Eigen-vector based leads for reconstruction of the 12-lead electrocardiogram

Abstract

In this study we have developed a limited lead system to allow reconstruction of the 12-lead ECG. This has been based upon the analysis of eigen-vectors calculated from body surface potential map (BSPM) data. Eigenvectors were calculated from a set of 117 lead BSPMs (normal=172, LVH=178, MI=209). The extrema of the first three eigen-vectors were used to determine the positions of recording sites for three bipolar leads. These three bipolar leads were then used to reconstruct 12-lead ECGs recorded from an unseen group of subjects (normal = 57, LVH = 59, MI = 69). It was found that the suggested leads could reconstruct the 12-lead ECG with a median RMS error of 57.0 muV and a correlation coefficient of 0.968. This was in comparison to the EASI leads, which, when tested on the same data, produced a median RMS error of 60.1 muV and a correlation coefficient of 0.955. Following comparison on a lead-by-lead basis the eigen-vector based leads were found to perform particularly well in the reconstruction of the precordial leads V1-V5. Further work is required to refine and to determine the practical utility of the proposed leads in practice.