Infrared and Raman Spectroscopy and Imaging of Breast Cancer - Vibrational Spectroscopy for Tissue Analysis - page 136

First, the spectra were measured in reflectance mode, whereas a number
of other researchers have used either transmission or reflection-absorbance
(reflectance of a thin film mounted on a mirror), which is essentially equiva-
lent to transmission. In theory, the reflectance spectrum should contain the
same information as the transmission spectrum. However, reflectance may
be more prone to artefacts due to scattering. More importantly, it is hard to
compare reflectance spectra to transmission spectra because the peaks are
shifted, and not by a constant amount. Using reflectance instead of trans-
mission is justified, however, if there is a significant advantage in doing so;
for example, does it reduce sample preparation time or enable analysis of
important samples that cannot be sectioned thinly enough for transmission?
If so, however, ATR would probably give the same benefits and would also
be more comparable to transmission data.
Second, the data reported here was run blind, and did not have detailed
histological information. In most of the studies reported in the literature,
samples have been analysed by a pathologist using light-microscopy tech-
niques. This provides Y- block information such as which regions of the sample
correspond to which kinds of tissue, and greatly facilitates the development
of classification models. The only Y-block information available in this study
was the overall tumour grade of the sample. In principle, it was only possible
to compare the obtained spectra (e.g., the cluster centroids) to spectra from
the literature to try to identify the tissue, but this may be extremely difficult
with reflectance spectra.
Obviously, a method that does not require further histological analysis has
advantages in terms of data acquisition time. But the challenge is to make
effective use of the imaging aspect of the data.

analysis of the reflectance Spectral Data

Structure of the Data

Table 5.4 lists the number of samples of different types of breast cancer. In
total, there were 38 samples, including 7 DCIS samples and 30 IDC samples.
There is only one sample of healthy or normal tissue.

Table 5.4
Sample Numbers Inferred from the File Names

Tissue Number of Samples

Healthy 1

Ductal carcinoma in situ Low nuclear grade 2

Intermediate nuclear grade 1

High nuclear grade 4

Invasive ductal carcinoma Grade I 6

Grade II 14

Grade III 10