Raman Spectroscopy for Early Cancer Detection, Diagnosis and Elucidation of Disease-Specific Biochemical Changes - Emerging Raman Applications and Techniques in Biomedical and Pharmaceutical Fields - page 320

Fig. 13.4. ( Top ) Mean PC scores for each of the three pathology groups. ( Middle )
The ANOVA F -value is plotted to indicate the PCs with the most significant differ-
ences between the groups. ( Bottom ) The weight of each PC used to construct the
LD functions
the combined contribution of LD weights and PC scores that make up the
mean scores for each group in the model. It can be seen that the most sig-
nificant contributions come from those PCs with high F -values in Fig. 13.4
(middle).
Perturbation of Spectra to Evaluate Effect on Predictive Power
of Model
The baseline model outlined above was used to evaluate the effect of sim-
ulated perturbations of the spectra on the predictive ability of the model.
These perturbations were selected to represent spectral changes that would
be caused by likely system variations, such as calibration changes, misalign-
ments, optical aberrations and temperature change, in addition to changes in
spectrometer design. For simplicity only a single perturbation has been applied
for each test; however, in reality it is likely that there will be a combination
of these instrumental effects. The pathology of the perturbed spectra have
been predicted by the baseline model and the effect investigated. Numerous
spectral model simulations have been performed. These include the follow-
ing perturbations: number of PCs, linear wave number x -shift, signal-to-noise