Multimodal Data Fusion for the Early Detection of Dementia

In our aging society, geriatric diseases are becoming increasingly common. One of these diseases is dementia, including its most common type: Alzheimer's disease. This chronic neurodegenerative disease is characterized by progressively declining cognitive capacities to the point of dependency in all areas of daily life.

As of yet, dementia can't be cured; only the progression of the disease can be decelerated by appropriate medication. Therefore it is vital to detect dementia early on and ideally even identify people who haven't yet developed the disease but one day will. State-of-the-art dementia diagnostics, however, are complex and time-consuming, and can only identify affected people once brain degeneration has progressed considerably and symptoms already show.

Therefore, with NAKO+ILSE we wish to research methods for early detection of dementia. To this end, we are working towards fusing two large scale epidemiological studies with our project partner Fraunhofer MEVIS:

  1. NAKO is a young longitudinal study all over Germany aiming at enabling research into common diseases: the German National Cohort, which is currently in the second period of acquiring health data from a representative sample of the population also in Bremen.
  2. ILSE, the "Interdisciplinary Longitudinal Study on Adult Development and Aging", on the other hand, has been following the social and health development of people from their working age to retirement ever since the nineties.

With NAKO+ILSE we will develop suitable tools and methods for automatically fusing such databases. In order to be able to apply these, we are currently planning a study to acquire data with methods of both studies from a subpopulation of the participants.

Based on these merged databases we want to analyze whether including multiple modalities in algorithms can improve early detection of dementia. Building on our longstanding research on (early) detection of dementia grounded solely on ILSE's speech data (see ALMED), we would now like to incorporate magnetic resonance imaging (MRI). The speech research and MRI research parts of the project in this project will be led by CSL and Fraunhofer MEVIS, respectively. We will further examine which fusion method optimizes early detection and how far into the future cognitive and brain aging effects of dementia can be predicted.

The outcomes we envision for NAKO+ILSE will also benefit reasearch outside our lab: on the one hand we will contribute to data fusion research with the tools and methods we will develop, on the other hand with the fused comprehensive health database we will enable more intensive research into not only neurodegenerative but also other common diseases like cancer and diabetes.

Funding: U Bremen Research Alliance (20212024)

Staff at CSL
Project Partners


Student assistent wanted!

Only available for students fluent in German.

For this project, we are currently preparing a study for which we would like to conduct, record and automatically analyze interviews of about 500 participants in the age range of 60+ years.

To accomplish this, we need your help! So if you're good with (elderly) people, are interested in conducting interviews, language is your thing and you would like to earn money along the way – get in touch with us.

Hopefully we can start this fall but you should be able to join our team in October latest.