Auditory cortical plasticity after cochlear implantation in asymmetric hearing loss is related to spatial hearing: a PET H215O study

Abstract In asymmetric hearing loss (AHL), the normal pattern of contralateral hemispheric dominance for monaural stimulation is modified, with a shift towards the hemisphere ipsilateral to the better ear. The extent of this shift has been shown to relate to sound localization deficits. In this study, we examined whether cochlear implantation to treat postlingual AHL can restore the normal functional pattern of auditory cortical activity and whether this relates to improved sound localization. The auditory cortical activity was found to be lower in the AHL cochlear implanted (AHL-CI) participants. A cortical asymmetry index was calculated and showed that a normal contralateral dominance was restored in the AHL-CI patients for the nonimplanted ear, but not for the ear with the cochlear implant. It was found that the contralateral dominance for the nonimplanted ear strongly correlated with sound localization performance (rho = 0.8, P < 0.05). We conclude that the reorganization of binaural mechanisms in AHL-CI subjects reverses the abnormal lateralization pattern induced by the deafness, and that this leads to improved spatial hearing. Our results suggest that cochlear implantation enables the reconstruction of the cortical mechanisms of spatial selectivity needed for sound localization.


Asymmetric hearing loss
Asymmetric hearing Loss (AHL) is defined as a binaural difference in hearing threshold greater than 15 dB, and is known to disrupt binaural processes, leading to impairments in sound localization and speech comprehension in noise. This has a significant negative impact on an individual's quality of life (Nelson et al. 2019). Studies have measured the extent of these binaural deficits, and in the case of sound localization, this has been expressed as an increase in the angular error of localization (Nelson et al. 2019) and as a right-left confusion in sound source lateralization. In a normal hearing subject, monaural stimulation leads to bilateral activation in the auditory cortex but with much stronger contralateral activation, referred to as contralateral dominance. In patients with AHL, studies have shown that the auditory cortical asymmetry is altered, with a dominance shift from contralateral to ipsilateral with respect to the better-hearing ear (Scheff ler 1998; Ponton et al. 2001) in line with animal studies. In children with congenital AHL, this shift is accompanied by a weaker cortical representation of the deaf ear (Gordon et al. 2015). In a recent fMRI study on adults with acquired AHL (Vannson et al. 2020), the dominance shift was observed in the nonprimary auditory cortex (NPAC), thus suggesting that the plasticity involves cortico-cortical interactions. Interestingly, the extent of this functional reorganization strongly correlated with the deficit in sound localization. We hypothesize that the dominance shift reflects a disruption in spatial hearing in line with the theory that the contralateral dominance seen in normally hearing subjects relates to the representation of the contralateral sound field rather than the contralateral ear input (Eisenman 1974;Middlebrooks and Pettigrew 1981;Phillips and Gates 1982). Our study aims to explore the neural correlates of spatial hearing improvement following cochlear implantation for AHL, and more specifically, the lateralization of brain activity in these patients.

Spatial hearing
In all species, one of the primary functions of hearing is to detect and locate brief sounds in order to direct visual attention and further analyze the sound sources (Heffner and Heffner 1992). The ability to locate sound sources involves detecting stimulus disparities between the right and left ear in terms of the intensity (interaural level difference) and timing (interaural time difference) of the sounds. It is known that the information from each ear converges at an early level in the auditory pathway and the combination of information from the two ears is known as binaural integration. This process enables spatial hearing as well as facilitates speech understanding in noisy environments. Studies have shown several advantages arising from binaural integration, including improved sound detection and identification in noise due to binaural redundancy, binaural unmasking, and head shadow effects (Avan et al. 2015).
In the case of unilateral hearing loss, the binaural response properties of the auditory neurons are strongly affected, as revealed by electrophysiological recordings using animal models of congenital (Kral et al. 2013;Tillein et al. 2016) and acquired unilateral deafness (Popelár et al. 1994;McAlpine et al. 1997). In these studies, it has been found that the neuronal activity shifts towards the hemisphere ipsilateral to the normally hearing ear, with both increased excitability/activity and reduced latency. The binaural responses display altered excitatory and inhibitory interactions (Kral et al. 2013;Tillein et al. 2016), which change the neuronal spatial properties at the cortical level (Clarey et al. 1992; Barone et al. 1996). These modifications induce profound functional reorganization at the areal level, affecting the auditory cortex's interhemispheric asymmetry (Scheff ler 1998; Ponton et al. 2001).
Though studies on both humans and animals have established that AHL has a deleterious effect on spatial hearing abilities, which is evident at both the behavioral and the neuronal levels (Keating and King 2013;Gordon et al. 2015;Vannson et al. 2015;Kumpik and King 2019), there is also evidence that some degree of auditory localization is still possible, even when the binaural cues are altered. For example, following early unilateral deafness, sound localization abilities can be developed using monaural spectral cue extraction (Slattery and Middlebrooks 1994;King et al. 2000), specific training (Firszt et al. 2015), and following cochlear implantation (Arndt et al. 2011;Távora-Vieira et al. 2015;Galvin et al. 2019). These results clearly demonstrate the plasticity of auditory spatial processing and show that it can be achieved by means of adaptive strategies or by restoring binaural inputs to the brain (Keating and King 2013).

