Practical example and current challenges
- From May 2020, the European legislator requires manufacturers of medical technology to provide performance and safety data about their products over the entire product lifecycle. The data is collected in clinics where the products are used. AI methods can help with the automated extraction and analysis of this data.
- A continuous data collection can benefit multiple stakeholders in the healthcare sector: manufacturers, doctors, health insurance companies, market surveillance authorities, and research institutes working in the field of medicine and supply research.
- A large portion of the clinical data collected is currently not accessible for statistical or scientific purposes due to different formats being used and the unstructured nature of the data. This means that no medical insights can be gained from this data.
- Anyone using healthcare data needs to comply with a host of rules, standards and statutes under national and international law. For this reason, ambitious technical measures are required that sometimes conflict with the goal of userfriendliness.
- The sector is highly reluctant to entrust commercial cloud-service providers operating closed-source systems with healthcare data.
What added value does the "GAIA-X project" offer?
- The project can form the basis for a reliable system for the use of healthcare data and for the technical implementation of international legal, data protection, and cyber security requirements. It provides storage and computing capacity in a safe environment, allowing hospitals and SMEs operating in sensitive areas to benefit from the economies of scale that result from cloud-based services.
- The project could include functionalities for standardised data anonymisation and pseudonymisation and for the classification of data to ensure that it can be legally used by different groups of users, e.g. for training AI models (secondary use).
- The project could make standardised interfaces available to allow different user groups and healthcare providers to access a trusted cloud environment.
- This would mean that bilateral, individual projects could turn into interconnected solutions developed by different project partners – solutions that make it possible to use the data, deploy AI, and create synergies.
- Modular, distributed solutions make it possible to separate data processing and data hosting services. Within a networked structure, anonymised data can be pooled together where it is needed for analysis. This makes it possible for the most sensitive data to be stored locally, e.g. in the hospital itself, whereas other data may be exchanged for processing and analysis.
- Frank Trautwein – Raylytic, for the ‘Artificial Intelligence for Clinical Studies’ consortium (KIKS), which is funded under the innovation competition ‘Artificial intelligence as a driver of ecosystems that are relevant for the economy’ launched by the Federal Ministry for Economic Affairs and Energy.