Level prediction of preterm birth using risk factor analysis and electrohysterogram signal classification

Abstract

As per the reports published by World Health Organization (WHO) in November, 2012, every year more than 15 million babies are born preterm and this number is rising [1]. Preterm labor is the major cause of neonatal deaths. Every year, pre term birth (PTB) complications leads to the death of almost 1 million babies [2][3]. Predicting the preterm labor well in advance can reduce the neonatal death considerably. There are some commonly attributed risk factors associated with preterm birth [4][5]. 33% of the women who deliver their babies prematurely have one or more of these risk factors. We propose to predict PTB by analyzing the historical data of patients who had one or more of the above risk factors. In addition to this, historic data of the patients who did not have any of the above risk factors but had PTB is also analyzed. Electrohysterogram (EHG) is the most commonly used clinical procedure which can reveal few indicators of preterm labor [6]. We analyze the EHG signals to predict the pre term labor by applying Feature Extraction coupled with semi-supervised learning (SSL). Predicting the preterm labor helps the health care professionals to make decisions about the treatment [7]. Hence the expectant mother undergoes minimal or no complications of preterm labor. On the other hand it also helps to avoid unnecessary hospitalization and treatment for women who are having a false labor pain.