Oscillatory activity in neocortical networks during tactile discrimination near the limit of spatial acuity

Abstract

Oscillatory interactions within functionally specialized but distributed brain regions are believed to be central to perceptual and cognitive functions. Here, using human scalp electroencephalography (EEG) recordings combined with source reconstruction techniques, we study how oscillatory activity functionally organizes different neocortical regions during a tactile discrimination task near the limit of spatial acuity. While undergoing EEG recordings, blindfolded participants felt a linear three-dot array presented electromechanically, under computer control, and reported whether the central dot was offset to the left or right. The average brain response differed significantly for trials with correct and incorrect perceptual responses in the timeframe approximately between 130 and 175ms. During trials with correct responses, source-level peak activity appeared in the left primary somatosensory cortex (SI) at around 45ms, in the right lateral occipital complex (LOC) at 130ms, in the right posterior intraparietal sulcus (pIPS) at 160ms, and finally in the left dorsolateral prefrontal cortex (dlPFC) at 175ms. Spectral interdependency analysis of activity in these nodes showed two distinct distributed networks, a dominantly feedforward network in the beta band (12-30Hz) that included all four nodes and a recurrent network in the gamma band (30-100Hz) that linked SI, pIPS and dlPFC. Measures of network activity in both bands were correlated with the accuracy of task performance. These findings suggest that beta and gamma band oscillatory networks coordinate activity between neocortical regions mediating sensory and cognitive processing to arrive at tactile perceptual decisions.

Keywords: Brain rhythms; Connectivity; EEG; ERP; Granger causality; Oscillations; Perceptual decision-making; Somatosensory; Touch.

Figure 1. Experimental set-up and behavioral performance

A. The middle dot of a raised 3-dot array was offset either to the right or left. A pneumatically driven stimulator presented stimuli to the right index fingerpad for 1s (on-interval) and participants responded within the next 2s (off-interval). B. Behavioral performance accuracy rates of better performers are displayed here. Better performers have at least 70 % correct responses in two runs.

Figure 2. Event-related potentials (ERPs)

The shaded regions covering two or three labeled circles, in the 68-electrode EEG recording montage, show the locations in sensor space where ERPs differed significantly for trials with correct and incorrect responses in the timeframe 140 - 175 ms. The average waveforms, at representative sites from the shaded regions, are also shown for correct and incorrect responses.

Figure 3. Spatiotemporal profiles of peak source-level electrophysiological activity during trials with correct responses

The top row represents minimum-norm estimate (MNE) sources with peak activity over L SI at 45 ms, R LOC at 130 ms, R pIPS at 160 ms and L dlPFC at 175 ms. The bottom row represents the corresponding fitted dipoles with their orientations. Abbreviations as in text.

Figure 4. Spectral power and coherence

Beta band (12- 30 Hz) (A), and gamma band (30-100 Hz) (B) spectral power during earlier (30-140 ms) and later (140 – 210 ms) periods. Power increases occurred only in the gamma band activity of pIPS and dlPFC. (C-D). For correct responses, pIPS-dlPFC coherence increased significantly in the later period.

Figure 5. Granger causality and net causal flow in the frequency range (0 – 30 Hz)

(A-F) Granger causality spectra when participants responded correctly; the peak causal influence is seen at roughly 13-16 Hz (low beta range). Significance thresholds (shown by dotted lines) as in text. (G) Schematic representation of the beta band Granger causality network graph associated with correct responses, based on the bivariate (pairwise) and trivariate (conditional) Granger causality results. (H) Significant changes in net causal inflow between earlier and later durations.

Figure 6. Granger causality and net causal flow in the gamma band (30 – 100 Hz)

(A-F). Granger causality spectra when participants responded correctly. Significance thresholds (shown by dotted lines) as in text. (G) The conditional Granger causality spectra SI→ pIPS| dlPFC (green line). (H) Schematic representation of the gamma band network activity associated with correct responses, based on the bivariate and trivariate (conditional) Granger causality results. (I) Changes in net causal inflow (total incoming causal flow minus total outgoing flow for a node) between earlier and later durations.

Figure 7. Relation between coherence and behavioral performance

Pairs of nodes showing positive correlations of beta band (A-B) and gamma band (C-D) coherence with behavioral performance (p < 0.05).

Figure 8. Relation between Granger causality and behavioral performance

Beta band (A) and gamma band (C-D) Granger causality relationships among node pairs (p < 0.05).