CEP-MI: a novel cue-execution dual-stage motor imagery paradigm with PGCANet decoding architecture
4 affiliations
- 1School of Electronic Information Engineering, Inner Mongolia University, Hohhot, China
- 2Key Laboratory of Intelligent Communication and Sensing and Signal Processing, Hohhot, China
- 3Inner Mongolia Vocational and Technical College of Transportation, Chifeng, China
- 4School of Biomedical Engineering, Capital Medical University, Beijing, China
Abstract
Background: Motor imagery-based brain-computer interface (MI-BCI) holds great potential in neurorehabilitation and motor assistance. Data quality and motor intention decoding effectiveness are closely associated with participants’ engagement. However, previous MI tasks ignored active engagement regulation, and neural mechanisms underlying the transition from motor imagery to execution remain unclear, with limited relevant research. Conventional separate-trial designs cannot capture continuous brain dynamics from imagery to execution and cannot objectively evaluate subject engagement, which hinders mechanism exploration and stable decoding.
Methods: To address these paradigm-related limitations, a novel cue-execution MI paradigm (CEP-MI) with countdown anchors was designed to quantify engagement. Motor execution sessions were added post-imagery to reveal neural differences during the transition and identify movement-related brain activities. Meanwhile, to boost MI-EEG decoding performance, a collaborative decoding architecture (PGCANet) integrating prior guidance and channel awareness was proposed, with dual branches and consistency constraint loss for decision fusion.
Results: CEP-MI effectively quantifies participant engagement, and PGCANet achieves superior decoding performance compared to traditional algorithms and advanced models, with the optimal average accuracy reaching 85.77% in binary classification and 82.47% in quadruple classification.
Conclusion: Targeting the above-mentioned research gaps, the CEP-MI paradigm provides valuable insights into the imagery-execution transition mechanism. PGCANet enhances MI decoding accuracy via multi-branch collaboration, laying a foundation for MI-BCI’s clinical application in neurorehabilitation.
Significance: This study introduces a novel paradigm and collaborative architecture that significantly advance MI-BCI decoding accuracy and engagement quantification, and helps to promote its clinical application in neurorehabilitation.
Related
- Decoding Graded Grip-Force Intensity From fMRI Data Reveals a Transformation From Abstract to Effector- and Movement-Specific Codes Prior to Execution.
- A Complementary Dual-Branch Receptive Field Network for Cross-Channel-Density Motor Imagery EEG Decoding.
- AVID: A 5T fMRI Dataset for Benchmarking Auditory-induced Visual Mental Imagery Decoding
- Decoding motor imagery EEG signals by a deep learning network integrating multiscale filtering and temporal-channel attention mechanisms
- Evidence for decoder-label-regime-dependent effects of complementary leaky integrate-and-fire temporal-state decoding in motor-imagery EEG
BibTeX
@article{li2026cepmi,
title = {{CEP-MI: a novel cue-execution dual-stage motor imagery paradigm with PGCANet decoding architecture}},
author = {Jianxiu Li and Pengda Yu and Huiyu Qiao and Zhiqiang Zheng and Changming Wang and Zhi Weng and Caili Gong},
journal = {Frontiers in Human Neuroscience},
year = {2026},
month = oct,
doi = {10.3389/fnhum.2026.1908608},
url = {https://doi.org/10.3389/fnhum.2026.1908608},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - CEP-MI: a novel cue-execution dual-stage motor imagery paradigm with PGCANet decoding architecture
AU - Jianxiu Li
AU - Pengda Yu
AU - Huiyu Qiao
AU - Zhiqiang Zheng
AU - Changming Wang
AU - Zhi Weng
AU - Caili Gong
PY - 2026
DA - 2026/10/22
JO - Frontiers in Human Neuroscience
DO - 10.3389/fnhum.2026.1908608
UR - https://doi.org/10.3389/fnhum.2026.1908608
AB - Background: Motor imagery-based brain-computer interface (MI-BCI) holds great potential in neurorehabilitation and motor assistance. Data quality and motor intention decoding effectiveness are closely associated with participants’ engagement. However, previous MI tasks ignored active engagement regulation, and neural mechanisms underlying the transition from motor imagery to execution remain unclear, with limited relevant research. Conventional separate-trial designs cannot capture continuous brain dynamics from imagery to execution and cannot objectively evaluate subject engagement, which hinders mechanism exploration and stable decoding. Methods: To address these paradigm-related limitations, a novel cue-execution MI paradigm (CEP-MI) with countdown anchors was designed to quantify engagement. Motor execution sessions were added post-imagery to reveal neural differences during the transition and identify movement-related brain activities. Meanwhile, to boost MI-EEG decoding performance, a collaborative decoding architecture (PGCANet) integrating prior guidance and channel awareness was proposed, with dual branches and consistency constraint loss for decision fusion. Results: CEP-MI effectively quantifies participant engagement, and PGCANet achieves superior decoding performance compared to traditional algorithms and advanced models, with the optimal average accuracy reaching 85.77% in binary classification and 82.47% in quadruple classification. Conclusion: Targeting the above-mentioned research gaps, the CEP-MI paradigm provides valuable insights into the imagery-execution transition mechanism. PGCANet enhances MI decoding accuracy via multi-branch collaboration, laying a foundation for MI-BCI’s clinical application in neurorehabilitation. Significance: This study introduces a novel paradigm and collaborative architecture that significantly advance MI-BCI decoding accuracy and engagement quantification, and helps to promote its clinical application in neurorehabilitation.
SN - 1662-5161
ER -