An enhanced sampling approach for Tabular Generative Adversarial Networks produced synthetic cardiovascular disease data with a higher degree of similarity to real datasets compared to CTGAN.
An enhanced sampling approach for Tabular GANs successfully generates high-quality synthetic cardiovascular disease data that closely resembles real patient data distributions, potentially improving CVD risk prediction models.
Cardiovascular diseases (CVDs) are a significant cause of global mortality; thus, early intervention of CVDs can prevent their progression and effectively lower their death rate. However, one of the main obstacles to this is the lack of high-quality datasets for CVDs, which are essential for efficiently training machine learning prediction models. The available datasets are limited in size and have missing values and a high-class imbalance. Generative adversarial networks (GANs) have been shown to have high potential in producing medical synthetic tabular data, but the heterogeneity and complex dependencies of such data present challenges. This study proposes an enhanced sampling approach that aims to improve the quality of synthetic data and their resemblance to the original data distribution. The proposed method is evaluated on a medical dataset containing demographic, clinical and laboratory data from patients with CVDs. As part of the examination process, the performance of the proposed method was evaluated by comparing it with the data generated from the CTGAN. Our results show a higher degree of similarity between the distributions of the generated data and the real datasets, demonstrating the efficacy of the proposed sampling approach. The synthetic data generated by our model closely resembles the statistical properties of real data and shows strong potential for improving data quality. By accurately capturing the patterns and features of cardiovascular disease data, our model can contribute to enhancing the CVD risk prediction model.
Alqulaity et al. (Wed,) conducted a other in Cardiovascular diseases. Enhanced sampling approach for Tabular Generative Adversarial Networks vs. CTGAN was evaluated on Similarity between the distributions of the generated data and the real datasets. An enhanced sampling approach for Tabular Generative Adversarial Networks produced synthetic cardiovascular disease data with a higher degree of similarity to real datasets compared to CTGAN.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: