Using AI methods in magnetic resonance imaging to untangle ageing and disease effects for better prediction of future diseases and their progression.
Helmholtz AI project call showcase: Deep Generative Models for Causal Prognosis using Neuroimaging Data
This week’s Helmholtz AI project showcase is about AI methods for the early detection of neurodegenerative diseases like Alzheimer's using magnetic resonance imaging (MRI) data. The project is a collaboration between researchers from the German Center for Neurodegenerative Diseases (DZNE) and Forschungszentrum Jülich (FZJ).
Could you introduce yourself, giving your affiliation, area of work, and of course, the project title?
The Helmholtz AI-funded project DeGen “Deep Generative Models for Causal Prognosis using Neuroimaging Data” is led by Prof. Dr. Martin Reuter from DZNE Bonn and Harvard Medical School in Boston, USA, as well as Dr. Kaustubh Patil from FZJ in Jülich. Dr. Reuter heads the Deep Medical Imaging Lab and works on advanced deep learning methods for the automatic extraction of sensitive preclinical biomarkers from non-invasive MRI. Dr. Patil runs the Juelich Applied Machine Learning group (JuAML) with a research interest in improving the understanding of biological systems.
In simple words, what specifically is your project about? And, how and why do you think it is a high risk, high gain endeavour?
The goal of the DeGen project is to devise AI methods for early detection of neurodegenerative diseases like Alzheimer’s disease using magnetic resonance imaging data. Such a method can help select candidates for inclusion into early intervention clinical studies, promote research into risk and preserving factors, and ultimately relieve the high social and economic burden of these debilitating diseases. Therefore, this endeavour is high gain, yet despite promising previous research, it comes with many challenges such as the entanglement of neurodegenerative and ageing effects, limited data availability especially from young/healthy brains with a known dementia prognosis, as well as the different manifestations of neurodegeneration in different individuals. We aim to tackle these challenges within DeGen by introducing FaderNets to help disentangle ageing and disease effects for better prediction of future disease or progression.
How important has the Helmholtz AI funding and platform been to carry out this project?
The Helmholtz AI platform and the funding has been crucial for conception and implementation of this project. It has granted us access to experts and resources, both with respect to computational- and human resources, without which this project and exciting collaboration would not have been possible, especially given its high risk nature.
What else would you like to share?
We believe that - on top of the potential high gain in this project - this funding mechanism in general encourages and fosters collaborations that ultimately lead to exciting new ideas and perspectives, needed to develop and translate the next generation of AI methods into medical applications.
Figure: Image showing three brains coloured by FastSurfer (Henschel et al., 2020).