A multi-subject, multi-modal human neuroimaging dataset

Abstract

We describe data acquired with multiple functional and structural neuroimaging modalities on the same nineteen healthy volunteers. The functional data include Electroencephalography (EEG), Magnetoencephalography (MEG) and functional Magnetic Resonance Imaging (fMRI) data, recorded while the volunteers performed multiple runs of hundreds of trials of a simple perceptual task on pictures of familiar, unfamiliar and scrambled faces during two visits to the laboratory. The structural data include T1-weighted MPRAGE, Multi-Echo FLASH and Diffusion-weighted MR sequences. Though only from a small sample of volunteers, these data can be used to develop methods for integrating multiple modalities from multiple runs on multiple participants, with the aim of increasing the spatial and temporal resolution above that of any one modality alone. They can also be used to integrate measures of functional and structural connectivity, and as a benchmark dataset to compare results across the many neuroimaging analysis packages. The data are freely available from https://openfmri.org/.

Spatio-temporal reconstruction of brain dynamics from EEG with a Markov prior

“…To further investigate MarkoVG we applied it to EEG recorded during a well studied 230 paradigm, namely the multi-subject multimodal dataset studying face recognition (Wakeman and Henson, 2015). Images of famous faces, unfamiliar faces and scrambled faces were presented to 19 subjects in six runs of 7.5 minutes.…”

“…1 and 2A). For further information on the experimental setup used in the data collection we refer to the documentation provided by Wakeman et al (Wakeman and Henson, 2015). We built forward models in 240 SPM8 using a three layered boundary element method head model (Phillips, 2000).…”

“…The face/scrambled face contrasted fMRI 385 recordings included in the multimodal study we extracted EEG from, have been analyzed at the group level in Fig. 3b in (Wakeman and Henson, 2015). The O/FFAs were also here dominating the face response as compared to the scrambled face condition.…”

Automated rejection and repair of bad trials in MEG/EEG

“…The first is the MNE sample dataset [13] , and RANSAC. Each point represents a subject from the Faces dataset [15]. Auto reject has a better RMSE whenever a point lies above the dotted red line.…”

“…The second is the EEGBCI motor dataset [14] with 104 subjects and the final one is the Faces dataset [15] with 19 subjects. The bad channel annotations for each run and subject was available from the authors of the Faces dataset [15].…”