Explore the vast range of titles available.

Assess the real-world performance of popular imputation algorithms on cochlear implant (CI) candidate audiometric data. 7,451 audiograms from patients undergoing CI candidacy evaluation were pooled from 32 institutions with complete case analysis yielding 1,304 audiograms. Imputation model performance was assessed with nested cross-validation on randomly generated sparse datasets with various amounts of missing data, distributions of sparsity, and dataset sizes. A threshold for safe imputation was defined as root mean square error (RMSE) <10dB. Models included univariate imputation, interpolation, multiple imputation by chained equations (MICE), k-nearest neighbors, gradient boosted trees, and neural networks. Greater quantities of missing data were associated with worse performance. Sparsity in audiometric data is not uniformly distributed, as inter-octave frequencies are less commonly tested. With 3-8 missing features per instance, a real-world sparsity distribution was associated with significantly better performance compared to other sparsity distributions (Δ RMSE 0.3 dB- 5.8 dB, non-overlapping 99% confidence intervals). With a real-world sparsity distribution, models were able to safely impute up to 6 missing datapoints in an 11-frequency audiogram. MICE consistently outperformed other models across all metrics and sparsity distributions (p < 0.01, Wilcoxon rank sum test). With sparsity capped at 6 missing features per audiogram but otherwise equivalent to the raw dataset, MICE imputed with RMSE of 7.83 dB [95% CI 7.81-7.86]. Imputing up to 6 missing features captures 99.3% of the audiograms in our dataset, allowing for a 5.7-fold increase in dataset size (1,304 to 7,399 audiograms) as compared with complete case analysis. Precision medicine will inevitably play an integral role in the future of hearing healthcare. These methods are data dependent, and rigorously validated imputation models are a key tool for maximizing datasets. Using the largest CI audiogram dataset to-date, we demonstrate that in a real-world scenario MICE can safely impute missing data for the vast majority (>99%) of audiograms with RMSE well below a clinically significant threshold of 10dB. Evaluation across a range of dataset sizes and sparsity distributions suggests a high degree of generalizability to future applications.

The electrically evoked compound action potential (eCAP) represents the synchronous firing of a population of electrically stimulated auditory nerve fibers. It can be directly recorded on a surgically exposed nerve trunk in animals or from an intra-cochlear electrode of a cochlear implant. In the past two decades, the eCAP has been widely recorded in both animals and clinical patient populations using different testing paradigms. This paper provides an overview of recording methodologies and response characteristics of the eCAP, as well as its potential applications in research and clinical situations. Relevant studies are reviewed and implications for clinicians are discussed.

Although significant progress has been made in understanding outcomes following cochlear implantation, predicting performance remains a challenge. Duration of hearing loss, age at implantation, and electrode positioning within the cochlea together explain ~ 25% of the variability in speech-perception scores in quiet using the cochlear implant (CI). Electrocochleography (ECochG) responses, prior to implantation, account for 47% of the variance in the same speech-perception measures. No study to date has explored CI performance in noise , a more realistic measure of natural listening. This study aimed to (1) validate ECochG total response (ECochG-TR) as a predictor of performance in quiet and (2) evaluate whether ECochG-TR explained variability in noise performance. Thirty-five adult CI recipients were enrolled with outcomes assessed at 3-months post-implantation. The results confirm previous studies showing a strong correlation of ECochG-TR with speech-perception in quiet ( r  = 0.77). ECochG-TR independently explained 34% of the variability in noise performance. Multivariate modeling using ECochG-TR and Montreal Cognitive Assessment (MoCA) scores explained 60% of the variability in speech-perception in noise. Thus, ECochG-TR, a measure of the cochlear substrate prior to implantation, is necessary but not sufficient for explaining performance in noise. Rather, a cognitive measure is also needed to improve prediction of noise performance.

Vestibular schwannomas (VS) are benign tumors that lead to significant neurologic and otologic morbidity. How VS heterogeneity and the tumor microenvironment (TME) contribute to VS pathogenesis remains poorly understood. In this study, we perform scRNA-seq on 15 VS, with paired scATAC-seq ( n  = 6) and exome sequencing ( n  = 12). We identify diverse Schwann cell (SC), stromal, and immune populations in the VS TME and find that repair-like and MHC-II antigen-presenting SCs are associated with myeloid cell infiltrate, implicating a nerve injury-like process. Deconvolution analysis of RNA-expression data from 175 tumors reveals Injury-like tumors are associated with larger tumor size, and scATAC-seq identifies transcription factors associated with nerve repair SCs from Injury-like tumors. Ligand-receptor analysis and in vitro experiments suggest that Injury-like VS-SCs recruit myeloid cells via CSF1 signaling. Our study indicates that Injury-like SCs may cause tumor growth via myeloid cell recruitment and identifies molecular pathways that may be therapeutically targeted. Vestibular schwannomas are benign tumours which can lead to neurological symptoms and morbidity. Here, the authors use single cell RNA-seq and ATAC-seq to identify Schwann cell subtypes in the tumour microenvironment which mimic a nerve injury phenotype.

