Data Processing Strategy of Raman Chemical Maps - Data Characteristics and Behavior

摘要

Raman maps, when acquired and processed successfully, produce Raman chemical images, which provide detailed information on the spatial distribution and morphology of individual chemical species in samples. The advantages of Raman chemical images are most significant when the sample is chemically and structurally complicated. In pharmaceutical applications, these Raman chemical images can be used to understand and develop drug formulations, drug delivery mechanisms, and drug-cellular interactions. Studies using Raman hyperspectral imaging - the term that encompasses the entire procedure from data measurement to processing and interpretation - is increasing and gaining a wider acceptance due to recent improvements in Raman instrumentation and software. Since Raman maps are a collection of numerous Raman spectra of different chemical species, within a single data set, spectral characteristics such as the scattering strength, fluorescence level, and baselines vary a great deal. To acquire and process a Raman map successfully, this heterogeneity must be taken into the consideration. This paper will show the impact of signal-to-noise ratio (S/N) on data processing strategies and their results. It will be demonstrated that the S/N of original data is critical for good classification and scientifically meaningful results regardless of the processing strategies.