Decoding Hearing-Related Changes in Older Adults' Spatiotemporal Neural Processing of Speech Using Machine Learning

Abstract

Speech perception in noisy environments depends on complex interactions between sensory and cognitive systems. In older adults, such interactions may be affected, especially in those individuals who have more severe age-related hearing loss. Using a data-driven approach, we assessed the temporal (when in time) and spatial (where in the brain) characteristics of cortical speech-evoked responses that distinguish older adults with or without mild hearing loss. We performed source analyses to estimate cortical surface signals from the EEG recordings during a phoneme discrimination task conducted under clear and noise-degraded conditions. We computed source-level ERPs (i.e., mean activation within each ROI) from each of the 68 ROIs of the Desikan-Killiany (DK) atlas, averaged over a randomly chosen 100 trials without replacement to form feature vectors. We adopted a multivariate feature selection method called stability selection and control to choose features that are consistent over a range of model parameters. We use parameter optimized support vector machine (SVM) as a classifiers to investigate the time course and brain regions that segregate groups and speech clarity. For clear speech perception, whole-brain data revealed a classification accuracy of 81.50% [area under the curve (AUC) 80.73%; F1-score 82.00%], distinguishing groups within ∼60 ms after speech onset (i.e., as early as the P1 wave). We observed lower accuracy of 78.12% [AUC 77.64%; F1-score 78.00%] and delayed classification performance when speech was embedded in noise, with group segregation at 80 ms. Separate analysis using left (LH) and right hemisphere (RH) regions showed that LH speech activity was better at distinguishing hearing groups than activity measured in the RH. Moreover, stability selection analysis identified 12 brain regions (among 1428 total spatiotemporal features from 68 regions) where source activity segregated groups with >80% accuracy (clear speech); whereas 16 regions were critical for noise-degraded speech to achieve a comparable level of group segregation (78.7% accuracy). Our results identify critical time-courses and brain regions that distinguish mild hearing loss from normal hearing in older adults and confirm a larger number of active areas, particularly in RH, when processing noise-degraded speech information.

Keywords: aging; event-related potentials; hearing loss; machine learning; speech perception; stability selection and control; support vector machine.

FIGURE 1

Behavioral audiograms (hearing thresholds) per group. NH, normal-hearing listeners; HI, mild hearing loss listeners; PTA, puretone average threshold.

FIGURE 2

Source-level ERPs for the NH and HI groups in representative ROIs. Solid lines = HI; dotted lines = NH. (A) Clear speech responses. (B) Noise-degraded speech responses. Baseline was corrected to the prestimulus interval. NH, normal hearing; HI, hearing impaired; L, Left; R, Right; lPT, parstriangularis L; lPRC, precentral L; rPRC, precentral R; rTRANS, transverse temporal R.

FIGURE 3

Time-varying group classification (NH vs. HI) as a function of neural data (clear and noise conditions) and hemisphere. Group classification accuracy from (A) Whole-brain data (all 68 ROIs), (B) LH data alone (34 ROIs), and (C) RH data alone (34 ROIs). LH, left hemisphere; RH, right hemisphere. 0 ms = stimulus onset. Green solid line indicates group segregation during clear speech perception, red dotted line indicates group segregation during noise-degraded speech perception.

FIGURE 4

Maximum classifier accuracy (y axis) and corresponding latency (x axis) for distinguishing NH and HI listeners using source amplitudes from the whole-brain (blue triangle), and LH (orange square) vs. RH (green circle) separately. (A) Clear speech responses. (B) Noise-degraded speech responses.

FIGURE 5

Effect of stability score threshold on model performance. The bottom of x-axis has four labels; Stability score, represents the stability score range of each bin (scores: 0∼1); Number of features, number of features under each bin; % features, corresponding percentage of selected features; ROIs, number of cumulative unique brain regions up to lower boundary of the bin. (A) Clear speech. (B) Noise-degraded speech.

FIGURE 6

Stable (most consistent) neural network distinguishing NH and HI listeners during clear speech processing. Visualization of brain ROIs corresponding to 0.50 stability threshold (12 top selected ROIs which segregate groups at 81.8%) for clear speech perception. (A) LH; (B) RH; (C) Posterior view; (D) Anterior view. Stability score (color legend): (0.70 ≤ pink ≤ 1.0); (0.60 ≤ blue < 0.70); (0.50 ≤ white < 0.60). L, Left; R, Right; rTP, temporal pole R; rFUS, fusiform; lSP, sperior parietal L; rPRC, precentral R; lPRC, precentral L; rCMF, caudal middle frontal R; lPREC, precuneus L; lMT, middle temporal L; rIST, isthmuscingulate R; lBKS, bankssts L; rBKS, bankssts R; lST, superior temporal L.

FIGURE 7

Stable (most consistent) neural network that distinguishes NH and HI listeners during noise-degraded speech processing. 16 top selected ROIs, 78.7% group classification. (A) LH; (B) RH; (C) Posterior view; (D) Anterior view. lRMF, rostral middle frontal L; rIT, inferior temporal R; lIP, inferior parietal L; rPARAC, paracentral R; lPHIP, para hippocampal L; rST, superior temporal R; lPERI, pericalcarine L; rIP, inferior parietal R. Otherwise as in Figure 6.