Page 187 - Vibrational Spectroscopic Imaging for Biomedical Applications

W idefield Raman Imaging of Cells and T issues 163

45
imaging to determine the viability of mammalian cells. From Raman
image data, they created a classification model to differentiate
between normal and apoptotic cells in a prostate cancer cell line.
Stewart and coworkers demonstrated widefield Raman imaging of
47
the waterborne pathogen Cryptosporidium parvum. They demonstrated
the capability to identify, from a Raman library of microorganism spectra,
a single C. parvum spore. They also discussed the potential for quan-
tification of samples by Raman imaging and demonstrated that
Raman signal increased with increasing C. parvum spores. Escoriza et al.
also illustrated quantification of waterborne bacteria using wide-
55
field Raman imaging. They observed that in higher concentrations, the
Raman signal of bacteria directly correlates with the number of cells pres-
ent in drops of water on sample slides. At very low concentrations (e.g., in
drinking water), a filtration step was necessary to concentrate the bacteria.
Escoriza’s group found that quantification results showed good correla-
tion between Raman measurements and more traditional methods such as
turbidity, plate counts, and dry weight.
Kalasinsky and coworkers reported on a comprehensive study of
widefield Raman imaging of a suite of pathogens with a view to detect-
48
ing and identifying them even within a complex background. They
noted that for complex, environmental samples, widefield Raman
imaging was their preferred method for Raman sensing over others
including SERS, UV-excited resonance Raman spectroscopy, and non-
linear Raman spectroscopy. In similar work, Tripathi et al. evaluated
widefield Raman imaging technology for waterborne pathogen detec-
49
tion. Taking into account key experimental and background interfer-
ences such as laser-induced photodamage threshold, composition of
water matrix and organism aging, the group reported the technique’s
capability of discriminate threat simulants down to the species level.
The information contained in widefield Raman images is immense.
In a typical Raman image there may be tens of thousands to hundreds
of thousands of pixels, each pixel of which contains spectral informa-
tion. In addition, because of the heterogeneous nature of their composi-
tion, biological materials often have very similar Raman spectra. As a
result, differences between biological Raman spectra are usually
difficult to see with the eye. In almost all cases, the large datasets gen-
erated by Raman imaging require the use of multivariate data analysis
methodologies in order to process and interpret the results.
Like all analytical methodology that requires the use of mechanical
equipment, several factors influence outputted data—both in quantity
and content. Equipment that contains high electronic noise results in
Raman data with lower signal-to-noise ratio (SNR). Artifacts that may
be produced by instrumentation may also cause misinterpretation of
data. For this reason, most imaging technology requires regular
implementation of calibration and preprocessing methodologies in
order to minimize factors such as instrument response and dark
current. These will be discussed in the following sections.