Direct visualization of auditory brainstem pathways in auditory implant candidates

Determining the integrity of the upper auditory pathways may help explain outcome variability following auditory device implantation, but conventional preoperative protocols do not characterize these structures. The fast gray matter acquisition T1 inversion recovery (FGATIR) sequence enhances subcortical gray–white matter contrast. This exploratory pilot study assessed whether FGATIR can technically resolve upper auditory brainstem pathway anatomy in auditory implant candidates. An exploratory prospective pilot case series was performed on patients ranging from 3 to 25 years in age who were undergoing evaluation for cochlear or auditory brainstem implantation between September 2022 and June 2024. After informed consent, patients underwent a standard preoperative magnetic resonance imaging protocol modified to include the FGATIR sequence with similar time under anesthesia. Two neuroradiologists with brainstem expertise who were blinded to clinical and audiometric data, conventional imaging findings, and implantation decisions determined the qualitative characteristics of upper auditory pathway structures by consensus. Four patients underwent FGATIR imaging. Upper auditory pathway structures were resolved in three of four patients, with the difference in size and/or signal intensity of auditory nuclei most consistently seen in the inferior colliculus. In a 5-month-old infant, the FGATIR contrast was suboptimal, which is thought to be due to incomplete myelination at this age. FGATIR is technically feasible for resolving upper auditory pathway anatomy in implant candidates older than approximately 2 years, although inadequate in the one patient below this age. Whether the observed structural differences carry diagnostic or prognostic significance is unknown. FGATIR should be considered investigational in this setting, pending prospective study after developing age-matched pediatric reference data to facilitate quantitative assessment.

Sensorineural hearing loss (SNHL) is a common hearing disorder in the pediatric population, with one in 7,000 neonates diagnosed with dysfunction on newborn hearing screening (Rance and Starr, 2015). Cochlear implantation (CI) can provide immense benefits for hearing rehabilitation in patients meeting CI criteria (Cottrell et al., 2024). For patients with contraindications to CI, such as an absent cochlear nerve, cochlear aplasia, or severe labyrinthitis ossificans, auditory brainstem implantation (ABI) may be helpful (Kaplan et al., 2015; Puram and Lee, 2015; Warren et al., 2010). The variability in performance seen in all auditory implants' performance can arise from a variety of factors, including hearing loss etiology, genetics, age at implantation, and postoperative rehabilitation (Cottrell et al., 2024; Hardie and Shepherd, 1999; Shepherd and Hardie, 2001). The anatomic integrity of the cochlear nuclei and other upper auditory pathways (defined here as the ascending pathways from the cochlear nucleus complex to the medial geniculate nucleus, comprising the cochlear nucleus, the trapezoid body, the superior olivary complex, the lateral lemniscus, the inferior colliculus, the brachium of the inferior colliculus, and the medial geniculate nucleus) also may affect outcome variability; however, standard computed tomography (CT) and magnetic resonance imaging (MRI) investigations for auditory implants are unable to characterize these structures (Cottrell et al., 2025).

Recent research has suggested diffusion tensor imaging (DTI) and diffusion tensor tractography (DTT) may characterize the myelination and directionality of nerve pathways (Muthusami et al., 2014; Lin et al., 2008). Myelination changes have been elucidated in various subgroups of patients with SNHL, including patients with unilateral hearing loss, older adults, and children with congenital deafness (Lin et al., 2008; Wu et al., 2009; Wang et al., 2019). In a 2018 systematic review by Tarabichi et al. (2018) evaluating DTI of central auditory pathways in sensorineural hearing loss, patients were found to have lower diffusion scalar values in several cortical and subcortical auditory structures when hearing loss was present.

