A comparison of multivariate curve resolution with endmember extraction methods in hyperspectral Raman imaging

Area of Science:

• Analytical Chemistry
• Spectroscopy
• Materials Science

Background:

• Hyperspectral imaging (HSI) generates large datasets, requiring pre- and post-processing for chemical information extraction.
• Endmember Extraction (EX) and Multivariate Curve Resolution (MCR) are common techniques for deriving pure spectral signatures and abundance maps from HSI data.

Purpose of the Study:

• To systematically compare EX and MCR methods for pure variable extraction in Raman mapping.
• To provide guidelines for selecting the appropriate pure variable extraction technique.

Main Methods:

• Numerical simulations and experimental Raman mapping on pharmaceutical powders and layered plastic foils.
• Comparison of Multivariate Curve Resolution (MCR) and Vertex Component Analysis (VCA) as a representative EX algorithm.

Main Results:

• Both MCR and EX (VCA) perform well when datasets contain pure pixels for each component.
• MCR demonstrated superior capability in extracting pure component spectra when pure pixels were absent in the dataset.

Conclusions:

• The presence or absence of pure pixels significantly impacts the performance of EX and MCR algorithms.
• MCR is the recommended method for pure spectral signature and abundance map retrieval in hyperspectral Raman imaging, especially when dealing with complex samples lacking pure pixels.