Bimodal electroencephalography-functional magnetic resonance imaging dataset for inner-speech recognition

Area of Science:

• Neuroscience
• Biomedical Engineering
• Computer Science

Background:

• Brain-computer interfaces (BCIs) face challenges in recognizing inner speech, limiting communication for non-verbal individuals.
• Existing datasets lack multimodal data integration, hindering performance improvements in inner speech decoding.
• Combining neuroimaging modalities like fMRI and EEG offers complementary spatial and temporal resolutions for enhanced BCI applications.

Purpose of the Study:

• To present the first publicly available bimodal dataset for inner speech recognition, integrating EEG and fMRI data.
• To facilitate advancements in speech prostheses and communication aids for patients with speech impairments.
• To enable research into multimodal fusion techniques for improved decoding of inner speech.

Main Methods:

• Acquisition of non-simultaneous electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) data during inner speech production.
• Data collected from four healthy, right-handed participants performing an inner-speech task involving social and numerical word categories.
• Each participant completed 320 trials per modality (EEG and fMRI) across 8 distinct word stimuli.

Main Results:

• The study successfully generated and curated a novel bimodal EEG-fMRI dataset for inner speech.
• The dataset comprises 320 trials per participant across two neuroimaging modalities.
• This resource is now publicly available for research purposes.

Conclusions:

• The presented bimodal EEG-fMRI dataset is a significant contribution to the field of inner speech recognition.
• This resource is expected to accelerate the development of more effective brain-computer interfaces and speech prostheses.
• Future research can leverage this multimodal dataset to explore advanced decoding algorithms and fusion strategies.