Artificial Intelligence
Project goals
The German National Cohort (NAKO) provides an extensive database of imaging and non-imaging information that can be used for research in the field of artificial intelligence (AI).
The expert group is working on networking different areas and applications. The aim is to develop proposals for a long-term, sustainable concept – taking into account all ethical and data protection requirements – for data exchange and access.
The expert group also supports AI-based projects such as the automated analysis of determinants of aortic morphology.
Speakers
Prof. Dr. Thomas Küstner
Prof. Dr. Christoph Lippert (Deputy)
Results
Automated analysis of MRI data: The researchers have trained a deep learning model with whole-body MRI imaging data from participants in the UK Biobank and the NAKO to automatically and reliably assess abdominal organs such as the liver, spleen, kidney or pancreas. The results obtained can be used as the basis for an automated analysis of thousands of data sets of MRI images.
Assessment of subcutaneous fatty tissue in whole-body MRI: Scientists have trained a deep learning model with the aim of fully automated assessment of visceral and subcutaneous fatty tissue in whole-body MRI.
Publications
Fay L, Hepp T, Winkelmann M T et al. Determinants of ascending aortic morphology: Cross-sectional deep learning-based analysis on 25,073 non-contrast-enhanced MRI of NAKO Preprint /medRxiv 2024.07.12.24310356; doi: https://doi.org/10.1101/2024.07.12.24310356
Fay L, Cobos E, Yang B, Gatidis S, Küstner T. Avoiding Shortcut-Learning by Mutual Information Minimization in Deep Learning-Based Image Processing. IEEE Access. 2023;11:64070-4086. http://doi.org/10.1109/ACCESS.2023.3289397
Kart T, Fischer M, Küstner T, et al. Deep Learning‐Based Automated Abdominal Organ Segmentation in the UK Biobank and German National Cohort Magnetic Resonance Imaging Studies. Invest Radiol. 2021;56(6):401-408. http://doi.org/10.1097/RLI.0000000000000755
Küstner T, Hepp T, Fischer M, et al. Fully Automated and Standardized Segmentation of Adipose Tissue Compartments via Deep Learning in 3D Whole-Body MRI of Epidemiologic Cohort Studies. Radiology: Artificial Intelligence. 2020;2(6):e200010. http://doi.org/10.1148/ryai.2020200010