Date of Award
Campus Access Dissertation
Chemistry and Biochemistry
Michael L Myrick
Unique algorithm development is vital for the success of novel instrumentation. Our lab has focused on the design of imaging systems based on molecular factor and multivariate optical computing. A simulation-driven design approach was utilized to develop a multimode infrared imaging system for chemical contrast enhancement. This infrared thermal imaging system is based on molecular factor (MFC) and lock-in computing methods. MFC was accomplished with filter elements made of thin organic films deposited on IR-transparent substrates and allows a system response to be tuned to a specific analyte. Unique algorithms were written in-house using MatLab (The Mathworks, Natick, MA). The algorithms used a lock-in computing technique to amplify the diffuse reflectance signal, which is only a few percent of the total signal. Intensive thin film studies were conducted to understand the effects of films on fabric to improve our simulation-driven design approach. A prototype instrument has been validated through the production of a real setup. We have shown that it is able to detect trace amounts of blood diluted in water (as small as 1:100) on fabric as well as differentiate blood from common false positives of other blood detection methods (i.e., luminol).
The second imaging system was designed for the differentiation of phytoplankton species in the ocean. Multivariate optical computing (MOC) was applied to the fluorescence excitation spectra of individual phytoplankton cells to design multivariate optical elements (MOEs). MOEs are filters fabricated to mimic linear discriminants analysis (LDA) results based on plankton spectroscopy. The imaging system uses these MOEs housed in a filter wheel to produce "streak" images of phytoplankton as they flow past a CCD camera, with each streak having the appearance of a barcode whose intensities are related to scores of the plankton spectra on linear discriminant functions. Algorithms for this system have been designed to automatically detect and analyze the "barcode-like" patterns as well as sizes/shapes of the plankton streaks and assign a class to the plankton where possible. While the design is still in its infancy, preliminary studies mimic predicted LDA results. With further instrumental and analytical improvements, we believe we will be able to classify phytoplankton species with LDA.
Pearl, M. R.(2011). Rapid Classification of Imaged Objects Using Molecular Factor and Multivariate Optical Computing. (Doctoral dissertation). Retrieved from https://scholarcommons.sc.edu/etd/718