Cortical contributions to the auditory frequency-following response revealed by MEG

Abstract

The auditory frequency-following response (FFR) to complex periodic sounds is used to study the subcortical auditory system, and has been proposed as a biomarker for disorders that feature abnormal sound processing. Despite its value in fundamental and clinical research, the neural origins of the FFR are unclear. Using magnetoencephalography, we observe a strong, right-asymmetric contribution to the FFR from the human auditory cortex at the fundamental frequency of the stimulus, in addition to signal from cochlear nucleus, inferior colliculus and medial geniculate. This finding is highly relevant for our understanding of plasticity and pathology in the auditory system, as well as higher-level cognition such as speech and music processing. It suggests that previous interpretations of the FFR may need re-examination using methods that allow for source separation.

Figure 1. Audio trace, EEG–ABR and single-channel MEG–ABR grand averages. A single MEG channel was selected by maximum correlation with the EEG channel.

(a) Time course of speech stimulus (syllable: /da/) and EEG/MEG responses, showing that the onset response and the FFR commonly studied with EEG are preserved in the single-channel MEG–ABR. The prestimulus baseline (−50 to 0 ms) and the frequency-following response (FFR) periods (30 to 130 ms) are marked in grey and blue, respectively, for the EEG and MEG responses. (b) Corresponding spectra of the periodic portion of the audio signal and the FFR of the responses are shown in blue. Baseline spectra are in grey (n=20).

Figure 2. Region of interest (ROI) locations and their spectra during FFR and baseline showing a large contribution at the fundamental frequency for brainstem nuclei and auditory cortex in the MEG data.

(a–c) show the spectra of the three subcortical ROIs. (d) Single-subject illustration of the locations of subcortical ROIs; and (e) the locations of the cortical ROIs including two control regions in the frontal and occipital poles. (f–h) Spectra of the cortical ROIs. Error bars indicate s.e. of the mean. All results are averaged across left and right ROIs and across subjects (n=20). AC, auditory cortex; CN, cochlear nucleus; IC, inferior colliculus; MGB, medial geniculate nucleus; controls: FP, frontal pole; OP, occipital pole.

Figure 3. Comparison of FFR topography to that of sensor projected source activity from anatomical ROIs and the cortical ERF.

All topographies are averaged across subjects (n=20). (a) Sensor distribution from bilateral region of interest (ROI) sources, in three-dimensional (side view) and two-dimensional (top view) sensor space. For each ROI, the colour map is scaled to minimum and maximum signal strength to show the distribution of sensor sensitivities to each source. (b) The topography of fundamental frequency magnitude in the FFR. (c) ERF topography at P1 (∼73 ms; independent data set) was significantly correlated with FFR topography for all subjects. The absolute value is taken before correlation with the FFR topography.

Figure 4. Respective timings of the four putative sources.

The explanatory power (coefficient estimates) of ROIs over successive 12 ms windows from sound onset to 72 ms. * Indicates significant increases between successive windows; CN and IC increase only up to 24–36 ms window, whereas MGB increases up to the 36–48 ms window and AC increases up to the 48–60 ms window; contributions from each structure peak successively (n=20). Results are scaled between 0 and 1 to better visualize the relative time courses of ROIs, and grey areas indicate the prestimulus period and the stable FFR period, which were not the subject of the statistical tests. Error bars show s.e.m. and are slightly offset for visibility. (MGB, medial geniculate body; IC, inferior colliculus; CN, cochlear nucleus; AC, auditory cortex.

Figure 5. Cortical asymmetry and behavioural correlations.

(a) Whole-brain MEG source results from a minimum-norm estimate (MNE) volume model (t-statistic parametric map, shown in blue–yellow scale) superimposed on the 1-mm MNI152 standard template in stereotaxic space. ABR–FFR signal strength was greater than baseline in two clusters centred on the auditory cortex. Clusters are significant applying cluster-corrected thresholds to control family-wise error rate (t>2.3, P<0.05); t>6 for visualization purposes. (b) Distribution of left–right amplitude differences in the auditory cortex ROIs for each individual using the mixed surface-volume MNE model (Fig. 2e) shows strong right-sided asymmetry (n=20). Correlations between behavioural variables and f0 amplitude in left and right ROIs: (c) training hours in musicians; more training correlated with stronger f0 in the right AC only (n=11). (d) Age training started in musicians; earlier start ages were correlated with stronger f0 representation in the right AC. (e) Fine pitch discrimination (n=20); finer pitch discrimination correlated with stronger f0 in the right AC. A non-significant trend is present in the left AC.