Decoding Finger-Level Tactile Imagery from Unilateral Hand for Non-invasive Fine-Grained Somatosensory BCI
OBJECTIVE: Fine-grained motor decoding alone is insufficient for functional dexterity, which also requires precise somatosensory representations. However, tactile imagery (TI) at the individual-finger level remains underexplored in non-invasive EEG-BCIs. In this study, we investigated whether endogenous TI could be decoded from non-invasive EEG at the individual-finger level within a unilateral hand. METHODS: Seventeen healthy participants performed a novel tactile imagery paradigm involving the endogenous mental simulation of touch at four intra-hand sites (thumb, middle finger, little finger, and wrist). We characterized the spatiotemporal neural representations and evaluated decoding performance across three frameworks: FBCSP-SVM, EEGNet, and IFNet. RESULTS: The IFNet architecture achieved the highest four-class classification accuracy of $46.2 \pm 17.3\%$, demonstrating the feasibility of fine-grained TI decoding. Further neurophysiological analysis showed that finger-level TI induced distinguishable cortical responses, characterized by low-gamma modulations and bilateral finger-specific ERD patterns. Moreover, systematic subcombination analysis yielded a best unilateral-hand binary classification accuracy of $71.8 \pm 14.2\%$, showing that intra-hand TI was decodable at the individual level, whereas wrist-based imagery was necessary to achieve robust group-level BCI usability. CONCLUSION: Fine-grained tactile imagery can be decoded from non-invasive EEG signals, indicating that endogenous TI elicits distinguishable somatosensory-related neural patterns and may serve as a feasible non-invasive BCI control paradigm. SIGNIFICANCE: The proposed tactile imagery-based decoding framework provides a promising approach for developing fine-grained non-invasive somatosensory BCIs, which may support finger-level command generation and complement conventional motor-based paradigms for rehabilitation and human-machine interaction.
Authors
- Lin Yao (ORCID: https://orcid.org/0000-0003-2065-7280)
- Jiaheng Wang
- Mingyu Wang
- Yueming Wang
Institutions
- Zhejiang University (CN)
Publication Details
- Journal
- IEEE Transactions on Biomedical Engineering
- Published
- 2026-11-17
- DOI
- https://doi.org/10.1109/tbme.2026.3754557
- Primary Topic
- EEG and Brain-Computer Interfaces
- Type
- article
- Field-Weighted Citation Impact
- 0.00