The DTI and DTT results are promising, and image quality continues to improve over time, but the diffusion contrast has some important limitations, including poor inherent resolution of inherent brainstem tracts on DTI source data, spatial misregistration and image distortion from skull base MRI artifacts and tractography, and lack of external validation for DTT segmentations (Shepherd and Hoch, 2022; Shepherd et al., 2020). As an alternative or complementary sequence for depicting subcortical anatomy, fast gray matter acquisition T1 inversion recovery (FGATIR) sequences use T1 relaxation properties to suppress white matter signal and enhance gray–white matter contrast in subcortical structures (Sudhyadhom et al., 2009). Previous reports demonstrated that FGATIR can depict size, orientation, and volume of auditory brainstem pathways and nuclei, including the medial geniculate nucleus, the brachium of the inferior colliculus, the inferior colliculus, the lateral lemniscus, the trapezoid body, and the cochlear nucleus (Shepherd and Hoch, 2022; Shepherd et al., 2020). The contrast resolution in the brainstem with FGATIR can be further enhanced by using a convolutional neural network (CNN) to denoise the image (Ades-Aron et al., 2022).

The present exploratory pilot study highlights our initial experiences utilizing CNN-denoised FGATIR sequences to evaluate the feasibility of visualizing structural differences within auditory brainstem pathways in a series of pediatric patients undergoing candidacy assessment for auditory implantation. This population was selected because congenital cochlear and cochlear nerve malformations tend to be seen more commonly during cochlear implant assessments, and the patient population is of particular interest for determining the potential utility of FGATIR sequences.

The study received approval from the NYU Langone Health Institutional Review Board (IRB# s18-01641). Patients were recruited prospectively for participation. Inclusion criteria included patients aged 3 months to 25 years with either pre- or postlingual SNHL who were undergoing preoperative evaluation for CI or ABI between September 2022 and June 2024. Patients were excluded if they had a contraindication to MRI. Not all patients underwent implantation. Patients who met the inclusion criteria and consented to participate in this research to receive FGATIR imaging had relevant data collected from their electronic chart, including demographic data (age, sex), comorbidities, duration of deafness, hearing loss etiology, auditory brainstem response (ABR), behavioral responses, and subsequent auditory rehabilitation. Eligible patients were approached during routine clinical visits. Whether an approached patient could subsequently be enrolled depended on scanner availability, anesthesia scheduling, and the availability of a research team member to coordinate the scan and confirm that research parameters were implemented. Patients were reassigned to standard imaging protocols if these logistics risked delayed care. Only patients who consented and were scanned had documentation gathered for analysis. This case series is therefore designed to assess technical feasibility, not diagnostic performance.

All scans were completed on a Food and Drug Administration (FDA)-approved 3T Magnetom Prisma MRI using a standard FDA-approved 20-channel head and neck coil (Siemens Healthcare, Erlangen, Germany). A modified MRI evaluation for CI and ABI candidacy assessment was developed to reduce standard sequence scan time to facilitate research sequences without significantly extending anesthesia time (Table 1). The modified protocol was judged to provide equivalent diagnostic capability through a collaborative review of the images by the neuroradiology and neurotology teams—lateral skull base surgical teams. Acquiring both the complete standard protocol and the full experimental protocol in a single session would extend anesthesia/scan time beyond what our institutional review board would approve, so a within-patient comparison of the two protocols was not possible. The modified MRI evaluation, including the new research sequences, increased the total scan time from 32 min 28 s to 36 min 11 s, a difference of 3 min 43 s, and the estimated anesthesia time by approximately 5 min (Table 1). The key sequence for this protocol was a 1-mm isotropic three-dimensional (3D) FGATIR sequence [repetition time (TR)/echo time (TE)/inversion time (TI) = 3,000/2.4/410 ms, flip angle 6 degrees, 256 mm field of view (FOV), 160 sagittal slices, partial Fourier = 7/8, time = 10 min 50 s]. A high-spatial, angular, and q -space resolution diffusion MRI sequence was also obtained (2-mm isotropic resolution, multiband 2-slice acceleration, b -values = 0, 1,000, 2,000, and 3,000 s/mm 2, 256 total unique gradient strengths and directions). Other MRI sequences included 3D volumetric T1, 3D high-resolution T2 of the posterior fossa and temporal bones [sagittal sampling perfection with application-optimized contrasts using different flip-angle evolutions (SPACE), axial, and sagittal constructive interference in steady state (CISS)], axial fluid-attenuated inversion recovery (FLAIR), and susceptibility-weighted imaging (SWI). All examinations of pediatric patients were performed under general anesthesia, which limited head motion. No motion correction was required, no examinations required repetition for motion artifact, and no anesthesia-related complications occurred.