AHL rehabilitation with cochlear implantation
Ever since the seminal work of Van de Vermeire and Van de Heyning (2009), patients with AHL have been considered candidates for cochlear implantation (Vermeire and Van de Heyning 2009;Arndt et al. 2011). Although the implants were initially used in these patients to target severe tinnitus, it was reasoned that the patients' binaural abilities could also be improved (Vermeire and Van de Heyning 2009;Arndt et al. 2011). This would involve combining the electrical cues conveyed by the implanted ear with the natural acoustic hearing in the other ear. It has since been found that the restoration of binaural integration is highly variable, presumably because of the difficulty involved in integrating the two types of information. Nevertheless, it has been found that the bilateral auditory input allows some degree of binaural hearing, as studies have shown improved sound localization (Távora-Vieira et al. 2015) and speech recognition in noise (Mertens et al. 2015;Dorbeau et al. 2018;Legris et al. 2018;Galvin et al. 2019;Arndt et al. 2020). These improvements have been found to relate to various factors, including the pre-implantation performance, the hearing level of the nonimplanted ear, the age at implantation, the duration of deafness and the age at the onset of deafness (Mertens et al. 2015;Távora-Vieira et al. 2015;Galvin et al. 2019;Nelson et al. 2019). We hypothesize that the restoration of binaural skills depends on an individual's brain plasticity to enable the integration of both electrical and acoustical information. So far, except for a few reports in children (Sharma and Glick 2016;Polonenko et al. 2018), studies have not fully explored the functional brain responses in these patients. This present study investigated the behavioral and cortical consequences of adult acquired AHL, which is most likely very different from what is reported in cases of early developmental hearing loss. From the sparse existing literature, EEG studies have shown that the neural responses elicited by electrical and acoustical stimulation share similar latencies (Wedekind et al. 2020), but there is a weaker global auditory response (Legris et al. 2018) compared to normally hearing controls, when both modalities are used.

Aim of the study
The present investigation was motivated by the scarcity of studies on brain plasticity following cochlear implantation for AHL, and our interest in the recovery of spatial hearing. We used PET brain imaging to study the pattern of brain activation in response to natural auditory stimuli in cochlear-implanted patients with AHL (AHL-CI); the cochlear implant precluded the use of other imaging techniques. We hypothesized that the stimulation of the deaf ear by a cochlear implant could restore, at least partially, the normal pattern of contralateral hemispheric dominance. Furthermore, as spatial hearing in AHL has been shown to relate to changes in hemispheric dominance (Vannson et al. 2020), we proposed that the restoration of contralateral dominance would relate to the improvement of spatial hearing.

Subjects
Twenty subjects were recruited: 10 patients with a unilateral cochlear implant for postlingual AHL (AHL-CI)  (Karoui 2019). All of the controls had normal thresholds over the range 0.5-4 kHz, with a PTA < 20 dB HL, they were sex and age-matched to the AHL-CI group, thus giving a total of five women and five men in each group, with ages ranging from 46 to 74 years. Our study used a patient group and an imaging technique that are rare. The patients all had cochlear implants to treat asymmetric hearing loss; this is a group for whom rehabilitation evaluations are still exploratory; most patients treated so far have been included in research protocols.
Our study used H 2 15 O PET imaging, an invasive and radioactive procedure, which imposes limitations on the number of subjects. It should be noted, however, that for PET H 2 15 O activations studies, by accepting a Type 1 risk α of 0.05 with correction for multiple comparisons and a statistical power 1 − β of 0.90, the sufficient sample size per group is estimated as 10-12 subjects (Andreasen et al. 1996). The study was approved by the French South West and Overseas ethics committee (approval number: 2016-A01442-49) and conducted according to the Declaration of Helsinki including the signed by all participants informed consent form before taking part in the study.

Localization task
Among the different audiological evaluations, we selected horizontal localization performance, which ref lects more directly the level of binaural integration. An array of 12 loudspeakers was set up in an anechoic chamber, each at a distance of one meter from the subjects, with an angular separation of 15 • (Chan et al. 2008) and located behind the subjects' ears at the same height. We used the same localization protocol as developed by Chan et al. (2008), also used in Aronoff et al. (2012), Dorbeau et al. (2018), Galvin et al. (2019). The stimulus is a broadband (230-11 kHz) impulse sound of 1.474 s duration (gunshot), presented at 65 dB SPL allowing the use of both ITDs and ILDs to localize the sound source. The subjects were asked to indicate where the noise had come from by clicking on one of the loudspeakers on the computer screen. For each run, there were 24 trials where the stimulus was presented randomly from one of the loudspeakers, with two presentations per loudspeaker. To minimize the use of loudness cues for the localization task, the amplitude of the stimulus could randomly increase or decrease by 6 dB from trial to trial (roving) (Galvin et al. 2019). The AHL-CI group completed the localization test under four different conditions: with the cochlear implant activated (CI-ON) or deactivated (CI-OFF), with a roving amplitude of 6 dB from trial to trial, and without a roving amplitude (nonroving). The NH group was only assessed using the roving condition. The localization accuracy was reported as the average root mean square (RMS) error in degrees.

PET procedure
The subjects were scanned in a shielded, darkened room. The head was immobilized and aligned transaxially to the orbitomeatal line using a laser beam. Regional distribution of radioactivity was obtained using a Biograph 6 TruePoint HiRez (Siemens Medical) PET scanner with full volume acquisition. The duration of each scan was 80 s; approximately 6 mCi of H 2 15 O was administered to each subject. A measure of auto-attenuation correction was obtained by performing a CT scan before the PET H 2 15 O.

