4,068 posters, 219 videos, 101 topics, 69 sessions, 14,365 authors, 6,128 institutions
ePostersLive by SciGen Technologies S.A. All rights reserved.
Alzheimer's Association International Conference 2026
July 12-15, 2026 | London, United Kingdom, and Online
4276
Predicting Longitudinal Progression Of Alzheimer’s Disease Brain Morphology From 3D MRI Via Latent Space Transformers
Poster Presenter
Part of Topic
Background:
•Alzheimer's disease (AD) has a prolonged progression, with brain changes believed to begin two decades or more before clinical symptoms.
•AD progression exhibits substantial inter-individual heterogeneity, with rates and patterns of brain atrophy varying widely across individuals, making individual-level longitudinal analysis essential.
To address this limitation, we propose a novel transformer-based deep learning model that leverages large-scale cross-sectional MRI data to generate individualized longitudinal MRI and predict disease progression for patients with Alzheimer's disease.
We utilized multi-timepoint T1-weighted MRI scans from the ADNI dataset. The proposed framework consists of two key stages.
Stage 1: a 3D Vector Quantized Variational Autoencoder (VQ-VAE) was trained to compress volumetric MRI images into compact, discrete latent codebooks, capturing time-independent anatomical representations of brain structure.
Stage 2: a Transformer-based temporal model was applied in this latent space to learn longitudinal temporal patterns of brain morphological changes.
By modeling sequences of latent representations across time, the Transformer captures disease-related temporal dynamics and predicts future brain morphology at the individual level.
Key Takeaways:
•Leverages large-scale cross-sectional MRI data to generate individualized longitudinal MRI trajectories.
Comments
Email is required but is not shown publicly.
Loading comments...