Figure 1 is a visualization of a normal brainstem auditory pathway in a healthy young adult control using the FGATIR sequence. For this study, it is important to note that in previous reports for healthy young adult subjects without hearing loss, no right–left asymmetries were detected (Shepherd and Hoch, 2022; Shepherd et al., 2020). Adult FGATIR findings cannot, however, serve as a normative standard for children because myelination and subcortical T1 relaxation properties change substantially throughout childhood. All subject FGATIR images were reviewed by two board-certified neuroradiologists with 2 and 13 years' postgraduate experience, respectively, and expertise in brainstem anatomy who were blinded to clinical and audiometric data (including laterality of hearing loss), conventional imaging findings, and implantation decisions. Assessment was qualitative and reached by consensus. Readings were not performed independently, and no formal interobserver reliability metric was calculated, as with four patients, any such coefficient would carry confidence intervals too wide to be interpretable. No additional quantitative criteria were established to define abnormality for these structures in this protocol because no pediatric normative FGATIR data set exists. Assessment therefore relied on expert qualitative comparison.

Four patients consented for participation in the study and are presented in this initial series. FGATIR was technically interpretable in three of four patients. FGATIR findings did not influence candidacy, implant type, implanted side, surgical planning, counseling, or rehabilitation for any patient. These decisions were determined by conventional imaging together with clinical and audiometric findings. In the one patient for whom FGATIR was not technically feasible, the remaining sequences in the imaging protocol allowed for conventional image analysis without concern for image degradation.

Case 1

A 4-year-old male born premature at 32 weeks gestation outside of North America presented to the clinic with concern for bilateral hearing loss. The pregnancy had been complicated by maternal malaria infection requiring treatment in the first and second trimesters, in addition to maternal hypertension prompting c-section delivery. The patient developed jaundice and anemia requiring 2 weeks in the neonatal intensive care unit (NICU) and blood transfusion. Comorbidities included choreoathetoid cerebral palsy, global developmental delay, strabismus, and prior umbilical hernia surgery.

The duration of deafness was unknown, as newborn hearing screening had not been completed before immigration, when a health assessment diagnosed hearing loss. The patient has no family history of hearing loss and had not trialed hearing amplification. ABR testing revealed absent waveform responses on the right. A cochlear microphonic was present with non-replicable wave V formation consistent with auditory neuropathy spectrum disorder. A moderately severe-to-profound SNHL at 500–4,000 Hz was noted in the left ear. There were no responses noted with visual reinforcement audiometry (VRA) on the right, while speech detection threshold (SDT) on the left was found to be 65 dB.

High-resolution T2-weighted MRI sequences demonstrated normal-appearing internal auditory canals (IACs), labyrinth, and mastoid. The posterior fossa and brain appeared grossly normal without hydrocephalus. An analysis of FGATIR sequences demonstrated a reduced size of the left medial geniculate nucleus, the inferior colliculus, and the brachium of the inferior colliculus. The right trapezoid body also appeared reduced in size (Figure 2).

The patient enrolled in speech therapy, physical therapy, and occupational therapy at their home. They utilized hearing aids for a year without adequate benefit and then underwent bilateral cochlear implantation (CI 632) without complication. At 3 months, the patient had aided SDT to “ba,” “sh,” or “s” of 35, 35, 30 dB hearing loss on the right and 25, 25, 35 dB hearing loss on the left.

Case 2

A 5-month-old female born full term by C-section for breech birth presented to clinic with concern for bilateral hearing loss. They had no history of NICU stay. They were noted to have hypertonia and poor balance, with evidence of an ear pit.

The patient passed their newborn hearing screening bilaterally; however, the parents grew concerned for hearing loss by 1 month of age. They were utilizing appropriately fitted hearing aids at the time of clinic evaluation. The patient had genetic testing completed that was unremarkable. ABR testing was consistent with severe-to-profound SNHL bilaterally. VRA demonstrated no responses, and no SDT was observed. Aided testing was consistent with bilateral profound hearing loss.

