A Dual-Stream Spatio-Temporal Network with Bidirectional Rectification for Differential Diagnosis of Dementia using EEG

Topic: Brain imaging, connectivity and assessment

Accurate differential diagnosis between Alzheimer’s disease (AD) and frontotemporal dementia (FTD) is of critical importance for the early screening of neurodegenerative diseases. Scalp electroencephalogram (EEG), as a non-invasive neurophysiological detection tool with excellent temporal resolution, has been widely adopted in dementia screening research. However, the inherent low signal-to-noise ratio and non-stationarity of scalp EEG signals, combined with the highly overlapping pathological rhythms between AD and FTD patients, render existing single-domain representation methods incapable of achieving precise pathological differentiation. To address this challenge, this paper proposes a dual-stream spatio-temporal network. In this framework, multiscale temporal features and effective connectivity graph features are extracted in parallel, followed by their integration. However, directly fusing the features from the two streams tends to amplify their respective noises. Therefore, the Spatio-Temporal Feature Fusion Based on Bidirectional Rectification module is proposed to execute dynamic noise suppression on the features and subsequently performs deep feature fusion via a cross-attention mechanism. Furthermore, this paper applies knowledge distillation using attention maps to address the issue of imbalanced sample distributions in the dataset. Experimental results on the ADFTD-RS dataset show that the proposed framework yields a subject-level accuracy of 60.00\% and an F1-score of 57.44\%, showing measurable improvements over several baseline methods. This work offers a feasible approach for the computer-aided differential diagnosis of dementia based on resting-state EEG.

Key findings

• Experimental results on the ADFTD-RS dataset show that the proposed framework yields a subject-level accuracy of 60.00\% and an F1-score of 57.44\%, showing measurable improvements over several baseline methods.

New to this subject? Start with our plain-English guide: Brain scans & tests →

Study details

• Evidence strengthLow

Read the original

• DOI: 10.21203/rs.3.rs-10724309/v1
• Original source →

Free copy unknown: no free-copy check has an answer for this record yet. That does not mean it is paywalled.

@article{yu2026-3df0a36b,
  author  = {Yu, L. and Meng, M. and Gao, Y. and Xi, X.},
  title   = {{A Dual-Stream Spatio-Temporal Network with Bidirectional Rectification for Differential Diagnosis of Dementia using EEG}},
  journal = {Research Square},
  year    = {2026},
  doi     = {10.21203/rs.3.rs-10724309/v1},
  url     = {https://doi.org/10.21203/rs.3.rs-10724309/v1}
}

Related studies in Brain scans and imaging

• MVMGNN: Multi-View Masked Graph Neural Network for Alzheimer’s Disease Diagnosis using Structural MRI Research Square · 2026-10-23 · usefulness score 16/100
• A Machine Learning Approach for the Differential Diagnosis of Alzheimer and Vascular Dementia Fed by MRI Selected Features Frontiers in neuroinformatics · 2020-06-11 · usefulness score 56/100
• Amyloid vs FDG-PET in the differential diagnosis of AD and FTLD Neurology · 2011-11-30 · usefulness score 52/100
• Alzheimer disease predicting from clinical and MRI data using DeepALZNET dual pathway framework Scientific Reports · 2025-12-04 · usefulness score 35/100
• Diagnosis of Alzheimer's disease from EEG signals: where are we standing? Current Alzheimer research · 2010-09-01 · usefulness score 36/100
• Differential default mode network trajectories in asymptomatic individuals at risk for Alzheimer's disease Alzheimer's & dementia : the journal of the Alzheimer's Association · 2019-05-18 · usefulness score 59/100

Similar studies

• Multimodal neuroimaging feature learning for multiclass diagnosis of Alzheimer's disease 2015 · Digital health tools
• DICE-Net: A Novel Convolution-Transformer Architecture for Alzheimer Detection in EEG Signals 2023 · Digital health tools
• Frequency-Resolved Connectivity Graphs for EEG-Based Classification of Alzheimer and Frontotemporal Dementia 2026 · Lewy body dementia
• Deep learning-based feature representation for AD/MCI classification 2013 · Digital health tools
• Joint Classification and Regression via Deep Multi-Task Multi-Channel Learning for Alzheimer's Disease Diagnosis 2019 · Digital health tools
• Interpretable Frequency-Aware Deep Learning for Multi-Class EEG Classification of Neurological Disorders 2026 · Brain scans and imaging

More on Brain scans and imaging

• A review that covers this ground Imaging β-amyloid using [(18)F]flutemetamol positron emission tomography: from dosimetry to clinical diagnosis European journal of nuclear medicine and molecular imaging · 2015-10-06 · usefulness score 38/100
• Newest in this subject Large-scale head-to-head comparison of on- and off-target profiles of [18F]Tauvid and [18F]MK6240 tau-PET in aging and Alzheimer’s disease Research Square · 2026-11-01 · Preprint — not yet peer-reviewed · usefulness score 16/100
• Newest in this subject Repetitive negative thinking is associated with amyloid accumulation over 3 years in cognitively healthy adults medRxiv · 2026-10-31 · Preprint — not yet peer-reviewed · usefulness score 16/100

See all studies in Brain scans and imaging →