Neural network model of noise-induced hearing loss

点击翻译按钮获取中文摘要

DESCRIPTION: The objective of the proposed work is to develop a prediction model using radial basis function neural network (RBFNN) that will predict the auditory consequences of excessive noise exposure in a chinchilla model. The RBFNN model will be fed training data from our existing database (chinchilla) consisting of noise exposure metrics and audiometric/histological/otoacoustic emission data acquired from previously exposed animals. Once trained, the model will be able to predict the auditory consequences of exposure to any noise environment characterized by ten noise metrics and five biological variables of the animals that are input to the model. There are two phases to this research. The first phase is the design of the system structure and implementation software (Year 1). The second phase involves training the system using an extensive database consisting of more than 2500 subjects on which comprehensive exposure and audiometric/histological data are available (Year 2). The training period will be an iterative process in which the RBFNN will be modified as training proceeds. The prediction model can also be used to design experimental conditions from which the model can be experimentally tested. The successful demonstration of the application of a RBFNN to the prediction of noise-induced auditory effects has considerable potential for application to the assessment of industrial and military noise environments for the protection of hearing in humans, and can result in a considerable savings in the amount of work/resources that are needed to develop new and improved noise standards. Specifically, given the many similarities in the response to noise of the chinchilla and the human, the model can be used to identify combinations of parameters of an exposure that are important determinants of damage that would also be applicable to human exposure conditions. The chinchilla model can be used as a template or a guide for developing a human model possibly from some of the existing human data from which current standards were developed.