Abstract
Gastric bio-electric slow-waves are in part responsible for generating motility. Extracellular recordings of the slow-wave activation phase have yielded significant physiological insight about its spatio-temporal characteristics. With the growth of multi-electrode data and long recording periods, there is a need for automated methods to detect the activation phase in a reliable and accurate manner. In this study, the Variable Threshold Wavelet (VTW) algorithm was developed, featuring wavelet decomposition, to compute the derivative to detect the slow-wave activation phase of monophasic signals. The performance of the VTW algorithm was compared against an existing Falling-Edge, Variable Threshold (FEVT) algorithm. Varying levels of synthetic noise representing ventilator and high-frequency noise were added to in vivo slow-wave recordings. Sensitivity, positive-predictive value, area under the curve (A(roc)) metric and percentage improvement metric (PIM) of activation phase identification accuracy were calculated. Compared to the existing FEVT algorithm, the VTW algorithm achieved similar performance in identifying the activation phase of slow-waves with various levels of ventilator noise. In the presence of high-frequency noise, the VTW algorithm improved the A(roc) of the existing FEVT algorithm by 11.1%. The VTW algorithm can now be applied to analyze normal and abnormal slow-wave recordings.