Efficient Evaluation of Fetal Wellbeing During Pregnancy Using Methods Based on Statistical Learning Principles

Recording and analysis of fetal heart rate signals is nowadays the primary biophysical method for the assessment of fetal state. Since the objective interpretation of crucial signal features is difficult, methods of automated quantitative signal evaluation are the subject of research. In the following paper we investigated some practical aspects of application of methods inspired by statistical learning theory for supporting the process of diagnosis of the fetal distress being a part of the FHR signal analysis with the computerized fetal monitoring system. Its primary objective is to increase the effectiveness of fetal screening tests during a high-risk pregnancy by introducing the automated recognition of fetal wellbeing based on the results of quantitative analysis of monitored signals. We studied the possibility of application of robust fuzzy clustering, different methods of logical interpretation of the fuzzy conditional statements as well as an ensemble learning based on bagging and boosting. The efficacy of the signal classification was evaluated using the data collected with the computerized fetal surveillance system. The learning performance was assessed with the number of correct classification and the overall quality index. The best obtained results, 88 and 87% respectively, show a high efficiency of the prediction of the risk of fetal hypoxia using the proposed new procedures.