We are pleased to announce that our latest research has been published as an Original Research article in Frontiers in Neuroscience (Open Access).
The study, titled “Olfactory EEG based Alzheimer disease classification through transformer based feature fusion with tunable Q-factor wavelet coefficient mapping,” proposes a deep learning framework that uses olfactory-evoked EEG and fuses spatial (CSP), covariance-tangent space, and time–frequency (TQWT-based wavelet coefficient mapping) features via a transformer-based feature fusion strategy.
Using an olfactory EEG dataset of 35 seniors (13 AD, 7 MCI, 15 healthy) with 3,877 samples and two odor conditions (lemon and rose), the model was evaluated with leave-one-subject-out (LOSO) validation. The proposed approach achieved up to 93.14% accuracy for rose trials and 90.86% accuracy for lemon trials, demonstrating strong potential for EEG-only, olfactory-driven AD/MCI/Healthy classification.
Authors:
B. Cansiz (Yıldız Technical University)
H.O. İlhan (Yıldız Technical University)
N. Aydin (Istanbul Technical University)
Gorkem Serbes (Yıldız Technical University)
