ISPBENDSARS

1
Integrated Sensing Processing Acoustic Resonance
Spectroscopy (ISP-ARS) for Rapid Tablet
Identification David J. Link, Thaddaeus Hannel
and Robert A. Lodder Department of Chemistry,
University of Kentucky, Lexington, KY 40506
Abstract Integrated sensing and processing
acoustic resonance spectroscopy (ISP-ARS) was
investigated to differentiate toll manufactured
tablets. In ISP-ARS, an ISP acoustic waveform
is created such that it comprises only the
distinguishing spectral details associated with
an analyte, enabling rapid acoustic analysis in
the detector. ISP-ARS is faster than
conventional FT-ARS acoustic methods because
post-detection data analysis is not needed. As a
process analytical technology (PAT), ISP-ARS
could have the ability to work in real-time to
eliminate mislabeling and contamination of
tablets on the production line and in clinical
trials, preventing recalls. Another commercial
application of ISP-ARS is in hospitals and
nursing homes. There are approximately 1.5
million preventable adverse drug events (ADEs)
each year in U.S. hospitals, costing an estimated
3.5 billion dollars1. ISP acoustic waveforms
could easily be downloaded from a database and
used in conjunction with a PDA (used for writing
e-prescriptions) for identification of
tablets. Objective Development of sensors for
the in-line detection of incorrect or
contaminated drugs to eliminate mass recalls
saving money and lives.

• Future Work
• Micro-ARS that utilizes ultrasonic frequencies
• Optimize ISP-ARS so that it is not so highly
  sensitive to external noise
• Inconsistent analyte placement due to a
  non-automated loading procedure
• Develop new chemometric approaches capable of
  dealing with high dimensional feature space for
  exploration with ISP-ARS
• Kernel principal component analysis
• Least-squares support vector machines
• Computational modeling of the instrumentation
  to determine the contribution of tablet density,
  viscosity, and other major constituents that
  affect such bulk properties.

Experimental Tablet Preparation. Tablets of
different over-the-counter pharmaceutical drugs
were obtained for scanning by the ARS. Tablets
included vitamin C (Spring Valley, 1000 mg),
vitamin B-12 (Spring Valley, 2000 mcg),
acetaminophen (Equate, 325 mg), aspirin (TopCare,
325 mg), ibuprofen (Equate, 200 mg) and
D-tagatose (Spherix Inc, 300 mg). The tablets
were scanned intact with no special preparation
and all data analysis were performed in Matlab
7.0 (Mathworks Inc).

Figure 3. Frequencies are selected for ISP
acoustic waveforms by comparing their frequency
spectrum to the PC loadings that separate the
tablets. PC loadings are higher where spectral
variability is greatest. A time domain acoustic
excitation waveform (AW) is created by
incorporating these regions of high spectral
variability. Separate acoustic waveforms must be
created for the positive and negative loading
data so the integrated detector response on a
waveform is not canceled.
Figure 1. A function generator is used to provide
broadband source to the transmitting piezo which
in turn excites the samples with white noise for
the initial FT-AR measurements. The sound
recorded at the receiving piezo is Fourier
transformed in Matlab to give the resulting full
spectra.
Figure 2. Once all samples are scanned in random
order and in triplicate, principal component
analysis (PCA) is performed on the FT spectra.
The first three PC scores depicted in the figure
above illustrates the separation of each tablet
from both the blank rod (control) and the other
tablets. The figure is depicted using ellipsoids
of one standard deviation of each dimension.
Introduction Many large pharmaceutical
manufacturers contract out their small-scale
manufacturing needs as a way of reducing cost or
meeting their production deadlines. As a result,
a contract manufacturer may make several kinds of
pills that are similar in appearance at almost
the same time, testing various dosages and
placebos for clinical trials. In addition, a
contract manufacturer may produce pills for
multiple outside firms. One way to reduce the
possibility that pills may inadvertently become
confused or contaminated is to employ a rapid and
nondestructive means of verifying tablet
identity. Such systems for identifying
contaminated or mislabeled products must be
strategically placed to prevent problems with
pills before they are shipped. Process
analytical technologies (PAT) on the production
line should have the ability to work in
real-time. Currently there are no fool-proof
processes to eliminate mislabeling or
contamination, and millions of pills can
sometimes be recalled. For example, in November
of 2006, 11 million bottles of contaminated
acetaminophen were recalled by the Perrigo
Company of Allegan, Michigan due to contamination
of the tablets with metal wire2. The FDA admits
that current good manufacturing processes (cGMP)
have reached their limits and better risk-based
scientific approaches are needed to insure
product safety3. PATs are designed to prevent
large recalls by detecting problems before they
occur.   ISP-ARS is
fast and non-destructive. Acoustic methods are
able to deeply penetrate many types of opaque
packaging, in contrast to near-infrared and other
optical methods. The ability to penetrate many
types of packaging can be a distinct advantage in
preparation of clinical trial lots, where drugs
and placebos must be blinded from users. As a
PAT, a series of ISP-ARS sensors could
potentially scan every pill produced by a
manufacturer, enabling the removal of only those
pills that did not meet quality standards. A
dynamic data-driven application system (DDDAS)
could control a manufacturers product line based
on measurements from a series of ISP-ARS sensors,
adjusting process conditions and ingredients in
real time based on actual process
measurements4-5.
Conclusion Integrated sensing processing
acoustic resonance spectroscopy has demonstrated
the ability to differentiate D-tagatose tablets
(an experimental toll-manufactured drug) from
different tablets including aspirin,
acetaminophen, vitamin C, vitamin B and
ibuprofen. With an experiment-specific ISP
waveform, the classification is far more rapid
than with conventional ARS. Simpler ISP
waveforms using fewer frequencies to represent
the factor loadings that separate the tablets may
outperform more complex waveforms using more
frequencies. By encoding waveforms on an MP3
player, ISP-ARS could become a method to quickly
identify different unlabeled tablets with a
similar appearance created in a
contract-manufacturing environment.

