Automatic Evaluation of FHR Recordings from CTU-UHB CTG Database

Fetal heart rate (FHR) provides information about fetal well-being during labor. The FHR is usually the sole direct information channel from the fetus – undergoing the stress of labor – to the clinician who tries to detect possible ongoing hypoxia. For this paper, new CTU-UHB CTG database was used to compute more than 50 features. Features came from different domains ranging from classical morphological features based on FIGO guidelines to frequency-domain and non-linear features. Features were selected using the RELIEF (RELevance In Estimating Features) technique, and classified after applying Synthetic Minority Oversampling Technique (SMOTE) to the pathological class of the data. Nearest mean classifier with adaboost was used to obtain the final results. In results section besides the direct outcome of classification the top ten ranked features are presented.