Integrating attractor dynamics and connectivity features for EEG-based dementia classification

The precise differentiation between Alzheimer’s disease (AD) and frontotemporal dementia (FTD) presents a clinical challenge, as both conditions share overlapping symptoms yet diverge in their pathophysiological mechanisms and treatment strategies. Electroencephalography (EEG), characterized by its higher temporal resolution and widespread applicability, provides the opportunity to uncover subtle discrepancies in brain dynamics that may be difficult to detect with conventional neuroimaging techniques. This paper presents a system that combines nonlinear attractor-based features derived from phase-space representations with phase-locking value connectivity features to encapsulate both local and global brain dynamics. In this regard, resting-state EEG recordings from 36 AD patients, 23 FTD patients, and 29 healthy controls (HC) were preprocessed and analyzed to extract dynamic and network features. Multiple classifiers were then used to assess these features under stratified 10-fold cross-validation. The results showed that the support vector machine achieved the highest performance for AD vs. FTD (81.7%), logistic regression performed well for FTD vs. HC (81.0%), and gradient boosting reached 82.9% for AD vs. HC. These findings illustrate the capability of EEG as a low-cost diagnostic technique, suggest that attractor dynamics and connectivity can offer complementary perspectives on the brain alterations linked to dementia, and enhance classification performance.