For the same exposure level, the prevalence of NIHL is greater in workers exposed to non-G noise environments than for workers exposed to G noise. The kurtosis metric may be a reasonable candidate for use in modifying exposure level calculations that are used to estimate the risk of NIHL from any type of noise exposure environment. However, studies involving a large number of workers with well-documented exposures are needed before a relation between a metric such as the kurtosis and the risk of hearing loss can be refined.
Seventeen groups of chinchillas with 11 to 16 animals/group (sigmaN = 207) were exposed for 5 days to either a Gaussian (G) noise or 1 of 16 different non-Gaussian (non-G) noises at 100 dB(A) SPL. All exposures had the same total energy and approximately the same flat spectrum but their statistical properties were varied to yield a series of exposure conditions that varied across a continuum from G through various non-G conditions to pure impact noise exposures. The non-G character of the noise was produced by inserting high level transients (impacts or noise bursts) into the otherwise G noise. The peak SPL of the transients, their bandwidth, and the intertransient intervals were varied, as was the rms level of the G noise. The statistical metric, kurtosis (beta), computed on the unfiltered noise beta(t), was varied 3 < or = beta(t) < or = 105. Brainstem auditory evoked responses were used to estimate hearing thresholds and surface preparation histology was used to determine sensory cell loss. Trauma, as measured by asymptotic and permanent threshold shifts (ATS, PTS) and by sensory cell loss, was greater for all of the non-G exposure conditions. Permanent effects of the exposures increased as beta(t) increased and reached an asymptote at beta(t) approximately 40. For beta(t) > 40 varying the interval or peak histograms did not alter the level of trauma, suggesting that, in the chinchilla model, for beta(t) > 40 an energy metric may be effective in evaluating the potential of non-G noise environments to produce hearing loss. Reducing the probability of a transient occurring could reduce the permanent effects of the non-G exposures. These results lend support to those standards documents that use an energy metric for gauging the hazard of exposure but only after applying a "correction factor" when high level transients are present. Computing beta on the filtered noise signal [beta(f)] provides a frequency specific metric for the non-G noises that is correlated with the additional frequency specific outer hair cell loss produced by the non-G noise. The data from the abundant and varied exposure conditions show that the kurtosis of the amplitude distribution of a noise environment is an important variable in determining the hazards to hearing posed by non-Gaussian noise environments.
The effects on hearing and the sensory cell population of four continuous, non-Gaussian noise exposures each having an A-weighted L(eq)=100 dB SPL were compared to the effects of an energy-equivalent Gaussian noise. The non-Gaussian noise conditions were characterized by the statistical metric, kurtosis (beta), computed on the unfiltered, beta(t), and the filtered, beta(f), time-domain signals. The chinchilla (n=58) was used as the animal model. Hearing thresholds were estimated using auditory-evoked potentials (AEP) recorded from the inferior colliculus and sensory cell populations were obtained from surface preparation histology. Despite equivalent exposure energies, the four non-Gaussian conditions produced considerably greater hearing and sensory cell loss than did the Gaussian condition. The magnitude of this excess trauma produced by the non-Gaussian noise was dependent on the frequency content, but not on the average energy content of the impacts which gave the noise its non-Gaussian character. These results indicate that beta(t) is an appropriate index of the increased hazard of exposure to non-Gaussian noises and that beta(f) may be useful in the prediction of the place-specific additional outer hair cell loss produced by non-Gaussian exposures. The results also suggest that energy-based metrics, while necessary for the prediction of noise-induced hearing loss, are not sufficient.
Named entity recognition (NER) is a well-studied task in natural language processing. However, the widely-used sequence labeling framework is usually difficult to detect entities with nested structures. The span-based method that can easily detect nested entities in different subsequences is naturally suitable for the nested NER problem. However, previous span-based methods have two main issues. First, classifying all subsequences is computationally expensive and very inefficient at inference. Second, the span-based methods mainly focus on learning span representations but lack of explicit boundary supervision. To tackle the above two issues, we propose a boundary enhanced neural span classification model. In addition to classifying the span, we propose incorporating an additional boundary detection task to predict those words that are boundaries of entities. The two tasks are jointly trained under a multitask learning framework, which enhances the span representation with additional boundary supervision. In addition, the boundary detection model has the ability to generate high-quality candidate spans, which greatly reduces the time complexity during inference. Experiments show that our approach outperforms all existing methods and achieves 85.3, 83.9, and 78.3 scores in terms of F1 on the ACE2004, ACE2005, and GENIA datasets, respectively.
