Agriculture Reference
In-Depth Information
to further exploitation is near-infrared spectros-
copy. Some of the current and potential uses of
NIRS have been discussed in earlier sections
dealing with specifi c components or functional
properties. A relatively recent review by Osborne
(2006) on the use of NIRS in cereal breeding
programs detailed the virtues of whole-grain
NIRS testing, its speed, its nondestructive nature,
and its ability to concurrently estimate multiple
analytes. The review cites studies that deal spe-
cifi cally with the problem of small sample size in
the early generations of inbreeding, speed and the
use of newer and faster diode-array instruments,
and other decisive factors related to the choice of
NIRS instrumentation.
Apart from the basic and necessary analyses of
protein, moisture, and hardness, NIRS calibra-
tions have been developed for a number of other
components. High-molecular-weight glutenin
content, as a percentage of total protein, was
predicted within an acceptable level of accuracy
(
r
2
= 0.88 and standard error of cross validation,
0.77; Bhandari et al., 2000). Elements of starch
composition including amylose-to-amylopectin
ratio (Wesley et al., 2003) and starch or fl our
swelling characteristics (Crosbie et al., 2007) have
also been predicted with an accuracy claimed to
be suffi cient for screening early-generation breed-
ing materials for their potential starch properties.
In another study Delwiche et al. (2006) indicated
that NIRS was
not
able to distinguish between
wild-type and single null partial-waxy lines in
durum wheat, although fully waxy genotypes
were correctly categorized.
The accuracy of NIRS was also assessed for
predicting numerous grain, milling, fl our, dough,
and breadmaking parameters from 198 US hard
red winter and spring wheat and fl our samples
(Dowell et al., 2006b). For factors other than
protein and moisture, NIRS only predicted fl our
yellowness with an accuracy claimed to be appli-
cable in process control (
r
2
> 0.97). Many other
parameters were predicted with accuracies
claimed to be suitable for coarse screening.
However, of these parameters the only ones pre-
dicted with
r
2
> 0.70 (once correlations with
protein were removed) were test weight, fl our
color, free lipids, fl our particle size, and the per-
centage of dark vitreous kernels. As many factors
including loaf volume and mixograph, farino-
graph, and alveograph parameters were correlated
with protein content, this relationship was con-
sidered to be the main infl uence on the NIRS
predictions.
It is considered by some that the future possi-
bilities for NIRS do not necessarily lie in the
detection of an increasing number of component
concentrations and/or functional traits but in
exploitation of the technique for real-time testing
(Scotter 2000), such as on loading conveyors, or
sorting of single kernels for chosen traits (Dowell
et al., 2006a). Delwiche (2005) summarized the
emerging uses and new instrumentation available
in NIRS. The issue of the correlation of many
quality attributes with protein content was again
highlighted as well as possible solutions to the
challenge of fi nding robust methods of predicting
“protein quality.” This article also suggested that
more physically robust NIRS instruments will
allow their deployment in harsher environ-
ments such as in harvesters or fl our mills.
Other opportunities arise from the use of single-
kernel NIRS instrumentation for high-speed
sorting of seeds in many circumstances: from
segregation of hard and soft kernels in breeding
programs to commercial-scale separation and
removal of fusarium-infected kernels from clean
kernels to achieve acceptably low levels of
deoxynivalenol.
Further possibilities lie in the use of other
spectroscopic techniques, alone, or in tandem
with NIRS. Raman spectroscopy has been used
in conjunction with NIRS to improve NIRS-
based predictions (Anderssen et al., 2005).
Seabourn et al. (2005) reported the use of
Fourier-transform (FT) midinfrared spectros-
copy to monitor dough mixing in a mixograph.
The ratio of α-helix to β-sheet structures as
determined in the mid-IR range in the fi rst
minute of mixing was highly correlated with
mixograph peak time. The authors considered
that the FT midinfrared technique could predict
mixograph peak time in a truncated mixing
process (<1 min) based on changes in the second-
