Automated Classification of CTG signals using Deep Learning based Scalogram Analysis

In this work, an attempt has been made to classify Cardiotocograph (CTG) signals using scalogram and deep learning approach. Preterm labor is the leading cause of death in prematurely born babies across the world. According to the World Health Organization (WHO), 15% of babies are born prematurely every year. Prediction of preterm labor could help in early maternal and neonatal healthcare thereby minimizing premature mortalities. In this work, the CTG signals consisting of Fetal Heart Rate (FHR) and Uterine Contraction (UC) are obtained from a publicly available database. The obtained FHR and UC signals are converted into two dimensional scalograms using Continuous Wavelet Transform (CWT). The scalogram images are labeled according to the gestational age and given as input to the deep learning network. GoogLeNet, a pre-trained Convolutional Neural Network (CNN) is used for classification of the CTG signals. The scalogram images are resized to match the input size of the GoogLeNet. The data is split into training and validation data to prevent overfitting of the deep neural network. The network is trained for 80 iterations and for each iteration, the training and validation data are split accordingly. The trained network is tested for three different gestational periods namely preterm, term and post-term labor. The proposed approach is able to differentiate gestational stages and produce classification accuracy of 88.23% for FHR signals and 87.50% for UC signals. Hence, this method could be used as a diagnostic tool for predicting preterm labor in pregnant women and provide appropriate health care services.