MediSynth

Most healthcare data remains unused due to privacy restrictions, even though it could greatly enhance patient care and research. Synthetic data, generated from original data, preserves patterns and characteristics without including sensitive information, enabling secure data sharing and improving dataset quality, especially for scarce or imbalanced cases. This project will focus on building a Multi Modal Synthetic Data Ecosystem (MM SDE). Within the ecosystem, new methodologies to generate realistic, complex synthetic data, will be proposed, as well as a coherent framework to evaluate the synthetic data (privacy, utility, and biases), and to facilitate regulatory compliance of synthetic data applications. The focus of the project will be on synthetic time series, text, audio, and multi modal data—areas with still insufficiently established techniques, and that have relevant applications within the healthcare sector. The collaboration with end users will allow to iteratively validate the developed techniques, and to tailor them to the specific needs of the healthcare domain.