Mattias Ohlsson, Bo Hedén and Lars Edenbrandt
High specificity - a necessity for automated detection of lead reversals in the 12-lead ECG


Abstract:
Recently a new rule based criteria for the detection of LA/LF lead reversals was presented. The authors reported a sensitivity of 90%, higher than that of the neural network, based on a data set of 70 recordings. However, the specificity level, which was very high using artificial neural networks, was not studied and discussed. The purpose of this study was to implement and test these criteria on our large database of ECG recordings in order to obtain both sensitivity and specificity levels. It is shown that the new criteria has a very low specificity and that artificial neural networks performs significantly better at this low specificity.

LU TP 99-24