Page 252 - Vibrational Spectroscopic Imaging for Biomedical Applications

228 Cha pte r Ei g h t

The accuracy of biopsies can be optimized by coupling with spectro-
scopic techniques. Spectroscopic diagnosis of tissue is also called optical
biopsy. The capability to perform optical biopsies is highly desired in
medical procedures, e.g., to guide microsurgery, to identify suspicious
tissue more sensitive and specific, and to collect tissue biopsies more
accurate. Unnecessary tissue excisions and the biopsy related costs and
risks could be avoided. Furthermore, these developments would enable
in the future to monitor the effects of therapies under real-time condi-
tions and to optimize the treatment using this information.

8.2.1 Mouse Brains
An important step in the development of new diagnostic methods is
their application in animal models. A mouse model was used to test the
detection of brain tumors by fiber optic Raman spectroscopic imaging. 8
Injection of tumor cells in the carotid artery induces tumors in mouse
brains. Primary brain tumors that originate from cells inside the brain
(see Sec. 8.2.2) as well as secondary brain tumors that are metastases
from tumors outside the brain can be induced using selected tumor cells.
After injection the tumor develops in one hemisphere whereas the other
hemisphere remains tumor free and serves as an internal control. This
constitutes an advantage compared with human brain tumor specimens
because normal human brain tissue is preserved as much as possible
during neurosurgery and consequently, no control specimen is usually
available. Figure 8.1a shows the experimental setup to study nondried
mouse brains with and without tumors. Before data acquisition, two
tissue sections were prepared from the mouse brains. The first tissue
section was stained by hematoxylin and eosin (H&E) for histopathologi-
cal inspection. The second tissue section was transferred onto a calcium-
fluoride slide for acquisition of FTIR images. The remaining tissue was
covered by a window to prevent the surface from drying. Such a subse-
quent preparation enables to compare the Raman, FTIR, and histopatho-
logical results. A Raman spectrometer (Kaiser Optical System, United
States) was coupled to a commercial fiber optic probe (Inphotonics,
United States) which focused the laser to a spot of approximately 60 μm
diameter. This setup is more flexible, more compact, and less expensive
than a microscope. The sample was placed on a motorized stage, a raster
was defined, and Raman spectra were collected from each raster point
with 8 seconds exposure time. The laser intensity of 100 mW at 785 nm
was found to be nondestructive. This is an important prerequisite to
apply the method to living animal or even patients in the future. The
obtained Raman image was segmented into groups of similar spectra
using k-means cluster analysis. Cluster analysis belongs to the unsu-
pervised algorithms by which objects are grouped into subsets, called
clusters, so that the differences between data within each cluster are
minimized and the differences between clusters are maximized accord-
ing to some defined distance measure. The k-means algorithm assigns
each point to one of k clusters whose center is nearest. The results of the