Local rank analysis for exploratory spectroscopic image analysis. Fixed Size Image Window-Evolving Factor Analysis

Many pharmaceutical and consumer products are composed of emulsion-based formulations that act as delivery systems for active ingredients or other benefit agents. The efficacy and performance of these products often rely on how the actives interact and are stabilized by the base emulsion formulation. It is therefore of great significance to the development and processing of these formulations to be able to accurately characterize the structure and chemical distribution of components in these systems. Recent advances in spectroscopic imaging technology have played an important role in bringing these methods into acceptance and perhaps even prominence in both the consumer product and pharmaceutical industries as standard methods of analyzing these kinds of chemical systems. Concurrently, this has kindled the interest in methodology capable of analyzing the large data sets generated by these multivariate imaging experiments. Efficient exploratory analysis of spectroscopic images is therefore crucial, the desired goal being to resolve the chemical information in the data into its pure component parts. An approximate description of the compositional complexity of images is not yet available through the usual exploratory procedures. Methods used for this purpose should be local and be able to take into account the complex spatial structure of the image. Fixed Size Moving Window-Evolving Factor Analysis has been a powerful approach to locally define the complexity of a process through the subsequent PCA analyses of data subsets built by moving a fixed size window along a unique process direction (e.g., time, pH). Spectroscopic images have two or three spatial directions (in surface or multilayer images, respectively). Algorithms based on local data analysis should be adapted to preserve this higher dimensionality in order to provide a representative description of the image complexity. Fixed Size Image Window-Evolving Factor Analysis (FSIW-EFA) modifies the parent local rank algorithm to achieve this purpose. (c) 2005 Elsevier B.V. All rights reserved.