Sound stimuli during the PET scan
We used an auditory discrimination task during PET scan based on natural environment sounds. The voice/ nonvoice task was chosen because it allows activation of a large auditory cortical network (see Belin et al. 2000). In a previous study, this task enabled us to obtain robust values of auditory lateralization indices in the UHL patients (Vannson et al. 2020). Finally, it was preferred to a linguistic task because it does not involve the adaptive strategies generally developed after hearing loss for speech processing. All of the stimuli were taken from our database developed in previous studies (Massida et al. 2011;Vannson et al. 2020). Subsets of 500-ms-long vocal and nonvocal stimuli were randomly presented using PsychoPy2 (Peirce et al. 2019). Vocal stimuli were 55 different sounds, including speech sounds (words in non-French languages and nonsemantic syllables) and nonspeech sounds (e.g. laughs, coughs). Nonvocal stimuli were different environmental sounds (alarms, car horns, bells . . . ). The subjects were given a two-alternative forced choice task to indicate whether they heard a vocal or a nonvocal sound by clicking on the left or right button of a mouse. The duration of each trial was determined by the subject's response time, but this was limited to a maximum of 8 s. During the PET scan, the subjects had their eyes covered, and the auditory stimulation began at the descending part of the bolus curve, 10-20 s before the tracer reached the brain. For the NH group, the stimuli were delivered using inserts; for the AHL-CI group, the sound was delivered either directly via a cable connected to the implant, or via an insert in the nonimplanted ear, or via the two systems simultaneously. As patients presented a variable hearing loss in the nonimplanted ear, we systematically corrected the stimulation level according to the PTA. According to the Fig. 6 prescription (Gitles and Niquette 1995), we applied a frequency gain for a 65 dB SPL presentation level to correct for hearing loss in the nonimplant ear. The input signal to the CI sound processor was calibrated to be equivalent to 65 dB SPL. In addition, subjects were asked to balance loudness to obtain a comfortable listening level for both ears. There were four different conditions: (i) a baseline condition, where the subjects were instructed to lie quietly in the scanner, (ii) a binaural condition, where the sounds were presented to both ears, (iii) a right-ear monaural, and (iv) a left-ear monaural conditions. There were two runs for each condition, each lasting 80 s, the order of the runs was randomized for each subject, as was the order of the trials within each run. The total duration of the PET scan was 80 min.

PET analysis
The PET data were analyzed using SPM12 software and included realignment and spatial normalization to the Montreal Neurological Institute template. The images of the patients with a right-sided CI (4 patients) were flipped so that all of the implants were on the left side in the final images. The global signal was normalized within and between the subjects; overall grand mean scaling scales the global blood f low to a physiologically realistic value of 50 mL/dL/min as proposed in SPM12. The analyses focused at the early cortical levels (primary [PAC] and nonprimary [NPAC] auditory cortex) based on our findings on UHL patients (Vannson et al. 2020); whole-brain analyses will be presented in a separate article. We used the combined cytoarchitectonically defined auditory areas Te1.0, Te1.1, Te1.2, and Te3.0 as the regions of interest, developed and implemented in the SPM Anatomy Toolbox (Eickhoff et al. 2006). We used the linear mixed-effects model ("lme4" package in R) with repeated measures for the comparisons between groups and conditions. An asymmetry index (AI) of the brain activity in the auditory regions was calculated to study the lateralization pattern of auditory activity with respect to the baseline: AI = (Contralateral − Ipsilateral)/(Contralateral + Ipsilateral). Contralateral and ipsilateral refer to the cortical side contralateral and ipsilateral to the stimulated ear.
In addition, the respective influence of the crossed (opposite) and uncrossed (same side) ear on the auditory activity in each cortical hemisphere was estimated using the aural preference index (API): API = (Crossed − Uncross ed)/(Crossed + Uncrossed), where Crossed refers to the activity elicited by the opposite ear and Uncrossed refers to the activity elicited by the ear on the same side as the cortical activity. For each auditory cortex, the values are positive in the case of crossed aural preference and negative in the case of uncrossed aural preference.

Data availability
The data that support the findings of this study are available from the corresponding author, upon reasonable request.

