Non-Conventional Measurements from ECG Recordings: Towards an Improvement of Diagnostic Properties

Anna M Bianchi, Luca T Mainardi, Sergio Cerutti

Department of Biomedical Engineering, Polytechnic University, Milano, Italy

Correspondence: Sergio Cerutti, Dipartimento di Bioingegneria, Politecnico di Milano, piazza Leonardo da Vinci 32, 20133 Milano, Italy.
E-mail: cerutti@biomed.polimi.it, phone 39 02 2399 3339, fax, 39 02 2399 3360

1.    Introduction

The recording and the analysis of the electrocardiographic (ECG) signal is widely employed in clinical practice. For this reason, some of the ECG features have been largely exploited in the past and their clinical relevance documented in large database of patients. Traditional ECG measurements basically include quantification of wave amplitude, duration and deflection, which are the basis for the development of conventional signal processing algorithm employed for: i) ECG filtering and artifact removal, ii) automatic classification and clustering of cardiac beat, iii) detection and classification of major arrhythmias and ischemia. More recently, methods for the quantitative analysis of repolarization period as well as late potentials, the assessment of pacemaker functions and implantable cardio-defibrillator (ICD) functions has been proposed [ACC/AHA Guidelines, 1999]. Most of these signal processing tools are available in many library of ECG analysis [Moody et al., 2001] and largely implemented in most of the commercially available electrocardiographs.

Beside these conventional measurements, a few emerging signal processing techniques are improving the amount of information which can be extracted from the ECG traces. These approaches range from modern techniques for Heart-Rate Variability (Task Force HRV, 1996) analysis (including short-term time-variant analysis [Mainardi et al., 2002] and long-term non-linear approaches [Baselli et al., 2002]), model-based analysis of QT and RR relationships [Porta et al., 1998], up to the quantification of respiratory rate and patterns from amplitude modulation of ECG waves. In this manuscript, a brief review of the more recent signal processing techniques applied to the analysis of ECG signal is presented. Emphasis is given on the addition information which can be extracted from the ECG traces and on the new perspective of these analyses for the actual diagnostic purposes and for fostering new ones.

2.    New Measurements from ECG Recordings

The diffusion of low-cost personal computers and the advances in microprocessor performances, widely contribute to the development of sophisticated signal processing algorithms for ECG analysis and their application to Holter analysis. Algorithms for automated arrhythmia detector and classification of each beat, measurements of ST-T displacement and detection of ischemia events and cluster analysis, are now available in most of commercial systems and are basically considered a standard nowadays.

Besides these standard products, there is a significant number of emerging methodologies which have progressively found solid applications. Among them we recall the analysis of RR (or NN) variability, both in time and in frequency domain, the study of T-wave alternans and QT-T interval dispersion, the estimation of respiratory parameters from the ECG signal. Table 1 summarizes some of the most innovative parameters that are now entering into the clinical evaluation, as they can provide further information on the general status of the patient.

Table 1.    Innovative parameters from the ECG signal.

2.1. HRV Time Domain Methods

In particular, analysis of NN variability may provide significant information on the status of autonomic nervous system control elicited in cardiovascular system, through indices evaluated in the time and in the frequency domain.

The following indices mainly provide a general indication of how the overall variability is distributed along the whole 24 hour period or, at least, during day and night.

The series of NN intervals can be converted into a geometric pattern such as the sample or the sample density distribution of differences between adjacent NN intervals. Simple indices are evaluated that measure the variability on the basis of the geometric and/or graphic properties of the resulting patterns. The HRV triangular index is the integral of the density distribution (the number of all NN intervals) divided by the maximum of the density distribution. The triangular interpolation of NN interval histogram (TINN) is the baseline width of the distribution measured as a base of a triangle approximating the NN interval distribution (the minimum square difference is used to find such a triangle). Both these measures express overall HRV measured over 24 hours and are more influenced by the lower than by the higher frequencies. The Differential Index is defined as the difference between the width of the histogram of differences between adjacent NN intervals measured at selected heights (e.g. at the level of 1000 and 10000 samples).

