Flexible and Explainable Graph Analysis for EEG-based Alzheimer's Disease Classification

TL;DR

This work tackles EEG-based Alzheimer's disease classification by proposing a Flexible and Explainable Gated Graph Convolutional Network (GGCN) whose hyperparameters are optimized with Multi-Objective Tree-Structured Parzen Estimator (MOTPE). The method constructs subject-specific graphs from PSD-derived features using connectivity measures (PLI/PLV) and adaptive Nearest Neighbor sparsification, then classifies with a gated graph network that includes ASAP-based pooling and LEConv-derived clustering. Across a dataset of 123 participants, the approach achieves high discrimination (often $AUC$ > 0.9) and yields explainable insights into frontal and parietal connectivity differences between healthy controls and AD patients, particularly in moderate-to-severe dementia. The combination of flexible graph architecture, multiobjective hyperparameter tuning, and interpretable adjacency representations offers a scalable framework for EEG-based AD screening with potential applicability to other neurological conditions.

Abstract

Alzheimer's Disease is a progressive neurological disorder that is one of the most common forms of dementia. It leads to a decline in memory, reasoning ability, and behavior, especially in older people. The cause of Alzheimer's Disease is still under exploration and there is no all-inclusive theory that can explain the pathologies in each individual patient. Nevertheless, early intervention has been found to be effective in managing symptoms and slowing down the disease's progression. Recent research has utilized electroencephalography (EEG) data to identify biomarkers that distinguish Alzheimer's Disease patients from healthy individuals. Prior studies have used various machine learning methods, including deep learning and graph neural networks, to examine electroencephalography-based signals for identifying Alzheimer's Disease patients. In our research, we proposed a Flexible and Explainable Gated Graph Convolutional Network (GGCN) with Multi-Objective Tree-Structured Parzen Estimator (MOTPE) hyperparameter tuning. This provides a flexible solution that efficiently identifies the optimal number of GGCN blocks to achieve the optimized precision, specificity, and recall outcomes, as well as the optimized area under the Receiver Operating Characteristic (AUC). Our findings demonstrated a high efficacy with an over 0.9 Receiver Operating Characteristic score, alongside precision, specificity, and recall scores in distinguishing health control with Alzheimer's Disease patients in Moderate to Severe Dementia using the power spectrum density (PSD) of electroencephalography signals across various frequency bands. Moreover, our research enhanced the interpretability of the embedded adjacency matrices, revealing connectivity differences in frontal and parietal brain regions between Alzheimer's patients and healthy individuals.

Figures (8)

• Figure 1: Workflow
• Figure 2: Architecture of the Workflow. (A) Preprocessing of EEG signals, which involves eliminating bad channels, artifact removal via ICA, and computing the Power Spectral Density (PSD) for each frequency band. (B) Construction of graphs begins with the computation of the functional connectivity matrix between channels, followed by the application of the Nearest Neighbors algorithm to form the adjacency matrix. (C) Implementation of the Gated Graph Convolutional Network (GGCN) module, followed by graph pooling and fully connected layers for classification purposes.
• Figure 3: This figure shows the averaged adjacency matrix for the HC and AD groups in Moderate to Severe Dementia, highlighting differences between them.
• Figure 4: This figure shows the Hyperparameter Settings for CDR of 0.5 - Very Mild Dementia.
• Figure 5: This figure shows the Hyperparameter Settings for CDR of 1 - Mild Dementia.