EEG-Based multimodal analysis for alzheimer's disease and frontotemporal dementia

Alzheimer's disease (AD) and frontotemporal dementia (FTD) are two major dementia subtypes with overlapping clinical symptoms, making differential diagnosis challenging. This study proposes an EEG-based hierarchical classification framework. In the first stage, subjects are classified as dementia or healthy controls, followed by a second stage distinguishing AD from FTD. After appropriate EEG preprocessing, multiple models including EEGNet, CNN+LSTM, CNN+BiLSTM, XGBoost, and LightGBM are evaluated. In addition, a decision-level ensemble model is developed by independently training EEGNet and XGBoost and averaging their output probabilities. Experimental results demonstrate that the proposed ensemble approach consistently outperforms individual models in both classification stages in terms of accuracy and F1-score. These findings highlight the effectiveness of combining deep learning and boosting-based methods for robust EEG-based dementia diagnosis.