Cochlear Implant Benefits in Deafness Rehabilitation: PET Study of Temporal Voice Activations

Subjects

Eighteen right-handed adults were included in this study; they gave informed, written consent, and the Xavier-Bichat Hospital ethics committee approved the protocol. They were divided into 3 groups of 6 persons each (Table 1).

A control group consisted of normal-hearing, male volunteers, without any otologic, neurologic, or psychiatric disorders.

Twelve implanted patients who had a bilateral postlingual hearing loss were chosen considering their intelligibility score to the “Lafon monosyllabic task” at 65 dB: 1 group (2 males) with a high intelligibility score (>80%, GOOD group) and 1 group (4 males) with a low intelligibility score (<20%, POOR group). Both groups had similar free-field warble tone audiometric thresholds indicating similar audibility of sounds (Table 1). All patients had been implanted (cochlear) between 1999 and 2003; 9 patients (5 in GOOD group and 4 in POOR group) had a Sprint (cochlear) processor; the others had an Esprit 3G (cochlear) processor. Four and 3 patients were left side implanted in the GOOD and POOR groups, respectively. The cochlear implant was used >8 h/d in all cases. The follow-up was >2 y. In the GOOD group, the etiology of the hearing loss was sudden deafness (n = 2), progressive sensorineural hearing loss (n = 3), and otosclerosis (n = 1); the duration of sound privation was from 0 to 7 y. In the POOR group, the etiology of the hearing loss was sudden deafness (n = 2), bacterial meningitis (n = 3), and viral deafness (n = 1); the duration of sound privation was from to 0 to 31 y.

Task and Stimuli

Twelve PET acquisitions during passive listening were obtained: Four measurements occurred during the rest condition, 4 occurred during passive listening to voice condition, and 4 occurred during passive listening of nonvoice condition. The listening of blocs had been randomized. The voice stimuli were those used in the original fMRI study (23) and were adapted to a PET paradigm. The proportion of voice stimuli heard in the fMRI paradigm was either speech (33%: words, non words, foreign language) or nonspeech (67%: laughs, sighs, various onomatopoeia). The PET adaptation modified these initial proportions (30% speech; 70% nonspeech). Initial nonvoice stimuli consisted of sounds from nature (14%: e.g., wind, streams), animals (29%: cries, gallops), modern human environment (37%: cars, telephones, airplanes), and musical instruments (20%: bells, harp, instrumental orchestra) that had also been modified by the PET paradigm (nature: 19%; animals: 24%; modern human environment: 47%; musical instruments: 10%). Stimuli were delivered binaurally, at a mean 65-dB SPL, by KOSS electrostatic headphones to normal-hearing patients, and monaurally, directly through the implant accessory input, at the most comfortable level to implanted patients.

PET Acquisition

The rCBF was determined from the distribution of radioactivity measured with PET (ECAT-EXACT-HR+; Siemens AG) after bolus intravenous H₂¹⁵O injections (26). Listeners received 12 H₂¹⁵O injections (333 MBq per injection) corresponding to 12 rCBF measurements, performed at 10-min intervals. Attenuation-corrected data were reconstructed into sixty-three 2.25-mm-thick axial slices, with a resulting resolution of 4.5-mm full width at half-maximum after reconstruction (27).

Data Analysis

The rCBF images were analyzed with statistical parametric mapping software (SPM99; Wellcome Neurological Laboratory, U.K.) used for image realignment, transformation into standard stereotactic anatomic space (28), smoothing (12 mm), and statistical analyses (29). State-dependent differences in global flow were covaried using proportional scaling. Comparisons across conditions were made with the t statistic subsequently transformed into the normally distributed z statistic by using a multistudy design. The resulting z maps were thresholded at P < 0.001. Three statistical analyses of activation were performed: A within-group comparison of activation for listening to voice or nonvoice sounds versus the rest condition, a within-group comparison between the voice condition versus the nonvoice condition, and a between-group comparison of these comparisons across conditions.

