Mapping tonotopic organization in human temporal cortex: representational similarity analysis in EMEG source space

Abstract

A wide variety of evidence, from neurophysiology, neuroanatomy, and imaging studies in humans and animals, suggests that human auditory cortex is in part tonotopically organized. Here we present a new means of resolving this spatial organization using a combination of non-invasive observables (EEG, MEG, and MRI), model-based estimates of spectrotemporal patterns of neural activation, and multivariate pattern analysis. The method exploits both the fine-grained temporal patterning of auditory cortical responses and the millisecond scale temporal resolution of EEG and MEG. Participants listened to 400 English words while MEG and scalp EEG were measured simultaneously. We estimated the location of cortical sources using the MRI anatomically constrained minimum norm estimate (MNE) procedure. We then combined a form of multivariate pattern analysis (representational similarity analysis) with a spatiotemporal searchlight approach to successfully decode information about patterns of neuronal frequency preference and selectivity in bilateral superior temporal cortex. Observed frequency preferences in and around Heschl's gyrus matched current proposals for the organization of tonotopic gradients in primary acoustic cortex, while the distribution of narrow frequency selectivity similarly matched results from the fMRI literature. The spatial maps generated by this novel combination of techniques seem comparable to those that have emerged from fMRI or ECOG studies, and a considerable advance over earlier MEG results.

Tonotopic representation of loudness in the human cortex

“…5 cf. Baumann et al., 2013, Moerel et al., 2014, Saenz and Langers, 2014, Su et al., 2014). Improvements in source reconstruction (through the gathering of more data or improved inverse techniques) may reveal this layout more accurately.…”

Relating dynamic brain states to dynamic machine states: human and machine solutions to the speech recognition problem

“…The possibility of engaging on this enterprise depends on new methods for non-invasively investigating the real-time electrophysiological activity of the human brain. We build here on earlier work [1] which used a novel multivariate pattern analysis method, called Spatiotemporal Searchlight Representational Similarity Analysis (ssRSA), to decode information about frequency preference and selectivity directly from the dynamic neural activity of the brain as reconstructed in combined MEG and EEG (EMEG) source space. This method is an extension of fMRI-based RSA [2,3] to time-resolved imaging modalities.…”

“…The brain data used in this study come from simultaneously recorded EEG and MEG (EMEG) measurements of the neural activity generated by each of 400 spoken words heard by the listeners. Critically, we use these EMEG data, as in the earlier Su et al [1] study, to generate a 'source space' reconstruction that localises at the cortical surface (specifically, the white matter/grey matter boundary) the electrophysiological activity that gives rise to the EMEG measurements recorded at sensors external to the skull. The combination of MEG and EEG delivers better source localisation than either of these modalities alone [23], using well established minimum norm estimation (MNE) techniques guided by neuroanatomical constraints from structural MR for each participant [24,25].…”

“…RSA was originally developed for fMRI [5], but one of its strengths is its agnosticism toward modality. In particular, RSA has more recently been applied to EEG and MEG data [1,31]. Methodological considerations for EMEG are detailed in Su et al's work [4], but we will give a brief overview here.…”

“…RSA involves the comparison of the similarity (or dissimilarity) structures found in the responses of neural systems to differing experimental conditions, and the modelling of those similarity structures by categorical or computational models. A basic component of the RSA analysis strategy is the construction of representational dissimilarity matrices (RDMs) based on the pairwise dissimilarity of all items or conditions entered into the analysis process [1,5]. Typically, the data RDM or set of data RDMs is computed from a neuroimaging data source, and the model RDM(s) from a specified model of the stimuli.…”