Afferent-efferent connectivity between auditory brainstem and cortex accounts for poorer speech-in-noise comprehension in older adults

Abstract

Speech-in-noise (SIN) comprehension deficits in older adults have been linked to changes in both subcortical and cortical auditory evoked responses. However, older adults' difficulty understanding SIN may also be related to an imbalance in signal transmission (i.e., functional connectivity) between brainstem and auditory cortices. By modeling high-density scalp recordings of speech-evoked responses with sources in brainstem (BS) and bilateral primary auditory cortices (PAC), we show that beyond attenuating neural activity, hearing loss in older adults compromises the transmission of speech information between subcortical and early cortical hubs of the speech network. We found that the strength of afferent BS→PAC neural signaling (but not the reverse efferent flow; PAC→BS) varied with mild declines in hearing acuity and this "bottom-up" functional connectivity robustly predicted older adults' performance in a SIN identification task. Connectivity was also a better predictor of SIN processing than unitary subcortical or cortical responses alone. Our neuroimaging findings suggest that in older adults (i) mild hearing loss differentially reduces neural output at several stages of auditory processing (PAC > BS), (ii) subcortical-cortical connectivity is more sensitive to peripheral hearing loss than top-down (cortical-subcortical) control, and (iii) reduced functional connectivity in afferent auditory pathways plays a significant role in SIN comprehension problems.

Keywords: Aging; Auditory cortex; Auditory evoked potentials; Frequency-following response (FFR); Functional connectivity; Neural speech processing; Source waveform analysis.

Figure 1:. Audiometric and perceptual results.

(A) Audiograms for listeners with normal hearing (NH) and hearing loss (HL) pooled across ears. Hearing was ~10 dB better in NH vs. HL listeners. (B) Behavioral accuracy for detecting infrequent /ta/ tokens in clear and noise-degraded conditions. Noise-related declines in behavioral performance were prominent but no group differences were observed. (C) Reaction times (RTs) for speech detection were similar between groups and speech SNRs. (D) HL listeners showed more variability and marginally poorer QuickSIN performance than NH listeners. errorbars = ± s.e.m., *p< 0.05.

Figure 2:. ERP (top traces) and FFR (bottom) source waveforms reflect the simultaneous encoding of speech within cortical and brainstem tiers of the auditory system.

(A) NH listeners show a leftward asymmetry in PAC responses compared to HL listeners (B), who show stronger activation in right PAC. Noise weakens the cortical ERPs to speech across the board, particularly in the timeframe of P1 and N1, reflecting the initial registration of sound in PAC. In contrast to cortical responses, BS FFRs are remarkably similar between groups and noise conditions. Shaded regions demarcate the 100 ms speech stimulus. BS, brainstem; PAC, primary auditory cortex.

Figure 3:. Cortical speech processing is modulated by noise interference, hearing status, and cerebral hemisphere.

(A) P2 amplitudes are stronger in NH listeners regardless of SNR. (B) Brain volumes show distributed source activation maps using Cortical Low resolution electromagnetic tomography Analysis Recursively Applied (CLARA; BESA v6.1) (Iordanov et al., 2014). Functional data are overlaid on the BESA brain template (Richards et al., 2016). (C) P2 latency revealed a group x hemispheric interaction. In HL listeners, responses were ~3 ms earlier in right compared to left hemisphere (R

Figure 4:. Brainstem speech processing as a function of noise and hearing loss.

(A) Source FFR spectra for response to clear and degraded speech. Strong energy is observed at the voice fundamental frequency (F0) but much weaker energy tagging the upper harmonics of speech, consistent with age-related declines in high-frequency spectral coding. Group and noise-related effects in FFRs were less apparent than in the cortical ERPs (cf. Fig. 3). errorbars = ± s.e.m.

Figure 5:. Functional connectivity between auditory brainstem and cortex varies with hearing loss and predicts SIN comprehension.

Neural responses are collapsed across hemispheres and SNRs. (A) Transfer entropy reflecting directed (casual) afferent neural signaling from BS→PAC. Afferent connectivity is stronger in normal-hearing compared to hearing-impaired listeners. (B) Afferent connectivity is weaker in listeners with poorer hearing (i.e., worse PTA thresholds) and predicts behavioral SIN performance (C). Individuals with stronger BS→PAC connectivity show better (i.e., lower) scores on the QuickSIN. (D) Efferent neural signaling from PAC→BS does not vary between NH and HL listeners, suggesting similar top-down processing between groups. Similarly, efferent connectivity did not covary with hearing loss (E) nor did it predict SIN comprehension (F). Solid lines=significant correlations; dotted lines=n.s. relationships. errorbars = ± s.e.m., ***p<0.001, ****p<0.0001.

Figure 6:. Afferent neural signaling from BS to PAC mediates the relation between hearing loss and SIN comprehension.

Sobel mediation analysis (Sobel, 1982) between listeners’ hearing loss (PTA thresholds), neural connectivity (BS→PAC signaling), and SIN comprehension (QuickSIN scores). Edges show significant relations between pairwise variables identified via linear regression. (A) Hearing loss by itself strongly predicts QuickSIN scores such that reduced hearing is associated with poorer SIN comprehension. (B) Accounting for BS→PAC afferent connectivity renders this relation insignificant (Sobel test: z=2.42, p=0.016; Sobel, 1982), indicating the strength of neural communication between BS and PAC, rather than hearing loss per se, mediates older adults’ SIN comprehension. **p <0.01, ***p<0.001.