Page 206 - Vibrational Spectroscopic Imaging for Biomedical Applications

182 Cha pte r S i x

Near-infrared Raman spectroscopy was used to map single cells. 3
Thousands of Raman spectra were assembled into a spectral image
creating a Raman map of the cell under examination. Raman spectral
maps of a freeze-dried human embryonic lung epithelial fibroblast,
human osteogenic sarcoma cells in medium, and human astrocytoma
cells in medium were created. PCA of the Raman spectral maps of the
-1
different cells in the spectral region of 2800-3000 cm revealed cellu-
lar details like organelles, the membrane, and the nucleus within the
image of the cell. The spectral features of the subcellular components
were more pronounced for cells in medium than for the freeze-dried
cells. It was concluded that the intensity of the spectra strongly
depend on the state of hydration of the cell because the conformation
of molecules within the cell change upon dehydration.
Similarly, Eliasson et al. also used Raman spectral mapping to
identify intracellular components of single living cells, specifically
70
human lymphocytes. In this case, SERS using gold colloids was
employed to image an introduced analyte, rhodamine 6G, together
with other intracellular components. An inverted confocal Raman
microscope was used to map the cells in three dimensions over the
-1
spectral range of 450 to 1700 cm . Several cells were mapped and the
images were analyzed with PCA in order to extract the signal of rhoda-
mine 6G from the image spectra. A score plot of PC1 and PC2 showed
that the spectra could be divided into three separate classes: (1) spectra
with spectral contributions from rhodamine 6G and native cellular
components; (2) spectra with only native cellular component contribu-
tions; and (3) spectra without any spectral band information. There-
fore, SERS in combination with PCA, makes it possible to measure
Raman-active analytes taken up by cells.
PCA of Raman dispersive spectra was used to discriminate normal,
inflammatory, premalignant, and malignant conditions in oral tissue. 71
PCA of the spectra from the four different types of tissue showed
differentiation between the normal and pathological states; how-
ever, there was poor discrimination among the three pathological
states. It is likely that the strong presence of proteins in the pathological
tissues versus the lipid content in normal tissue accounts for some
of the difference. To diagnose correctly the pathological conditions,
PCA combined with match/mismatch criteria using three parameters
(MD, sum of squared spectral residuals, and scores) was able to
discriminate between the different types of tissue. In a similar analytical
procedure, discrimination of normal, benign, and malignant breast
72
tissues by Raman spectroscopy was possible. Likewise, the same
process was used to discriminate normal and malignant stomach
73
mucosal tissue, normal and malignant mucosal tissues of the colon; 74
and normal and malignant ovarian tissue. 75
Shafer-Peltier et al. applied morphological modeling to Raman
images of human colonic carcinoma cells, human breast tissue, and
76
arterial tissue. Morphological modeling is a technique for analyzing