Outliers Detection and Processing in CTG Monitoring

Cardiotocography is a technique used to assess the foetal wellbeing using the recording of foetal heart rate and uterine contractions. Generally these signals are observed to get information about the foetal development and wellbeing and it represents the only medical report that has a legal value to testify the foetal health. Unfortunately, really accurate prediction of foetal wellbeing is still a goal quite difficult to reach. Some problems are related to the visual interpretation of cardiotocographic traces, others to the acquisition system that, in some conditions, can cause the degradation or the loss of the signal. These anomalies, cardiac arrhythmia and artifacts in foetal heart rate signals do not represent physiological variations, so that they can be defined outliers. In cardiotocography literature the problem of outliers processing is underestimated even if they affect both time and frequency analysis. When faced, the outlier problem is solved with a pre-processing phase that uses algorithms to detect and remove spikes.

In this work, we firstly present in detail the updated version of an algorithm for detection and correction of outliers. Then, we evaluate the impact of outliers and of different correction strategies on foetal heart rate analysis. Obtained results demonstrate that the proper outliers detection is fundamental because they heavily influence the estimation of foetal heart rate variability, calculated by the short term variability, a parameter well known for its diagnostic value.