ORIGINAL RESEARCH article
Front. Neurosci.
Sec. Neuromorphic Engineering
Ghost-LENet: A Robust Convolutional Neural Network for EEG-based Brain-Computer Interfaces
Ruidong Ma 1
Zhenyu Wang 2
Lingfan Zhang 2
Shangbin Chen 1
Qiying Song 3
1. Northwest Normal University, Lanzhou, China
2. Fudan University, Shanghai, China
3. 1st Medical Center of Chinese PLA General Hospital, Beijing, China
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
Abstract
In this paper, we propose a novel convolutional neural network (CNN) architecture named Ghost-LENet, specifically designed for EEG-based brain-computer interfaces (BCIs), particularly for motor imagery (MI) tasks. This model incorporates several key innovations aimed at enhancing performance and generalizability across different motor imagery paradigms. Specifically, it combines dilated convolutions and wavelet transforms in the initial block for feature extraction, thereby enabling multi-scale temporal analysis. Furthermore, we introduce a dynamic residual fusion mechanism through the use of Efficient Channel Attention (ECA) modules, referred to as DR-RCA. This attention mechanism allows the model to adaptively adjust the contribution of residuals. Lastly, the incorporation of the Ghost module improves the model's efficiency by reducing computational complexity while maintaining its feature extraction capabilities. Experimental results demonstrate that Ghost-LENet, which utilizes only 3,305 parameters, outperforms state-of-the-art models across multiple EEG datasets, highlighting its robustness and versatility in various BCI paradigms. Specifically, Ghost-LENet achieves classification accuracies of 82.18% and 83.05% on two motor imagery (MI) datasets (BCI Competition IV-2a and IV-2b). In summary, Ghost-LENet exhibits significant potential as a general decoding model for EEG-based motor imagery tasks.
Summary
Keywords
brain–computer interface, Convolutional Neural Network, EEG, Efficient Channel Attention, Ghost module, Motor Imagery, Wavelet Transform
Received
06 November 2025
Accepted
04 January 2026
Copyright
© 2026 Ma, Wang, Zhang, Chen and Song. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Qiying Song
Disclaimer
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.