A doctor needs a great deal of experience, knowledge and years of training to be able to expertly analyze an ECG. Automated procedures based on AI can offer effective support here. And an AI system in this field also has to be trained. This is namely undertaken (in this case) with very large amounts of high-quality ECG data. Such training means that, for instance, even the smallest abnormalities in ECGs can be found and that reliable data can then be used to make a diagnosis. But what are the high-quality data that are needed for this? And how do we know if an AI system is reliable or not? These are the questions that PTB is going to address in the TEF Health project. Daniel Schwabe, one of PTB’s researchers, is heading a dynamic project team made up of experts from PTB, the Fraunhofer Institute for Telecommunications and the TÜV AI Lab. This team is responsible for work packages focusing on “Standards and Quality” and “Certification”. “PTB wants to use this project to significantly advance the quality assurance of data as the foundation of AI systems in medicine,” explains Dr. Schwabe. “This will mean that we can create a basis for using AI in a trustworthy way within the EU,” adds Prof. Tobias Schäffter. He is the head of PTB’s Medical Physics and Metrological Information Technology Division, who initiated the project group and is accompanying it strategically.
Credit: PTB
A doctor needs a great deal of experience, knowledge and years of training to be able to expertly analyze an ECG. Automated procedures based on AI can offer effective support here. And an AI system in this field also has to be trained. This is namely undertaken (in this case) with very large amounts of high-quality ECG data. Such training means that, for instance, even the smallest abnormalities in ECGs can be found and that reliable data can then be used to make a diagnosis. But what are the high-quality data that are needed for this? And how do we know if an AI system is reliable or not? These are the questions that PTB is going to address in the TEF Health project. Daniel Schwabe, one of PTB’s researchers, is heading a dynamic project team made up of experts from PTB, the Fraunhofer Institute for Telecommunications and the TÜV AI Lab. This team is responsible for work packages focusing on “Standards and Quality” and “Certification”. “PTB wants to use this project to significantly advance the quality assurance of data as the foundation of AI systems in medicine,” explains Dr. Schwabe. “This will mean that we can create a basis for using AI in a trustworthy way within the EU,” adds Prof. Tobias Schäffter. He is the head of PTB’s Medical Physics and Metrological Information Technology Division, who initiated the project group and is accompanying it strategically.
Providing trust in measurements – that is what PTB fundamentally stands for. And it does this by evaluating the quality of measurements and data sets. PTB will also guarantee reliability in the digital metrology of the future. It will do this by evaluating data and explaining algorithms which will then contribute to certifying artificial intelligence. PTB is working with other central players in the field of quality infrastructure in Germany (BAM, DAkkS, DIN, DKE). Together they are bringing forward a quality infrastructure that is digitally transformed, interoperable and geared to the future – and that is QI Digital. The German Federal Ministry for Economic Affairs and Climate Action (BMWK) brought the QI Digital project into being. AI in medicine is one of three central themes in the German initiative, and this means that it will cooperate closely with the European TEF Health project.
Practically speaking, digital data in medicine can never be measured and evaluated as individual values. Instead, complex systems of measurement procedures and sensors more often have to be understood and characterized. Thinking holistically and cross-contextually is crucial too. And this is also something that will be required by metrology and the testing systems that have to guarantee trust, reliability and security in the future. Just as the international prototypes of the meter and the kilogram served as the standards for metrology in the past, so should the quality standards for networked data and AI be traceable to quantities which can be measured. This research will be advanced at PTB within the scope of an innovation cluster that is currently addressing three thematic fields: Smart Mobility, the Smart City and Smart Health.
The TEF Health project
The EU’s TEF Health project takes into account the breathtaking speed of technical advancements in AI and robotics in various fields including the healthcare sector. A total of 51 institutions are involved in TEF Health (with TEF standing for Testing and Experimenting Facilities). They are going to test novel approaches to AI in environments that are close to reality. This applies to new software, such as that used when diagnosing patients, and it also applies to intelligent devices that work directly on or even inside patients – such as surgical robots, carebots or intelligent implants. Apart from this, new regulatory requirements – like standardized test protocols – will be drawn up within the scope of the project. For this reason, not only leading hospitals and medical establishments (such as the Charité in Berlin and the Karolinska Institutet in Stockholm) along with industrial research institutes are represented in this project. Regulatory bodies like TÜV and, not least, national metrology institutes such as PTB and its French counterpart LNE are on board too. The project is led by Prof. Dr. Petra Ritter of the Charité – Universitätsmedizin Berlin. One important long-term goal of TEF Health is to establish sustainable collaborations between the economy, academic research and other players. The newly created evaluation resources and infrastructures will be available to industry in the form of services in the future. This will make it possible for us to employ AI in healthcare for the whole of Europe, and this AI will be trustworthy and secure as well as having been tested.
es/ptb
Subject of Research
Not applicable