A FDA clinical trial was carried out to evaluate the potential benefit of cochlear implant (CI) use for adults with unilateral moderate-to-profound sensorineural hearing loss. Subjects were 20 adults with moderate-to-profound unilateral sensorineural hearing loss and normal or near-normal hearing on the other side. A MED-EL standard electrode was implanted in the impaired ear. Outcome measures included: (a) sound localization on the horizontal plane (11 positions, −90° to 90°), (b) word recognition in quiet with the CI alone, and (c) masked sentence recognition with the target at 0° and the masker at −90°, 0°, or 90°. This battery was completed preoperatively and at 1, 3, 6, 9, and 12 months after CI activation. Normative data were also collected for 20 age-matched control subjects with normal or near-normal hearing bilaterally. The CI improved localization accuracy and reduced side bias. Word recognition with the CI alone was similar to performance of traditional CI recipients. The CI improved masked sentence recognition when the masker was presented from the front or from the side of normal or near-normal hearing. The binaural benefits observed with the CI increased between the 1- and 3-month intervals but appeared stable thereafter. In contrast to previous reports on localization and speech perception in patients with unilateral sensorineural hearing loss, CI benefits were consistently observed across individual subjects, and performance was at asymptote by the 3-month test interval. Cochlear implant settings, consistent CI use, and short duration of deafness could play a role in this result.

Auditory neuropathy spectrum disorder (ANSD) is characterized by an apparent discrepancy between measures of cochlear and neural function based on auditory brainstem response (ABR) testing. Clinical indicators of ANSD are a present cochlear microphonic (CM) with small or absent wave V. Many identified ANSD patients have speech impairment severe enough that cochlear implantation (CI) is indicated. To better understand the cochleae identified with ANSD that lead to a CI, we performed intraoperative round window electrocochleography (ECochG) to tone bursts in children ( = 167) and adults ( = 163). Magnitudes of the responses to tones of different frequencies were summed to measure the \"total response\" (ECochG-TR), a metric often dominated by hair cell activity, and auditory nerve activity was estimated visually from the compound action potential (CAP) and auditory nerve neurophonic (ANN) as a ranked \"Nerve Score\". Subjects identified as ANSD (45 ears in children, 3 in adults) had higher values of ECochG-TR than adult and pediatric subjects also receiving CIs not identified as ANSD. However, nerve scores of the ANSD group were similar to the other cohorts, although dominated by the ANN to low frequencies more than in the non-ANSD groups. To high frequencies, the common morphology of ANSD cases was a large CM and summating potential, and small or absent CAP. Common morphologies in other groups were either only a CM, or a combination of CM and CAP. These results indicate that responses to high frequencies, derived primarily from hair cells, are the main source of the CM used to evaluate ANSD in the clinical setting. However, the clinical tests do not capture the wide range of neural activity seen to low frequency sounds.

Objectives: Tip fold-over (TFO) is a rare but critical occurrence in cochlear implant procedures where the electrode array folds back on itself within the cochlea, compromising programming and device performance. Timely intraoperative detection is essential for immediate correction and optimal placement. Electric field imaging (EFI) has shown promise for identifying TFO both intra- and post-operatively. This study evaluates the performance of a TFO detection algorithm implemented in Target CI (Version 1.6) using Advanced Bionics’ cochlear implant systems, validated through bench and patient datasets. Methods: Sample data included (1) bench testing with a plastic cochlea and human temporal bones with and without induced TFOs, confirmed visually or radiographically; (2) intraoperative EFI measurements recorded using the AIM™ system, with electrode placement confirmed through imaging; and (3) historical EFI recordings from the Target CI DataLake, which lacks imaging and programming metadata. The TFO algorithm’s performance was evaluated by assessing its sensitivity and specificity using these datasets. Results: The TFO algorithm achieved 100% sensitivity and specificity in bench models and intraoperative EFI with imaging-confirmed placements. Among 226 intra-op cases, four TFOs were confirmed by imaging, and all were correctly identified by the algorithm. In the large set of DataLake cases (14,734 implants), 0.80% were flagged as potential TFOs. TFO prevalence was higher with pre-curved arrays (1.22%) than straight lateral wall arrays (0.32%). Conclusions: The TFO algorithm showed high reliability with 100% sensitivity and specificity using routine clinical EFI data. While not a replacement for imaging, the TFO algorithm serves as a fast, accessible tool to alert clinicians to potential TFOs.

