For science, medical data are very valuable. To acquire insights into diseases and to further develop medical technologies, large amounts of data must be recorded and analysed. Alongside their considerable value for society, however, the collection of such data also harbours certain risks. Especially biosignals can enable conclusions to be drawn about individual persons through what is known as “re-identification analysis” and disclose potentially sensitive information about them. For this reason, the use of personal data is strictly regulated by the General Data Protection Regulation (GDPR). In the coming three years, the consortium of the newly launched NEMO project (Non-identifiability of Electroencephalograms (EEG) and similar sensor signals from Medical care for Open science) will examine the extent to which a person can be clearly identified from EEG data. It will then work on technical solutions for data anonymisation in order to prevent identification while at the same time safeguarding the usability of the data for scientific questions.
As an example of biosignal data, the project consortium is looking at electroencephalograms recorded during sleep. An EEG shows brain activity recorded via electrodes attached to the head, in this case parameters related to sleep. The Fraunhofer Institute for Digital Media Technology IDMT in Oldenburg is coordinating the “NEMO” project and has been conducting research into mobile EEG recording for many years, especially for use in sleep research.
Dr Insa Wolf, Head of Mobile Neurotechnologies, explains: “Recent publications indicate that personally identifiable information can be extracted from an EEG and that in this way re-identification can be possible. At the same time, significant information about the person’s state of health can be derived from the data and, in the future, possibly also markers for the early detection of diseases. Devices that collect EEG data can meanwhile also be found in the consumer sector, which underlines the particular importance of data protection in our ‘NEMO’ project.”
Within the project, the experts in Oldenburg are contributing their neurophysiological expertise in the recognition of individual sleep phases and events. They are also defining the exact requirements for EEG data from a research perspective.
Her colleagues at Fraunhofer IDMT in Ilmenau have been conducting research into data security and technical data protection for many years and developing corresponding techniques, such as for the anonymisation of audio data. They have the necessary know-how to develop methods that can quantify the risks of re-identification in EEG data. On this basis, they will then develop new algorithms for the anonymisation of EEG data.
“Our goal is to develop techniques that on the one hand prevent the unintentional disclosure of people’s identities and of sensitive information, but on the other hand also allow an analysis and use of sleep data that is as uncompromised as possible. To achieve this, we will look at various data protection technologies vis-à-vis their application to EEG data, including our methods for anonymising audio data. If we are successful, the results can also be extremely helpful for comparable biosignal data and other areas of health research,” says Patrick Aichroth, Head of Media Distribution and Security at Fraunhofer IDMT in Ilmenau.
Within the project, Kiel University and University Hospital Schleswig-Holstein (Kiel Campus) are delivering the clinical and medical perspective on the topics under study. They are also providing important sleep data for the development of the algorithms and their evaluation. Ascora GmbH, a company in Ganderkesee, is in charge of developing an application-related platform, which builds on the anonymisation algorithms developed in the project and will be used for data exploration and analysis as well as proof-of-concept. The aim is to integrate users in the anonymisation process and to explain to them the mode of action and the added value of the anonymisation variants deployed.
In a nutshell, the “NEMO” project takes both data protection requirements and the need for usable data in research and development into account. The aim is to illustrate specific risk scenarios and to test anonymisation techniques in the sensitive area of health data. A technical infrastructure in the sense of “open data” should create the foundation for the comprehensive use of secure health data in research and development.
MEDICA-tradefair.com; Source: Fraunhofer-Institut für Digitale Medientechnologie IDMT
