Biomedical Engineering Reference
In-Depth Information
Figure
12.2. Real-time capacitance probe trace showing the correlationbetweenoff-line viable
cell density measurements by hemacytometer (CEDEX). Note that the spikes in the viable cell
density trace correlate with feed days during a mammalian cell culture fed-batch process.
polarized, there is no further capacitance increase related to charge accumulation. The
increase in permittivity from high to low frequencies has a sigmoidal shape and is known
as the
b
-dispersion [22].
The NIR probe is one of the quintessential PAT tools in the process scientist and
engineer's toolbox. With the recent advances in fiber optics and the broad availability of
chemometrics software, it is now common practice to perform in situ NIR analysis by
submerging the probe into the reactor. It is based on the absorption of electromagnetic
energy in the region of 700-2500 nm, caused by the overtone and combination bands of
the fundamental bending and stretching vibrations as seen in the mid-IR region [23].
These NIR absorptions are generally 10-100 times weaker than the fundamental bands of
mid-IR [25], but this is in fact an advantage, as it enables the direct analysis of samples
without any sample preparation. Therefore, it is ideally suited for real-time measure-
ments, and although the spectra appear broad and overlapping, much information can be
elucidated from them by using sophisticated chemometric techniques [24]. From first
principles, therefore, NIRS is perhaps more suited for application in situ than any
competing technology [56].
In situ NIR has been implemented successfully in both microbial fermentations and
cell cultures. As with most of these spectroscopic techniques, chemometrics is critical to
building predictive, quantitative models. Parameters fall into two main categories:
pretreatments and diagnostics. Pretreatment of data usually takes the form of derivati-
zation, smoothing, and path-length treatments. Diagnostics include the tools necessary
for demonstrating or quantifying the success of a calibration. Typical calibrations involve
an “add back” model, as described by Yeung et al. [4], where the analyte(s) of interest is
(are) added to the matrix in which they would be found. Validation runs use “unknown”
samples and spiking of unknowns with knowns and quantifying them against the
calibration model and correlating the values to an orthogonal off-line analysis.
For fermentation processes, NIR has been extensively used in antibiotic production,
recovery of a yeast alcohol dehydrogenase (ADH) from an unclarified yeast cell
