Page 304 - Vibrational Spectroscopic Imaging for Biomedical Applications

278 Cha pte r Ni ne

techniques automated analysis applications can be built which for
instance can discriminate between tumor and non tumor cells.

9.4.2 Quantification Techniques
Quantification techniques are mostly supervised techniques;
although there are also unsupervised techniques, like independent
component analysis (ICA) and multivariate curve resolution (MCR),
they are difficult to apply for generating Raman images because of the
complexity of the spectra and the generally low signal-to-noise ratio of
the spectra. Supervised quantification methods like least squares (LS)
or partial least squares (PLS) can extract direct concentration informa-
tion from the spectra. LS fitting is often used to determine and map the
34
biochemical composition of cells. The use of PLS to generate Raman
images is not widespread but has been applied to predict the concen-
tration of necrotic tissue in a section of brain tumor tissue. 35

9.5 Raman Mapping and Imaging in Bioscience
In this paragraph a number of applications are reviewed that show
the current state of art of (nonresonant Stokes shift) Raman imaging.
This overview is in no way complete—the applications were chosen
to illustrate the broad range of possible applications and to illustrate
different methods to extract useful data from the images. The appli-
cations are divided into two sections: Secs. 9.5.1 and 9.5.2.

9.5.1 Single Cells
Raman spectroscopy provides a way of studying cellular processes
on a molecular level and in vivo, without having to change the cel-
lular interior and environment. Raman imaging adds spatial informa-
tion to the chemical information, so that the molecular concentrations
and changes therein can be localized, providing unique insight in the
biochemical functioning of cells. 5,6,9,36–43
For intracellular Raman imaging a confocal detection setup is
used with a high-spatial resolution of 1 μm or less and a laser
power in the order of 10 to 50 mW (although higher powers are
used in the NIR without measurable damage). Because of the high-
spatial resolution, spectra obtained from within a cell resemble
pure compound spectra more than spectra obtained from whole
tissues. Often the spectra are a simple mixture of a limited number
of compounds.
To determine the biochemical composition of the cell, a “basic
set” of single molecule Raman spectra is often used, consisting of the
spectra of DNA, RNA, and some sugars, proteins, and lipids. An
example of such a spectral basis set is given in next figures. Figure 9.5
−1
shows part of the so called “fingerprint region” (from 400 to 1800 cm ,
a very molecule specific spectral region) and Fig. 9.6 shows the high