency electrical transients.

  • N-Acylation Reactions of Amines

    2014, Comprehensive Organic Synthesis: Second Edition
  • A versatile approach for the immobilization of amines via copper-free "click" reaction between azido self-assembled monolayer and alkynyl Fischer carbene complex. Application to the detection of staphylococcal enterotoxin A antibody

    2011, Journal of Organometallic Chemistry
    Citation Excerpt :

    The formation of the alkoxycarbene-terminated SAM on gold chips is illustrated in Scheme 1. A glass slide coated with a thin layer of gold was flame-annealed then treated with a 1 mM ethanolic solution of bis(11-azidoundecyl)disulfide synthesized according to the literature [5,15] for 60 h at RT. After extensive washing, the gold surface (step A) was analyzed by PM-IRRAS (Fig. 1A, spectrum a).

View all citing articles on Scopus
View full text