Fig. 2 provides a visual depiction of the differences
between the relaxation decays, when the blue line is longer
than the orange one. It implies differences in sample composition. Higher values of T2 are related with less content of fat or
more water content, and the opposite is true.
It is standard practice to calculate only T2 values to
determine the fat content, but this information is insufficient
for tenue differences between the signals.
With chemometrics tools such as PLS, the relation for each
signal is computed based on the reference values obtained
from standard methods, and a robust model can be developed.
Wet analysis cannot be replaced or fully eliminated, as the
multivariate model must occasionally be validated. However,
the number of wet analyses requiring solvent extractions may
be drastically reduced.
Our research showed that the determination of fat content
in emulsion-based foods through the packets used for mayonnaise and dressing salads was possible with a TD-NMR instrument and multivariate models [ 4].
The permanent Halbach magnet of 0.23 T ( 9 MHz for 1H)
and a 50-cm-long analytical magnet allowed us to measure the
entire package contents [ 4]. The reference values from Bligh
and Dyerś method for fat content were the dependent variable
of matrix y, as depicted in Fig. 2. The relaxation decays shown
in this figure were comprised with the samples on rows and
the instrumental response on columns (time).
The main advantage of the multivariate model over standard methods of wet chemical analysis was better repeatability
in determining fat content in emulsions, as repeatability is a key
limitation in lipid extraction due to variations in droplet size.
Our solution for measuring fat in foods that are intact in their
packaging was demonstrated to be feasible. Once each
model was developed, few chemical wet analyses need to be
performed. The models described were highly correlated
(r > 0.9), and correlated between the reference values and the
values predicted by PLS.
Fabíola Manhas Verbi Pereira is an assistant professor in the
Department of Analytical Chemistry, Institute of Chemistry,
Unesp, Araraquara SP, Brazil. He can be contacted at
FIG. 2. TD-NMR signals evaluated with chemometric tools