Motivation
Modern methods of artificial intelligence (AI) offer the opportunity to fundamentally support and even revolutionize knowledge-intensive activities in many fields. While AI applications have already brought about significant changes in business and many scientific disciplines, medicine is still in its infancy when it comes to the use of AI. Medical data is highly complex and, at the same time, particularly sensitive and worthy of protection. This complicates the use of AI in development and research activities. The AI-AIM project aims to help overcome these hurdles and, through anonymization, make medical data more usable for research and industry.
Goals and approach
As part of the “AI-based anonymization in medicine” (AI-AIM) project, the team is developing an anonymization platform for providing large amounts of realistic data. This data does not contain any personal references and can therefore be used more easily in terms of data protection law. The aim is to facilitate access to data for research and commercial development of data-based medical solutions. Important research areas addressed by the project include methods for flexibly combining anonymization techniques for sensitive patient data and synthesis methods for generating data sets. The developers are also focusing on the potential transferability to various medical fields. In addition, the researchers assess privacy risks using innovative models and evaluate the realism of the output data for anonymized and synthesized data.
Innovations and perspectives
As part of “AI-AIM,” innovative methods for anonymizing personal data are being developed and evaluated in light of residual risks using a specific use case from oncology. The availability of verifiably anonymous data will facilitate both AI research and the development of AI-based commercial products for medical institutions. This will open up further innovation opportunities in the field of AI applications. By making the developed platform available as open-source software and establishing a user community, it will be possible to create a locational advantage for Germany and Europe in the medium to long term through improved data availability.