• References
• Institute of Medicine (IOM), PREVENTING
  MEDICATION ERRORS, 2006. http//www.iom.edu/CMS/3
  809/22526/35939.aspx
• FDA, http//www.fda.gov/bbs/topics/NEWS/2006/NEW01
  507.html, September 3, 2007
• Woodcock, Janet. US Food and Drug Administration.
  http//www.fda.gov/ohrms/dockets/ac/02/briefing/38
  69B1_08_woodcock/sld001.htm. September 12, 2007
• M. Parashar et al., Towards Dynamic Data-Driven
  Management of the Ruby Gulch Waste Repository,
  Lecture Notes in Computer Science, Springer
  Berlin / Heidelberg, 2006, 3993, 384 392
• NSF, January 2006 DDDAS Workshop Report,
  http//www.nsf.gov/cise/cns/dddas/2006_Workshop/in
  dex.jsp, November 4, 2007
• Credit
• This research was supported in part by
• The Kentucky Science and Engineering Foundation
  through KSEF 914-RDE-008
• The National Science Foundation through
  0405349
• The National Institutes of Health through
  N01AA 33003.
•  
• Acknowledgement
•  
• The authors would like to thank Spherix for the
  gift of the tagatose tablets. One of the
  authors, Robert Lodder, serves as president and a
  member of the board of directors of the company.

Figure 5. The positive and negative (AW1 and AW2)
voltages were combined using canonical variables
(CV). The three CVs are plotted above with
ellipsoid radii representing three standard
deviations in each direction. Using BEST
distances in a leave-one-out cross validation,
the tablets were classified and the results are
depicted in tables 1-3.
Figure 4. The ISP-Waveforms were used to excite
new tablets and characterize them via the voltage
at the receiving piezo. Voltage Pos 1 represents
the voltage received while exciting with the ISP
waveform created from the positive loadings (AW1
in figure 3) from PC 1. Similarly, Voltage Neg 1
represetns the voltage received while exciting
with the ISP waveform created from the negative
loadings (AW2 in figure 3) from PC 1. The
ellipsoids are plotted similar to figure 2 where
the radius is one standard deviation in each
direction.
10 Frequency ISP-Waveform Model
100 Frequency ISP-Waveform Model
1000 Frequency ISP-Waveform Model
Table 1. Method statistics for ISP-ARS utilizing
both the 10 frequencies with the greatest
variance according to the positive loads and also
the 10 frequencies with the greatest variance
according to the negative loads. Each tablet was
classified to any group within three standard
deviations in hyperspace.
Table 2. Method statistics for ISP-ARS utilizing
both the 100 frequencies with the greatest
variance according to the positive loads and also
the 100 frequencies with the greatest variance
according to the negative loads. Each tablet was
classified to any group within three standard
deviations in hyperspace.
Table 3. Method statistics for ISP-ARS utilizing
both the 1000 frequencies with the greatest
variance according to the positive loads and also
the 1000 frequencies with the greatest variance
according to the negative loads. Each tablet was
classified to any group within three standard
deviations in hyperspace.
Accuracy (TP TN) / (TP TN FP
FN) Precision TP / (TP FP) Recall TP / (TP
FN) TP True Positive TN True Negative FP
False Positive FN False Negative.