Critical Reviews in Food Science and Nutrition

ISSN: 1040-8398 (Print) 1549-7852 (Online) Journal homepage: https://www.tandfonline.com/loi/bfsn20

Terahertz spectroscopy technology as an

innovative technique for food: Current state-of-
the-Art research advances

Chao-Hui Feng & Chiko Otani

To cite this article: Chao-Hui Feng & Chiko Otani (2020): Terahertz spectroscopy technology as
an innovative technique for food: Current state-of-the-Art research advances, Critical Reviews in
Food Science and Nutrition, DOI: 10.1080/10408398.2020.1779649

To link to this article: https://doi.org/10.1080/10408398.2020.1779649

Published online: 25 Jun 2020.

Submit your article to this journal

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=bfsn20
CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION
https://doi.org/10.1080/10408398.2020.1779649

REVIEW

Terahertz spectroscopy technology as an innovative technique for food: Current

state-of-the-Art research advances
Chao-Hui Fenga and Chiko Otania,b
a
RIKEN Centre for Advanced Photonics, RIKEN, Sendai, Japan; bDepartment of Physics, Tohoku University, Sendai, Miyagi, Japan

ABSTRACT KEYWORDS
With the dramatic development of source and detector components, terahertz (THz) spectroscopy THz spectroscopy; imaging;
technology has recently shown a renaissance in various fields such as medical, material, biosensing nondestructive;
and pharmaceutical industry. As a rapid and noninvasive technology, it has been extensively adulteration;
chemometrics analysis
exploited to evaluate food quality and ensure food safety. In this review, the principles and proc-
esses of THz spectroscopy are first discussed. The current state-of-the-art applications of THz and
imaging technologies focused on foodstuffs are then discussed. The advantages and challenges
are also covered. This review offers detailed information for recent efforts dedicated to THz for
monitoring the quality and safety of various food commodities and the feasibility of its wide-
spread application. THz technology, as an emerging and unique method, is potentially applied for
detecting food processing and maintaining quality and safety.

