Page 46 - FoodFocusThailand No.246 October 2026

STRONG QC & QA

on nutritionally complete medical foods, FT-NIR was able Raman Spectroscopy:
to analyze protein, carbohydrates, and fat simultaneously Looking Deeper into Molecular Structure
using chemometric models to estimate each component Raman Spectroscopy is based on the scattering of light
from the measured spectrum. This differs from traditional by molecules. Changes in the energy of scattered light are
methods, which require a separate analytical procedure for related to the vibrations of chemical bonds and molecular
each component. Portable NIR has also been tested on pasta structures, producing a unique spectral pattern known
with sauce and found capable of rapidly estimating energy, as a “molecular fingerprint.” Raman produces narrow
carbohydrates, fat, dietary fiber, protein, and sugar. In the spectral bands with detailed structural information, allowing
grain industry, NIR is used to analyze nutritional and quality substances with similar structures to be distinguished more
parameters such as protein, fat, starch, and beta-glucan in precisely. It can also analyze samples in aqueous solutions,
oats. As a result, NIR has become one of the most widely where mid-infrared (MIR/FTIR) techniques can be more
used techniques on food production lines worldwide, covering challenging to use.
everything from grains and meat to dairy and processed In nutritional analysis, Raman has been used to detect
foods. specific compounds such as carotenoids in fruits and
However, NIR has limited ability to penetrate thick or vegetables, whose conjugated double-bond systems
highly opaque samples. Its prediction accuracy for some produce particularly strong Raman signals. It has also been
parameters, such as protein in complex food products, may used to rapidly estimate total protein in meat, protein and
also be lower than that of traditional chemical analysis. fat in milk powder, and gluten in flour and bakery products
Careful calibration is therefore essential. with accuracy.
A key limitation is that Raman signals are naturally weak,
Hyperspectral Imaging: as only a very small proportion of scattered light is produced
When Every Pixel Reveals Food Composition through the Raman effect. In many applications, sensitivity-
If spectroscopy can be thought of as measuring a spectrum, enhancing techniques such as surface-enhanced Raman
Hyperspectral Imaging (HSI) combines that concept with spectroscopy (SERS) are therefore required. The laser
imaging. It captures both spatial and spectral information equipment is also relatively expensive, so Raman has not
at the same time. The result is a three-dimensional dataset yet achieved the same level of industrial adoption as NIR.
known as a hypercube, which makes it possible to see how Overall, compared with traditional methods such as
chemical components are distributed across different points Kjeldahl and Soxhlet, all three techniques offer clear
of a food sample, rather than simply obtaining an average advantages: analysis time can be reduced from hours
value for the entire sample as with conventional NIR. to just seconds or minutes, no chemicals are required,
The major strength of HSI is its ability to detect variations samples are not destroyed, and multiple parameters can
within the same sample, such as fat distribution in meat or be analyzed simultaneously. Although accuracy may still
differences in moisture between individual grains. Research lag behind standard methods in some cases, this gap is
has shown that HSI can distinguish fresh beef from previously narrowing as machine learning (ML) continues to improve
frozen and thawed beef with nearly 89% accuracy when predictive models.
1
combined with appropriate signal-processing techniques.
The main challenges are the huge amount of data The Future of Non-Destructive Technologies
generated by each image, which requires greater computing The future of these technologies lies in deeper integration
power and more complex algorithms than conventional with AI and ML. Advanced approaches such as deep learning
NIR. Equipment costs also remain relatively high, while (DL) and neural networks are increasingly being used to
there is still a gap between laboratory results and real-time interpret complex spectral data and improve prediction
implementation on production lines. accuracy.
At the same time, smaller and more affordable devices
are making non-destructive technologies more accessible.
Portable sensors can be used directly on production lines
or even with smartphones. Combining data from techniques
such as NIR and HSI could further improve analytical
accuracy, while better model validation and transfer between
instruments will be important for reliable industrial application.
Ultimately, real-time integration with factory data systems
could shift quality control from occasional sampling to
continuous monitoring, helping the food industry move from
end-of-line quality testing to advanced quality prediction.

More Information Service Info C006

46 FOOD FOCUS THAILAND OCT 2026

22/9/2569 BE 19:10
42-46_Strong QC&QA_Non.indd 46 22/9/2569 BE 19:10
42-46_Strong QC&QA_Non.indd 46