High-resolution T2-weighted MRI sequences demonstrated bilateral labyrinthine malformations and IAC stenosis. Visualization of the right cochlear nerve was difficult, with only a thin left cochlear nerve visualized. The right lateral semicircular canal (SCC) was absent, and the left lateral SCC was dysplastic. FGATIR sequences showed poor myelin contrast (Figure 3) because the current sequence inversion, echo, and repetition times are optimized for subcortical T1 values in patients 2 years and older (mature subcortical myelination).

The patient was connected with early intervention support and is receiving speech therapy, physical therapy, occupational therapy, and American Sign Language (ASL) education. At 8 months of age, the patient underwent bilateral CI with the Cochlear 632 electrode, with good placement confirmed on X-ray. Neural response telemetry thresholds were questionable at the time of surgery. At the 3-month follow-up, the patient demonstrated bilateral sound detection, with aided SDT to “ba” and “sh” of 45 dB and 70 dB hearing loss on the right and 60 dB hearing loss on the left. They continue to wear the device regularly.

Case 3

The patient presented at 3 years of age with concern for bilateral hearing loss. They were born at term and required a stay in the NICU for several days due to respiratory distress. They were noted to have strabismus but were otherwise healthy. The patient failed the newborn hearing screening and was subsequently found to have bilateral microtia and aural atresia. They were fitted with binaural bone conduction amplification for 6 months until ABR and behavioral testing demonstrated no responses.

High-resolution T2-weighted MRI sequences demonstrated bilateral cochlear hypoplasia and absent cochlear nerves within hypoplastic IACs. Possible fascicles of aberrant eighth cranial nerves were noted within the cerebellopontine angle (CPA). The labyrinths demonstrated hypoplasia in the right lateral SCC, which was absent in the left lateral SCC. The FGATIR sequence demonstrated subtly diminished myelin signal intensity in the left inferior colliculus and its brachium (Figure 4).

After the patient was deemed to not be a CI candidate, the risks and reasonable expectations of ABI, with a low expectation of oral communication, were discussed with the family at length. Speech-language pathology (SLP) and ASL supports were initiated. The family ultimately decided to undergo ABI placement on the right side, which occurred without complication. One month following ABI activation, they were vocalizing more and recognizing environmental sounds and are continuing to work through progressive programs with SLP.

Case 4

A 15-year-old female had presented earlier in life for consideration of CI/ABI. At the time, the family deferred surgical management. The patient became curious about hearing when she got older and wished to revisit treatment options. She had been born full term without complication. The mother had reported a history of gestational diabetes. She has multiple comorbidities, including Klippel–Feil syndrome with fused vertebrae, Duane syndrome with limited peripheral vision, and SNHL (the overall pattern was consistent with Wildervanck syndrome).

The patient failed her newborn hearing screening bilaterally and at the time of this study communicated using ASL. She wore binaural hearing aids until fourth grade and then stopped using hearing aids for 3 years before starting to wear a hearing aid in her right ear. Behavioral responses demonstrated profound SNHL bilaterally, with aided testing demonstrating a possible response at 60–80 dB for 250–500 Hz, although this may have been vibrotactile.

High-resolution T2-weighted MRI sequences detected no cochlea bilaterally and demonstrated hypoplastic IACs. Hypoplastic seventh and eighth cranial nerves were present bilaterally in the CPA. There were bilateral congenital anomalies of the inner ear structures, with a 4-mm cystic structure in the left petrous apex and a more complex cystic structure in the right petrous apex, with no well-defined cochlea, vestibule, or SCCs. The FGATIR sequence showed a reduced inferior colliculi size bilaterally (right smaller than left; Figure 5).

Extensive counseling was provided to the patient, in addition to a guarded prognosis for ABI due to age and lack of oral language. The patient planned to continue ASL with educational supports.