Auditory test performance
The participants were given a voice/nonvoice discrimination task during the PET scan. Performance was assessed by calculating the hit rate under each condition. Using a linear mixed-effects model with the group and the condition as fixed effects, significant results were found for the group, the condition, and the interaction. Overall, the AHL-CI patients had poorer scores than the controls, with an average hit rate of 63 ± 27% (SD), compared to 93 ± 6% for the controls (P = 3 × 10 −9 ; Fig. 1A). Although the patients' performance depended on the condition (binaural stimulation, acoustic stimulation or cochlear implant [CI] stimulation), it remained above chance (bootstrap, P taken at 0.05). The AHL-CI patients' hit rates for the CI-only condition (49 ± 27%) were significantly lower than for the binaural and acoustic conditions (77 ± 19 vs 68 ± 28%, respectively, P < 1 × 10 −4 ). We observed a weak difference between the binaural and acoustic conditions (P = 0.014), suggesting that the information provided by the implant slightly improved global auditory performance. The level of residual Hit rates on the voice/nonvoice discrimination task and spatial hearing scores. Bin-Binaural condition. NI ear-Nonimplanted ear. CI-Cochlear implant stimulation. RMS-Average RMS error in degrees. (A) Patients had significantly lower hit rates than the controls for all of the stimulation conditions; their hit rates for the CI condition were significantly lower than for the binaural and acoustic conditions. A small difference was observed between the binaural and acoustic conditions in the patients. (B) Patients made significantly more errors during the spatial perception task than the controls. Without amplitude roving, there was a significant difference between the CI-ON and CI-OFF conditions, showing that the implant improved spatial localization. However, when amplitude roving was introduced, there was only a trend towards better performance in the CI-ON condition (P = 0.08).
hearing, (PTA values), correlated with the scores for the binaural (r = 0.61, P = 2.3 × 10 −6 ) and the acoustic (r = 0.63, P = 1.6 × 10 −4 ) conditions. We did not find a correlation between the hit rate and CI experience.
The spatial hearing abilities were evaluated just before the PET scan using a free-field sound localization task. Performance was assessed with the cochlear implant on (CI-ON; i.e. a binaural, bimodal condition) and with the cochlear implant off (CI-OFF; i.e. an acoustic, monaural condition). On average, the AHL-CI subjects had poorer sound localization than the controls (P < 0.05; Fig. 1B).
This was observed, as expected, in the CI-OFF condition, where the RMS errors often exceeded the range of 50-60 • . When the amplitude of the sounds was constant, performance was significantly better in the CI-ON condition than in the CI-OFF condition (paired t-test; P = 0.01); in the more challenging roving amplitude condition, there was only a trend towards better performance in the CI-ON condition (P = 0.08). However, for this latter condition, seven of the ten subjects had superior sound localization in the CI-ON condition, with a binaural benefit ranging from +9 to +33%. Thus, there was a beneficial effect, albeit limited, of the implant on spatial hearing, similar to the results found for discrimination task.

Average activity in the auditory areas
The auditory activity in the PAC/NPAC regions was assessed for the different stimulation conditions ( Fig. 2A and B). A linear mixed effects model was used to analyze the differences between the subject groups, the stimulation conditions, and the side of the auditory cortical activity (contralateral and ipsilateral to the stimulated ear). The results showed significant differences for all three ( Table 2, Fig. 2A), the activity in the auditory regions being reduced by about 5% in the AHL-CI subjects compared with the controls (group effect; P = 0.024). The differences between the conditions are about 2-5% of the average activity level in the areas of interest ( Fig. 2A). During acoustic stimulation, there were higher levels of activity in the hemisphere contralateral to the stimulation, both in the AHL-CI patients and the controls (side effect; P = 2.2 × 10 −16 ). In contrast, during CI stimulation, the activity was higher in the hemisphere ipsilateral to the implanted ear (condition: side interaction, P = 0.002), and was found to be higher than the ipsilateral activity to acoustic stimulation (post hoc comparisons P = 4 × 10 −5 ). The opposite pattern was found for contralateral activity: the auditory activation levels were higher for the acoustic than for the CI stimulation (P = 2 × 10 −12 ). Altogether, these results suggest that the implant only partially restores a normal pattern of contralateral auditory cortical activation.
In addition, we computed a binaural integration index (BII) comparing for each hemisphere the activity derived from the stimuli simultaneously presented to both ears (Bin) with the summed activity obtained after separate stimulation of each ear: Bin/(Crossed + Uncrossed ear). Binaural integration mechanisms occur when the BII is significantly different from 1.0. In both the NHS and the patients, we observed BII values lower than 1.0, ref lecting a binaural suppression mechanism. BII values are similar when comparing NHS and patients, whatever the side considered (in NHs, the mean for both hemispheres is 0.50 ± 0.007; in patients, 0.51 ± 0.008 for the hemisphere ipsilateral to the implant and 0.51 ± 0.01 for the hemisphere contralateral to the implant (bootstrap, P < 0.05) ( Fig. 2A). We carried out further analyses to determine whether the auditory cortical activity relates to the level of residual hearing in the nonimplanted ear. Significant negative correlations with PTA (Fig. 3) were observed, but only for the auditory cortex contralateral to the acoustic stimulation. This was found for both the binaural and the acoustic conditions (r = 0.62 (P = 1.3 × 10 −6 ) and 0.52 (P = 3.6 × 10 −5 ), respectively; Fig. 3A and B), and it implies that the level of hearing loss negatively impacts the level of auditory cortical activity. We also found a similar negative correlation when the cochlear implant was stimulated, but this was for the cortex ipsilateral to the implant (r = −0.54 (P = 8 × 10 −6 ); Fig. 3C). This suggests that the level of hearing loss affects the overall cortical excitability and influences the response to peripheral stimulation, whether acoustic or electrical.
We also carried out analyses to determine whether the duration of the hearing loss relates to the cortical activity. We found a weak, negative correlation for the activity contralateral to the implant during binaural stimulation (r = −0.36, P = 0.009) and CI stimulation (r = −0.37, P = 0.004), but not during the acoustic stimulation (r = −0.13, P = 0.93). This implies that a longer duration of deafness weakens the restoration of auditory activity in the hemisphere contralateral to the implant.
The influence of post-implantation experience was also assessed. The duration of CI experience correlated with activity in the cortex contralateral to the implant during binaural (r = 0.32, P = 0.023) and acoustic stimulation (r = 0.36, P = 0.007), but not during the CI-only condition (r = −0.08, P = 0.55). Thus, there is a slightly higher activity contralateral to the implant with CI experience. Our hypothesis is that the treatment of the profound hearing loss by cochlear implantation affects the overall cortical excitability and influences the response to peripheral stimulation, whether acoustic or electrical. No significant correlations were found for the ipsilateral to CI cortex.
To summarize, the auditory cortical activity was found to be lower, on average, in the AHL-CI subjects than in the controls. This can be attributed to weaker activation from the cochlear implant as well as hearing loss in the nonimplanted ear.

