A computational approach to optimal deactivation of cochlear implant electrodes

项目摘要 耳蜗植入术已成功地恢复了全世界数十万人的听力。 然而，语音理解性能与CI仍然是高度可变的。尽管这种变化， 结果，CI通常使用默认设置进行编程，只有少数受试者特定编程 参数调整。这可能是因为有太多可能的CI设备设置组合 缺乏可用的和数据驱动的指导，以实现个人CI用户的最大利益。为 例如，大多数CI都是在所有可行电极都处于活动状态的情况下进行编程的，即使有些电极可能 不适合传达语音信息。以前的研究评估了是否选择性失活的一些 这些电极改善了语言理解，但效果有限。然而，这些研究并没有 考虑语音信息如何在剩余电极上表示，或者最佳数量的 电极去激活。本研究的目标是使用计算驱动的语音模型 CI用户的理解，以指导搜索哪种有源电极组合可以产生最佳效果 对特定患者的言语理解。 目标1是量化语音理解和声音质量与模型推荐的组合的积极 电极与临床标准护理设置相比，以及与两个对照有源电极相比 组合。这些替代条件将使用与模型相同数量的电极- 推荐的条件下，但电极选择与以前的研究类似的方式。受试者将 1.5在每个实验活性电极条件下定期使用数月。性能与实验 将使用重复测量设计比较临床活性电极条件。据推测 模型推荐的条件将导致比另一个更好的语音理解 条件目标2是将目标1中的模型驱动建议转化为以下方面的实际指导： 去激活多少个（以及可能哪些）CI电极。受试者在实验和 目标1的临床活性电极条件将用于构建分层线性模型。这种模式的 将具有实验活性电极设置的性能与以下自变量相关联： 由模型为受试者选择的有源电极的数量、有源电极的物理跨度、受试者的语音- 线索分辨率与每个实验条件，和三个人口统计变量。它是假设 受试者的表现将与有源电极的数量和受试者的 语音提示分辨率，并适度相关的物理跨度的积极电极。 在当前提案的支持下获得的数据将是更大规模研究的基础，以提供数据- 用于CI器械最佳适配的驱动指导。

PROJECT SUMMARY Cochlear implantation has successfully restored hearing to hundreds of thousands of individuals worldwide. However, speech understanding performance with CIs remains highly variable. Despite this variability in outcomes, CIs are typically programmed with default settings with only a few subject-specific programming parameters adjusted. This likely occurs because too many possible CI device setting combinations are available and data-driven guidance to achieve maximal benefit for an individual CI user is lacking. For example, most CIs are programmed with all viable electrodes active even though some electrodes may be poorly suited to convey speech information. Previous studies evaluated whether selective deactivation of some of these electrodes improved speech understanding, with limited success. However, these studies did not consider how speech information is represented across remaining electrodes, or the optimal number of electrodes to deactivate. The goal of the present study is to use computationally driven models of speech understanding in CI users to guide the search for which combination of active electrodes can yield the best speech understanding for a specific patient. Aim 1 is to quantify speech understanding and sound quality with model-recommended combinations of active electrodes compared to clinical standard-of-care settings, and compared to two control active electrode combinations. These alternative conditions will use the same number of electrodes as the model- recommended condition, but with electrodes selected in a similar way as previous studies. Subjects will have 1.5 months of regular use with each experimental active electrode conditions. Performance with experimental and clinical active electrode conditions will be compared using a repeated-measures design. It is hypothesized that the model-recommended condition will result in significantly better speech understanding than the other conditions. Aim 2 is to translate the model-driven recommendations from Aim 1 into practical guidance about how many (and possibly which) CI electrodes to deactivate. Subjects’ performance with the experimental and clinical active electrode conditions of Aim 1 will be used to build a hierarchical linear model. This model will relate performance with experimental active electrode settings to the following independent variables: number of active electrodes selected for subjects by the model, physical span of active electrodes, subjects’ speech- cue resolution with each experimental condition, and three demographic variables. It is hypothesized that subjects’ performance will be more strongly correlated with the number of active electrodes and subjects’ speech-cue resolution, and moderately correlated to physical span of active electrodes. Data obtained under the auspices of the current proposal will be foundational for larger studies to provide data- driven guidance for optimal fitting of CI devices.