Characterizing the subtle changes of functional brain networks associated with the pathological cascade of Alzheimer’s disease (AD) is important for early diagnosis and prediction of disease progression prior to clinical symptoms. We developed a new deep learning method, termed multiple graph Gaussian embedding model, which can learn highly informative network features by mapping high-dimensional resting-state brain networks into a low-dimensional latent space. These latent distribution-based embeddings enable a quantitative characterization of subtle and heterogeneous brain connectivity patterns at different regions, and can be used as input to traditional classifiers for various downstream graph analytic tasks, such as AD early stage prediction, and statistical evaluation of between-group significant alterations across brain regions. We used our proposed method to detect the intrinsic latent dimensionality of MEG brain networks, predict the progression of patients with mild cognitive impairment (MCI) to AD, and identify brain regions with network alterations related to MCI.
Format: 
Format: 
Format: 
Format: 
If you use this repository, please cite our paper in the below.
@article{xu2021graph, title={A Graph Gaussian Embedding Method for Predicting Alzheimer's Disease Progression With MEG Brain Networks}, author={Xu, Mengjia and Sanz, David Lopez and Garces, Pilar and Maestu, Fernando and Li, Quanzheng and Pantazis, Dimitrios}, journal={IEEE Transactions on Biomedical Engineering}, volume={68}, number={5}, pages={1579--1588}, year={2021}, publisher={IEEE} }