Synthetic health data: ethical development and deployment via deep learning approaches

There is a large potential for using artificial intelligence (AI) for improving diagnosis and treatment of many diseases. Large amounts of clinical data are needed to train useful AI systems. However, patient data is person-sensitive and only select individuals can obtain access, which can be a huge roadblock for researchers and students. The aim of this project is to develop methods for making synthetic data, which looks like the real, but with no person-sensitive information. This data can be used to develop AI models with no danger of leaking sensitive information, and the successful models can later be trained on the real data after proper authorization. The challenges in the project are to generate synthetic data that is realistic and at the same time ensure that there is no way to reconstruct the original sensitive data. Therefore, we have assembled researchers that are experts in AI, statistics, and all the legal aspects of health data.