Understanding neurodegeneration in Down syndrome through AI-driven neuroimaging analysis
Research Lines
Down syndrome (DS), also known as trisomy 21, is a genetic condition caused by an extra copy of chromosome 21 that affects how the brain and body develop from birth. It is typically associated with lifelong developmental differences and mild-to-moderate intellectual disability, often accompanied by characteristic physical features and reduced muscle tone, especially in infancy. DS can also be linked to additional medical conditions such as an ultra-high risk of early‑onset Alzheimer’s disease (AD) after age 40.
In this context, brain magnetic resonance imaging (MRI) is particularly valuable because it is non-invasive, radiation-free, and provides detailed 3D information about brain structure. However, understanding how neuroanatomy relates to cognitive function, neuroinflammation, and comorbidities such as Alzheimer’s disease remains challenging in persons with Down syndrome.
The AI4DS project aims to overcome current analysis bottlenecks by developing unsupervised AI models that can learn from large collections of 3D brain MRI scans with minimal annotation. We will train autoencoders and diffusion models to generate high-resolution 3D alteration maps comparing euploid and DS populations, including DS subgroups across AD-related neurodegeneration stages, and we will exploit latent representations learned by variational autoencoders to build lightweight machine-learning models that stratify individuals with DS according to neurodegeneration severity.
All key outputs—alteration maps and stratification results—will be clinically evaluated and validated by neurologists and neuroradiologists at the Sant Pau Memory Unit to ensure interpretability and medical relevance.
Thanks to the funding from the Fondation Jérôme Lejeune, we have acquired a dedicated compute node, Geronimo, equipped with two high-capacity GPUs —NVIDIA RTX PRO 6000 Blackwell Server Edition with 96GB memory—which will enable large-scale training of 3D AI models essential to achieving the project’s objectives.
The AI4DS project (Ref.: 2544_GRT-2025B) is financed by the Fondation Jérôme Lejeune Grants session 2025B.
Collaborators:
Sant Pau Memory Unit
Faculty of Biology, Universitat de Barcelona
Team Leader