The Logarithmic Index is the coefficient f of the negative exponential curve ke^-ft, which is the best approximation of the histogram of absolute differences between adjacent NN intervals.

Return Maps o Delay Maps are constituted by the graphic display of the temporal series of apparently erratic values (consecutive NN intervals), which can highlight the presence of orderlying structure, and reveal the nature of the dynamic system generating the series. Return maps represent the relation between a point N(i) on x axis and the following N(i+τ) on y axis. The τ index is the temporal delay (τ=1, 2..N). Return maps can indicate the different complexity of the signal in various pathophysiological conditions. Also quantitative indexes can be evaluated to describe the pattern of complexity of the graphic display. Return maps may provide specific conditions with marked variation of subsequent R-R intervals, such as atrial fibrillation.

2.2. HRV Frequency Domain Methods

Power spectral density (PSD) analysis provides the basic information of how power (variance) distributes as a function of frequency. Methods for the calculation of PSD are generally classified as non-parametric (based on FFT) and parametric (based on a signal generation model). Three main spectral components are distinguished in a spectrum evaluated from short-term recordings (2 to 5 minutes): very low frequency (VLF), low frequency (LF) and high frequency (HF) components. LF component is centered around 0.1 Hz, and many experimental results both on humans and on animals, have demonstrated a power increase following a sympathetic activation; The HF component is synchronous with respiration and is modulated by the vagal tone. Therefore, the LF/HF ratio is considered a measure of sympatho-vagal balance. The VLF component (below 0.04 Hz) does not have a physiological interpretation in the short-term analysis. The investigation of the lower frequency content is reliable only on long-term recordings. In such case, the PSD can be represented in a log-log scale and the slope of the spectrum, the a-slope index, in the range of the VLF or even in the ultra low frequency (ULF) band, was found to be effective in the discrimination among different outcome in ICU patients or to have prognostic value in post myocardial infarction patients.

A limitation in the use of PSD both on short and on long time periods is the hypothesis of stationarity of the sequence of the NN intervals under analysis. The introduction of time-frequency and time-variant spectral analysis, in the recent years, allowed the dynamical evaluation of the classical spectral parameters also in transient conditions of clinical relevance, such as during ischemia and syncope [Bianchi, et al., 1992; Furlan, et al., 1998].

2.3. QT Analysis

This analysis is intended to explore the repolarization phase of the cardiac cycle [Porta, 1998]. The analysis of the static beat-to-beat QT-NN relation is generally performed on median beats. Beside the linear regression, the “best fit” function is evaluated, which can better describe the relation between the QT and RR parameters. For the quantification of the dynamic QT-NN relations a parametric linear model is proposed to quantify the dependency of QT variability as a function of the cycle length variability. The QT interval variability is decomposed, by spectral techniques, in two components, one dependent on NN interval and one independent from NN interval. Therefore, both the short-term dependence, i.e. the direct effect of the R-R interval immediately preceding the QT interval, and the long-term dependence, generated by the preceding cycle lengths, are considered in time and in frequency domain (estimation of the transfer function between QT and NN intervals).

The presence of ectopic beats can be utilized in order to obtain information on the transient response of QT interval following sudden cycle length changes. An ectopic beat generates a shorter cardiac cycle (coupling interval), followed by a longer cardiac cycle (post-extrasystolic pause). The transient response of QT interval following those sudden R-R interval changes can provide further information on QT-NN relation.

