Viacheslav Antsiperov - Kotelnikov Institute of Radioengineering and Electronics of Russian Academy of Sciences, Mokhovaya 11–7, Moscow, Russia
The nature and characteristics of ECG signals have been the subject of research for more than 100 years. This is because ECG signals contain a valuable diagnostic infor- mation about the functioning of the heart and, indirectly, about some other organs. Now, despite the impressive track record of the diagnostics and recommendations already de- veloped , ECG signals continue to be the subject of in- tensive study both in the field of cardiology and biomed- ical engineering. Cardiologists search for new diagnostic properties of ECG signals, while engineers are interested in new approaches and methods for such signals process- ing and analysing. New methods imply, in particular, new techniques of noise suppression, efficient signal represen- tation and feature extraction.
Feature extraction is an essential step in ECG processing. It consists in formation of some feature patterns – a set of descriptors most adequately describing the signal. Among these features / descriptors the most important are those with diagnostic properties. As a rule, they represent time domain parameters, though sometimes they can be defined in the frequency domain. Basing on ECG patterns, one can make some initial diagnosis . For example, an “irregu- lar” QRS complex without P wave is the hallmark of atrial fibrillation, the shape of the QRS with a left bundle branch block indicates a risk of cardiomyopathy, etc.
ECGpatterns, especiallyP–QRS–Tcomplexes, arealsothe basis for the subsequent heart rate analysis. Usually it is a more complex phase of ECG signal processing. However, now there are many methods and approaches related, for example, to the heart rate variability (HRV) analysis . Quantitative evaluation of the HRV is usually based on the calculation of some indices connected with the variability of sequential NN intervals (normal–to–normal intervals be- tween adjacent QRS complexes). Some of these indices are determined in the time domain – SDNN, CV, RMSSD, PNN5O, etc. Others – in the frequency domain – VLF, LF, HF spectral power components, IC index, etc. It should be noted that, due to the complexity of the EEG signal, its fractal–like nature, the values of almost all indices can es- sentially diverge for different lengths of time intervals anal- ysed, even for the same record. Therefore, the European Society of Cardiology suggests differentiating the diagnos- tic methodologies based on HRV indices obtained from the short–term recordings (of 2 to 5 min) and from the long– term recordings (of entire 24–h period) .
The above discussion is illustrated in Figure 1 by cardioin- tervalograms (CIG, the NN interval dependences on time) for a healthy person (A) and a heart disease person (B). It is clear from the figure that both CIGs are qualitatively dif- ferent. For example, at a small temporal scale, the variability (spread) of NN intervals in a healthy person is greater, which is a well known fact  – but on large scales the vari- ability becomes equal in both cases. This observation sug- gests that the diagnosis based on HRV is adequte, but di- agnostic procedures (automatic in particular) may be non- trivial. This circumstance gives high relevance to the de- velopment of new ECG processing methods based on long records – more than 10 minutes, for example, Holter.
- J.S. Coviello. ECG interpretation made incredibly easy! Philadelphia: Wolters Kluwer, 3 edition, 2017.
- R.M. Baevsky and A.G. Chernikova. Heart rate variabil- ity analysis: physiological foundations and main methods. Cardiometry, (10):66–76, 2017.
- Task Force of the European Society of Cardiology, the North American Society of Pacing, and Electrophysiol- ogy. Heart rate variability. standards of measurement, physiological interpretation, and clinical use. Eur Heart J, 7(3):354–381, 1996.
- A.L. Goldberger and H.E. Stanley et al. Long-range an- ticorrelations and non-gaussian behavior of the heartbeat. Phys Rev Letters, 70(9):1343–1346, 1996.
- S.S.Yamamoto J.F. Thayer and J.F. Brosschot. The re- lationship of autonomic imbalance, heart rate variability and cardiovascular disease risk factors. Phys Rev Letters, 141(2):122–131, 2010.
- M. Malik. Geometrical methods for heart rate variability assessment. In M. Malik and A. Camm, editors, Heart Rate Variability, pages 47–61. Armonk, Futura Publishing Company -New York, 1995.
- V.E. Antsiperov and I.V. Zabrosaev. New results for the pvc / spb detection using based on the mca heart rhythm estimation method. In Proceedings of the 12th Russian- German Conference on Biomedical Engineering, pages 182–186. Vladimir State University, jul 2016.
- V.E. Antsiperov. Syclic renewal point processes for heart rate variability modeling. In Extended abstracts of the sec- ond Russian Conference with international participation "Physics for Life Sciences", page 61. Ioffe Institute sience- tech information department, sen 2017.
- R. Serfozo. Basics of Applied Stochastic Processes. Springer, -Berlin, -Heidelberg, 2009.  L.A.N. Amaral A.L. Goldberger and et al. P.h. Ivanov. Physiobank, physiotoolkit, and physionet: Components of a new research resource for complex physiologic signals. Circulation, 101(23):e215–e220, 2000