Logo of the Physikalisch-Technische Bundesanstalt

PTB’s strategy for artificial intelligence in medicine

Annual Report 2021

Illustration of the complex interrelationships when using artificial intelligence in medicine. The quality aspects (left) are the prerequisite for obtaining the approval for using them in the different application areas (right).

The development of digital products and services that use artificial intelligence (AI) has greatly accelerated since the start of the digital transformation. AI techniques could be the solution to many medical challenges: For example, imaging procedures generate large volumes of data which are unmanageable for physicians whereas an algorithm can assess them within seconds. But however high the benefits for patients and the economic potential may be – as application areas of AI are growing, the need for clear rules and consideration in quality infrastructure is also rising. PTB has the legal mandate to ensure the uniformity of metrology in medical science and has therefore drawn up guidelines in a strategy paper titled “Metrology for AI in Medicine”.

PTB’s research activities are not aimed at developing new AI methods, but at new metrics and procedures to assess important quality aspects of AI methods in a quantifiable manner. These aspects include: explainability, generalizability and uncertainty. Other aspects consist in developing and providing validated reference data sets to form the basis of binding legal requirements concerning training and test data for AI systems, along with the confidentiality of personal data.