Introduction O’Sullivan, and Donnell 2012; Liu, Liu, Tang, et al. 2019;
Ren et al. 2019).
Food is very crucial source to provide essential amino acid,
In contrast to sophisticated used optical and near infrared
carbohydrates, proteins, vitamins, minerals, and other nutri-
spectroscopy, THz spectroscopy offers numerous advantages
tional compounds, which is supporting human’s daily activ-
for food science and technology, the most significant being,
ity, energy and nutritional requirement (Wang, Sun, and Pu
without doubt, its longer THz spectral band (Yang et al.
2017; Feng, Makino, Oshita, et al. 2018; Wang, Zhou,
2016; Ren et al. 2019). THz falls in the frequency range of
Huang, et al. 2019). The increasing concerns about quality
0.1–30 THz (1 THz ¼ 1012 Hz) (Tonouchi 2007; Dhillon
and safety of foodstuffs have been pointed out (Zhai et al.
et al. 2017) and the radiation possesses much lower photon
2018; Ren et al. 2019). Besides processing techniques like
cooling (Feng, Drummond, Zhang, et al. 2013; Feng, Sun, energies (4 meV for 1 THz) compared to X-ray photons
Garc
ıa-Mart
ın, et al. 2013; Feng and Sun 2014; Feng, (100 eV–100 keV) (Mittleman et al. 1999; Wang, Sun, and
Drummond, Zhang, et al. 2014; Feng, Drummond, Sun, Pu 2017). According to US Federal Communications
et al. 2014; Feng, Drummond, and Sun 2014; Feng and Li Commission (FCC), a photon energy greater than 10 eV is
2015; Feng et al. 2016; Feng, Liu, Makino, et al. 2017; Feng, defined as ionizing radiation. THz wave, with a non-ionizing
Li, Garc
ıa-Mart
ın, et al. 2017; Feng, Wang, Makino, et al. form, is biomolecules friendly. Tissue on food or other live
2019), drying (Sotome et al. 2009; Xiao et al. 2015; Pu and cells will not be negatively affected (Hu and Nuss 1995;
Sun 2015; Yang, Sun, and Cheng 2017; Wang, Law, Nema, Mittleman et al. 1999; Ren et al. 2019). Furthermore, abun-
et al. 2018; Ando et al. 2019), freezing (Cheng et al. 2017; dant useful information of both intra- and inter-molecular
Bilbao-Sainz et al. 2019) and packaging (Siripatrawan et al. connected by weak conformation-related interactions can be
2011; Siripatrawan and Makino 2018), a rapid and reliable obtained. For that reason, many low-energy physical phe-
detection method is also highly desirable for meeting the nomena (such as vibrational modes, phonon, rotational, and
ever-growing global demand for food commodities (Feng, vibrational energy levels) and biomolecular activities (e.g.
Makino, Oshita, et al. 2018; Feng et al. 2018a, 2018b; Feng, nucleic acids, amino acids, carbohydrates, peptides, and pro-
Makino, Yoshimura, et al. 2018; Ok et al. 2019). From the teins) can be observed in the THz frequency range
food safety point of view, any adulteration, contamination, (Markelz, Roitberg and Heilweil 2000; Brucherseifer et al.
and pathogenic infection not only reduces the nutritive and 2000; Walther, Fischer and Uhd Jepsen 2003; Markelz et al.
functional properties, but also generates a serious foodborne 2007; Haddad et al. 2013; Yang et al. 2016). In addition, the
illness (Kamruzzaman, Makino, and Oshita 2016a, 2016b, dynamics related to the hydrated water showed the time-
2016c; Teixeira and Sousa 2019). Therefore, technologies scales around sub picoseconds to picoseconds, correspond-
that can significantly improve or noninvasively detect the ing to THz frequency (Pal et al. 2002; Fukasawa et al. 2005;
quality of the foodstuffs are highly required (Gowen, Yada, Nagai and Tanaka 2008, 2009; Shiraga, Ogawa and

CONTACT Chao-Hui Feng chaohui.feng@riken.jp RIKEN Centre for Advanced Photonics, RIKEN, 519-1399 Aramaki-Aoba, Aoba-ku, Sendai 980-0845, Japan.
ß 2020 Taylor & Francis Group, LLC
2 C.-H. FENG AND C. OTANI

Kondo 2016). Because water is a strong THz wave absorber, different fields of food products. In spite of the fact that
skin, fat and lean tissues even in the same animal can be there are available reviews in literature on development of
characterized (He et al. 2006; Gowen, O’Sullivan, and THz spectroscopy (Hillger et al. 2019) and applications of
Donnell 2012). The functions of biomolecules and the con- THz spectroscopy in food (Qin, Ying, and Xie 2013), agri-
formations of proteins were associated with their THz spec- culture (Mathanker, Weckler, and Wang 2013; Ren et al.
tra (Yang et al. 2016). Due to the sensitivity to 2019), chemistry (McIntosh et al. 2012), astronomy
macromolecular structures, it has been considered as a (Rogalski and Sizov 2011), modern web technology (Notake
promised and emerging detection method to inspect foreign et al. 2014) and biology (Yang et al. 2016), scarce reviews
bodies such as metal, paper, plastics (Nagai and Fukasawa have addressed the assessment of the safety and quality of
2004; Wietzke et al. 2008; Haddad et al. 2013), leather and agricultural and food products using THz spectroscopy
wood (Koch et al. 1998) in packaged foodstuffs (Mittleman coupled with chemometrics since 2017. With the rapid
et al. 1999; Kawase et al. 2003; Kawase 2004; Dobroiu, technological THz sources and detectors development, the
Otani, and Kawase 2006; Ogawa et al. 2006; Ariyoshi et al. application of THz spectral has been intensively exploited to
2006; Hoshina et al. 2009; Shin, Choi and Ok 2018; Jiang, the agri-food products: A Web of Science recent search with
Ge and Zhang 2019; Wang, Zhou, Huang, et al. 2019). It keywords of “THz and food” resulted in 120 published
was also comprehensively used to determine pesticide and papers over the years 2002–2019, with approximately 62
antibiotic residues in agri-food industry (Qin, Xie, and Ying from 2017 to 2019. The aim of this paper is thus firstly to
2015, 2017), to classify edible oils (Zhan et al. 2016; Liu comprehensively review the latest applications of THz spec-
et al. 2018), and to identify genetically modified food (Liu, troscopy to foodstuffs. Following this, the future trends of
Liu, Hu, et al. 2016; Liu et al. 2017; Qin et al. 2017; Liu and THz spectroscopy are also discussed.
Kan 2018).
As a very useful procedure to analyze the data, chemo-
metrics analysis is regarded as an important access to mine Principles and processes of THz spectroscopy
the spectral data and to explain the relationship between the
At the early age, the application of spectroscopy measure-
measured parameters and spectral data (Nakajima et al.
ment in THz region was not well developed due to the lack
2007; Haddad et al. 2013). It has been widely applied to
of appropriate source and detector (Shen 2011). At that
Raman spectroscopy (Karacaglar et al. 2019; Luna et al.
period, THz gas laser was the source of pulsed THz. It was
2019; Nunes et al. 2019), near infrared spectroscopy (Garc
ıa-