2.4. ECG Derived Respiration

The possibility of analyzing adjunctive signals in addition to the ECG, is acquiring increasing relevance in the diagnosis of many pathologies related to the cardiovascular system. In particular, recent studies, have demonstrated how dysfunctions in respiration may be correlated to a high risk factor in patient populations with or without cardiopathy. Further, disturbances in respiration during sleep have been found in large populations of hypertensive subjects. For sure, information on the respiratory pattern related to ECG morphological abnormalities (ectopic beats, arrhythmias, etc.) may lead to a better understanding of the generating mechanisms and on the possible clinical consequences on the subject. On the other hand, a reliable ambulatory monitoring of the respiration signal is not feasible under the hypothesis of not interfering with the patient day-life. In case of Holter recording, information on the respiratory activity can be derived directly from the ECG signal according to proper signal processing procedures [Moody, et al., 1986; Travaglini, et al., 1998], mainly based on the acquisition of three orthogonal leads. The respiratory activity can be estimated both on short and on long time periods, and some parameters (respiratory frequency, presence of apneas, etc.) are related to the more classical ECG or HRV parameters. Fig. 1 shows an example of respiration trace derived from the X, Y and Z orthogonal leads of the ECG, in a period of 5 minutes. The calculated trace is superimposed to the real respiratory signal (obtained through an airflow measurements), and the goodness of the estimation is well evident also in case of simulated apneas.

3.    Conclusions

The advances in processing methodologies, supported by clinical evaluation on large populations, contribute in the development of new quantitative indices useful in extracting the information hidden in the ECG signal. Besides classical morphological content, the ECG brings great information related not only to the heart, but also to many other organs or functions or processes somehow affecting the cardiovascular functions. Thus, simple ECG recordings are a real source for a widespread physiological and clinical evaluation. Holter systems, from simple ECG portable recording systems, are rapidly evolving to “intelligent” monitoring devices, which will be able to produce a complete and real-time report on the subject conditions, non only in case of life-threatening pathologies or events, but also for a more valuable long-term evaluation of therapeutic interventions or to implement bio-feedback in cardiovascular rehabilitation processes. The future may be in integrated systems, minimally interfering with the subject day-life, eventually wearable system, for a continuous and complete monitoring of the subject functions.

Figure 1.    Example of ECG derived respiration. The estimated trace is superimposed to the real trace obtained through airflow measurement.

References

ACC/AHA Guidelines for Ambulatory Electrocardiography, a report of the American Collage of Cardiology/American heart Association Task Force on Practice Guidelines(Committee to revise the guidelines for ambulatory electrocardiography), Journal of Amer. Collage of Cardiol., 34, 912-948, 1999.

Baselli G, Cerutti S, Porta A, Signorini MG. Short and long term non-linear analysis of RR variability series. Med Eng Phys. 2002 Jan;24(1):21-32.

Bianchi, A. M., Mainardi, L., Petrucci, E., Signorini, M. G., Mainardi, M., and Cerutti, S., 1993, “Time-variant power spectrum analysis for the detection of transient episodes in HRV signal”, IEEE Trans. Biomed. Eng,vol.40, no.2, pp.136-144

Furlan, R., Piazza, S., Dell'Orto, S., Barbic, F., Bianchi, A., Mainardi, L., Cerutti, S., Pagani, M., and Malliani, A. 1998, “Cardiac autonomic patterns preceding occasional vasovagal reactions in healthy humans”, Circulation, vol.98, no.17, pp.1756-1761.

Mainardi L. T. Bianchi., A. M and Cerutti S., "Time-Frequency and Time-Varying Analysis for Assesing the Dynamic Responses of Cardiovascular Control", Critical Reviews in Biomedical Engineering, 30(1-2):181-223 (2002).

Moody GB, Mark RG, Bump MA, Weinstein JS, Berman AD, Mietus JE, and Goldberger AL, Clinical Validation of the ECG_Derived Respoiration (EDR) Technique, Computers in Cardiology 1986, vol 13, pp. 507-510

Moody GB, Mark RG, Goldberger AL. PhysioNet: a Web-based resource for the study of physiologic signals. IEEE Eng Med Biol Mag. 2001 May-Jun;20(3):70-5.

Porta A, Baselli G, Caiani E, Malliani A, Lombardi F, Cerutti S. Quantifying electrocardiogram RT-RR variability interactions. Med Biol Eng Comput. 1998 Jan;36(1):27-34.

Task force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology, Heart rate variability: Standards of Measuremet, hysiological interpretation, and Clinical Use, 93:1043-1065, 1996.

Travaglini A., Lamberti C., De Bie J., Ferri M, Respiratory signal derived from eight-lead ECG, Computers in Cardiology, 25: 65-68, 1998.