/ EN
2026-08-21 00:00 China Papers Foundations & Methods Translated from EN

EFS-Net Fuses EEG and fNIRS for Hybrid BCI Decoding

Summary Researchers proposed EFS-Net, an end-to-end network that aligns fast EEG activity with slower fNIRS haemodynamic signals through temporal, spatial and cross-modal branches. In subject-specific leave-one-session-out validation, the model reached 77.71% accuracy on a word-generation dataset and 81.69% on a mental-arithmetic dataset.
Why it matters The architecture directly tackles the temporal and spatial mismatch at the heart of EEG-fNIRS hybrid BCIs and reports comparable results on public datasets.

BCIwiki (bciwiki.com) – EFS-Net, an end-to-end fusion network for hybrid brain-computer interfaces, reached 81.69% ± 9.49% classification accuracy on a mental arithmetic task and 77.71% ± 8.23% on word generation, results the authors report as ahead of state-of-the-art unimodal algorithms and conventional fusion models. The work was published in Cognitive Neurodynamics on August 21, 2026.

Hybrid interfaces combine neuroimaging modalities so that one covers what another misses. For the pairing of electroencephalography and functional near-infrared spectroscopy, the obstacle is that fNIRS localises activity well but carries a delayed haemodynamic response, while EEG offers millisecond timing with limited spatial detail, leaving the two streams misaligned in both time and space. EFS-Net addresses that mismatch with a multi-scale spatio-temporal fusion architecture built on spatial calibration.

The network runs three complementary branches: a multi-scale temporal convolution branch that captures rapidly changing cortical electrophysiological features in EEG, an EEG spatial branch that constructs latency-compensated cortical topographies to accommodate the slower haemodynamic response, and a spatially calibrated fNIRS branch that dynamically fuses spatial feature maps with their EEG counterparts to produce temporally aligned, spatially enhanced neural representations. Evaluation used subject-specific leave-one-session-out cross-validation on the public Word Generation and Mental Arithmetic datasets, and visualisations indicate the alignment strategy recovers realistic cortical spatial distributions. The authors are based at East China University of Science and Technology, the University of Essex, and partner institutions.

Compiled by BCIwiki from public sources

Sources · 1

From these organizations

2026-09 University of Essex Speech Imagery BCIs May Be Overrated as Only 36% of Participants Reach Significance 2026-08 East China University of Science and Technology AutoMI: Hands-Free Motor Imagery EEG Classification via LLM Multi-Agents 2026-06 Ministry of Education China's 8 Ministries Push AI-BCI Integration in Consumer Policy 2026-04 University of Essex Enhancing the Capability and Accuracy of Motor Imagery Classification: A Deep Neural Network-Powered Multifaceted Strategy Model
Read original ↗
Suggest a correction Revisions · none / EN
© 2026 BCIwiki.com Digest Topics Tips Subscribe Revisions About