Work

Önceden Eğitilmiş Bir EEG Temel Modeli Ve Öğrenilebilir Kanaldan-Kanala Dikkat Mekanizması Kullanılarak EEG Sinyallerinden Denek-Bağımsız Alzheimer Hastalığı Sınıflandırması

Year: 2026
Type: article
Abstract: Accurate classification of Alzheimer’s disease (AD) in its early stages is of critical importance for clinical decision support systems. Although EEG is a promising biomarker due to its high temporal resolution and non-invasive nature, factors such as low signal-to-noise ratio and high inter-subject variability hinder reliable classification. In this study, we propose a novel approach based on channel-to-channel attention (C2C) built upon the frozen REVE EEG foundation model for the discrimination of AD and healthy controls (HC). The proposed module adaptively models inter-channel relationships by operating directly on raw EEG windows, and the resulting reweighted signals are provided as input to the REVE encoder. Model performance was evaluated on a dataset consisting of 36 AD and 29 HC individuals using a subject-independent protocol. Compared to the baseline approach using only frozen REVE representations and a linear classifier, the proposed method achieved improvements across all performance metrics, including accuracy (ACC: 0.7538→0.7846) and Matthews correlation coefficient (MCC: 0.5019→0.5682). These results demonstrate that the C2C module enhances discriminative spatial information, thereby increasing the performance of foundation models.
Source: DergiPark
Language: English
Open Access status: green
OpenAlex ID: W7160471778
FWCI:
Cites: 38
Cited by:
Related to: 10