Date of Award

5-2023

Degree Type

Dissertation

Degree Name

Doctor of Philosophy in Clinical-Bioanalytical Chemistry

Department

Chemistry

First Advisor

Turner, John F., II

Second Advisor

Boyd, Warren C.

Third Advisor

Guo, Baochuan

Abstract

Poly-l-lactic acid has been of a great interest to the medical profession in recent years because of its biodegradability and biocompatibility. The biodegradation rate can be controlled by its crystallinity content. One method to modify its crystallinity content is by cold drawing. Raman spectroscopy is utilized to distinguish between the different crystalline content on the sample. However, a chemical imaging method is needed to characterize the polymer on the macro level. Chemical imaging using Raman spectroscopy is an important method of characterization that is non-invasive and non-destructive. The emphasis of my research has been to develop a method to characterize the crystalline content of poly-l-lactic acid using Raman chemical imaging and multivariate analysis to construct a robust image of the crystalline map. No Technique exists for this purpose.

Multivariate analysis has been beneficial for reducing the time spent on finding correlations in a vast amount of data. Principal component analysis has been one of the most common methods for reducing the data dimensions to those that are the most responsible for the observed variation with the data. However, the method is scale variant and blends qualitative and quantitative information in a way that can render misleading results. Classical least squares has also been used in chemical imaging, but it requires substantial training data. In this document, we introduce a modified version of reduction of spectral images (ROSI), a multivariate analysis method that heavily modifies principal component analysis to reduce the amount of data while still prioritizing the minority pixel population on the image data to retain more of the analytically meaningful characters of the data.

Preprocessing of the data is an important step to obtaining robust results and can involve multiple steps depending on the application. It often serves as an essential step for removing irrelevant information from the raw data. One important preprocessing step is background removal. Acquired Raman spectra often have contributions from other phenomena such as Rayleigh scattering and fluorescence, among others. It is common for the raw spectrum to have a sloped or undulating baseline that is not a part of the analytically relevant features. Thus, it must be removed.

Conventional baseline removal methods are not well suited to large fluctuations in the signal-to-noise level of the spectra or to large datasets that must be processed without human assistance. While there are many background removal algorithms, the production of artifacts or subtle changes the shape and position of the Raman bands are common disadvantage. Sliding kernels estimated Bezier background (SKEBB) is a novel background removal method, introduced here, that utilizes a sliding window Bezier function and 2nd order fits over small intervals to effectively remove the background while leaving the Raman signal intact. The method will be compared to some of the common baseline correction techniques to evaluate the effectiveness of SKEBB.

Download

Share

COinS