/ EN
2026-08-26 00:00 China Papers Foundations & Methods Translated from EN

Tianjin University's MTGNet Denoises EEG, Lifting Fatigue Detection by Over 6 Points

Summary EEG signals are only microvolts strong, so blinks, jaw clenching and muscle activity easily contaminate them. Researchers at Tianjin University and Tiangong University in northern China proposed MTGNet, a framework that suppresses these artifacts while preserving the information downstream tasks need. On the public EEGDenoiseNet dataset, it cut spectral relative root-mean-square error by 18.9%, 31.5% and 14.0% for EMG, EOG and mixed artifacts respectively; on a real-world fatigue EEG dataset, it raised classification accuracy by 6.20 to 6.69 percentage points over unprocessed input. Adapting the framework to a new task takes only 0.33 million low-rank adaptation (LoRA) parameters and no paired clean EEG reference.
Why it matters EEG denoising has long faced a dilemma: the cleaner the filtering, the more task-relevant information may be stripped out with the noise. This work ties denoising to the downstream objective and replaces paired clean EEG references with 0.33 million lightweight adaptation parameters, sidestepping the lack of a ground truth in real-world recordings. The 18.9%, 31.5% and 14.0% reductions and the 6.20–6.69 percentage-point accuracy gain all come from one public dataset and a single fatigue dataset; performance across tasks and recording conditions has yet to be validated.

BCIwiki (bciwiki.com) — Researchers at Tianjin University and Tiangong University proposed MTGNet, a task-oriented and spectrally guided EEG denoising framework that cut spectral relative root-mean-square error by 18.9%, 31.5% and 14.0% for EMG, EOG and hybrid artifacts on the public EEGDenoiseNet dataset, and raised classification accuracy by 6.20 to 6.69 percentage points over unprocessed input on a real-world fatigue EEG dataset. The results were reported in Journal of Neural Engineering on August 26, 2026.

EEG sits at the microvolt level, so electromyography, electrooculography and mixed physiological artifacts readily contaminate it, while aggressive filtering can strip information that downstream tasks need. MTGNet binds denoising to a specific task: a pretrained 11.97 million-parameter backbone learns to preserve intrinsic EEG characteristics from paired noisy-clean data, while 0.33 million LoRA parameters handle task-specific adaptation, removing the need for paired clean EEG references.

The team reported the artifact suppression results on EEGDenoiseNet (p<0.001) and the accuracy gain on a real-world fatigue EEG dataset (p<0.05). Ablation, cross-classifier and cross-dataset analyses were used to validate the components and support the transferability of MTGNet across the evaluated settings. The authors note that its applicability across diverse tasks, artifact types and acquisition conditions still requires further validation.

Compiled by BCIwiki from public sources

Sources · 1

Tianjin University timeline

2026-08 Tianjin University Team Cuts EEG Channels Without Losing Decoding Accuracy 2026-08 Sixth-Finger BCI Neurofeedback Aids Stroke 2026-08 Integrated Decoding of Local and Prospective Spatial Representations for Future Decision Prediction 2026-07 A Multi-Paradigm Longitudinal EEG Dataset Including 'Sixth-Finger' and 'Affected-Hand' Motor Imagery of Stroke Patients All entries →
Read original ↗
Suggest a correction Revisions · none / EN
© 2026 BCIwiki.com Digest Topics Tips Subscribe Revisions About