Preoperative candidacy assessment for CI and ABI typically begins with evaluation for behavioral responses to auditory stimuli, in conjunction with ABR testing (Cottrell et al., 2024). Conventional MRI, often supplemented with CT imaging, is performed to assess inner ear and cochlear nerve anatomy; however, the sensitivity of imaging to identify the cochlear nerve is imperfect (Warren et al., 2010). Moreover, the presence of an intact cochlear nerve does not reliably predict favorable CI outcomes, as many additional patient-specific factors, including age at implantation and rehabilitation supports, influence postoperative performance variability. Earlier implantation has been shown to be safe and effective in children as young as 6 months, emphasizing the importance of minimizing delays in determining candidacy and interventions (Cottrell et al., 2024; Warner-Czyz et al., 2022). Interestingly, some patients presumed to have cochlear nerve aplasia on imaging still achieve meaningful outcomes following implantation (Warren et al., 2010). When there truly is a non-functional cochlea or auditory nerve, an ABI can bypass the structures to directly stimulate the brainstem. Substantial variability in postoperative auditory performance persists in each case, underscoring the need for improved predictors of postoperative auditory performance (Kaplan et al., 2015; Puram and Lee, 2015).

An improved understanding of auditory brainstem pathway integrity may provide additional insight into the expected performance with implantable auditory devices. Animal studies have demonstrated that SNHL can reduce neuron soma size in the cochlear nucleus, the medial nucleus of the trapezoid body, the superior olivary complex, the lateral lemniscus, and the inferior colliculus (Hardie and Shepherd, 1999; Shepherd and Hardie, 2001; Nishiyama and Hardie, 2000; Pasic et al., 1994). Non-invasive imaging techniques capable of detecting structural changes in these regions for human patients is an important area of current investigation. Prior work using DTI in patients with unilateral and bilateral profound deafness reported different diffusion scalar values in several cortical and putative subcortical auditory structures compared with normal-hearing controls (Chang et al., 2004; Rachakonda et al., 2014; Aksoy and Karagöz, 2023; Huang et al., 2015; Reiman et al., 2009). Additionally, positive correlations have been reported between cochlear implant Categories of Auditory Performance scores and fractional anisotropy values within the auditory radiation, the superior olivary nucleus, the inferior colliculus, the medial geniculate body, the trapezoid body, Broca's area, and the genu of the corpus callosum (Tarabichi et al., 2018).

Previous studies have reported the visualization of changes to auditory brainstem anatomy in patients with hearing loss using color-encoded fractional anisotropy maps from DTI or diffusion tractography (Muthusami et al., 2014; Lin et al., 2008; Wu et al., 2009; Wang et al., 2019; Tarabichi et al., 2018). Diffusion MRI has three important limitations: (1) low spatial resolution (e.g., 2-mm isotropic resolution diffusion MRI is state of the art at 3T, but the lateral lemniscus is < 1 mm in thickness); (2) significant spatial misregistration and image distortion, especially in the skull base near the brainstem; and (3) poor direct contrast resolution of adjacent brainstem structures (particularly in the tegmentum; Figure 6) (Shepherd and Hoch, 2022; Shepherd et al., 2020). Previous DTI investigations risk analytical circularity and confirmation bias by using diffusion image contrast or tractography to segment structures and then report diffusion MRI changes associated with hearing loss. The FGATIR sequence, which more closely resembles postmortem histology contrast of myelination in the brainstem (Shepherd and Hoch, 2022; Shepherd et al., 2020), can provide complementary contrast and/or an independent reference standard for extracting diffusion data. The FGATIR sequence is derived from gradient-based 3D volumetric sequences that are the gold standard for surgical planning because of higher resolution, accurate co-registration, and the absence of skull base artifacts (additional advantages over diffusion MRI).

In three of the four patients in our series (Cases 1, 3, and 4) ages 2 years or older with more adult-like brainstem myelination, we were able to identify most relevant upper auditory brainstem pathway structures using the FGATIR sequence. Among these patients, an asymmetric reduction in the size of multiple auditory pathway nuclei was observed (i.e., left medial geniculate nucleus, the left inferior colliculus, and the brachium and right trapezoid body in Case 1). These findings are anatomically concordant rather than discordant, as the principal ascending auditory projection decussates at the trapezoid body level. Predominantly right-sided deprivation would be expected to affect the right trapezoid body, together with the contralateral left inferior colliculus, the brachium of the inferior colliculus, and the medial geniculate nucleus. These changes most often included reduced signal intensity in the inferior colliculus. Interestingly, Case 3 has also completed a positron-emission tomography CT, which also demonstrated asymmetric fluorodeoxyglucose (FDG) uptake of the inferior colliculus ipsilateral to the FGATIR abnormality, a finding previously associated with hearing loss (Speck et al., 2020).

