Deriving the 12-lead Electrocardiogram
From four Standard Leads Using information
redundancy in the 12-lead system

Daming Wei
Graduate School of Information System, The University of Aizu
Ikki-machi, Aizu-Wakamatsu City, Fukushima 965-8580, Japan

Abstract: We previously proposed a lead method for monitoring the 12-lead electrocardiogram (ECG) with four standard leads. We also reported the implementation of deriving 12-lead ECG using lead vectors based on the Frank image surface. In this paper, we present an alternative approach that derives the 12-lead ECG using information redundancy in the 12-lead system. Performances of the two approaches are compared. Correlation coefficient between the original and derived waveforms is used as a measure of similarity. The potential difference in the middle ST segment between the original and derived waveforms is measured with an automatic ECG interpretation program for a quantitative evaluation. The results showed that our new approach significantly improved the accuracy of derived ECGs.

INTRODUCTION

The 12-lead electrocardiogram (ECG) is the gold standard for the diagnosis of myocardial ischemia and acute infarction that are main targets in clinical ECG monitoring. However, most current cardiac monitors do not provide the 12-lead ECG capacity because mounting and maintaining ten electrodes (among these four are placed on the extremities of the patient) are impractical for ambulatory monitoring and long-term bedside monitoring.

For monitoring the 12-lead ECG, we have proposed a lead method that derived the 12-lead ECG from four lead, and reported the implementation in deriving the 12-lead ECG using lead vectors based on the Frank image surface [1]. In the evaluation study, we found that the Frank torso model was not sufficiently satisfactory in deriving 12-lead ECG for many reasons. In this paper, we report a new approach to derive the 12-lead ECG using information redundancy in the 12-lead ECG system.

METHODS

The proposed lead system for derived 12-lead ECG

The electrode placement proposed by Wei [1] is shown in Fig. 1. The proposed lead method totally uses six electrodes to record four-channel ECG signals. The electrodes of RA, LA, and LL in the M-L lead system [2] are used to obtain lead I and II, from which all limb leads, as well as the Wilson terminal, can be correctly derived. In addition to leads I and II, two chest leads, V1 and V6 are recorded. Based on leads I, II, V1, and V6, other four chest leads, V2 through V5 are derived using relationship among the standard leads as described below. Besides, a sixth electrode, RL, is used for grounding.

Figure 1. The lead system for deriving 12-lead ECG.

The electrode configuration is sufficiently convenient and relatively comfortable for ambulatory or long-term bedside monitoring. In addition, it provides sufficient information in three spatial dimensions for deriving ECGs of the unrecorded leads.

Deriving 12-lead ECG from four standard leads

For deriving the potentials on an unrecorded lead i (i refers to leads V2 through V5), we suppose the potentials v_i is a linear combination of those on the recorded leads, as

                              (1)

where j refers to leads I, II, V1, and V6. This assumption is based on the fact that the potentials among the 12-lead ECG system are not independent. According to the fixed dipole theory, three independent variables can uniformly determine v_i. However, considering the independency among lead I, II, V1, and V6 may not be sufficient, we used four variables in (1). In certain sense, in (1) is equivalent to the lead vector in the lead theory, and should be a constant. Similar to the lead vectors, values of have to be determined in advance in order to use (1) for deriving the 12-lead ECG. To obtained average values of that apply to any subjects as parameters of an average model, an ECG database was developed, which consisted of ECGs recorded from 113 normal young adults. For each of these subjects, the 12-lead ECG with the electrode positioned in the ML lead configuration as shown in Fig. 1 was recorded. Half of the data was used to develop the values of and another half is used for the evaluation of derived ECGs.

Values of  were solved with (2) in the sense of the least square errors, expressed in matrix notation as

                      (2)

where , , , i referred to leads V2, V3, V4 and V5, j referred to leads I, II, V1 and V6, and s referred to data samples. For each individual’s ECG, typical data samples in P, QRS and T waves were selected to yield an overall good fitting to the original values. A special priority was given to a good fitness in the ST segment.

RESULTS

Table 1 shows the correlation coefficients between the original ECGs and ECGs derived based on the present approach (new) and Frank image surface (Frank). While the derived ECGs by both approaches show high similarity to the original ECGs, the present approach gives higher values. Fig. 2 shows an example of original (a) and derived (b) ECG waveforms using the present approach.

Table 1

Because the derived ECG was mainly aimed at ST monitoring, the ST changes (depression or elevation) were quantitatively measured using an automatic ECG interpretation program (ECAPS, Nihon Kohden Corp.). The potential difference more than 0.05mv in the middle ST segment was considered as significant change. Table2 shows the percentages of evaluated ECGs, where the potentials difference in the middle ST segment between the original and derived ECGs is less than 0.05mv. It shows that the present approach significantly improves the accuracy of derived ECGs.

Table 2

DISCUSSION

Recently, physicians have shown increasing interest in the derived 12-ECG because of the need in ST monitoring. The lead method we proposed uses four standard lead to derive the 12-lead ECG. To the EASI lead [2], the advantage of our method is that it provides both measured waveforms used as the main information and derived waveforms used as the secondary information. The present study shows that the accuracy of derived ECGs with our new approach is very close to the requirement in practical use.

(a)

(b)

Figure 2. The measured (a) and derived (b) ECGs.

REFERENCES

[1]        D. Wei: “Deriving the 12-lead Electrocardiogram From Four Standard Leads Based on the Frank Torso Model”, in Proceedings of 23rd Annual International Conference of the IEEE EMBS, 2001, CD-ROM.

[2]        B. J. Drew, M. G. Adams, M. M. Petter, et al, “Comparison of Standard and derived 12-lead electrocardiograms for diagnosis of coronary angioplasty-induced myocardial ischemia,” Am J Cardiol, Vol. 79:639-44, 1997.