392
H. J. Byrne et al.
each component are described by the corresponding loadings plot. unlike KmCA
and FCm, PCA identifies quite accurately the boundaries between each feature.
however, the images generated suffer from inferior contrast and in some instances
interpretation may be difficult as biochemical features may be spread across different loadings.
vCA is another method which has been used for similar analytical purposes. In
brief, vCA computes a linear combination of supposed pure component spectra
which are termed endmember spectra. As described in miljkovic et al., the endmember spectra are acquired under the assumption that the most extreme data
points in the dataset are representative of pure component spectra [173]. however,
it has been pointed out that the endmembers generated are not truly representative
of the pure component they describe in the data set and can often contain a mixture
of biochemical constituents i.e. dNA and proteins. While this is representative of
the true nature of nucleic acids in-situ, it could lead to inaccuracies in interpretation.
A more recent study introduced the technique of spectral cross correlation to
screen spectral maps for the spatial distribution of specific spectral signatures, including those of polystyrene nanoparticles, RNA and lipids, and argued that it was
more effective than Classical least squared analysis, both supervised and unsupervised, which is commonly available on instrument software [174].
As in the case of multivariate techniques for disease diagnostics, there is a myriad of potential techniques to explore the spatial variations of the spectral profiles
with heterogeneous structures such as biological cells. A truly quantitative comparison of the different techniques is lacking, however.
13.7 Conclusions
Since their origins in the nineteenth and twentieth centuries, the techniques of Infrared absorption and Raman scattering spectroscopy have come a long way. technological developments have rendered them routine benchtop techniques for chemical
and process analysis. Although their applications for the analysis of complex mixtures have raised some challenges, an understanding of the fundamental physical
processes and interactions has helped the scientific community address these challenges. In biomedical applications, the chemical and physical inhomogeneities of
cells and tissues have presented many such challenges, not easily understood by
a multidisciplinary community. however, concerted efforts through collaborative
programmes such as the Eu FP6 Special Support Action dASIm [175] have helped
understand the origins of spectral artefacts which can mask or obscure the true
information contained within a spectral profile, such that such effects can be minimised by improved sample presentation and/or data preprocessing.
over the past decades, the potential of vibrational spectroscopy for biomedical
applications has been extensively explored and demonstrated. the prospect is of
a truly label free, non-invasive, non destructive, rapid, automated and objective
technology for disease diagnostics and beyond. disease diagnostics can be made on
H. J. Byrne et al.
each component are described by the corresponding loadings plot. unlike KmCA
and FCm, PCA identifies quite accurately the boundaries between each feature.
however, the images generated suffer from inferior contrast and in some instances
interpretation may be difficult as biochemical features may be spread across different loadings.
vCA is another method which has been used for similar analytical purposes. In
brief, vCA computes a linear combination of supposed pure component spectra
which are termed endmember spectra. As described in miljkovic et al., the endmember spectra are acquired under the assumption that the most extreme data
points in the dataset are representative of pure component spectra [173]. however,
it has been pointed out that the endmembers generated are not truly representative
of the pure component they describe in the data set and can often contain a mixture
of biochemical constituents i.e. dNA and proteins. While this is representative of
the true nature of nucleic acids in-situ, it could lead to inaccuracies in interpretation.
A more recent study introduced the technique of spectral cross correlation to
screen spectral maps for the spatial distribution of specific spectral signatures, including those of polystyrene nanoparticles, RNA and lipids, and argued that it was
more effective than Classical least squared analysis, both supervised and unsupervised, which is commonly available on instrument software [174].
As in the case of multivariate techniques for disease diagnostics, there is a myriad of potential techniques to explore the spatial variations of the spectral profiles
with heterogeneous structures such as biological cells. A truly quantitative comparison of the different techniques is lacking, however.
13.7 Conclusions
Since their origins in the nineteenth and twentieth centuries, the techniques of Infrared absorption and Raman scattering spectroscopy have come a long way. technological developments have rendered them routine benchtop techniques for chemical
and process analysis. Although their applications for the analysis of complex mixtures have raised some challenges, an understanding of the fundamental physical
processes and interactions has helped the scientific community address these challenges. In biomedical applications, the chemical and physical inhomogeneities of
cells and tissues have presented many such challenges, not easily understood by
a multidisciplinary community. however, concerted efforts through collaborative
programmes such as the Eu FP6 Special Support Action dASIm [175] have helped
understand the origins of spectral artefacts which can mask or obscure the true
information contained within a spectral profile, such that such effects can be minimised by improved sample presentation and/or data preprocessing.
over the past decades, the potential of vibrational spectroscopy for biomedical
applications has been extensively explored and demonstrated. the prospect is of
a truly label free, non-invasive, non destructive, rapid, automated and objective
technology for disease diagnostics and beyond. disease diagnostics can be made on