Biomedical Engineering Reference
In-Depth Information
This barrier of about 0.2 mM in 100 s can be overcome in limited ways.
If the target analyte has a substantially larger cross section than those men-
tioned above, it will be correspondingly easier to detect with more accuracy.
Occasionally, the molecule simply happens to have an unusually large cross
section for its molecular weight. In this case, this is a true advantage over
glucose and the other analytes mentioned above. Larger molecules with many
subunits, such as proteins and DNA, tend to have correspondingly larger
cross sections per molecule as well. In such cases, while the minimum error
in millimolar can drop well below 0.2 mM, the error in milligram per deciliter
remains relatively unchanged. In terms of parameters the experimenter can
control, most simplistically, one can increase the laser power on the sample,
as long as damage is not done. More elegantly, one can revise the collection
geometry to increase eciency. In either case, the benefit in sensitivity only
grows as the square root of the increase in signal, as the Raman peak and the
spectral background will grow in tandem.
If an application calls for the quantification of a single analyte rather
than a broad panel, then another possibility can be resonance Raman spec-
troscopy. In resonance Raman, the laser is tuned to an electronic absorption
band of the target species, greatly increasing the species' Raman scattering
cross section while leaving other chemicals largely unaffected. Typically, this
is performed with UV lasers, although unusual level structure in carotenoids
and hemoglobin make resonance Raman practical in the visible for these im-
portant molecules. To date, no major applications of resonance Raman to
biofluid analysis have been reported to the author's knowledge.
16.5.2 Blood Serum
Representative spectra of blood serum samples, processed as described above
to remove broad fluorescent background via a polynomial, are shown in
Fig. 16.6, along with similar samples of urine (data from Qi and Berger [5]).
Serum spectra contain many more visible Raman peaks than urine, whose
spectrum is dominated by a single urea peak.
The first major reports of analyte quantification in blood serum sam-
ples came out in 1999. (As noted above, most conventional blood analysis
is performed upon blood serum, not whole blood.) Berger et al. [1] and Qu
et al. [18] used similar near-infrared laser wavelengths (830 and 785 nm), pow-
ers (250 and 300 mW), spectral integration times (1-5 min), total number of
patients' samples (60-70), and multivariate calibration methods (PLS). Both
experiments analyzed samples in cuvettes using non-equivalent excitation and
collection volumes, one employing a Cassegrain collecting lens to view through
the cuvette wall, the other 90
◦
scattering from a collimated beam passing up
through the cuvette's floor. In both cases, significant correlations between
reference concentrations and PLS predictions were obtained for multiple an-
alytes in blood serum, including total protein, albumin, triglyceride, choles-
terol, blood urea nitrogen (BUN), and glucose. In Qu's study, spectra were
Search WWH ::
Custom Search