Auditory activity and test performance
For the AHL-CI patients, we examined whether the auditory activity was related to performance on the sound discrimination task (hit rates). Correlation analyses were run and revealed several significant positive correlations. These were found for the auditory cortex contralateral to the nonimplanted ear during binaural stimulation (r = 0.4, P = 0.0019; Fig. 4A), for both the contralateral and ipsilateral to acoustic stimulation auditory regions (r = 0.63 (P = 1.5 × 10 −7 ) and r = 0.54 (P = 1.7 × 10 −5 ), respectively; Fig. 4B and C), and for the ipsilateral to CI auditory cortex during CI-only stimulation (r = 0.40, P = 0.0017; Fig. 4D).
A positive relationship between auditory activity and performance levels in all cases indicates that auditory activity in our study is related to the efficiency of auditory perception in patients and not to the task difficulty.
Relationships were also examined between the auditory cortical activity and auditory spatial abilities, expressed by the RMS values on the sound localization task. A significant correlation was only found for . The values are arbitrary units (a.u.) of regional cerebral blood f low (rCBF) after the normalization per subject. For the binaural bar chart, the average of both auditory cortices is presented. AHL-CI subjects had significantly lower activity in the auditory regions than the controls. In both patients and controls, during acoustic stimulation, activity was higher on the side contralateral to the stimulation. However, during CI stimulation in patients, activity was higher in the ipsilateral than the contralateral auditory cortex. Binaural integration index (indicated as binaural index) compares for each hemisphere the activity caused by the stimuli simultaneously presented to both ears (bin) with the summed activity obtained after separate stimulation of each ear: Bin/(Crossed+Uncrossed ear). The mean of both hemispheres is presented for NHS. (B) Illustration of activity in patients during binaural stimulation, better acoustic ear stimulation (right ear for this image) and CI (left side) stimulation. It can be seen that the activity during binaural stimulation is similar to the acoustic stimulation. The images have been f lipped for the four patients with right CI. T-values of the statistical comparison with the baseline condition are presented. the activity in the hemisphere contralateral to the nonimplanted ear (r = 0.31 P = 0.018). This was confirmed by a regression analysis at the whole-brain level between the sound localization scores and brain activity during binaural stimulation (acoustic and CI stimulation); this pointed to a restricted area located within the contralateral to the nonimplanted (NI) ear auditory cortex (Fig. 5). This result indicates that the auditory cortex (PAC/NPAC) contralateral to the nonimplanted ear is implicated in spatial hearing processing in AHL-CI subjects.
To summarize, the relationship between the performance levels and the auditory cortical activity was mainly driven by the nonimplanted ear during both the acoustic and the binaural conditions. Conversely, the performance levels only related to the CI evoked activity in the auditory cortex ipsilateral to the CI.

Asymmetries in the auditory activity
The asymmetry index (AI; see materials and methods) indicates which hemisphere is more active according to the side of the auditory stimulation (Fig. 6A). AI values are positive when there is greater contralateral activity, and negative when there is greater ipsilateral activity. We calculated them with respect to the stimulated ear for the monaural conditions. For the binaural stimulation, we calculated the AI with respect to the NI ear in patients and to the right ear in controls. We found that the control subjects had greater contralateral activity for both of the monaural stimulation conditions; during binaural stimulation, the index did not differ significantly from zero, thus indicating similar levels of activity in both the left and the right auditory cortices (Fig. 6A).
In the AHL-CI subjects, there were significant positive AI values when the nonimplanted ear was stimulated (Fig. 6B). This demonstrates that the acoustic stimulation induced a contralateral to the NI ear dominance, as observed in normally hearing subjects. Following stimulation through the cochlear implant, the AI index values also tended to be positive, but this was not statistically significant (bootstrap, P = 0.05). The AI differences between the AHL-CI group and the controls were not found to be statistically significant for any of the stimulation conditions. We suspect that is probably due to the lower responsiveness of the cortical activity to the CI in the contralateral hemisphere.
These results indicate that cochlear implantation restores the normal pattern of interhemispheric asymmetry, with a reactivation of the contralateral auditory cortex in response to stimulation at the nonimplanted ear. This can be attributed to the partial recovery of binaural hearing through cochlear implantation. As discussed previously, UHL patients display ipsilateral to the better-hearing ear activation in response to its stimulation (Vannson et al. 2020). In contrast, the AHL-CI patients displayed contralateral activation that was similar to the pattern seen in normally hearing controls.
We also examined whether the AHL-CI subjects' asymmetry index related to performance on the sound localization task. We found that the AI following stimulation of the nonimplanted ear was highly correlated (rho = −0.8, P = 0.0014) with the localization performance, i.e. RMS errors in the roving condition. Superior spatial performance therefore may be related to the restoration of a normal pattern of interhemispheric asymmetry (Fig. 6C). No correlation of AI with CI experience was found.

