Evaluating cortical responses to speech in children: A functional near-infrared spectroscopy (fNIRS) study

Abstract

Functional neuroimaging of speech processing has both research and clinical potential. This work is facilitating an ever-increasing understanding of the complex neural mechanisms involved in the processing of speech. Neural correlates of speech understanding also have potential clinical value, especially for infants and children, in whom behavioural assessments can be unreliable. Such measures would not only benefit normally hearing children experiencing speech and language delay, but also hearing impaired children with and without hearing devices. In the current study, we examined cortical correlates of speech intelligibility in normally hearing paediatric listeners. Cortical responses were measured using functional near-infrared spectroscopy (fNIRS), a non-invasive neuroimaging technique that is fully compatible with hearing devices, including cochlear implants. In nineteen normally hearing children (aged 6 - 13 years) we measured activity in temporal and frontal cortex bilaterally whilst participants listened to both clear- and noise-vocoded sentences targeting four levels of speech intelligibility. Cortical activation in superior temporal and inferior frontal cortex was generally stronger in the left hemisphere than in the right. Activation in left superior temporal cortex grew monotonically with increasing speech intelligibility. In the same region, we identified a trend towards greater activation on correctly vs. incorrectly perceived trials, suggesting a possible sensitivity to speech intelligibility per se, beyond sensitivity to changing acoustic properties across stimulation conditions. Outside superior temporal cortex, we identified other regions in which fNIRS responses varied with speech intelligibility. For example, channels overlying posterior middle temporal regions in the right hemisphere exhibited relative deactivation during sentence processing (compared to a silent baseline condition), with the amplitude of that deactivation being greater in more difficult listening conditions. This finding may represent sensitivity to components of the default mode network in lateral temporal regions, and hence effortful listening in normally hearing paediatric listeners. Our results indicate that fNIRS has the potential to provide an objective marker of speech intelligibility in normally hearing children. Should these results be found to apply to individuals experiencing language delay or to those listening through a hearing device, such as a cochlear implant, fNIRS may form the basis of a clinically useful measure of speech understanding.

Keywords: Auditory cortex; Children; Functional near-infrared spectroscopy; Neuroimaging; Speech comprehension; fNIRS.

Fig. 1

Behavioural results collected concurrently with fNIRS imaging. Mean accuracy (a) and response time (b) is shown for each of the four stimulation conditions. The bars show results for children in the present study, while the inset markers show comparative data for normally hearing adults acquired in our previous study (Lawrence et al., 2018). Error bars show ±1 standard error of the mean (corrected for repeated measures). The dashed horizontal line in (a) represents chance performance. Asterisks indicate significant pairwise differences in post-hoc tests following a significant RM-ANOVA omnibus test result (* p < .05, ** p < .01, *** p < .001, Bonferroni-corrected).

Fig. 2

Channel-wise relationship between fNIRS response amplitude and speech stimulus intelligibility. Rows a to c show the results of statistical significance testing (uncorrected p-values, thresholded at p < .05) for 0th-order, 1st-order (linear) and 2nd-order (quadratic) effects, respectively. Individual channels exhibiting significant effects after correction for multiple comparisons (q < 0.05, FDR corrected) are highlighted. Note the maps are interpolated from single-channel results and the overlay on the cortical surface is for illustrative purposes only.

Fig. 3

Mean contrast values (i.e. estimated response amplitude relative to silence; arbitrary units) in select regions of interest. Error bars show ±1 standard error of the mean (corrected for repeated measures). Bold lines indicate the overall LMM fit for each region including polynomial expansion of stimulus condition up to 2nd (quadratic) order.

Fig. 4

Grand average event-related haemodynamic time courses. The red and blue traces show estimated changes in the concentration of HbO and HbR, respectively. Shading indicates ±1 standard error of the mean across participants. Note that, prior to averaging across participants, the mean response to silent trials was subtracted out to derive overlap-reduced event-related responses. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5

Secondary group-level fNIRS analyses. (a) Main effect of perceptual veracity (i.e. the contrast between correct vs. incorrect trials) for the NVLow and NVHigh conditions. (b) Main effect of hemisphere (assessment of laterality across all stimulation conditions). In each case, the colourmap shows uncorrected p-values, thresholded at p < .05. Individual channels showing a significant effect after correction for multiple comparisons (q < 0.05, FDR corrected) are highlighted. Note that the maps are interpolated from single-channel results and the overlay on the cortical surface is for illustrative purposes only.