Decoding Visual Imagination and Perception from EEG via Topomap Sequences

Area of Science:

• Neuroscience
• Cognitive Science
• Biomedical Engineering

Background:

• Distinguishing between visual imagination and perception is crucial for understanding cognitive processes.
• Existing electroencephalography (EEG) decoding methods may overlook critical spatiotemporal patterns.
• Data scarcity poses a significant challenge in developing robust EEG-based brain-computer interfaces (BCIs).

Purpose of the Study:

• To develop and validate a Topomap-based EEG decoding framework for differentiating pictorial imagination from perception.
• To assess the framework's performance in data-scarce scenarios using a leave-one-subject-out (LOSO) cross-validation.
• To explore the potential of Topomaps as effective EEG feature representations for cognitive state decoding.

Main Methods:

• EEG signals were converted into dense sequences of scalp voltage maps (Topomaps).
• A convolutional neural network (CNN) with squeeze-and-excitation (SE) blocks was applied to Topomap sequences.
• A leave-one-subject-out (LOSO) cross-validation scheme was employed with a single trial per subject.

Main Results:

• The proposed framework achieved 95.1% accuracy in distinguishing imagination from perception under data-scarce conditions.
• Results demonstrated clear neural distinctions between imagination and perception states.
• The study confirmed the viability and generalizability of Topomaps for EEG feature representation.

Conclusions:

• Topomap-based EEG decoding offers a robust method for differentiating visual imagination and perception, even with limited data.
• The framework has potential applications in enhancing diagnostic tools for cognitive disorders.
• Future work could extend this approach to other modalities and advanced deep learning architectures for improved BCI applications.