Aural preference indices of cortical activity
The aural preference index (API; see materials and methods) is used to assess the respective influence of the Fig. 4. Mean auditory cortical activity and hit rates on the voice/nonvoice task. NI ear-Nonimplanted ear. CI-Cochlear implant. There was a significant correlation between the hit rate and activity in the auditory cortex contralateral to the nonimplanted ear during binaural stimulation. The highest significant correlations with performance were for the acoustic stimulation (for both auditory cortices). Following CI stimulation, there was a significant correlation between the hit rate and activity in the ipsilateral to the CI stimulation. crossed and uncrossed ear on the auditory cortical activity in one hemisphere (Kral et al. 2013; Fig. 6A).
API values are positive when the crossed ear has a greater inf luence. For the control subjects, we found that the API were positive for both auditory cortices, with values that differed significantly from zero (P < 0.05; Fig. 6D). For the AHL-CI subjects with a left-sided cochlear implant, there was a significant crossed aural preference for the left auditory cortex (API of about 1%), ref lecting a preference for the nonimplanted right ear; for the right auditory cortex, the API values did not differ significantly from zero, thus implying that the implanted ear did not strongly inf luence the contralateral hemisphere (Fig. 6D). For the patients with a right-sided implant, the pattern of results was similar, with a significant crossed aural preference for the right auditory cortex, but not for the left auditory cortex (Fig. 6D). These results imply that the hemisphere contralateral to the nonimplanted ear shows a crossed aural preference, similar to that seen in controls.
The API values for the left-implanted and rightimplanted AHL-CI patients were pooled into a single group to assess correlations clinical factors. Correlation analyses revealed that the longer the duration of the deafness, the stronger the crossed aural preference for the auditory cortex ipsilateral to the implant (rho = 0.67, P = 0.031). No correlation of API with CI experience was found.

Discussion
This study aimed to explore how the restoration of binaural hearing in AHL through cochlear implantation can restore both spatial hearing processing and a normal pattern of brain activity in auditory areas. We used a sound localization task and PET brain imaging to study a group of AHL-CI subjects and matched controls. The results provide evidence that the improvement of binaural inputs re-establishes the physiological lateralization of auditory cortical activity and that this relates to the recovery of auditory spatial abilities.

Spatial hearing after CI
Unilateral or asymmetric hearing loss greatly affects the auditory processes involved in spatial hearing. For instance, AHL impedes the analysis of binaural disparities, which leads to deficits in sound localization and speech comprehension in noisy environments (review in Kumpik and King 2019). In this study, we showed that cochlear implantation reintroduces binaural indices, leading to improved sound localization in AHL. We found a substantial binaural benefit (20-30%) for spatial performance (lower RMS), with better scores on the CI-ON condition compared with the CI-OFF condition. However, the performance remained impaired and was particularly low when the stimulus amplitude was roved, with the difference between the CI-OFF and CI-ON conditions becoming only a tendency. Nevertheless, our results suggest that AHL-CI subjects learn to use binaural cues, most probably related to interaural level differences (Seeber et al. 2004;Távora-Vieira et al. 2015) in a head shadow-like effect (Wanrooij and Opstal 2004), which become ineffective in the amplitude roving condition. These results provide further evidence that AHL-CI subjects can recover substantial spatial hearing abilities when binaural inputs are restored (Zeitler et al. 2015;Dorman et al. 2016;Polonenko et al. 2018;Litovsky et al. 2019;Legris et al. 2020), even though the electrical information is distorted and temporally offset (Zirn et al. 2015).

Restoration of cortical activity after CI
In adult subjects with acquired AHL, acoustic stimulation of the deaf ear leads to lower levels of cortical activity compared with normally hearing subjects, depending on the level of the hearing loss (Langers et al. 2005;Burton et al. 2012;Vannson et al. 2020). In the present study, we showed that auditory input through a CI was able to restore cortical activity. This has also been shown in studies on CI-recipients with bilateral profound deafness, where there is evidence for progressive reactivation of the auditory cortex in response to natural sounds as well as reactivation of the neural networks involved in language processing (Naito et al. 1995;Giraud et al. 2001;Green et al. 2005;Mortensen et al. 2006;Stropahl and Debener 2017). The rather strong activity observed in both hemispheres following CI stimulation in our study (about 80% of the controls) related to the amount of CI experience and the duration of the hearing loss, as is classically reported after CI rehabilitation (Giraud et al. 2001;Rouger et al. 2011;Strelnikov et al. 2013Strelnikov et al. , 2015. However, overall, we found a lower level of auditory cortical activity compared with controls, which may reflect a weaker efficiency of the electric stimulation following a relatively restricted period of activation (Kral et al. 2006. This may also reflect a lower response to the nonimplanted ear, as the AHL-CI subjects had a certain amount of hearing loss in this ear (PTA 15-68 dB); there may also be fewer cortical interactions between the inputs from each ear, as the activity levels for acoustic and binaural stimulation depended on the duration of CI experience.

