Prediction of labor using non-invasive laplacian EHG recordings

Non-invasive electrohysterogram (EHG) recordings could be used as an alternative technique for monitoring uterine dynamics. Bipolar recordings of EHG have proven to provide valuable information to predict labor. Recently it has been stated that uterine EHG bursts could also be identified in Laplacian recordings on abdominal surface. Taking into account that Laplacian potential technique permits to acquire more localized electrical activity than conventional recordings; these recordings could also be helpful for deducing uterine contraction efficiency. The aim of this paper is to examine the feasibility of Laplacian potential EHG recording for labor prediction and to compare it with monopolar recordings. To this purpose, a total of 42 EHG recordings were acquired from women of similar gestational age: 29 antepartum patients, and 13 patients in labor. Then linear and non-linear classifiers have been implemented using EHG burst parameters as input features. Experimental results show significant differences in temporal and spectral parameters in both monopolar and Laplacian potential recordings between the two groups. In addition, support vector machine based classifier achieved an accuracy of 93% for labor prediction for monopolar recordings, 92% for bipolar recordings and 91% for Laplacian potential.