Investigation and Classification of Cardiotocography Data using Correlation Coefficient and PCA based Features

In this work, an attempt is made to classify the Cardiotocography data using correlation coefficient and PCA based features. Cardiotocography (CTG) is the monitoring tool to assess the foetal status. It is potential tool used to record changes in fetal heart rate (FHR) and Uterine Contractions (UC). To improve the diagnostic accuracy an automated analysis is essential. Cardiotocography records and monitors 23 vital features continuously during the intrapartum period. Cardiotocography data is obtained from public online database with 3 classes such as normal, suspect and pathology. To detect the foetal distress condition accurately an automated analysis with significant features are essential. As a pre-processing step of automated analysis, correlation coefficient and descriptive statistical features are extracted from CTG data. Results showed among the normal and pathology group only 6 out of 23 features are observed to be significant. Similarly, among the normal and suspect group 4 out of 23 features are observed to be significant. The feature extracted from PCA showed 6 significant features among normal, suspect and pathology. They classified normal and pathology more accurately compared to normal and suspect. Hence PCA based features shall be considered with more number of samples as one of the efficient pre-processing features to classify the normal and abnormal data.