Restoration of the functional lateralization of cortical activity following CI
The contralateral lateralization of auditory information processing is present at all levels of the auditory pathway (Schönwiesner et al. 2007) and is referred to as the contralateral ear dominance. The disruption of binaural inputs in unilateral or asymmetric hearing loss leads to a shift in hemispheric dominance ipsilateral to the betterhearing ear. This has been found in humans (Scheffler 1998;Ponton et al. 2001;Langers et al. 2005;Burton et al. 2012;Vannson et al. 2020) and also in animal models of both congenital (Kral et al. 2013;Tillein et al. 2016) and acquired unilateral deafness (Popelár et al. 1994;McAlpine et al. 1997).
It has been suggested that the lateralization of auditory information processing may relate to the contralateral sound field rather than the contralateral ear per se (Eisenman 1974;Middlebrooks and Pettigrew 1981;Phillips and Gates 1982). This hypothesis was originally proposed because of the inhibitory influence of ipsilateral stimulation on auditory cortical neurons (Brugge and Merzenich 1973;Irvine 1979, 1981)  For binaural stimulation, the asymmetry index in patients was calculated with respect to the nonimplanted ear. In controls the binaural asymmetry index is presented with respect to the right ear. AI is marked with an asterisk when significantly different from zero (bootstrap, P < 0.05). AI values are in the range of that observed in similar PET scan studies (see Berman and Weinberger 1990). In AHL-CI patients, there were significant positive (contralateral) AI values for stimulation of the nonimplanted ear, as seen in normally hearing subjects. For stimulation through the CI, the AI index was not significantly different from zero; and for binaural stimulation, the AI index showed a contralateral shift with respect to the nonimplanted ear. (C) the AI for acoustic stimulation correlated highly with the sound localization scores, implying that smaller errors in spatial localization were related to the restoration of contralateral hemispheric dominance. (D) AHL-CI-Patients with a cochlear implant for asymmetric hearing loss. API-Aural preference index. The API is marked with an asterisk when significantly different from zero (bootstrap, P < 0.05). In normally hearing controls, a crossed aural preference was seen for both auditory cortices (API values significantly different from zero). In the left-implanted subjects, there was a crossed (right) aural preference for the left auditory cortex; in the right-implanted participants, there was a crossed (left) aural preference for the right auditory cortex. and the low proportion of such neurons unresponsive to ipsilateral stimulation (7% according to Hall and Goldstein 1968). The theory has been supported by several unilateral lesion studies, which showed that sound localization was mainly impaired in the sound field contralateral to the lesion, irrespective of whether the lesion affected a lower or higher stage of brain processing (see (Thompson and Masterton 1978) for the brainstem; (Sanchez-Longo and Forster 1958) for the temporal lobe; (Clarey et al. 1992) for a review). However, whether it reflects ear dominance or the sound field, it is clear that the usual pattern of lateralization is disrupted following unilateral and asymmetric hearing loss. The first aim of our study was to determine whether cochlear implantation could restore this functional lateralization in AHL.
The aural preference index (API, see Fig. 6D) is a commonly used quantitative measure to assess functional lateralization. It has been used in both human (Gordon et al. 2015) and animal studies (Kral et al. 2013), and has characterized Aural Preference Syndrome, which can occur in unilateral deafness (Gordon et al. 2015). API values reflect the strength of activation resulting from each ear at the level of a single hemisphere. In normally hearing controls, the API reveals a crossed aural preference for both hemispheres (Fig. 6D); in unilateral and AHL, this preference shifts towards the better-hearing ear for both hemispheres (Kral et al. 2013). In our study, we showed that crossed aural preference is not restored following cochlear implantation in AHL, as CI stimulation activates the contralateral auditory cortex as much as the ipsilateral cortex (see Fig. 6D). These results are in agreement with a study on AHL-CI children (Polonenko et al. 2018;Lee et al. 2020), where there was only a slight, nonsignificant tendency towards a crossed preference for the implanted ear.
It is important to be aware that the API compares the level of cortical activation resulting from two auditory pathways (from the cochlea to the cortex) that differ in their functional integrity. It is known that deafness leads to a structural alteration of the anatomical pathway (Moore et al. 1994;Syka 2002;Eggermont 2017), and this, coupled with the degraded sound representation provided by the neuroprosthesis (Sato et al. 2017), could weaken the efficiency of the cochlear implant, leading to a weak or absent crossed aural preference for the implanted ear. However, the crossed aural preference for the nonimplanted ear is preserved, both before and after cochlear implantation (present study and Lee et al. 2020). When considering these results, it should be noted that the API does not ref lect functional lateralization concerning representations of the contralateral sound field.
Another quantitative index, the asymmetry index (AI; or lateralization index in Vannson et al. 2020) (see Fig. 6) is also frequently used to assess auditory cortical processing in normal hearing (Jamison et al. 2006;Schönwiesner et al. 2007;Stefanatos et al. 2008) as well as in deafness and its treatment (Lee et al. 2020). The AI provides a direct comparison of the strength of activation in the auditory cortical areas of the two hemispheres following stimulation at a single ear. The lateralization in normally hearing controls is generally reported to be contralateral, meaning that the auditory pathways onl activate weaker the auditory cortex ipsilateral to the stimulated ear. This weaker ipsilateral activation is partly due to the small number of projections in the ipsilateral pathways (Schönwiesner et al. 2007) and partly due to the various excitatory/inhibitory ipsilateral inputs that reach the auditory cortex (see Gutschalk and Steinmann 2015. These neuronal interactions are important because they lead to contralateral sound field representation (Phillips and Gates 1982;Clarey et al. 1992;Samson et al. 2000), which is thereby ref lected in the AI values.
In a previous study on adult subjects with AHL (Vannson et al. 2020), we demonstrated that the AI shifts towards the ipsilateral to the better-hearing ear auditory cortex following its stimulation. This leads to disruption of the neuronal interactions in this auditory cortex. We found that the AI shift was related to the deficit in sound localization, thus suggesting that a change in activation towards the ipsilateral auditory cortex disrupts representations of the sound field. In the present study, we present data that mirrors this finding, showing that the restoration of auditory input to the deaf ear through the CI re-establishes the contralateral dominance for the NI ear. The was also a trend towards contralateral dominance for the implanted ear, although this was not statistically significant. In addition, the extent of the contralateral to the NI ear lateralization was found to be associated with better performance on the sound localization task, thus suggesting a normalization of the cortical representation of the sound field. This strengthens our previous claim that the asymmetry index can predict the recovery of spatial hearing.
At present, there is no clear evidence concerning the mechanisms that could underlie the loss and the recovery of contralateral activation in AHL and AHL-CI subjects, respectively. It is possible that there are changes in the balance of excitatory and inhibitory interactions between the ipsilateral and contralateral inputs; this may also affect the spatial receptive fields for auditory localization (Imig et al. 1990).

