Single lead ECG informativity for wearable application.
White paper by RocketBody Technologies, Inc.
This white paper describes potential of Lead I ECG signal measurements for wearable on wrist application. The ECG limb leads signal futures extraction and informativity versus precordial leads are discussed.
Introduction
The history of ECG acquisition by device wearable on wrist began in 1970th (for example [1]). Since then many manufacturers of both nameless and top brand devices began using ECG modules in their products [2].
Largely due to appearance and further development of specialized embedded electronics, provided by leading manufacturers like Analog Devices and Maxim Integrated, the ECG method is widely available now [3,4]. Manufacturers not only provide ready-for-use modules adapted for the needs of ECG signal acquisition, but also give ample opportunities for their tuning, from recording of pulse during training, to monitoring of cardiac activity during the state of complete rest.
However, most of the devices with ECG sensors are limited to detecting pulse, heartrate variability or arrhythmia (the same functionality could be easily achieved by pulse oximetry method), or sending the recorded files to medical specialists.
At the same time, the classic multi-channel ECG is capable to provide huge amount of specific information about complex phenomena in cardiac muscle and neural regulation of the heart.
So, the main question is whether it possible to obtain more information from single-lead ECG signal recorded by wrist-worn devices rather than HR only.
12 Lead ECG
The classical gold-standard electrocardiography (ECG) is the method of recording the electrical activity of the heart using electrodes placed on the body. The standard ECG uses 12 leads (Limb, Augmented, and Precordial) formed with 10 electrodes. Each of the 12 leads detect electrical activity of the different areas of the heart (figure 1).
The typical ECG signals from different leads include P-wave, QRS-complex, and T-wave (figure 2) that form PR, QT intervals and ST segment.
Origin of ECG signals
As shown in Plonsey’s works the electrical activity of the heart can be represented as a circular motion of a single electrical vector (dipole) of variable length with the initial point in the sinoatrial node (figure 3) [5]. All ECG signals recorded from different leads are projections of this vector onto corresponding axes. It is known that only three projections (not lying in the same plane) are required to provide an unambiguous description of any vector. Thus, the classical 12-lead ECG provides a lot of redundant information. This redundancy is used for detailed characterization of the heart’s chambers functioning and the individual signal phases visualization for medical purposes.
Signals correlation inside ECG
Due to the redundancy of the entire 12-lead ECG system, it can be assumed that signals from different leads should be highly correlated. From a practical point of view, the correlation between limb leads and precordial leads is of the greatest interest for use with wearable devices. Highly informative chest leads give specific information about functional state (or pathologies) of the heart.
Let’s consider, for example, leads V6 (reflects activity of the lateral wall of left ventricle) and V2 (reflects activity of the septal surface of the heart) which are located at a large distance from each other and at a sufficiently large angle between their vectors.
In order to justify and find the correlation between Lead I signal and signals V6 and V2, it is first necessary to convince that vectors V6 and V2 have projections (vector components) on the axis of limb Lead I vector
As shown in Figures 1,2 and 4, vector V6 is parallel (in the sagittal plane) to Lead I that indicates considerable correlation between them. At the same time, there is a significant angle between vector V2 and Lead I so the length of vector V2 projection on the axis of Lead I is quite short, and hence their correlation should be less significant.
Since both vectors V6 and V2 have projections on the axis of Lead I, it is possible to compare the Lead I signal with the sum of the projections of V2 and V6 signals onto the Lead I axis:
Lead I = a × V6 + b × V2 ?
In order to establish the values of the constants a andb, it is sufficient to take standard ECG record containing signals from the corresponding leads. The values of the constants a and b can be determined by minimizing the standard deviation of the difference between the Lead I signal and the sum a×V6+b×V2.
Figure 5 shows the recordings of V6 and V2 signals, as well as the sum a×V6+b×V2 with the coefficients a = 0.23 and b = 0.08 compared against Lead I signal for a randomly selected ECG record. As appears from the figure 5, coincidence between Lead I signal and the sum a×V6+b×V2 is quite adequate for assessment of informational content and correlation between the leads.
Correlation between signals recorded from different leads does not automatically provides any additional significant information. From the graphs shown in
Figure 5, it is difficult to imagine how to extract useful information from the sum of the signals, since none of the peaks coincide in time in each of the three signals. However, not only the signal peaks carry useful information, but also their slopes. This can be easily verified by taking the first derivative of the signals (which characterizes their slope).
As shown in Figure 6, many peaks of the first derivative of ECG signals from different leads coincide in time.And it wouldn’t be difficult to see that the relationshipbetween amplitudes of these peaks is similar to the relationship between signals described above:
d1I ≈ a × d1V6 + b × d1V2,
d2I ≈ a × d1V6,
d3I ≈ b × d3V2.
Thus, the reasonings given above can be considered as the basis for identifying specific characteristics from single-lead ECG and using their informational content for fitness-related applications.
References
1. Wrist mounted heartbeat measurement system — has two electrodes on housing detecting differential ECG signals, Patent application DE2753165A1.
3. Analog Devices. AD8233. Heart Rate Monitor for Wearable Products.
5. Malmivuo J., Plonsey R. Bioelectromagnetism — Principles and Applications of Bioelectric and Biomagnetic Fields. Oxford University Press, New York, 1995.
Single lead ECG informativity for wearable application Published: Sep 2018
© 2018 RocketBody Technologies, Inc.All rights reserved