mainly based on CO2 laser to stimulate the roto-vibrational
Mart
ın 2015; Garc
ıa-Mart
ın, Al
es-Alvarez, Carmen L
opez-
levels of gas molecule, to provide watts of power in a con-
Barrerz, et al. 2019, Garc
ıa-Mart
ın, Carmen L
opez-Barrerz,
tinuous-wave single-frequency beam (Wang, Sun, and Pu
Garc
ıa, 2019), hyperspectral imaging (Cheng and Sun, 2015;
2017). With the development of optics and electronics tech-
Cheng, Sun and Cheng 2016; Cheng et al. 2017; Huang, Liu
nology, THz systems experienced an extensive revolution in
and Ngadi 2017; Feng, Makino, Oshita, et al. 2018; Feng
et al. 2018a, 2018b; Feng, Makino, Yoshimura, et al. 2018; past decade and many efficient THz sources have been
Feng and Makino 2019; Jiang, Yoon, Zhuang, et al. 2019), exploited (Wang, Sun, and Pu 2017). THz spectroscopy sys-
fluorescence spectroscopy (Ahmad et al. 2016; Monago- tems can be divided into two types, namely continuous and
Mara~ na et al. 2018; Elmas et al. 2019; Wang, Wu, Long, pulsed wave systems. For continuous wave system, it does
et al. 2019) and so on. Due to the vast data generated by not require femtosecond (fs) lasers because the THz wave is
THz spectroscopic imaging systems, chemometrics methods generated by combining two frequency stabilized lasers
were utilized to reduce high dimensionality to obtain the which differences are in the THz region. For that reason,
most meaningful dimension (generating a simplified data) those systems are probably less expensive and complex in
without compromising the information contained in the ori- comparison with the pulsed systems (Gowen, O’Sullivan,
ginal data (Wang, Sun, and Pu 2017). Besides, the predictive and Donnell 2012). The generation of THz wave from
capability of the process could be improved because some pulsed system based on ultrafast lasers that emit sub 100 fs
random noise or redundant information could be eliminated pulses (Hu and Nuss 1995). Pulsed THz radiation (PTR) is
(Haddad et al. 2013; Garc
ıa-Mart
ın 2015). able to directly measure the transient electric field, obtaining
To date, THz technology, as an innovative, effectively THz spectrum with far better sensitivity and dynamic range
and quantitatively inspection method, has been intensively in compared with Fourier transform infrared (FTIR) method
exploited to water content determination (Sun, Wang, and (Han et al. 2001). Furthermore, the ambient noise generated
Zhang 2012; Abdul-Munaim et al. 2016), biology threats and from the incoherent blackbody radiation from the sample
defect inspection (Liu et al. 2007), residue detection (Suzuki, and surrounding was decreased owing to the time-gated
Ogawa, and Kondo 2011; Chen et al. 2015; Qin, Xie, and coherent detection technology used in PTR, which facilitated
Ying 2017; Xu et al. 2017), foreign body detection (Ok et al. heated sample characterization under extreme conditions
2014; Wang, Zhou, Huang, et al. 2019), transgenic food (Cheville and Grischkowsky 1995). Last but not least, the
detection (Liu and Li 2014; Chen et al. 2016; Liu et al. 2016) time-gated phase information preserved by pulsed radiation
and so on. All these studies accentuate an unyielding inter- and associated coherent detection scheme will allow THz
est in the application of THz spectroscopy as a powerful imaging for quantitatively and noninvasively on charactering
and promising detection method that can be applied to the inner structures of a sample (Shen 2011).
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 3

