Target of selective auditory attention can be robustly followed with MEG

Abstract

Selective auditory attention enables filtering of relevant acoustic information from irrelevant. Specific auditory responses, measurable by magneto- and electroencephalography (MEG/EEG), are known to be modulated by attention to the evoking stimuli. However, such attention effects have typically been studied in unnatural conditions (e.g. during dichotic listening of pure tones) and have been demonstrated mostly in averaged auditory evoked responses. To test how reliably we can detect the attention target from unaveraged brain responses, we recorded MEG data from 15 healthy subjects that were presented with two human speakers uttering continuously the words "Yes" and "No" in an interleaved manner. The subjects were asked to attend to one speaker. To investigate which temporal and spatial aspects of the responses carry the most information about the target of auditory attention, we performed spatially and temporally resolved classification of the unaveraged MEG responses using a support vector machine. Sensor-level decoding of the responses to attended vs. unattended words resulted in a mean accuracy of [Formula: see text] (N = 14) for both stimulus words. The discriminating information was mostly available 200-400 ms after the stimulus onset. Spatially-resolved source-level decoding indicated that the most informative sources were in the auditory cortices, in both the left and right hemisphere. Our result corroborates attention modulation of auditory evoked responses and shows that such modulations are detectable in unaveraged MEG responses at high accuracy, which could be exploited e.g. in an intuitive brain-computer interface.

Figure 1

Temporally and spatially resolved decoding reveals highest decoding accuracy around 160 ms after each stimulus word and in the MEG channels above the auditory cortices. (a) Time-resolved decoding in a 2-s time window of attended vs. unattended word stimulus plotted for a representative subject (top) and for the group (bottom). The mean decoding accuracy is shown as a dark blue line for the “Yes” and as a light blue line for the “No” stimulus word. The standard deviation (SD), computed across the cross-validation folds of the classifier training and testing, is shown as dark/light blue shading. (b) Spatially-resolved decoding accuracy maps in a representative subject (top) and at the group level (bottom) for attended vs. unattended stimulus words. Prior to the decoding, epochs of the high- and low-pitch words were concatenated.

Figure 2

Target of selective auditory attention could be reliably detected in all subjects. The dark blue circles and light blue squares indicate the accuracy of the entire-epoch (all data points in a 1-s window and all channels given to the decoder) classification of responses to attended vs. unattended “Yes” and “No” word-stimuli for all subjects. The standard deviation (SD) was computed over the five cross-validation folds of the decoder and is shown as plot whiskers for each subject.

Figure 3

Source estimation of the MEG evoked responses corroborates attention modulation in auditory cortical regions. Top: Source estimates of the evoked responses to attended and unattended word stimuli; estimates averaged across the group (N = 11). The colour represents the source amplitude first normalized to the absolute peak value of each individual source estimate and then averaged across subjects. Bottom: The temporal dynamics of left (LH) and right (RH) auditory-cortex activation to attended and unattended stimulus words (“Yes”/“No”), extracted from the source estimate at the coloured dots (green/blue in the top panel).

Figure 4

Auditory cortical regions generate the signals most informative of the attended stimulus stream. The colour gradient (yellow highest) represents the source-space spatial searchlight decoding result averaged across the subject group (N = 11). Each color dot represents the accuracy peak in one subject.

Figure 5

Experimental design: By a text cue on the screen (“LEFT-YES” or “RIGHT-NO”), subjects were instructed to attend either the “Yes” or the “No” stimulus stream. (a) Stimulus timing in 2-s time window with the onsets of the stimulus words marked as black triangles. (b) The virtual arrangement of the speakers uttering the words “Yes” and “No” with respect to the subject. (c) The structure of each stimulus stream. The low- and high-pitch versions of the word stimuli alternate (standard) but occasionally three high-pitch versions are presented (deviant).