Abstract BACKGROUND Neurofibromatosis type 2 (NF2) often results in profound hearing loss and cochlear implantation is an emerging hearing rehabilitation option. However, cochlear implant (CI) outcomes in this population vary, and intraoperative monitoring to predict cochlear nerve viability and subsequent outcomes is not well-established. OBJECTIVE To review the use of intraoperative electrically evoked cochlear nerve monitoring in patients with NF2 simultaneous translabyrinthine (TL) vestibular schwannoma (VS) resection and cochlear implantation. METHODS A retrospective review was performed of 3 patients with NF2 that underwent simultaneous TL VS resection and cochlear implantation with electrical auditory brainstem response (eABR) measured throughout tumor resection. Patient demographics, preoperative assessments, surgical procedures, and outcomes were reviewed. RESULTS Patients 1 and 3 had a reliable eABR throughout tumor removal. Patient 2 had eABR pretumor removal, but post-tumor removal eABR presence could not be reliably determined because of electrical artifact interference. All patients achieved auditory percepts upon CI activation. Patients 1 and 2 experienced a decline in CI performance after 1 yr and after 3 mo, respectively. Patient 3 continues to perform well at 9 mo. Patients 2 and 3 are daily users of their CI. CONCLUSION Cochlear implantation is attainable in cases of NF2-associated VS resection. Intraoperative eABR may facilitate cochlear nerve preservation during tumor removal, though more data and long-term outcomes are needed to refine eABR methodology and predictive value for this population.

The goal of this work was to better understand speech perception for cochlear implant (CI) users with bilateral residual hearing, including consideration of effects related to listening conditions and test measures. Of interest was the role of acoustic hearing for speech perception in a complex background, the role of listening experience for CI-alone conditions, and whether performance with electric-acoustic stimulation (EAS) was improved by a contralateral hearing aid (HA). Eleven subjects provided data on Consonant-Nucleus-Consonant (CNC; Peterson & Lehiste, 1962) words in quiet, City University of New York (CUNY; Boothroyd, Hanin, & Hnath, 1985) sentences in steady noise, and Bamford-Kowal-Bench (Bench, Kowal, & Bamford, 1979) sentences in multitalker babble. Listening conditions included: CI with a full-frequency map, CI with a truncated-frequency map, EAS, and EAS+HA (EAS plus contralateral HA). Sounds were presented at 0° azimuth. For CNC words and CUNY sentences, performance was better with the truncated-frequency than the full-frequency map, and performance with EAS was better than for either CI-alone condition. For Bench-Kowal-Bamford sentences, EAS+HA was better than EAS. As demonstrated previously, performance was better in the EAS condition than either CI-alone condition. Better performance in the truncated-frequency than full-frequency CI-alone condition suggests that listening experience may be important. A contralateral HA improved performance over unilateral EAS under some conditions.

Assess the real-world performance of popular imputation algorithms on cochlear implant (CI) candidate audiometric data. 7,451 audiograms from patients undergoing CI candidacy evaluation were pooled from 32 institutions with complete case analysis yielding 1,304 audiograms. Imputation model performance was assessed with nested cross-validation on randomly generated sparse datasets with various amounts of missing data, distributions of sparsity, and dataset sizes. A threshold for safe imputation was defined as root mean square error (RMSE) <10dB. Models included univariate imputation, interpolation, multiple imputation by chained equations (MICE), k-nearest neighbors, gradient boosted trees, and neural networks. Greater quantities of missing data were associated with worse performance. Sparsity in audiometric data is not uniformly distributed, as inter-octave frequencies are less commonly tested. With 3-8 missing features per instance, a real-world sparsity distribution was associated with significantly better performance compared to other sparsity distributions ([DELTA] RMSE 0.3 dB- 5.8 dB, non-overlapping 99% confidence intervals). With a real-world sparsity distribution, models were able to safely impute up to 6 missing datapoints in an 11-frequency audiogram. MICE consistently outperformed other models across all metrics and sparsity distributions (p < 0.01, Wilcoxon rank sum test). With sparsity capped at 6 missing features per audiogram but otherwise equivalent to the raw dataset, MICE imputed with RMSE of 7.83 dB [95% CI 7.81-7.86]. Imputing up to 6 missing features captures 99.3% of the audiograms in our dataset, allowing for a 5.7-fold increase in dataset size (1,304 to 7,399 audiograms) as compared with complete case analysis. Precision medicine will inevitably play an integral role in the future of hearing healthcare. These methods are data dependent, and rigorously validated imputation models are a key tool for maximizing datasets. Using the largest CI audiogram dataset to-date, we demonstrate that in a real-world scenario MICE can safely impute missing data for the vast majority (>99%) of audiograms with RMSE well below a clinically significant threshold of 10dB. Evaluation across a range of dataset sizes and sparsity distributions suggests a high degree of generalizability to future applications.