Figure 1. Components of a typical THz-TDS imaging system with transmission mode (a), a schematic illustration a transmission THz pulsed spectroscopy instrument
(b) (reproduced from Shen 2011), a time-resolved THz spectroscopy (c) (reproduced from Schmuttenmaer Research Group 2019a), THz emission spectroscopy (d)
(reproduced from Schmuttenmaer Research Group 2019b), and a THz pulsed imaging system (e) (reproduced from Shen 2011).
Note: BS: beam splitter; M1-M2: metallic mirrors; OEM1–OEM2: off-axis elliptic mirrors; SL: silicon lens system

There are three types of THz spectroscopy, namely THz includes a femtosecond laser, delay line, THz emitter, THz
time-domain spectroscopy (THz-TDS) (Figure 1a and b), detector and lock-in amplifier. As depicted in Figure 1b, the
time-resolved THz spectroscopy (TRTS) (Figure 1c) and laser light (usually produced by titanium sapphire laser) was
THz emission spectroscopy (TES) that is based on the con- split into pump beam and probe beam by a beam splitter
cept of Laser THz Emission Microscope (LTEM, Tonouchi, (Shen 2011). After excited by pump beam, the THz emitter
Yamashita, and Hangyo 2000; Kiwa et al. 2003) (Figure 1 d). generated a broadband pulsed of THz radiation and then
As an emerging and cutting-edge technology, THz-TDS is collimated and focused onto the sample by an off-axis ellip-
the most commonly used technique in current commercial tic mirror (OEM1). Again, the transmitted THz pulses, car-
THz spectroscopy (Baxter and Guglietta 2011). The typical rying sample information, were collected and focused using
setup of THz-TDS is shown in Figure 1a and b, which another off-axis elliptic mirror (OEM2) and finally reached
4 C.-H. FENG AND C. OTANI

Figure 2. Results from a THz-TDS scan and a time-resolved THz spectroscopy (b) scan (reproduced from Schmuttenmaer Research Group 2019a).
Note: blue circle in (a) is centered on the time-domain peak of the THz pulse.

