Complex approach to fetal heart rate analysis: A hierarchical classification model

The electronic fetal monitoring (EFM) is used for fetal behaviour surveillance via measurement of fetal heart rate. In the course of labour, fetus can suffer by severe hypoxic insults that might lead to possible adverse long term consequences. The main goal of EFM is to provide indirect information about fetal well-being and help obstetricians to indicate timely intervention to prevent adverse consequences. Nowadays, the EFM is an integral part of every day obstetrics practice. The EFM most commonly refers to cardiotocography (CTG) that is a measurement of fetal heart rate and uterine contractions. The reliability, validity, and efficiency of CTG have neither been confirmed nor disproved. Also all the computerized systems have yet to prove its efficiency. The reasons for stagnating or even non-existent technical progress are: i) use of very small and ad-hoc created databases, ii) neglect of the high intra/inter observer variability of clinical evaluation of CTG, iii) unclear definition of pathological labour outcome (usually the outcome is imprecisely defined by a pH value), iv) strict technical approach disconnected from the clinical reality. Therefore, in this thesis, we introduce the first open access database of intrapartum CTG. The database that enables other researchers to developed and test new algorithms for CTG analysis and classification. We show that it is possible to overcome the high inter observer variability using a model of clinical annotation. We also show that there is a group of pathological CTG records on which the clinicians have good agreement. Finally, we develop a novel approach for CTG evaluation using a hierarchical model. The model considers different outcome measures as a mixture of individual components. Thus is able to overcome discrepancies between biochemical markers (pH, base excess, base deficit), Apgar score, and clinical evaluation of CTG. The developed model is able to answer the difficult question, whom to trust when you are given multiple noisy and imprecise information about labour outcome.