Lightweight Alzheimer’s Disease Diagnosis Based on Gaussian Graphical Lasso Estimation

Alzheimer’s Disease (AD) is the primary cause of dementia in the elderly. Utilizing advanced Graph Neural Networks (GNNs) to map the relationships between various brain regions and the AD using structural magnetic resonance imaging (sMRI) is one of the effective diagnostic methods. However, current GNN-based methods suffer from significant data redundancy, resulting in high computational costs and overlooking the critical role of connectivity between different brain regions. To solve this, in this paper we propose a lightweight Alzheimer’s disease diagnosis method based on Gaussian Graphical Lasso (GLasso) estimation. Our method first compresses features from traditional MRI and maps them to corresponding Regions of Interest (ROIs) using the Automated Anatomical Labeling (AAL) template. Subsequently, a Gaussian Graphical Model (GGM) is employed to construct the graph network, and iterative GLasso estimation is applied to retain significant ROI connections while removing redundant and irrelevant connections. Finally, a graph transformer network with edge features is used to achieve efficient diagnosis of AD symptoms. Extensive testing on the open-source dataset ADNI demonstrates that our method achieves precise diagnostic results, outperforming state-of-the-art approaches, and significantly reduces training costs. Furthermore, it effectively identifies critical region connections associated with AD, providing a feasible solution for the early diagnosis of Alzheimer’s disease.