/ EN
2026-09-11 00:00 Papers Foundations & Methods Translated from EN

fMRI-Guided Training Lifts EEG Individual-Finger Decoding Accuracy to 74.53%

Summary Because the fingers' representations sit close together in the motor cortex, telling individual finger movements apart from scalp EEG has long been difficult. Researchers first learned a set of spectral projections from simultaneously recorded EEG and functional MRI (fMRI), then used the class geometry derived from fMRI to correct EEG predictions. In tests on 12 healthy participants, group average accuracy for two-class movement execution rose from 66.93% to 74.53%, and for three-class execution from 44.83% to 56.58%; for two-class motor imagery it rose from 80.78% to 85.63%. At inference the system uses EEG alone, with no paired fMRI data required.
Why it matters Finger representations sit close together in the cortex and volume conduction blurs EEG signals, so non-invasive devices have struggled to tell which finger is moving. This study takes a different tack, borrowing fMRI's high spatial resolution to constrain the EEG model during training while requiring only an EEG cap in use. Two-class movement execution accuracy improved by about 7.6 percentage points and motor imagery by nearly 5; but with just 12 healthy participants and offline, retrospective decoding, the approach is still some distance from patient use.

BCIwiki (bciwiki.com) — A preprint proposes using fMRI-derived representation information to constrain an EEG decoding model, letting a non-invasive brain-computer interface tell individual fingers apart without adding electrodes. The work was posted to the arXiv preprint server on September 11, 2026, and has not been peer reviewed. The framework, called FRIST (fMRI Representation-Informed Shared-Space Training), works in two stages: it first learns fMRI-informed spectral projections from simultaneously recorded EEG and fMRI, then uses fMRI-derived class geometry to guide residual refinement of EEG predictions.

Finger representations sit close together in the sensorimotor cortex, and volume conduction blurs the signal, making individual-finger decoding from scalp EEG difficult. FRIST transfers information across recordings through shared finger labels without requiring paired trials, and uses only EEG at inference. The study enrolled 12 able-bodied participants and evaluated two-class and three-class chronological session-held-out decoding during movement execution and motor imagery, simulating an online scenario.

With EEGNet as the EEG feature extractor, FRIST raised group average accuracy from 66.93% to 74.53% for two-class movement execution, from 44.83% to 56.58% for three-class movement execution, from 80.78% to 85.63% for two-class motor imagery, and from 60.93% to 69.90% for three-class motor imagery. The method also improved EEG-only decoding when the target participant's own fMRI data were unavailable, and generalized across decoding backbones, reaching 87.40% for two-class motor imagery and 72.54% for three-class motor imagery with EEG Conformer. The authors conclude that fMRI provides useful spatial constraints for EEG representation learning, offering a multimodal strategy that combines fMRI's spatial specificity with EEG's real-time applicability.

Compiled by BCIwiki from public sources

Sources · 1
arxiv.org 2026-09-11
Read original ↗
Suggest a correction Revisions · none / EN
© 2026 BCIwiki.com Digest Topics Tips Subscribe Revisions About