Cardiotocogram Data Generation using CT-GANs for Enhancement of Performance metrics used for Classification Models

\nAutonomous Fetal distress classification is crucial to monitor the Cardiotocogram (CTG) signals for fetal well-being and ensure timely medical intervention during pregnancy. However, the imbalanced nature of the existing CTG datasets impose significant challenges in developing autonomous models. To address this issue, the study provides a novel approach in generating synthetic CTG data using Conditional Tabular Generative Adversarial Networks (CT-GANs). By augmenting the existing CTG dataset with high- quality synthetic samples from GANs, the performance metrics are enhanced with this approach. The obtained results of the synthetic CTG data classification with algorithms namely k-NN and Random Forest with and without Grey Wolf optimisation are compared with the same classification models of Imbalanced CTG data. Hence the approach of Random Forest model with Greywolf optimization for CT-GANs generated balanced data outperforms most widely used techniques with an Overall accuracy of 94.49%, mean sensitivity of 95.23%, mean specificity of 97.7%, mean precision of 95.31%, mean MCC of 91.68% mean Kappa score of 91.63%, Weighted F1-score of 94.50% and averaged AUC-ROC of 99.40% with Random Forest algorithm for multiclass classification of CTG synthetic data.\n