EEG-Based Alzheimer’s Disease Detection Using Deep Learning Techniques

Alzheimer’s disease (AD) is a neurological condition characterized by the degeneration of neuron cells, leading to cognitive impairment. This project investigates the effectiveness of EEG-based features for detecting Alzheimer’s Disease using deep learning techniques. The dataset comprises EEG recordings from healthy controls, Frontotemporal Dementia (FTD) patients, and patients diagnosed with AD. Data preprocessing involved segmenting the EEG signals into fixed-length epochs and filtering the baseline EEG into alpha, beta, and theta frequency bands. Features such as Mean Phase Coherence (MPC), Differential Entropy (DE), and Mean Squared Coherence (MSC) were computed pairwise for each epoch and frequency band. A feature matrix was constructed for each subject by stacking these features, which were then converted into images and used as input data for a ResNet50-based pre-trained model. The model was tested in leave-one-out cross-validation setting and achieved results comparable to the state-of-the-art in distinguishing between EEG recordings from healthy and AD-affected subjects.