Extended multiple linear regression in the derivation of electrocardiographic leads

In this study we investigate the performance of an approach for deriving electrocardiographic leads with the aim of improving derivation accuracy. We focus our attention on a limited lead system that uses leads I, II, V2 and V5 to derive the remaining precordial leads. Our extended multiple linear regression based lead transformation (EMLRLT) approach extends the standard multiple linear regression based lead transformation (MLRLT) approach by combining the data from the recorded leads with quadratic and cross product terms from the same leads. It was found that all missing leads were more accurately derived using an EMLRLT approach in comparison with the MIRIT approach. Using the standard MLRLT approach, the median RMSEs for the QRST were found to be 44.2 μV, 42.7 μV, 40.3 μV and 19.3 μVfor leads V1, V3, V4 and V6, respectively. Using the EMLRLT approach, the median RMSEs for the QRST were found to be 28.2 μV, 29.3 μV, 25.1 μVand 13.4 μV for leads V1, V3, V4 and V6, respectively. According to the sign test, all differences were statistically significant with p <; 0.05. In conclusion, it has been shown that alternative methods for lead transformation have the potential to improve derivation accuracy.