Automatic analysis of the REpolarization segment

Jean Philippe Couderc, PhD, Wojciech Zareba, MD, PhD, Arthur J. Moss, MD
Department of Cardiology/HRFUP, University of Rochester
601 Elmwood av.Box 653,14642 NY  USA

Abstract: Classical methods for the automatic or semi-automatic measurement of QT intervals in digital Holter and bedside ECGs suffer from methodological problems decreasing their use for clinical and research purposes. T wave morphology, low amplitude T waves, presence of U waves are the main factors resulting in discrepancies between algorithms identifying the end of the repolarization.

We reviewed the literature and described the factors affecting the quality of algorithms outputs used to measure the repolarization interval. We briefly introduce the area-based approach we are currently investigating which is close to the concept used for measuring the length of monophasic action potential recordings. The area-based parameters could provide more stable measurements of the repolarization segment when the T wave shape is abnormal.

Introduction

Since European and American governmental agencies request from pharmaceutical companies to check potential QT prolongation in their new drug application process including non-cardiac related drugs. There is an increasing need for fast, accurate and stable method for measuring the repolarization interval in short term ECG recordings as well as in less clean signals from Holter systems.

The quality of the manual QT-interval measurements for intra and extra observers from ECG tracing has been shown to be low (1) . The inter-observer differences between measurements can be reduced using digital ECG recording technology and computer-based methods due to their  methodological consistency (2-4) . The computerized methods show limitations related to the large variety in the T wave shapes. The difficulty of obtaining stable estimation of T wave did not led constructors to a common approach but rather to a set of different approaches we shall briefly review in this paper.

Also, we describe the method we implemented. This method was designed in accordance with the recommendations of the ISHNE guidelines (5) .

Method

Classical approaches

We define 3 classes of algorithms for the identification of the end of the T wave as described in Figure 1.

Figure 1: Classical methods for the determination of the QT intervals.

These classes differ by their definitions of the end of the T wave : 1) when the amplitude of the T wave signal decreases below a certain amplitude level (Panel A), 2) when the first derivative of the T wave pass below a threshold value (Panel B), and 3) when the slope of the final descending portion of the T wave crosses the baseline, the slope being either the maximum slope of this terminal portion (Panel C) or a least squares model fitting this terminal portion (Panel D) (6-8) . This list is not exhaustive but rather describes the most popular approaches. Other interesting methods identifying the end of the T wave and quantify QT dispersion using different approaches have been published (9;10) .

The source of instability of computerized approaches are explained by the large variety of T wave shapes recorded from the body surface (6) . These main factors characterizing the T wave morphology are the following: 

1) The amplitude constraint: the common limitation to all these approaches is their need for having T wave with sufficient amplitude to identify a terminal portion and then identify a slope and/or a significant decrease in this slope. If the T wave is flat most of the techniques fail determining the end of the T wave (3;11) .

2) The shape constraint: having a biphasic T wave can create unexpected results with certain types of algorithm trying to identify the terminal descending slope of the T wave. The occurrence of U waves is also a noticeable shape-related issue for computer and manual measurements (12) . Encompassing the U wave in the repolarization measurement is still a controversial issue and the occurrence of U wave may affect the stability of computerized approaches.

3) The baseline, the respiratory and body movement factors: these specific factors are quite accurately addressed in the current commercial systems (13) ,and in research-oriented software (14) . It is however of concern in Holter ECG data where signal to noise ratio and signal quality in general are lower (15) .

Finally the noise level has impact on the determination of the end of the T wave and more profoundly in Holter ECG recordings than in bedside.

Area-based approach:

The area-based approach is based on the cumulative integration of the T wave (Figure 2). It provides a curve describing the evolution of the area under the T waves across time. The curve is normalized in amplitude. Then, percentages of the maximum amplitude are used to define time intervals (Figure 3) (16) .

The method does not need an accurate determination of the end of the T wave rather it requires the T wave to have a right tail rapidly converging to the baseline level.

Figure 2: examples of repolarization measurements on a biphasic T-wave (measurements done on the bold signal).

Figure 3: area-based approach based on the normalized total area of the T wave.

The area-based approach provides a different approach for the measurements of the repolarization segment. Its main interest is linked to its ability to consider the entire repolarization morphology instead of the terminal portion for the identification of the end of the T wave. Under good signal condition, such approach should allow for more stable measurements and could allow measurements in low-amplitude T wave. More investigations are needed and are currently in process for demonstrating the benefits of this approach.

Reference List

   1.   Murray A, McLaughlin NB, Bourke JP et al. Errors in manual measurement of QT intervals. Br Heart J. 1994;71:386-390.

   2.   Puddu PE, Bernard PM, Chaitman BR et al. QT interval measurement by a computer assisted program: a potentially useful clinical parameter. J Electrocardiol. 1982;15:15-21.

   3.   Murray A, McLaughlin NB, Campbell RW. Measuring QT dispersion: man versus machine. Heart. 1997;77:539-542.

   4.   Kors JA, van Herpen G. The coming of age of computerized ECG processing: can it replace the cardiologist in epidemiological studies and clinical trials? Medinfo. 2001;10:1161-1167.

   5.   Moss AJ, Zareba W, Benhorin J et al. ISHNE guidelines for electrocardiographic evaluation of drug-related QT prolongation and other alterations in ventricular repolarization: task force summary. A report of the Task Force of the International Society for Holter and Noninvasive Electrocardiology (ISHNE), Committee on Ventricular Repolarization. Ann Noninvasive Electrocardiol. 2001;6:333-341.

   6.   Xue Q, Reddy S. Algorithms for computerized QT analysis. J Electrocardiol. 1998;30 Suppl:181-186.

   7.   van Bemmel JH, Zywietz C, Kors JA. Signal analysis for ECG interpretation. Methods Inf Med. 1990;29:317-329.

   8.   McLaughlin NB, Campbell RW, Murray A. Comparison of automatic QT measurement techniques in the normal 12 lead electrocardiogram. Br Heart J. 1995;74:84-89.

   9.   Acar B, Yi G, Hnatkova K et al. Spatial, temporal and wavefront direction characteristics of 12-lead T-wave morphology. Med Biol Eng Comput. 1999;37:574-584.

  10.   Lund K, Nygaard H, Kirstein PA. Weighing the QT intervals with the slope or the amplitude of the T wave. Ann Noninvasive Electrocardiol. 2002;7:4-9.

  11.   McLaughlin NB, Campbell RW, Murray A. Accuracy of four automatic QT measurement techniques in cardiac patients and healthy subjects. Heart. 1996;76:422-426.

  12.   Ireland RH, Robinson RT, Heller SR et al. Measurement of high resolution ECG QT interval during controlled euglycaemia and hypoglycaemia. Physiol Meas. 2000;21:295-303.

  13.   Gang Y, Guo XH, Crook R et al. Computerised measurements of QT dispersion in healthy subjects. Heart. 1998;80:459-466.

  14.   Coumel P, Fayn J, Maison-Blanche P et al. Clinical relevance of assessing QT dynamicity in Holter recordings. J Electrocardiol. 1994;27 Suppl:62-66.

  15.   Tikkanen PE, Sellin LC, Kinnunen HO et al. Using simulated noise to define optimal QT intervals for computer analysis of ambulatory ECG. Med Eng Phys. 1999;21:15-25.

  16.   Merri M, Benhorin J, Alberti M et al. Electrocardiographic quantitation of ventricular repolarization. Circulation. 1989;80:1301-1308.