Within the scope of digitalization, the importance of mathematical and physical simulations and computer-based experiments is increasing rapidly. If such simulations are used to imitate real measuring devices and measurements, these can be designated as a "virtual measuring instrument". Here, the task of metrology is to ensure confidence in simulation results if these are to be used in the same way as real measurements. For this reason, PTB has established the "Metrology for Virtual Measuring Instruments" (VirtMet) Competence Center. Here, it carries out interdisciplinary research and pools expertise in order to develop concrete virtual measurement procedures and to work on overarching questions, for example: How can the comparability of virtual and real measurements be established? Which standards are needed for interfaces, metadata and data formats? And how can virtual experiments for complex measurement systems be handled with machine learning methods? 

Contact

Dr. Sascha Eichstädt, FB 9.4 "Metrologie für die digitale Transformation", Phone:(030) 3481-2008, Sascha.Eichstaedt(at)ptb.de

Digital change continuously allows new products and processes to emerge that push proven quality-assurance measures to their limits. This is particularly evident in the case of complex products which dynamically change their state after they have been placed on the market. To be reliable, a product would need to be tested several times during its life cycle, sometimes continuously. As of today, there are still no practicable solutions for this. An example of this is applications of machine learning (as a type of artificial intelligence – AI) in medical devices. Although numerous innovative medical products are currently being developed with a high share of software, only a fraction of them make the leap into the healthcare market. A major reason for this is the lack of structured processes as well as the lack of a quality infrastructure for an objective, verifiable and reproducible validation of AI technologies. For the use of machine learning, PTB is working on defining objective evaluation methods to assess the quality of algorithms; together with medical experts, PTB is developing procedures for the provision of reference data; and it collaborates on norms and standards for quantifiable and verifiable data quality criteria. For example, researchers at PTB are training an artificial neural network in such a way that it is able to assess the quality of individual mammography images. A virtual mammography developed by PTB which has led to a huge database of images, each with precisely known characteristics, is particularly helpful. In this context, PTB is also involved in the development and establishment of a digital quality platform with which AI-based solutions can be validated. This platform is to provide comprehensible criteria for the determination of data quality and to ensure that the evaluation by a computer is objective and cannot be manipulated.

Departments involved

6.2 Dosimetry for Radiation Therapy and Diagnostic Radiology

8.4 Mathematical Modelling and Data Analysis