An Optimized Statistical Channel Minimization Framework for Automated Preterm and Term Labor Detection

Regular and continuous monitoring of uterine con-tractions aid in the diagnosis of critical pregnancies. Maternal and fetal complications can be timely intervened by obstetri-cians with uterine activity monitoring at homes. The aim of this paper is to reduce the number of channels for Uterine Magnetomyography (MMG) recording in order to facilitate its use in home uterine activity monitoring (HUAM). In this work, automatic channel minimization is presented by integrating Point Biserial Correlation with a comparative Merit based strategy and Local Channel Inclusion. An additional scheme of Channel Repeatability has been introduced in order to further reduce the channel set. As a result, the minimized array of 14 achieves a superior performance in comparison to the existing methods. Using Support Vector Machine (SVM) a classification accuracy, specificity and sensitivity of 78.33%, 85% and 75%, respectively, has been achieved.