An automated measure of P-wave duration from surface ECG maps

The aim of this paper is to introduce and statistically characterize an estimator of the P-wave duration. The estimator is based on the computation of the P-wave area of an average P-wave. No fiducial points are needed for the estimation of P-wave duration. The computation of the total area is performed by summing up the absolute value of the average P-wave. The absolute value is used since P-wave can be polyphasic. The estimator is not biased even when the signal is corrupted by 0-mean noise, as a corrective term is introduced. The corrective term represents the mean error due to have taken the absolute value of the signal. The error depends on the value of the original signal, thus the corrective term has to be calculated for each sample. The computation of the corrective term, under the hypothesis of Gaussian noise, is based on the computation of the residual noise after the averaging procedure. The proposed estimator is characterised in terms of confidence interval in order to be used in a prediction model of atrial fibrillation. The accuracy obtained for the estimation of P-wave duration was -3.8% +7.5%. This accuracy is lower than that obtained with previously proposed methods.