a THz receiver. The probe beam, after passing through the parameter extraction is more difficult (Baxter and Guglietta
optical delay line, enhanced the optical path of this radiation 2011). In addition, the limitation to the highest absorbance
and gated the THz detector (Shen 2011). The THz receiver still occurs in a reflection measurement (Jepsen, Cooke, and
acted as a photoconductive switch or an optical rectification Koch 2011). With regard to ATR mode, it was developed by
in an electro-optic crystal (Yin, Tang, and Tong 2016a). Tanaka group to measure the strongly absorbing samples
Finally, the lock-in amplification was utilized to increase the such as water (Hirori et al. 2004). THz-TDS-ATR was
signal-to-noise of the measurement and a simple Fourier widely utilized to determine the hydration states of L-threo-
transformation was used to convert the time-domain signal nine in aqueous solutions (Huang et al. 2019), solvated dis-
into the frequency spectrum (Leahy-Hoppa et al. 2009). The accharides (Arikawa, Nagai, and Tanaka 2008), glucose
dielectric properties of the sample were expressed by the concentration (Suhandy et al. 2012) and so on. Jepsen,
absorption coefficient and the refractive index (Wang, Sun, Cooke and Koch (2011) stated ATR THz-TDS will be more
and Pu 2017). suitable for measuring high moisture content samples as it
For mirrors used in THz region, the thickness of the is more sensitive than that in transmission and reflection
metal coating should be at least two skin depths at the fre- THz-TDS. Shen (2011) mentioned that THz ATR measure-
quency of the incident beam to maximize reflectivity charac- ment will be more suitable for rapid screening of many sam-
teristics. Traditionally, they were made of metals such as the ples. Compared to reflection mode, the advantage of ATR
aluminum, silver, copper and gold. With the development of mode is that the crystal and sample can be put directly in a
THz mirrors technology, semiconductors (Ye, Zhang, and
transmission spectrometer. However, it should be noticed
Shen 2006), the tunable mirrors (Zhang et al. 2015) and the
that the temperature of the prism must be carefully main-
hybrid mirrors (Krumbholz et al. 2006) are commonly
tained because a change of 1 K leads to a 6-fs shift (Nagai
applied in current THz spectrometers.
et al. 2006). A reference spectrum is required to obtain the
In addition to the strong absorbing nature of THz radi-
THz spectrum of the sample during the measurement.
ation by liquid water, it would also negatively influence the
According to different modes, the operations for obtaining
result if the water vapor in the sample chamber. The sample
reference spectrum are different. The reference was obtained
chamber is therefore either purged with dry nitrogen gas or
vacuum throughout the measurement (Chen et al. 2015). without putting any sample when using transmission (Yang
There are major three modes in the THz-TDS system et al. 2016) and ATR modes (Suhandy et al. 2012) while
namely transmission, reflection and attenuated total reflec- putting a material of known reflectance (e.g. mirror)
tion (ATR) modes. Transmission mode, as a main mode (Gowen et al. 2012) for reflection mode. It was reported that
performed in THz-TDS, allows for more straightforward reliable and quantitative THz spectra were generated by
alignment and data analysis (Baxter and Guglietta 2011). It THz pulsed spectroscopy (TPS) transmission measurement
performed well when samples are transparent or weakly or and useful for studying opaque samples (Shen 2011) or bac-
moderate absorbing (Qin, Ying, and Xie 2013). In transmis- terial detection (Yang et al. 2016).
sion, although a strongly absorbing sample can exceed the Similar to THz-TDS, TRTS requires the pulsed laser
available dynamic range, parameter can be accurately deter- beam to be split into three parts (Figure 1c). Different from
mined by comparing to the maximum measurable absorp- the THz-TDS that provides information about the static
tion (Baxter and Guglietta 2011). Reflection mode is properties of the sample, TRTS probes the dynamic, evolv-
recommended if samples are highly absorbing or opaque. ing properties of the material and exerts an important
The reflection measurement is applicable to the wide band- impact on semiconductor photophysics (Baxter and
width sources and detectors for characterization of phonon Guglietta 2011). As illustrated in Figure 2, the result of
resonances (Ho, Guo, and Zhang 2010). However, the THz-TDS transmission is a function of the time delay
dynamic range and noise limitations are different in reflec- between the pump and probe beam (Figure 2a) whilst the
tion in comparison to transmission mode and material change in peak absorption for TRTS is a function of the
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 5

