Magnetoencephalography (MEG)-based Brain-Computer Interface for detecting Motor and Cognitive Imagery

We release a 306-channel MEG-BCI data recorded at 1KHz sampling frequency during four mental imagery tasks (i.e. hand imagery, feet imagery, subtraction imagery, and word generation imagery). The current dataset is the only publicly available MEG imagery BCI dataset as per our knowledge.
Published in Research Data
Like

What is Magnetoencephalography (MEG)?

Magnetoencephalography (MEG) is a non-invasive neurophysiological technique that measures the magnetic fields generated by the neuronal activity of the brain. MEG recording provides timing as well as spatial information about brain activity, which makes it beneficial over electroencephalography (EEG). 

Elekta Neuromag TRIUX, Northern Ireland Functional Brain Mapping (NIFBM), University of Ulster, UK

What are MEG-based BCI Systems?

Brain-computer interfaces (BCIs) are often used for neurorehabilitation therapies and for developing control and communication systems for patients suffering from various neurological disorders. Motor imagery (MI) and Cognitive imagery (CI) practices through BCI have been found to be useful as a therapeutic substitute for standard rehabilitation practices for post-stroke patients. Current BCI systems may use MEG, electroencephalography (EEG), functional magnetic resonance imaging (fMRI) or electrocorticography (ECoG) approaches for mapping brain responses. 

MEG allows us to record temporal characteristics of brain activations with sub-millisecond precision which is a big advantage over fMRI technology where poor temporal resolution posses significant constraints. On the other hand, MEG provides better spatial localization of neural activities as compared to EEG, thus providing the right balance of spatial and temporal resolutions.

Where and why this Data was recorded?

The dataset was recorded by Dr Dheeraj Rathee at Northern Ireland Functional Brain Mapping (NIFBM) Facility at the Ulster University under the supervision of Professor Girijesh Prasad. The NIFBM is the only brain imaging facility on the whole island of Ireland and one of nine in the whole UK. This is one out of only 170 active MEG labs worldwide. We have processed and analysed the dataset at the BCI lab of the University of Essex with the support from ESRC Business and Local Government Data Research Centre

To understand the reason behind recording this dataset, let me first preface it with some background information. Multiple research groups, such as the Berlin Brain-Computer Interface (BBCI), Laboratory of BCI at the Graz University of Technology, BCI group at the Wadsworth centre and others, have open-sourced datasets such as BNCI Horizon 2020 EEG database, BCI competition II, BCI competition III, and BCI competition IV. These open BCI datasets have created growing research interest in BCI system development for different experimental paradigms. However, the majority of these datasets are recorded with EEG and there is a lack of an open MEG-based BCI dataset, which may be due to the high cost of installation and maintenance of MEG systems. This dataset provides a new opportunity to the BCI community and can be used to address the following research challenges:

  1. Evaluation of existing BCI-related data analysis pipeline using MEG dataset.
  2. Development of new machine learning and pattern recognition algorithms for detecting MI and CI tasks in the MEG dataset.
  3. Development of new signal processing methods for MEG dataset to enhance single-trial classification performance.
  4. Development of techniques to correct head movement and enhance the signal-to-noise ratio in the MEG dataset.
  5. Analysis of the non-stationarity in the data. Can we develop novel methods to measure the amount of non-stationarity in the MEG dataset? 

We strongly believe that non-stationarity in MEG signals can be a big factor for low accuracy in detecting single-trial events. It has been noticed that while transitioning between MEG recording sessions the feature representation changes significantly. Even when different sessions on the same participant are considered, BCI systems need recalibration due to the non-stationary nature of the MEG signals leading to high inter-session inconsistency. In the EEG-based BCI system, a lot of research has focused on either development of novel feature extraction methods or the improvement of current classification methods to tackle the issue of non-stationarity. Thus, we encourage the BCI community to utilise this dataset to develop and evaluate similar methods for MEG-based BCI systems. Furthermore, this dataset can be used by the scientific community towards the development of novel pattern recognition machine learning methods to detect brain activities related to MI and CI tasks using MEG signals.

To help facilitate future work, we open-sourced our codes at GitHub to process the raw MEG BIDS format dataset. The codes can be used to reproduce common spatial pattern (CSP) algorithm-based features and evaluate the single-trial classification performance for six pair-wise binary combinations. The data descriptor is available on Scientific Data [link to publication].

Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in

Subscribe to the Topic

Research Data
Research Communities > Community > Research Data

Related Collections

With collections, you can get published faster and increase your visibility.

Remote sensing data for changes in land use

This Collection comprises a series of articles presenting data on changes to land use in urban areas, farmland, forests, and natural environments, as determined using remote sensing techniques.

Publishing Model: Open Access

Deadline: Jan 31, 2024

Medical imaging data for digital diagnostics

This Collection presents a series of articles describing annotated datasets of medical images and video. All medical specialities are considered and data can be derived from study participants, tissue samples, electronic health records (EHRs) or other sources.

Publishing Model: Open Access

Deadline: Dec 20, 2023