Brain plasticity in the primary auditory cortex in deafness and its rehabilitation
Only a few studies have investigated whether the reactivation of a deaf ear can counteract the cortical changes induced by AHL (reviewed by Vanderauwera et al. 2020). However, none of these studies established a link between brain plasticity and performance on auditory tasks, as we have shown here. In addition, some of the studies only investigated single cases (Firszt et al. 2013;Sharma and Glick 2016), while others had highly variable data (Lee et al. 2020), thus precluding any robust conclusions. Most of these previous studies used EEG to study the neural changes (Legris et al. 2018(Legris et al. , 2020Polonenko et al. 2018;Lee et al. 2020), which limits the spatial resolution of any findings. They were therefore unable to precisely locate the functional reorganization that takes place, as EEG relies on algorithms of source reconstruction, which lack accuracy. In our study, the PET methodology was able to precisely locate the restoration of functional lateralization to the auditory cortex (primary and nonprimary) and outside in higher integrative levels (Strelnikov et al. in preparation). One of the limitations is the difficulty in obtaining cortical activation maps through an implant, which is incompatible with fMRI imaging or creates artifacts in EEG imaging. On the other hand, PET imaging provides the unique advantage of being compatible with stimulation through the cochlear implant. As in most clinical studies in deaf patients, our results are relatively variable. Still, positive AI values were obtained in the vast majority of patients (8 out of 10). This result makes us confident about the validity of the restoration of cortical lateralization after implantation and allows us to show that the auditory cortical activity relates to sound localization performance. It was also possible to show that the auditory cortical activity relates to sound localization performance. It was also possible to show, using a whole-brain regression analysis, that the function integrity (capacity to respond) of the auditory cortex is predictive of auditory spatial perception. These data highlight the early stages of auditory processing in the cortex as a key level that links brain plasticity to spatial hearing skills. However, brain plasticity may also take place in areas outside of the auditory cortex, such as when acoustical and electrical information are combined to process nonspatial auditory information (Strelnikov et al. in preparation). Developmental studies in animals or humans have led to the notion of an aural preference syndrome after unilateral deafness. The studies on the impact of sequential bilateral cochlear implantation showed that the restoration of lateralization was more or less complete depending on the age at implantation, the time between each CI, and the duration of CI experience (Sharma et al. 2005;Polonenko et al. 2018;Lee et al. 2020), in line with the notion of critical periods during the development . Our present results are somewhat striking as they are obtained in adultacquired deafness, which is supposed to present a lower plasticity potential compared to early developmental stages. However, adult plasticity has been demonstrated in the case of profound deafness, particularly crossmodal plasticity Anderson et al. 2017;Stropahl and Debener 2017). Crossmodal plasticity can be reversed by restoring hearing functions through a CI , indicating that intra-and crossmodal reversal plasticity can occur in the adult auditory cortex. However, while the recovery of language or spatial functions in children is quite convincing (Polonenko et al. 2018;Illg et al. 2019), the capacity to restore spatial hearing processing in adults SSD-CI is still relatively restricted and mostly relies on behavioral studies (Dillon et al. 2017;Buss et al. 2018). Thus, the relationships between the level of cortical restoration (at the macroscopic level of the lateralization pattern) and the level of functional recovery differ between the adult and the early developmental stages, suggesting the existence of additional age-dependent fine-grained neural interaction processes.

Conclusion
In this study, we demonstrated that the restoration of binaural stimulation in AHL-CI patients reverses the abnormal lateralization pattern induced by AHL. In addition, the hemispheric dominance contralateral to the nonimplanted ear was found to be associated with superior sound localization. This supports the hypothesis that aural dominance relates to representations of the spatial sound field; this would be disrupted in unilateral deafness and restored in AHL-CI patients. The results suggest that cochlear implantation rehabilitates the binaural excitatory/inhibitory cortical interactions, and that these enable the recovery of the spatial selectivity involved in sound localization.