358
H. J. Byrne et al.
analysis techniques, Raman spectroscopic analysis of cervical biopsies and cytological samples can provide high specificity and sensitivity for the detection of cervical
cancer and early pre-malignant states. the following provides an illustration of such
applications combining Raman analysis and Principal Component Analysis/Linear
discriminate Analysis [34].
Formalin fixed paraffin preserved (FFPP) cervical tissue samples were obtained
from the National Maternity Hospital, Holles St., Dublin. Two parallel 10 μm FFPP
sections were cut from each block using a microtome, mounted on glass slides and
dried. Samples were dewaxed in-house prior to investigation by immersion in baths
of Xylene (Bdh), Ethanol Absolut (merck) and Industrial methylated Spirits 95 %
(Lennox) and air-dried. Cervical tissue was dewaxed using the same procedure outlined above, but xylene was substituted with hexane, due to the improved level of dewaxing [49]. one section from each sample (the reference section) was stained with
haematoxylin and Eosin and the other kept unstained for spectroscopic examination.
FFPP cervical tissue sections were characterised by consultant pathologist dr.
Kelehan P, National maternity hospital, holles St., dublin, and the samples consisted of 20 normal and 20 invasive carcinoma sections from 40 individuals. of the
20 carcinoma samples, 10 samples were identified as having various grades of CIN,
which were also marked for examination.
An Instruments S. A. Labram Raman spectroscopic confocal microscope was
used, with an Argon Ion laser operating at an excitation wavelength of 514.5 nm. A
x50 objective lens was used. the laser power at the sample was measured and found
to be 6.5 ± 0.05 mW, focused to a spot size of ≈ 2 μm at the tissue surface. Raman
spectra of the various tissues were accumulated for 150 s. A total of 10 spectra were
recorded from different spots on each sample.
Selected spectra were baseline corrected, dark current subtracted and were
smoothed using a 10 point moving average before being normalised to the spectral
maximum, from 0 to 1. Basic matrix manipulations and data reduction was carried out in microsoft Excel Professional 2003 (v. 11.0), before being exported into
minitab to perform Principal Component Analysis (PCA) and Linear discriminant
Analysis (LdA). minitab Release 14.1 Statistical Software Analysis Programme
was used to produce PCA scores and LdA plots, as well as to carry out leave-oneout cross validation.
Figure 13.1 shows the layer structure of normal cervical tissue in an unstained
FFPP tissue section together with the Raman spectra recorded from basal cells, epithelial cells and connective tissue. the spectra of the three different regions do have
a degree of similarity, but notably there are many differentiating spectral features.
the spectra of basal cells show strong bands at 724, 779 and 1,578 cm
−1
which are
characteristic of nucleic acids. the morphology of basal cells consists of a single
line of tightly packed cells, with large nuclei in relation to the compacted surrounding cytoplasm. In addition, these cells are constantly dividing, providing cells to the
parabasal layer. For both of these reasons, a high concentration of dNA would be
expected in the basal cells. Spectra of epithelial cells have characteristic glycogen
bands at 482, 849, 938, 1,082 and 1,336 cm
−1
. Collagen contributions can be clearly
seen in the spectra of connective tissue at 850, 940 and 1,245 cm
−1
.
H. J. Byrne et al.
analysis techniques, Raman spectroscopic analysis of cervical biopsies and cytological samples can provide high specificity and sensitivity for the detection of cervical
cancer and early pre-malignant states. the following provides an illustration of such
applications combining Raman analysis and Principal Component Analysis/Linear
discriminate Analysis [34].
Formalin fixed paraffin preserved (FFPP) cervical tissue samples were obtained
from the National Maternity Hospital, Holles St., Dublin. Two parallel 10 μm FFPP
sections were cut from each block using a microtome, mounted on glass slides and
dried. Samples were dewaxed in-house prior to investigation by immersion in baths
of Xylene (Bdh), Ethanol Absolut (merck) and Industrial methylated Spirits 95 %
(Lennox) and air-dried. Cervical tissue was dewaxed using the same procedure outlined above, but xylene was substituted with hexane, due to the improved level of dewaxing [49]. one section from each sample (the reference section) was stained with
haematoxylin and Eosin and the other kept unstained for spectroscopic examination.
FFPP cervical tissue sections were characterised by consultant pathologist dr.
Kelehan P, National maternity hospital, holles St., dublin, and the samples consisted of 20 normal and 20 invasive carcinoma sections from 40 individuals. of the
20 carcinoma samples, 10 samples were identified as having various grades of CIN,
which were also marked for examination.
An Instruments S. A. Labram Raman spectroscopic confocal microscope was
used, with an Argon Ion laser operating at an excitation wavelength of 514.5 nm. A
x50 objective lens was used. the laser power at the sample was measured and found
to be 6.5 ± 0.05 mW, focused to a spot size of ≈ 2 μm at the tissue surface. Raman
spectra of the various tissues were accumulated for 150 s. A total of 10 spectra were
recorded from different spots on each sample.
Selected spectra were baseline corrected, dark current subtracted and were
smoothed using a 10 point moving average before being normalised to the spectral
maximum, from 0 to 1. Basic matrix manipulations and data reduction was carried out in microsoft Excel Professional 2003 (v. 11.0), before being exported into
minitab to perform Principal Component Analysis (PCA) and Linear discriminant
Analysis (LdA). minitab Release 14.1 Statistical Software Analysis Programme
was used to produce PCA scores and LdA plots, as well as to carry out leave-oneout cross validation.
Figure 13.1 shows the layer structure of normal cervical tissue in an unstained
FFPP tissue section together with the Raman spectra recorded from basal cells, epithelial cells and connective tissue. the spectra of the three different regions do have
a degree of similarity, but notably there are many differentiating spectral features.
the spectra of basal cells show strong bands at 724, 779 and 1,578 cm
−1
which are
characteristic of nucleic acids. the morphology of basal cells consists of a single
line of tightly packed cells, with large nuclei in relation to the compacted surrounding cytoplasm. In addition, these cells are constantly dividing, providing cells to the
parabasal layer. For both of these reasons, a high concentration of dNA would be
expected in the basal cells. Spectra of epithelial cells have characteristic glycogen
bands at 482, 849, 938, 1,082 and 1,336 cm
−1
. Collagen contributions can be clearly
seen in the spectra of connective tissue at 850, 940 and 1,245 cm
−1
.