The project will be carried out by 16 partner organizations from seven European countries. The project is coordinated by the German Federal Institute for Physics and Metrology (PTB), and the other partners are the metrology institutes of France, Portugal, the Czech Republic, and Bosnia and Herzegovina, as well as six universities in Belgium, Germany, and Lithuania (KTU), three universities in the United Kingdom, and one company.
Photoplethysmography (PPG) sensors are already widely used in smart wearable devices, wristbands, and watches because of their user-friendliness, simplicity, and low cost. Until now, the usage of PPG has been mostly limited to heart rate monitoring, but these signals have the potential to provide much more information about human physiology and health than is currently available. Researchers hope to use machine learning models, which belong to the class of artificial intelligence techniques, to extract much more information about cardiac arrhythmias, vascular conditions, acute pain levels, and even psychophysiological conditions such as psychological stress and depression. To ensure that clinicians have confidence in the information extracted from PPG signals, it is necessary to specify the level of reliability, which is opposite to uncertainty, of the new information. This project has ambitious objectives: (i) to develop reference PPG datasets to test machine learning algorithms; (ii) to select or develop machine learning algorithms for selected applications of PPG signals and to develop methodologies suitable for assessing the uncertainty of machine learning algorithms; and (iii) to develop and publish a good practice guide with software tools that will allow medical wearable device manufacturers to assess the uncertainty of the PPG signal analysis models they develop.
BMII has more than ten years of experience in the development of PPG sensors, signal processing algorithms, and PPG signal databases, and will contribute to the selection of three important PPG signal use cases, the development of reference PPG signal sets, the development of machine learning algorithms for the selected PPG signal applications, and the development of a good practice guide and a software system for the independent assessment of the robustness of the PPG signal analysis models. The BMII will contribute to six project reports, lead the preparation of one report, and produce three scientific publications and one presentation for an international conference.
The project WEB site can be found at https://www.qumphy.ptb.de/
The project’s first meeting in Berlin on 11 July was attended by the project leader, BMII Director Vaidotas Marozas.
