Protect and Extend - Using GANs for Synthetic Data Generation of Time-Series Medical Records
Published in 15th International Conference on Quality of Multimedia Experience (QoMEX 2023), Ghent, Belgium, 2023
Ashrafi, N., Schmitt, V., Spang, R. P., Möller, S. & Voigt-Antons, J.-N.
We introduce a GAN-based pipeline for generating realistic, privacy-preserving medical time-series to support model development when access to sensitive data is limited. Using fidelity, utility, and privacy metrics, we demonstrate that synthetic records can retain clinically meaningful patterns while mitigating re-identification risk. We discuss deployment considerations and ethical safeguards.
Recommended citation: Ashrafi, N., Schmitt, V., Spang, R. P., Möller, S. & Voigt-Antons, J.-N. (2023, June). Protect and Extend - Using GANs for Synthetic Data Generation of Time-Series Medical Records. Paper presented at the 15th International Conference on Quality of Multimedia Experience (QoMEX 2023), Ghent, Belgium. Diversity and Societal Impact Award. https://doi.org/10.1109/QoMEX58391.2023.10178496