Picosecond Spectral Cars Imaging with Principal Component Analysis

Abstract

Imaging methods based on vibrational Raman contrast offer label-free chemical imaging of tissues. Video rate imaging acquisition times can be achieved by coherent Raman techniques, however, they do so only at a single vibrational frequency - limiting a more in-depth chemical analysis of the tissue. The most direct solution is to extend the measurement in the vibrational frequency domain, which has been manifested in the development of broadband coherent Raman methods. Unfortunately, the attributed acquisition time renders them to be impractical in examining large tissue areas. In this study, we demonstrate that picosecond spectral coherent anti-Stokes Raman scattering (CARS) imaging combined with principal component analysis (PCA) can be utilized to achieve fast high-resolution maps of lipids in tissues. Spectral CARS data sets were generated by imaging tissue sections at consecutive Raman shifts in the CH2 stretching vibrational range. By avoiding spectral redundancy, the acquisition time was shortened, whereas loss of spectral information was compensated by a much higher information density in the spatial domain via PCA. This resulted in sufficient spectral contrast, which was illustrated in our recent application in characterizing lipids distribution and composition of the meibomian glands in the context of dry eye disease. Spectral differences between CARS spectra obtained from specific regions were confirmed with a Raman microspectroscopic analysis. The resulting multi-component chemical maps revealed a lipid maturation process throughout the gland structure, which was not observed before and may provide important clues toward understanding meibomian gland dysfunction.