Generating Synthetic Fetal Cardiotocography Data with Conditional Generative Adversarial Networks

Ahmed, Halal AbdulrahmanNepomuceno Chamorro, Juan AntonioVega Márquez, BelénNepomuceno Chamorro, Isabel de los Ángeles2025-01-162025-01-162023Ahmed, H.A., Nepomuceno Chamorro, J.A., Vega Márquez, B. y Nepomuceno Chamorro, I.d.l.Á. (2023). Generating Synthetic Fetal Cardiotocography Data with Conditional Generative Adversarial Networks. En 18th International Conference on Soft Computing Models in Industrial and Environmental Applications (SOCO 2023). Volum. 2. Lecture Notes in Networks and Systems (111-120), Salamanca, España: Springer.978-3-031-42535-6978-3-031-42536-32367-33702367-3389https://hdl.handle.net/11441/166797In recent years, the use of machine learning models has become increasingly common, and the availability of large datasets is an essential for achieving good predictive model performance. However, acquiring medical datasets can be a challenging and expensive task. To address this issue, generating synthetic data has emerge as a viable alternative. This paper proposes using a Conditional Generative Adversarial Network (CGAN) to generate synthetic data for predicting fetal health diagnosis from a publicly available Fetal Cardiotocography (CTG) dataset. The study also evaluates the efficacy of the Generative Adversarial Network (GAN), specifically Conditional GAN, in the clinical problem. We analyzed 2126 fetal cardiotocogram samples that were labeled by medical doctors. We used CGAN-generated data together with Support Vector Machines (SVM) and Extreme Gradient Boosting (XGBoost) as classifiers to show the performance of classifiers using the real and the synthetic dataset. The experiment results revealed that the synthetic dataset behave similarly to real data in terms of classifier performance.application/pdf10engAttribution-NonCommercial-NoDerivatives 4.0 Internationalhttp://creativecommons.org/licenses/by-nc-nd/4.0/Synthetic dataConditional Generative Adversarial Network (CGAN)Fetal Cardiotocography (CTG) datasetGenerating Synthetic Fetal Cardiotocography Data with Conditional Generative Adversarial Networksinfo:eu-repo/semantics/conferenceObjectinfo:eu-repo/semantics/openAccesshttps://doi.org/10.1007/978-3-031-42536-3_11