NAKO-MNA: Multimodal norm atlas and synthetic data model based on the NAKO health study

The AI Center for Health Care project NAKO-MNA aims at the AI-based development of a multimodal implicit data model based on combined image data and complex tabular data from the NAKO health study. One application objective is the improved ability to sensitively detect deviations from the norm and previously undetected incidental findings.

Activation required

By clicking on this thumbnail, you consent to the downloading of content from YouTube (USA). This provides Google (USA) with the information that you have accessed our site as well as the technically necessary data in this context. We have no influence on the further data processing by Google. Please note that Google and YouTube do not provide an adequate level of data protection.
Further information / data protection

NAKO-MNA is a joint project of Leibniz BIPS, Fraunhofer MEVIS and the University of Bremen. It aims at the AI-based development of a multimodal implicit data model based on combined image data and complex tabular data from the NAKO health study. The goal is to realistically synthesize this combined data, including norm variants and incidental findings. The aim is to better understand subpopulations, norm variants and the correlations between incidental findings and epidemiologic parameters.

To achieve this goal, MEVIS will combine AI methods for medical imaging and BIPS methods for epidemiological data. This data is used to create a model of a "healthy" population. The random findings from the NAKO Imaging Study can also be used to model important early disease patterns.

The project investigates different methods for data synthesis, including generative adversarial networks (GANs), normalizing flow models (NFs) and variational autoencoder networks (VAEs). These models are used to generate synthetic data that is privacy-compliant and usable for further research projects.

For patients, this means an improved ability to detect abnormalities and previously undetected incidental findings at an early stage. This enables an earlier diagnosis of potentially dangerous diseases, allowing treatment at an earlier and gentler stage of the disease. The use of the synthetic data model provides immediate decision support and can directly influence healthcare.