Development of an Open, Synthetic Medical Record Dataset in Swedish for AI Advancement in Healthcare

Background:The increasing demand for effective healthcare solutions highlights the need for innovative technologies to support healthcare professionals and improve patient outcomes.

However, the development of artificial intelligence (AI) applications in Swedish healthcare faces significant challenges due to the lack of accessible data, particularly when it comes to sensitive data, such as electronic health records (EHRs).Purpose:The purpose of this project is to create an open, synthetic, and realistic dataset of Swedish medical records for fictional patients, but based on real ones.

This dataset can primarily be used for training AI models to understand and process the language of Swedish EHRs, thereby accelerating the development of AI applications in healthcare.Research Questions:The project aims to evaluate whether the generated synthetic EHRs are linguistically similar to real EHRs, whether they are realistic enough to be indistinguishable from real records by physicians, and whether they are sufficiently de-identified to ensure patient privacy.Method:The project will involve developing a process for generating synthetic EHRs inspired by real patient data, ensuring that the synthetic records are linguistically representative while protecting patient privacy.

The algorithm will be initially tested and refined using mock data before being applied to real EHRs.

The synthetic EHRs generated will undergo rigorous testing to ensure they meet the required standards for linguistic similarity and de-identification.

This evaluation will include the use of advanced machine learning and natural language processing techniques as well as human evaluation.Expected Result:The project is expected to produce a dataset of 2,000 synthetic EHRs that are both realistic and secure, with minimal risk of re-identification. This dataset will be made available to researchers and developers through publication in a public repository.

It is anticipated that this resource will accelerate the research and development of AI for EHR processing in Swedish healthcare.