Analyzing Current Fetus Risk Conditions Through Fetal Heart Rate (FHR) and Uterine Constructions (UC) Values by Using Machine Learning Algorithms

Fetal Heart Rate (FHR) is often used to assess situation of fetal health. The aim of this thesis is to infer current fetus risk conditions by monitoring and computing FHR and Uterine Constructions (UC) values. Doppler devices are generally used for gathering FHR and UC values from labors instantly. Doppler devices produce Non-Stress Test (NST) graphs. In this thesis, we used the CTU-UHB intrapartum Cardiotocography database (CTU) which is gathered by Prague and Brno University Hospital (UHB). The database contains 552 Cardiotocography records (CTG) and each record involves a FHR and an UC signal. In this thesis, several machine learning algorithms are developed to perform feature extractions and classification to analyze NST graphs. MATLAB and R tools are used for signal processing, feature extraction and classification steps. With the proposed system, instant interpretation of FHR and UC signals on a NST records: (1) value of instant baseline, (2) variable baseline signal, (3) baseline variability, (4) type and number of accelerations, (5) type and number of decelerations, (6) classification of the NST traces. In experimental studies of this thesis, CTU-UHB Cardiotocography records are interpreted by an expert obstetrician. Experimental results are evaluated and compared with expert obstetrician’s observations and some of related works.