Tapping Into a Reseource Hidden Behind MR Images: Learning Quantitative Imaging Biomarkers from Raw Big Data (MRI Big Data)

Today “Radionomics” is an often used approach to extract image features, which are then correlated with clinically relevant parameters. Since MR images are recorded in a so called k-space they need to be reconstructed first, before those approaches can be used. In a novel approach within this project, neural networks shall be introduced into the reconstruction processes and extract relevant biomarkers directly from MR signals to learn relevant information from the raw, and not only from the processed data.

Project Goals

The goal of this project is for us to gather first hand experience in regard to building, maintaining and working on a data lake. Hereby, one expected outcome is to get closer to a standardized and generic data lake setup which can be used by various researchers.

GWDG’s Role in the project

Within this project we build a prototypical data lake and continously re-evaluate the applicability of the current approach based on user feedback within the project. In addition, a number of necessary services is maintained.

Project Partners


Prof. Dr. Philipp Wieder
Hendrik Nolte



01.11.2019 - 31.03.2023

Funded by

VW Stiftung Logo

Volkswagen Stiftung, Niedersächsisches Vorab