Auditory Activation: Nonvoice Sounds Versus Silence

When compared with silence, the nonvoice stimuli elicited bilateral activation in the temporal regions. In the 3 groups (Table 2; Fig. 1A), the stimuli were located from the anterior to the posterior part of STS in the bank of the superior temporal gyrus (STG) and the middle temporal gyrus (MTG) (Brodmann areas: BA21, BA22, BA42).

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1. 

Intragroup contrasts. Location of activation peaks when listening to nonvoice compared with silence (A) and voice compared with nonvoice sound (B) in 3 groups (normal hearing and cochlear implanted patients) at P < 0.001 are shown in a lateral view of both hemispheres.

Comparison of the auditory activation maps between the 3 groups showed supplementary bilateral activation in control patients compared with implanted patients (P < 0.001; Table 3; Fig. 2A). The main plots were located bilaterally in the posterior part of the STS. No difference was observed between the GOOD and POOR groups of implanted patients.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2. 

Intergroup contrasts. Location of differences of activation peaks between 3 groups (normal hearing and cochlear implanted patients) when listening to nonvoice compared with silence (A) and voice compared with nonvoice sounds (B) at P < 0.001 are shown in a lateral view of both hemispheres.

Voice-Selective Activations: Voice Versus Nonvoice Sounds

In the control group, the activation maps elicited by the voice versus nonvoice contrast (P < 0.001) were bilaterally located, along the middle (±66, −10, −6) and the posterior part of the STS (±64, −30, −2) (Table 4; Fig. 1B), close to the TVA identified by previous fMRI results ([±62, −14, 0] and [56, −30, 6]) (23). In the GOOD group, the activations were very similar to those of the control subjects. As shown in Table 4 and Figure 1B, the activations were also bilateral and located in the middle and the posterior part of the STS. Conversely, in the POOR group, the activations were restricted to a single left peak of activation in the middle part of the STS.

The group comparison (Table 5; Fig. 2B) confirmed that TVA activation was significantly and bilaterally reduced in the POOR group compared with that of the control subjects, whereas there were no significant differences between the control and GOOD groups. In addition, when the activations between the 2 implanted groups were compared, there was a significant right supplementary activation in the anterior STS in the GOOD group compared with that of the POOR group (P < 0.001).

TVA and Brain Imaging

A meta-analysis of neuroimaging studies with normal-hearing subjects (30) suggested that acoustical parameters of speech useful to language intelligibility are specifically able to induce bilateral temporal activations in a region that has been called the “speech-sensitive auditory cortex.” These activations appear to be unrelated to linguistic analysis but, rather, to be associated with the speech signal per se (31,32). The TVA has been defined as the cortical areas activated by the contrast of carefully matched voice versus nonvoice stimuli (23) and belongs to this associative auditory cortical network, which is centered on the upper bank of the STS.

The activation of the TVA was observed in the control subjects of this PET study, but the pattern differed from that of the original fMRI study. In the latter (23), a larger activation of the right TVA was reported, whereas we found a slight reverse left asymmetry. This discrepancy may be explained by the modified repartition of acoustic cues of the initial fMRI paradigm in the time and in the spectral domain, which is known to induce different patterns of activation (33). Indeed, no hemisphere differences were recorded in the fMRI study when stimuli were spectral filtered (23).

Interestingly, the GOOD group had the same pattern of activation as that of the control subjects, suggesting that the cochlear implant transduction had correctly transmitted the acoustic cue proportions even if the stimulation was monaural in the implanted group and binaural in the control group. Some studies (34) found stronger activation in the contralateral size of the implant. In this study, in the GOOD group, 66% of patients were left-implanted and the main effect was also left-sided as in the control group. Moreover, the POOR group with 50% of left-implanted patients presented only left activations. The TVA activations did not seem to depend on the size of the implantation. This paradigm did not require any language knowledge and any active participation. Rather, it avoided testing of the entire audiophonatory loop in contrast to the intelligibility task. It allowed more specific testing of the role of the cochlear implant in the transduction of sound information without testing increased activity in areas of the inferior prefrontal cortex when listening to speech and nonspeech stimuli (35). This PET paradigm, without speech stimuli, would allow the direct testing of the processing of “structural encoding” (36), specific for the auditory system, even with infants. TVA activations were preliminary to the oral intelligibility brain network and were close to the activations induced by intelligible speech in normal-hearing subjects (37).

