Related Experiment Video
Updated: Jun 13, 2026
07:21
Electroencephalographic Signal Acquisition Framework for Neurodiverse: A Case Study of Dolphin-Assisted Therapy
Published on: June 27, 2025
A Time-Frequency Decoupled Contrastive Learning Framework for Electroencephalography-Based Parkinson's Disease
Shunwu Xu1,2, Lanxiang Chen1,3, Hao Yan1,2
1Faculty of Data Science, City University of Macau, Macau 999078, P. R. China.
International Journal of Neural Systems
|June 12, 2026
Summary
A new framework called TDCNet uses electroencephalography (EEG) signals for accurate Parkinson's disease (PD) diagnosis. This AI approach significantly improves early detection rates by analyzing brainwave patterns.
Area of Science:
- Neuroscience
- Artificial Intelligence
- Biomedical Engineering
Background:
- Parkinson's disease (PD) diagnosis remains challenging, particularly in early stages.
- Electroencephalography (EEG) offers a non-invasive method for brain activity monitoring.
- Advances in representation learning can enhance diagnostic accuracy for neurodegenerative disorders.
Purpose of the Study:
- To introduce a novel EEG-based diagnostic framework for Parkinson's disease (PD).
- To leverage self-supervised representation learning for improved PD detection.
- To develop a framework outperforming existing methods in PD diagnosis using EEG.
Main Methods:
- Developed the time-frequency decoupling contrastive learning network (TDCNet) using collaborative contrastive learning.
- Incorporated temporal-evolution and frequency-aware contrastive losses for EEG signal analysis.
- Utilized subject-level pseudo-labels for latent space aggregation and downstream supervised evaluation.
Main Results:
- TDCNet achieved high accuracies (95%-99% subject-dependent, 73%-89% subject-independent) across multiple datasets.
- The framework consistently outperformed existing EEG-based PD diagnostic approaches.
- Central electrodes were identified as dominant brain regions for diagnosis, and medication status may influence performance.
Conclusions:
- TDCNet demonstrates significant potential for accurate and early diagnosis of Parkinson's disease using EEG.
- The proposed self-supervised learning framework effectively deciphers complex EEG patterns for PD detection.
- Further research into medication effects and electrode importance can refine EEG-based PD diagnostics.