The purpose of this study is to assist in the use and interpretation of intrapartum cardiotocography (CTG), as well as in the clinical management of specific CTG patterns. Unexpected complications may occur during labor, even in patients without prior evidence of risk, so maternity hospitals need to ensure the presence of trained staff, as well as appropriate facilities and equipment for an expedite delivery. Unnecessary obstetric intervention confers additional risks for the mother and newborn and the former may result from poor CTG interpretation, limited knowledge of the pathophysiology of fetal oxygenation, and inadequate clinical management.
Cardiotocography has well‐documented limitations, and it is necessary to be aware of these for safe use of the technology.
The main aspects that are prone to observer disagreement are the identification and classification of decelerations, the evaluation of variability, and the classification of tracings as suspicious and pathological. The subjectivity of observer analysis has also been demonstrated in retrospective audit of tracings, where CTG features are frequently assessed to be more abnormal in cases with known adverse neonatal outcome.
2126 fetal cardiotocograms (CTGs) were processed and the respective diagnostic features measured. The CTGs were also classified by three expert obstetricians and a consensus classification label assigned to each of them. Classification was both with respect to a fetal state (N, S, P). The following analysis asks what predictive model best encompasses the features and target data within the CTG dataset.
A successful model will be highly accurate with minimal required processing time, as well as minimal difference between the training and testing datasets by resulting type one and type two errors.