Electrical Cochlear Implant Stimulation and TVA

In both cochlear implant groups the activations induced by nonvoice stimuli were reduced, a possible correlate of diminished audibility, which was altered to the same extent in both groups as shown by the subjective audiometric results. Despite this reduced audibility, human voice stimuli without lexical or semantic information were able to induce TVA activation along the upper banks of both hemispheres of the STS, only when intelligibility was restored by a cochlear implant. Bilateral peaks of activations were found within 9.8 mm of those observed in the control subjects, and the level of activation was not significantly reduced when compared with that of the control subjects. A previous neuroimaging study (15) also presented evidence of significant supplementary temporal activations to words compared with noise in patients with cochlear implants. Conversely, the selective TVA activation was markedly reduced in the POOR group. A single very small cluster of activation was found in the left STS, and the activation was bilaterally reduced significantly when compared with that of the control subjects. However, it should be noted that the direct comparison of the GOOD and POOR groups revealed only a single peak of significant difference in the left STG. This may be due to limited statistical power. Nevertheless, interestingly, a study (23) found a decreased response to voice when stimuli were degraded by spectral filtering in normal-hearing subjects. Therefore, activation of voice recognition networks of the STS may depend on sufficient acoustic information. It may not be available to an extent that gives rise to TVA activity in the POOR group.

Poor Results Analysis

Subjects with low comprehension perceived sounds but lacked the capacity for sufficient sound analysis to extract human voice acoustic parameters from other sounds. A cochlear implant patient would also fail to understand speech if the element of the speech signal could not be routed correctly to the region executing the fit. Because of the low activation of TVA described in this group of cochlear-implanted patients, one could assume that the poor intelligibility score of these patients might be related to the low voice recognition. This insufficient quality of sound-parameter discrimination could be linked to neural dysfunction between the cochlear implant and the TVA, as suggested by the anatomopathologic studies about ganglion cell degeneration induced by the hearing loss. Spiral ganglion cell counts were highest in individuals who were deafened by aminoglycoside toxicity and by sudden idiopathic deafness and were lowest in those deafened by genetic deafness or bacterial meningitis (38). This is in line with recent clinical cochlear implant analysis (39) showing worse results for cochlear implantation after bacterial meningitis. In the present study, the main etiology in the POOR group was effectively meningitis. Another explanation must be considered. Vocal information processing might be dissociated in different functionally independent systems. Cortical regions involved in processing of different types of vocal information is likely to interact to build increasingly abstract representations (36), and TVA activation may be dependent on the recruitment of other speech-processing systems. This is similar to the functional cortical map reorganization study of Wernicke area (15), in which the contribution of visual areas to speech comprehension (14) and an enhanced intentional resources network in rehabilitated cochlear implant patients contributed to speech processing. Our patients were all postlingually deaf with implanted cochlea of at least 2 y, and they all communicated orally without sign language. They all had strong lip-reading communication skills but, in the POOR group, the oral communication in daily life depended much more strongly on this ability, especially for patients with a duration of 30-y sound privation. There were no reasons to suspect that cochlear-implanted patients had a problem in TVA activation, unless one hypothesized a reorganization of the TVA due to the longer period of deafness in this group of patients. Also, the abstract representation of voice in the TVA could depend, in the POOR group, on supplementary sensory input helpful to complete degraded voice information along the auditory neural pathways. This PET paradigm tested more directly the auditory system per se and its ability to deal with voice perception analysis without testing other complementary modalities or intentional brain networks, as suggested by the localization of activations in the TVA, even if these postlingual deaf patients had developed robust lip-reading skills to communicate orally.