Decoding reach-to-grasp from EEG using classifiers trained with data from the contralateral limb.

Area of Science:

• Neuroscience
• Motor Control
• Human-Computer Interaction

Background:

• Human movement relies on object interaction, with reach-to-grasp intentions decoded from electroencephalography (EEG) signals.
• Understanding shared neural information between limbs is crucial for decoding motor intentions across hands.

Purpose of the Study:

• To investigate the extent of shared information in EEG signals between two limbs for cross-hand decoding.
• To determine if motor intentions for one limb can be decoded using EEG data from the contralateral limb.

Main Methods:

• Ten subjects interacted with a novel object, performing cued actions (lift/touch) with either the left or right hand.
• EEG data was recorded from 30 channels in bilateral frontal-central-parietal regions.
• Linear Discriminant Analysis (LDA) classifiers were trained on data from one limb and tested on the contralateral limb.

Main Results:

• Hand-object interaction types were decoded with peak accuracies of 59% (planning) and 69% (execution).
• Decoding accuracy for reaching directions varied based on spatial mirroring of EEG channels and coordinate system (extrinsic vs. intrinsic).

Conclusions:

• EEG signals contain shared information between limbs, allowing for cross-hand decoding of motor intentions.
• The effectiveness of cross-hand decoding depends on spatial representation and coordinate systems used for analyzing reaching directions.