Data Mining for the Prediction of Fetal Malformation Through Cardiotocography Data

Abstract

Despite advances in technology and health, the number of maternal and fetal deaths during and after pregnancy and childbirth remains significant. Most of these deaths could be avoided if there was prenatal care before and during pregnancy, which could assist in monitoring the fetal heart rate (FHR). Thus, medical methods have been developed for assisting fetal monitoring, such as cardiotocography (CTG). To collaborate with the methods developed, advances in the field of machine learning and computational intelligence made it possible to increase the effectiveness of classification and recognition systems and, thus, to predict possible maternal or fetal death. To this end, this paper tries to predict fetal well-being, through the classification of data resulting from fetal CTGs using two different types of classification, fetal state and morphological pattern. The classification by fetal state, using methods such as Decision Tree (DT) and k-Nearest Neighbors (kNN), presented high accuracy values, achieving values that range from 93% to 98%. However, although not expected, the classification by morphological standards also showed high accuracy values, achieving the best model a value of 93% of accuracy with the kNN. Therefore, the complementary between both classifications may guarantee success in predicting fetal well-being.