Cardiotocography Graph Classification Using Deep Learning and Ensemble Methods

Abstract

Fetal monitoring, essential for obstetric assessment and prenatal health, enables timely detection of abnormalities, reducing birth defects and mortality. Cardiotocography (CTG) is an effective method that records fetal heart rate changes and their relationship to movements and uterine contractions, allowing for the assessment of fetal condition. This paper presents a novel CTG-based classification method using deep learning and ensemble classification, utilizing eight selected characteristics from a cloud-based CTG analysis system.Key contributions include the collection and feature extraction of fetal heart rate data, deep learning model training, creation of a dataset with four reshuffled fetal data features, and ensemble classification. An empirical study using five-fold minus one cross-validation on a dataset of 21,570 records shows the proposed method achieves an average accuracy of 96.83%, compared to 94.88% from traditional methods.