Fetal Health State Monitoring Using Decision Tree Classifier from Cardiotocography Measurements

Perinatal Mortality is an alarming issue in the world which needs immediate attention for the sustainable growth of nation. To address this issue, both intrapartum and antepartum fetal health state monitoring is essential. Fetal and maternal risk can be assessed by monitoring the fetal heart rate. Cardiotocography records the fetal heart rate and Uterine contractions of mother simultaneously. This signal data can support the physicians to understand the fetal health status. In this paper, an efficient method to predict the high-risk pregnancy based on the fetal health status using CART is proposed. A CTG datasets having 2126 recordings from UCI machine learning repository is used for classification. 5- fold cross validation is done, and the proposed methodology using CART was quantified using precision, recall and F-score.

I. Introduction

Safe pregnancy with normal delivery, physically and mentally sound baby is the yearning of almost all mothers. Obstetrics complications during labor and delivery adversely affect the fetal and mother wellbeing. So perinatal care is of utmost importance for determining the appropriate delivery method, preparing the mother psychologically and physically. Cardiotocography (CTG) is the most popularly used diagnostic tool to determine the fetal distress during antepartum and intrapartum stage. Anabolic Acidosis and Hypoxic injury can be found from these readings. There are four fundamental and crucial factors in the CTG data such as Fetal Heart rate baseline (BL), Accelerations (ACC), Decelerations (DCL) and Variability. Using the abovementioned constraints, doctors classify the fetal state as normal, suspicious and pathological. In this work, a CART algorithm is used for the prediction problem. The paper is structured as follows. Section II throws light on the humble decision tree algorithm and section III provides the description of the CTG datasets, section IV contains the proposed work and the results are tabulated in section V.