Application of Similarity Measures to Magnetoencephalography data

Richard Gault, TM McGinnity, SA Coleman

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review


Magnetoencephalography (MEG) is a non-invasive neural imaging technique which passively measures the miniscule magnetic fields produced by neuronal activity without any risk to the subject. Its superior temporal resolution has led to its use in multimodal studies alongside neural imaging techniques, such as functional magnetic resonance imaging (fMRI), which provide high spatial resolution. Multivariate analysis (MVA) is a well-established field of statistics and is currently applied to MEG data for the purposes of artefact identification using principal component analysis (PCA) and independent component analysis (ICA). This paper considers how the similarity measures of Frobenius norm, PCA similarity measure (SPCA) and Eros can directly analyse the multivariate data produced from MEG recordings. These techniques are applied to auditory stimuli to evaluate to what extent different stimuli can be distinguished by the corresponding neural activity. The results show that Frobenius norm finds dissimilarity between the neural response to distinct tones while Eros and SPCA show similarity between neural responses to certain pairs of tones. The results will be used to inform future studies where the measures identified in this study can be used as part of classification algorithms as well as provide a basic measure to map the similarities in the conditions to the similarity of the neural responses.
Original languageEnglish
Title of host publicationUnknown Host Publication
PublisherUlster University
Number of pages6
Publication statusPublished (in print/issue) - 27 Aug 2014
EventIrish Machine Vision and Image Processing 2014 -
Duration: 27 Aug 2014 → …


ConferenceIrish Machine Vision and Image Processing 2014
Period27/08/14 → …


Dive into the research topics of 'Application of Similarity Measures to Magnetoencephalography data'. Together they form a unique fingerprint.

Cite this