Objective To test a kurtosis-adjusted cumulative noise exposure (CNE) metric for use in evaluating the risk of hearing loss among workers exposed to industrial noises. Specifically: to evaluate if the kurtosis-adjusted CNE (1) provides a better association with observed industrial noise-induced hearing loss; (2) provides a single metric applicable to both complex (non-Gaussian) and continuous or steady-state (Gaussian) noise exposures for predicting noise induced hearing loss (dose-response curves). Design Audiometric and noise exposure data were acquired on a population of screened workers (N = 341) from two steel manufacturing plants located in Zhejiang province, and a textile manufacturing plant located in Henan province, China. All the subjects from the two steel manufacturing plants (N=178) were exposed to complex noise while the subjects from textile manufacturing plant (N=163) were exposed to a Gaussian (G) continuous noise. Each subject was given an otologic examination to determine their pure tone hearing threshold levels (HTL); and had their personal 8-hour equivalent A-weighted noise exposure (LAeq) and full shift noise kurtosis statistic (which is sensitive to the peaks and temporal characteristics of noise exposures) measured. For each subject an unadjusted and kurtosis-adjusted cumulative noise exposure (CNE) index for the years worked was created. Multiple linear regression analysis controlling for age was used to determine the relationship between CNE (unadjusted and kurtosis-adjusted) and the mean HTL at 3, 4 and 6 kHz (HTL346) among the complex noise exposed group. In addition, each subjects' HTLs from 0.5 - 8.0 kHz were age and gender adjusted using ANNEX A (ISO-1999) to determine whether they had adjusted high frequency noise induced hearing loss (AHFNIHL), defined as an adjusted HTL shift of 30 dB or greater at 3.0, 4.0 or 6.0 kHz in either ear. Dose-response curves for AHFNIHL were developed separately for workers exposed to G and non-G noise using both unadjusted and adjusted CNE as the exposure matric. Results Multiple linear regression analysis among complex exposed workers demonstrated that the correlation between HTL3,4,6 and CNE controlling for age was improved when using the kurtosis-adjusted CNE compared to the unadjusted CNE (R2=0.386 vs. 0.350), and that noise accounted for a greater proportion of hearing loss. In addition, while dose-response curves for AHFNIHL were distinctly different when using unadjusted CNE, they overlapped when using the kurtosis-adjusted CNE. Conclusions For the same exposure level, the prevalence of NIHL is greater in workers exposed to complex noise environments than for workers exposed to a continuous noise. Kurtosis adjustment of CNE both improved the correlation with NIHL and provides a single metric for dose response effects across different types of noise. The kurtosis-adjusted CNE may be a reasonable candidate for use in NIHL risk assessment across a wide variety of noise environments.
This pilot study demonstrated that machine learning algorithms are potential tools for the evaluation and prediction of noise-induced hearing impairment in workers exposed to diverse complex industrial noises.This is an open-access article distributed under the terms of the Creative Commons Attribution-Non Commercial-No Derivatives License 4.0 (CCBY-NC-ND), where it is permissible to download and share the work provided it is properly cited. The work cannot be changed in any way or used commercially without permission from the journal.
Data from an earlier study [Hamernik et al. (2003). J. Acoust. Soc. Am. 114, 386-395] were consistent in showing that, for equivalent energy [Leq= 100 dB(A)] and spectra, exposure to a continuous, nonGaussian (nonG) noise could produce substantially greater hearing and sensory cell loss in the chinchilla model than a Gaussian (G) noise exposure and that the statistical metric, kurtosis, computed on the amplitude distribution of the noise could order the extent of the trauma. This paper extends these results to Leq= 90 and 110 dB(A), and to nonG noises that are generated using broadband noise bursts, and band limited impacts within a continuous G background noise. Data from nine new experimental groups with 11 or 12 chinchillas/group is presented. Evoked response audiometry established hearing thresholds and surface preparation histology quantified sensory cell loss. At the lowest level [Leq=90 dB(A)] there were no differences in the trauma produced by G and nonG exposures. For Leq >90 dB(A) nonG exposures produced increased trauma relative to equivalent G exposures. Removing energy from the impacts by limiting their bandwidth reduced trauma. The use of noise bursts to produce the nonG noise instead of impacts also reduced the amount of trauma.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.