Single-Trial Classification of MEG Recordings

Abstract

While magnetoencephalography (MEG) is widely used to identify spatial locations of brain activations associated with various tasks, classification of single trials in stimulus-locked experiments remains an open subject. Very significant single-trial classification results have been published using electroencephalogram (EEG) data, but in the MEG case, the weakness of the magnetic fields originating from the relevant sources relative to external noise, and the high dimensionality of the data are difficult obstacles to overcome. We present here very significant MEG single-trial mean classification rates of words. The number of words classified varied from seven to nine and both visual and auditory modalities were studied. These results were obtained by using a variety of blind sources separation methods: spatial principal components analysis (PCA), Infomax independent components analysis (Infomax ICA) and second-order blind identification (SOBI). The sources obtained were classified using two methods, linear discriminant classification (LDC) and v-support vector machine (v-SVM). The data used here, auditory and visual presentations of words, presented nontrivial classification problems, but with Infomax ICA associated with LDC we obtained high classification rates. Our best single-trial mean classification rate was 60.1% for classification of 900 single trials of nine auditory words. On two-class problems rates were as high as 97.5%.

Tracking neural coding of perceptual and semantic features of concrete nouns

“…While extensive work decoding nouns and verbs from MEG data has been published (Guimaraes et al, 2007; Pulvermüller et al, 2005; Suppes et al, 1999), only a few studies have looked at the different semantic features that describe a noun. For example, Chan et al (2010) have successfully decoded whether a subject is considering a living vs. nonliving stimulus based on MEG activity, but this leaves open the question of what other semantic features may be encoded by the MEG signal at different times and cortical locations.…”

Partial Orders of Similarity Differences Invariant Between EEG-Recorded Brain and Perceptual Representations of Language

“…When brain wave recordings are time-locked to the presentation of words in a sentential context, certain brain waves are representations of these words. The extensive computational analyses of four experiments presented here are intended to provide evidence for this claim, given in section 4, together with reference to some of our already published results on recognition of brain representations of isolated words (Suppes, Lu, & Han, 1997;Suppes, Han, & Lu, 1998;Suppes, Han, Epelboim, & Lu, 1999;Suppes & Han, 2000;Wong, Perreau-Guimaraes, Uy, & Suppes, 2004;Perreau-Guimaraes, Wong, Uy, Grosenick, & Suppes, 2007) and on recognition of brain representations of sentences (Suppes et al, 1998(Suppes et al, , 1999Wong et al, 2004).…”

“…The classifications presented here involve a large number of features and classes with a relatively small number of trials per class. Even with a good feature reduction scheme, optimizing nonlinear models in the kind of experiments we analyze is certainly a challenge and, in our experience, does not significantly improve classification results (Perreau-Guimaraes et al, 2007). Finally, the aim of this work is not the classification method itself, but to use confusion matrices to study invariant partial orders of similarity differences.…”

EEG classification by ICA source selection of Laplacian-filtered data

“…Single-trials were classified into either "stop" or "go" by using the linear discriminant classification (LDC) procedure described in [11]. The classification accuracy was defined as the ratio of number of trials correctly classified over the total number of trials.…”