An adaptive real-time beat detection method for continuous pressure signals.

Abstract

A novel adaptive real-time beat detection method for pressure related signals is proposed by virtue of an enhanced mean shift (EMS) algorithm. This EMS method consists of three components: spectral estimates of the heart rate, enhanced mean shift algorithm and classification logic. The Welch power spectral density method is employed to estimate the heart rate. An enhanced mean shift algorithm is then applied to improve the morphologic features of the blood pressure signals and detect the maxima of the blood pressure signals effectively. Finally, according to estimated heart rate, the classification logic is established to detect the locations of misdetections and over detections within the accepted heart rate limits. The parameters of the algorithm are adaptively tuned for ensuring its robustness in various heart rate conditions. The performance of the EMS method is validated with expert annotations of two standard databases and a non-invasive dataset. The results from this method show that the sensitivity (Se) and positive predictivity (+P) are significantly improved (i.e., Se>99.45%, +P>98.28%, and p value 0.0474) by comparison with the existing scheme from the previously published literature.