کنفرانس مهندسی برق ایران

صفحه اصلی / سی و چهارمین کنفرانس بین المللی مهندسی برق

A Self-Supervised Deep Learning Framework for Alzheimer’s Disease Stage Classification Using Resting-State fMRI

نویسندگان :

Mohammad Ahmadabadi¹ Ali Sadr²

1- دانشگاه علم و صنعت ایران 2- دانشگاه علم و صنعت ایران

کلمات کلیدی :

Functional magnetic resonance imaging (fMRI)،Mild cognitive impairment (MCI)،SimCLR،Deep Learning،t-SNE

چکیده :

Early and reliable staging of Alzheimer’s disease (AD) plays a crucial role in enabling timely treatment and effective disease management. Resting-state functional magnetic resonance imaging (rs-fMRI), as a non-invasive neuroimaging technique, provides valuable information about intrinsic brain functional activity and has therefore gained increasing interest in AD research. Nevertheless, the scarcity of available fMRI data remains a major obstacle for training deep learning–based models. To address this limitation, this study proposes a hybrid learning framework that combines self-supervised and supervised strategies for the classification of multiple AD stages, including healthy control (HC), early mild cognitive impairment (EMCI), late mild cognitive impairment (LMCI), and AD. In the first stage, preprocessed fMRI time-series signals are utilized for self-supervised pretraining using the SimCLR algorithm, which learns robust and discriminative representations without relying on labeled data. The pretrained network is then fine-tuned in a supervised manner using class labels. Moreover, a novel encoder architecture, named sRRLATT (simplified ResNet with residual LSTM and Attention), is introduced. Experimental evaluations demonstrate that the proposed framework consistently outperforms state-of-the-art approaches across various binary classification tasks. In addition, feature visualization results verify that self-supervised pretraining substantially improves feature separability, further confirming the effectiveness and robustness of the proposed method.