Infrared and Raman Spectroscopy and Imaging of Breast Cancer - Vibrational Spectroscopy for Tissue Analysis - page 144

The samples of GI, GII, GIII, and HNG types were chosen for analysis.
In each case, the steps were as follows:

1. Average in 2 × 2 squares to reduce size and reduce noise

2. Second derivative via Savitzky-Golay filter with mild smoothing

3. Selection of wavelength range ~1000-2000 cm −1

4. Principal component decomposition using singular value decompo-
sition (SVD)

5. Selection of clusters from score plots

6. Spatial plot of clusters and mean cluster spectra

IDC Grade III

This section will help in understanding how calculations are carried
out and the way these help in performing statistical analysis. The plot in
Figure 5.28 of the singular values suggests that two PCs explain a large
percentage of the variance in the spectra. The PC1/PC2 score plot suggests
two clusters and is presented in Figure 5.29. The clusters of these two PCs
are more obvious if the data are plotted as a histogram instead, as described
in Figure 5.30.
It is interesting to see whether these clusters correspond to different spa-
tial regions of the sample. To test this, points in rectangular regions around
the samples were selected and their spatial ( x , y ) coordinates calculated and
plotted (Figure 5.31).
The blue asterisks correspond to the upper cluster in step 5 above, the
green triangles to the lower one, and the red circles to the one on the far
right. It can be seen that the blue points are generally on the right side of the
image (large x ) while the red are on the left.

× 10 -3

6

4

2

0

0 2 4 6 8 10

Factor Index

Figure 5.28
Two PCs explaining a large percentage of the variance in the spectra.