Date of Graduation

1996

Document Type

Dissertation/Thesis

Abstract

Published research has shown that a simple SCUBA-type mouthpiece may effectively be used in conjunction with an acoustic impedance tube to perform an acoustic reflection test which is thought to well characterize cross-sectional area changes in the larger airways of human subjects. The test, however, has only been performed on the researchers themselves. To test the applicability of the technique in the clinical setting, acoustic testing was conducted on a population of known asthmatic subjects and age-matched controls. It is shown that the SCUBA-type mouthpiece is not appropriate for general population testing. An enhanced mouthpiece is required which both prevents strong acoustic reflections from the tube end and also minimizes variation in the shape of the mouth and pharynx. An assumption is made, when performing the acoustic reflection test, that the system under study is at least wide sense stationary. In addition, once the test has been completed there is always some degree of random error present in the subsequent estimation of the frequency response. The only available technique for reducing this random error is the obtainment of a large number of readings and the calculation of their ensemble average. In testing human subjects, however, it has been found to be difficult, if not impossible, to obtain a sufficient number of readings for the ensemble average while still maintaining a stationary system. A solution was found by devising a novel signal processing algorithm based on a simple back-propagation neural network. When tested on lung models, this algorithm has been shown to decrease the random error of the frequency response to 50% of its original level when fewer than 10 readings were obtained for the ensemble average.

Share

COinS