A statistical learning model for predicting noise-induced hearing loss in humans

用于预测人类噪声引起的听力损失的统计学习模型

• 批准号: 7527104
• 负责人: Roger P Hamernik
• 金额: $ 36.63万
• 依托单位: PLATTSBURGH STATE UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2008
• 资助国家: 美国
• 起止时间: 2008-09-01 至 2011-08-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/7527104
• 关键词: statistical; learning; model; predicting; noise

DESCRIPTION (provided by applicant): Existing damage risk criteria for noise exposure are based entirely on an energy metric. Our animal model experiments have shown that energy alone is not sufficient to characterize a complex noise for hearing conservation purposes. These data suggest that energy and the statistical metric kurtosis may constitute necessary and sufficient metrics for the prediction of hearing loss that will result from long-term industrial exposures. This proposal is a continuation of this effort but focuses on the human condition. It is important to verify the animal results with comparable human exposure and hearing loss data. The main aim is to develop a noise measurement/analysis strategy that can be used to estimate the hearing loss that will develop in workers exposed to complex industrial noise* environments. Statistical learning models will be developed to achieve this goal. Model development will require that the following data be obtained: (1) Stable audiograms on a population of highly screened workers exposed to Gaussian and nonGaussian noise environments. (2) Record, archive and analyze the noise waveforms that each subject is exposed to over the course of an entire work shift. The models developed from this database will identify the noise variables that are important in the production of hearing loss and can also be used to develop exposure criteria. In addition, in response to ethical concerns of collecting data from unprotected workers, elements of a hearing conservation program will be introduced. The experimental paradigm will generate data needed for the development of an empirical basis for using in addition to an energy metric to predict the hazards of an exposure, variables that quantify the temporal and peak characteristics of a noise such as the kurtosis, impact interval and peak histograms, and transient durations. Data will be collected in collaboration with researchers at Peking University and include a population of at least 1,250 workers employed in various complex, high noise level industrial settings in China. This noise exposure and cross-sectional audiometric data will constitute a unique database that will fill a NIOSH acknowledged void in the data available for developing exposure criteria for hearing conservation purposes. [*Throughout this proposal our use of the term 'complex noise' refers to non-Gaussian noise, which in the course of a work cycle, may be intermittent, interrupted and of variable level. Non-Gaussian noise is very common in industry and the military where it consists of a background Gaussian noise with embedded high- level transients (impacts or noise bursts).] Existing safety criteria for noise exposure are based entirely on an energy metric. New data indicate that energy alone is not a suitable metric to use for this purpose. The main aim of this proposal is to develop a noise measurement/analysis strategy that can be used to estimate the hearing loss that will develop in workers exposed to complex industrial noise environments. Statistical learning models will be developed to achieve this goal.

描述（由申请人提供）：现有的噪声暴露损害风险标准完全基于能量指标。我们的动物模型实验表明，仅能量不足以表征用于听力保护目的的复杂噪声。这些数据表明，能量和统计指标峰度可能构成预测长期工业暴露导致的听力损失的必要且充分的指标。该提案是这一努力的延续，但重点关注人类状况。重要的是用可比较的人类暴露和听力损失数据来验证动物结果。主要目的是开发一种噪声测量/分析策略，可用于估计暴露于复杂工业噪声*环境的工人的听力损失。将开发统计学习模型来实现这一目标。模型开发需要获得以下数据：（1）暴露于高斯和非高斯噪声环境的经过严格筛选的工人群体的稳定听力图。 (2)记录、存档并分析每个受试者在整个工作班次过程中所接触到的噪声波形。从该数据库开发的模型将识别对听力损失产生很重要的噪声变量，也可用于制定暴露标准。此外，为了回应从未受保护的工人那里收集数据的道德问题，将引入听力保护计划的要素。实验范式将生成开发经验基础所需的数据，除了使用能量指标来预测暴露的危害外，还使用量化噪声的时间和峰值特征的变量，例如峰度、冲击间隔和峰值直方图以及瞬态持续时间。数据将与北京大学的研究人员合作收集，包括中国各种复杂、高噪音工业环境中至少 1,250 名工人。这种噪声暴露和横截面听力测定数据将构成一个独特的数据库，该数据库将填补 NIOSH 承认的可用于制定听力保护暴露标准的数据空白。 [*在本提案中，我们使用的术语“复杂噪声”是指非高斯噪声，其在工作周期过程中可能是间歇性的、中断的和可变级别的。非高斯噪声在工业和军事中非常常见，它由背景高斯噪声和嵌入的高级瞬变（冲击或噪声突发）组成。]现有的噪声暴露安全标准完全基于能量指标。新数据表明，仅能源并不是用于此目的的合适指标。该提案的主要目的是开发一种噪声测量/分析策略，可用于估计暴露于复杂工业噪声环境的工人的听力损失。将开发统计学习模型来实现这一目标。