Reduction in the Number of Electrodes for
Body Surface Potential Map (BSPM) by
Classifying BSPM in the Direction of Vectorcardiogram

Yasuaki Teramachi^a, Etsuzo Terauchi^a, Kazuhiro Tsuchimoto^a, Yoshiwo Okamoto^b, Masafumi Kasama^c, Takeshi Tsutsumi^c, and Youichi Takeyama^c

^aDepartment of Computer Engineering, University of Industrial Technology, Kanagawa, Japan
^bDepartment of Electrical Engineering, Chiba Institute of Technology, Chiba, Japan
^cShowa University Fujigaoka Hospital, Kanagawa, Japan

Correspondence: Y Teramachi, Department of Computer Engineering, University of Industrial Technology, 4-1-1, Hashimotodai, Sagamihara, Kanagawa, Japan. E-mail: teramati@cs.uitec.ac.jp, phone +42 763 9174, fax +42 763 9186

1.    Introduction

BSPM did not spread so well because it required much time and effort to install many electrodes and equipments became expensive. Lux et al. proposed a limited lead system with 30 electrode sites [Lux et al., 1978]. This method is based on the fact that there is some correlation between potentials on different electrode sites. Potentials on not selected electrodes are estimated from potentials on selected electrodes by means of a transfer matrix. The transfer matrix can be estimated from a number of full BSPMs in a statistical way. This method did not assume any prior cardiological knowledge. Okamoto et al. proposed a new method to reduce the number of electrodes by classifying BSPMs into several types following after the direction of vectorcardiogram. Different transfer matrices are prepared depending on the direction of vectorcardiogram. Simple example is depicted in Fig.1 as in the case of eight partitionings. For each quadrant, one transfer matrix is prepared. Eight transfer matrices are different wether the direction of the heart vector belongs to which quadrant. It became clear that it is possible to reduce the number of electrodes compared with Lux’s method by this method. But it required huge amount of computing time, and further possibilities of this method were not examined yet. Then re-evaluation of this method was carried out.

2.    Material

Our BSPM system has 128 inputs. Total of sixty sample BSPMs are used for calculation, ten normals, ten complete right bundle branch blocks (CRBBBs), ten Complete Left BBBs, ten myocardial infarctions, ten inferior myocardial infarctions, and ten Wolf Parkinson White syndromes. Method of calculations is omitted here.

3.    Results

Dependence of Root mean square (RMS) errors is depicted in Fig. 2, number of partitionings as a parameter.

4.    Discussion and Conclusions

It was found that RMS error decreased with increasing the number of partitioning. Not only RMS error but also singular error evaluation has carried out. It also supports the usefulness of this method. But total number of samples may be not sufficient for this method, further detailed evaluation will be necessary.

	Figure 1. Space partinging

 
Figure 2. RMS error dependency of the number of electrodes, number of partioning as a parameter.

References

Lux RL, Smith CR, Wyatt RF, Abildskov JA. Limited lead selection for estimation of body surface potential maps in electrocardioraphy. IEEE Transactions on Biomedical Engineering, BME-25(3), 270-276,1978.

Okamoto Y, Musha T, Harumi K. Reduction in the number of electrodes required for the measurement of body surface potential distribution. Japanese Journal of Medical Electronics and Biological Engineering, 25(4), 269-276,1987. ( in Japanese)