The Acoustic Change Complex in Response to Frequency Changes and Its Correlation to Cochlear Implant Speech Outcomes

Abstract

One of the biggest challenges that face cochlear implant (CI) users is the highly variable hearing outcomes of implantation across patients. Since speech perception requires the detection of various dynamic changes in acoustic features (e.g., frequency, intensity, timing) in speech sounds, it is critical to examine the ability to detect the within-stimulus acoustic changes in CI users. The primary objective of this study was to examine the auditory event-related potential (ERP) evoked by the within-stimulus frequency changes (F-changes), one type of the acoustic change complex (ACC), in adult CI users, and its correlation to speech outcomes. Twenty-one adult CI users (29 individual CI ears) were tested with psychoacoustic frequency change detection tasks, speech tests including the Consonant-Nucleus-Consonant (CNC) word recognition, Arizona Biomedical Sentence Recognition in quiet and noise (AzBio-Q and AzBio-N), and the Digit-in-Noise (DIN) tests, and electroencephalographic (EEG) recordings. The stimuli for the psychoacoustic tests and EEG recordings were pure tones at three different base frequencies (0.25, 1, and 4 kHz) that contained a F-change at the midpoint of the tone. Results showed that the frequency change detection threshold (FCDT), ACC N1' latency, and P2' latency did not differ across frequencies (p > 0.05). ACC N1'-P2 amplitude was significantly larger for 0.25 kHz than for other base frequencies (p < 0.05). The mean N1' latency across three base frequencies was negatively correlated with CNC word recognition (r = -0.40, p < 0.05) and CNC phoneme (r = -0.40, p < 0.05), and positively correlated with mean FCDT (r = 0.46, p < 0.05). The P2' latency was positively correlated with DIN (r = 0.47, p < 0.05) and mean FCDT (r = 0.47, p < 0.05). There was no statistically significant correlation between N1'-P2' amplitude and speech outcomes (all ps > 0.05). Results of this study indicated that variability in CI speech outcomes assessed with the CNC, AzBio-Q, and DIN tests can be partially explained (approximately 16-21%) by the variability of cortical sensory encoding of F-changes reflected by the ACC.

Keywords: acoustic change complex; cochlear implant; frequency change detection; hearing loss; speech perception.

FIGURE 1

An example of event-related-potentials (ERPs) before (top) and after (bottom) artifact removal using the independent component (ICA) analysis. After removing the artifacts, the onset-CAEP and ACC are revealed. The response peaks (N1 and P2 for the onset-CAEP and the N1′ and P2′ for the ACC, as shown in the two boxes) are marked.

FIGURE 2

The ACC measures (N1′ and P2′ latencies, N1′-P2′ amplitude) at three base frequencies. Error bars indicate the standard error. A one-way repeated ANOVA was conducted separately for N1′ latency, P2′ latency, and N1′-P2′ amplitude. Results showed that there was a statistical significance for N1′-P2′ amplitude (p < 0.01). Bonferroni follow-up test showed a larger ACC amplitude for 0.25 kHz than for other frequencies (p < 0.01).

FIGURE 3

Mean ERPs from good CI ears (mean FCDT < 10%, n = 21) and poor CI ears (mean FCDT ≥ 10%, n = 8). The ACCs were more prominent for the f_base at 0.25 kHz than for 1 and 4 kHz. The ACCs were worse in the poor CI ears compared to the good CI ears.

FIGURE 4

The mean ACC measures (N1′ and P2′ latencies as well as the N1′-P2′ amplitude) in good and poor CI ears at the three f_bases. Mann–Whitney Rank Sum Test showed that the difference in ACC measures between these two subgroups did not reach statistical significance (p > 0.05).

FIGURE 5

Mean ERPs from Left (left) and Right CI ears (right) in eight bilateral CI users. The ACCs were larger for Right CI ears for the f_base at 0.25 kHz.

FIGURE 6

The behavioral performance for Left (left) and Right CI ears (right) in eight bilateral CI users. Wilcoxon Signed Rank Tests showed Right CI ears had better CNC score than Left CI ears (p < 0.05). After correcting for multiple pairs of comparisons, the difference did not reach statistical significance (p > 0.05).