Symbolic AI to Improve Certification of Medical Software
News of August 2, 2022
Symbolic AI to Improve Certification of Medical Software
While software-based medical products promise advances over traditional medical products, their certification requires a complicated and time-consuming, extensive documentation of safety claims. Such a proof is even harder to obtain for complex and highly interconnected software. In addition, the software development itself is electronically decoupled from the creation of the safety documentation and verification. Oftentimes, documents are created manually and, therefore, need to be reviewed manually by regulatory authorities. This is true for the certification as well as for post-market surveillance (where problems with the product are reported).
The goal of the KIMEDS partners is to make the certification of medical software more efficient and more transparent. Beyond ICCL, the Federal Institute for Drugs and Medical Devices (BfArM), the Else Kröner Fresenius Center for Digital Health (EKFZ), iSAX, Siemens Healthineers, and TÜV Süd are working together towards that goal. An initial solution shall be developed that can be used as the basis for an international standard. Symbolic methods help in checking complex specifications for completeness and consistency. The creation of the safety documentation starts with the software development and will be supported by developer tools. Regulatory affairs specialists create requirements that are tackled during architecture design and software development. Due to the early integration, regulatory requirements are continuously checked and validated.
The BMBF is funding KIMEDS through the "Medical Technology Solutions for Digital Healthcare" programme with 1.46 million euros over 36 months. The overarching goal is to provide best possible medical care for patients.
