Text

theses

Currently, there is a lack of an FDA approved bio-equivalence test for generic aerosol drug products. The prime candidate for such a test is a device known as the cascade impactor. The cascade impactor uses a momentum-based impaction mechanism to classify incoming aerosols into different size bins. As the deposition of an aerosol in the lung is also driven by a momentum-based impaction, in principle, the cascade impactor is expected to be ideally suited for the task. In practice the cascade impactor has been limited by a major shortcoming: it does not sharply divide incoming aerosol into various size bins. To make impactors as accurate as possible, the current development methodology uses data from computational fluid dynamics (CFD) models to produce stages with maximally sharp cut-off sizes. However, these models do not account for the propensity for solid particles to rebound off the collection plate. As a result, cascade impactors have fairly straight impaction curves, but these curves come at the cost of increased variability. It is hypothesized that a superior impactor could be created by eliminating particle bounce, and then developing methods to accurately recover the data for the non-ideal impactor stage performance. In this work, tools are developed to help make such a device a reality. First, two inference-based inversion techniques, maximum entropy and fisher information, are formulated for use with the cascade impactor and tested with a model Andersen Cascade Impactor. Both techniques are shown to be capable of recovering accurate distributions from non-ideal impactors. The maximum entropy technique is found to be mathematically less complex, but also less accurate. The fisher information technique demonstrated superior accuracy, but it is much more mathematically complex and difficult to implement. For both inversion techniques, the relationship between neighboring impactor stages is found to be important to the accuracy of the inversion technique. In the second part, the ability of CFD tools to predict the impactor curve shape was tested. It was found that this approach lead to an over prediction of the sharpness of the impactor curves.

ETD_3220

Ph.D.

Includes bibliographical references

Includes vita

by Eric Jayjock

http://hdl.rutgers.edu/1782.1/rucore10001600001.ETD.000061293

doi:10.7282/T3KK9B34

ETD doctoral

The author owns the copyright to this work.

1403392

windows xp

ETD

application/pdf

application/x-tar

1413120

21406c2dc6d237442ee0bd63f878f0841f9fcc00