In Case 2, FGATIR was unable to reliably demonstrate myelin contrast, which was attributed to the patient's age at the time of imaging, when brain myelination remains incomplete. Prior experience suggests that the ability of the existing FGATIR to resolve small brainstem structures improves after 2 years of age. Further work could potentially adjust the FGATIR sequence's inversion, echo, and repetition times to known T1 values for children younger than 2 years (e.g., specific FGATIR sequences tuned to 6, 12, and 18 months of age). This is an important future direction because considering ABI in patients before the age of 2 years may be ideal (Sennaroglu et al., 2013). Comparing anatomic FGATIR developmental findings to prior research evaluating the evolution of DTI/DTT images in premature newborns, patients with perinatal injury, patients with congenital hearing loss, and in the aging brain also would be of interest (Miller et al., 2002; Forbes et al., 2002; Anjari et al., 2007; Schneider, 2004; Hüppi et al., 2001; Berman et al., 2005; Wu et al., 2016; Profant et al., 2014).

This series demonstrates the technical feasibility of using FGATIR to evaluate upper auditory pathway structures in adult and pediatric patients undergoing assessment for auditory implantation. As this is a small series, with each patient representing a variety of complex hearing pathologies, it is not possible to determine the mechanisms underlying each imaging findings. At least four explanations could be considered that may not be mutually exclusive, including deafferentation-related degeneration of a structure that had matured normally, developmental hypoplasia arising from absent afferent input during a critical period, primary central nervous system injury independent of auditory deprivation (e.g., bilirubin-mediated injury and/or choreoathetoid cerebral palsy), and independent delayed or atypical myelination development. In previous published studies of healthy adults, the auditory brainstem pathways are consistently identified and appear consistently symmetric (Shepherd and Hoch, 2022; Shepherd et al., 2020). Asymmetric differences in structure visualization were observed with FGATIR in three of the four initial patients with hearing loss. Whether these differences identify the more profoundly affected upper auditory pathway is a hypothesis that requires prospective testing and is not demonstrated by this series. Whether choosing the less affected side on FGATIR images for CI or ABI would result in better outcomes likewise remains an open question for future prospective study. Because central auditory projections are extensively bilateral and predominantly crossed, an asymmetry observed within an isolated brainstem nucleus cannot be interpreted as evidence favoring implantation of one ear over the other. Future studies examining whether FGATIR-derived structural findings complement DTI findings will be another important area of future investigation (Tarabichi et al., 2018).

This study has important limitations. Four patients were enrolled, and three were technically evaluable with FGATIR, so no inference regarding diagnostic performance, prognostic value, or clinical utility is possible. The FGATIR acquisition and CNN postprocessing were utilized together in all cases; therefore, this series is limited in its ability to attribute observed visualization advantages solely to FGATIR acquisition when the contribution of CNN denoising has not been independently assessed. Recruitment approaches were not recorded prospectively, so the denominator required to assess selection bias is unavailable. Normal appearance has been referenced to prior experience in healthy adult patients; however, adult FGATIR findings cannot serve as a pediatric normative standard, given ongoing myelination throughout childhood. That some observed asymmetries in this series fall within normal developmental variation is possible. In addition, the assessment was qualitative, and image reviews were by consensus, without predefined quantitative criteria and without a formal interobserver reliability metric. Quantitative volumetric and signal-intensity comparisons against pediatric norms would be required to establish that reported asymmetries exceed developmental variability. Expanding this technique to generate age-matched pediatric normative FGATIR data is a necessary next step to help in this regard.

The FGATIR MRI sequence is a promising novel structural neuroimaging tool capable of detecting morphologic differences within upper auditory pathways. These morphologic differences were not apparent on the conventional imaging that was available in these cases. The sequence is currently limited in younger patients (< 2 years) for whom central myelination has not matured. While the incremental diagnostic value of FGATIR MRI sequences and prognostic value of the findings presented have yet to be established, this preliminary series supports further investigation of FGATIR as a complementary modality alongside diffusion-based techniques.