delay between the optical pump pulse and THz probe pulse prediction: 1.8–2.0; very good model/prediction: 2.0–2.5;
(Figure 2b). excellent model/prediction: over 2.5) (Kamruzzaman
As for TES, the source of THz emission is the sample et al. 2016a).
itself: pulsed light irradiates a sample that generates THz
radiation (Tonouchi, Yamashita, and Hangyo 2000; Kiwa
et al. 2003; Baxter and Guglietta 2011) (Figure 1d). The Applications of THz technique to foodstuffs
amplitude and shape of the transient electric field emitted Hitherto, THz showed a significant breakthrough in applica-
from the sample can be analyzed via TES (Baxter and tion for material property study (Yamazaki et al. 2018;
Guglietta 2011), which allows researchers to understand the Notake et al. 2019), medical field (Nakajima et al. 2007;
properties of materials. It was applied to the materials such Yang et al. 2016), pharmaceutical industry (Taday et al.
as semiconductors, superconductors, and magnetic films 2003; Ajito 2015), wireless communication (Garc
ıa-Mu~ noz
(Qin, Ying, and Xie 2013). et al. 2019; Liaskos et al. 2019), biosensing (Deng et al.
Regarding THz imaging system, the core technology is 2019; Li et al. 2019), art (Fukunaga et al. 2007) and so on.
the same as the aforementioned spectroscopy (Figure 1e). Compared to sophisticated and prevalent spectroscopic and
THz-TDS imaging simultaneously provides the spectrum imaging techniques, its application in foodstuffs is still at
and image corresponding to intrinsic properties and mor- the early stage. Figure 3 and Table 1 summarized the THz
phological characteristics of the sample (Wang, Sun, and Pu application in different types of food products.
2017). THz waveform is captured at many points mapped
over the surface of a sample. Each pixel of THz waveform is
recorded as a function of optical time delay. As a result, Meat
with a temporal scanning, three-dimensional information, Meat, containing large amount of iron, B-vitamins (thia-
i.e. vertical (x-axis), horizontal (y-axis) dimensions of the mine, niacin, vitamin B12 and riboflavin), minerals, amino
sample and the time-delay (z-axis) dimension, can be acids and lipoic acids, plays an important role in people’s
obtained (Shen 2011; Wang, Sun, and Pu 2017). A reflection daily life (Feng, Makino, Oshita, et al. 2018). However, the
configuration is highly recommended for THz imaging food safety of the meat products is an increasing concern
(Wallace et al. 2008) because it not only renders thick sam- among consumers. In order to prevent disease in animals,
ple to be imaged but also allows the use of the time-of-flight veterinary drugs are commonly used when feeding the ani-
of the technique (Shen 2011). The most attractive and mals, leading to drug residues in livestock products like
unique capability for THz imaging is its penetration prop- pork, beef, and mutton. THz spectroscopy, as an innovative,
erty, which can offer spectral information of interior pack- noninvasive and real-time technique, has been potentially
aged materials without damaging the exterior packaging applied to the antibiotic residue detection in animal origin
(Ren et al. 2019). Numerous literatures have pointed out products (Redo-Sanchez et al. 2011). In addition, the foreign
that the THz spectral band is long and thus will not be eas- material in sausages was investigated by Wang, Zhou,
ily affected by scattering (Lu et al. 2016; Heshmat et al. Huang, et al. (2019). Aluminum sheets with different shapes
2017; Wang, Sun, and Pu 2017) or the scattering losses in (polygon and strip) were enrolled in lean and fat. THz signal
biological tissues are negligible because of its relatively long confronted a severe damping of time-domain electric field
wavelength (Yang et al. 2016). However, Shen (2011) stated and frequency-domain amplitude when the sausages were
that the effects of scattering on TPS measurement should be contaminated by the aluminum sheets. The amplitude of fat
attended, especially for granulated materials with particle piece with a thickness of 1 mm was larger than that of 2 mm
size equivalent to the wavelength of the THz radiation. (Wang, Zhou, Huang, et al. 2019). However, accurately
The output of the THz spectroscopy contains a large determining the image contour of the foreign body still
number of spectral data sets. Consequently, chemometrics meets a big challenge: if the THz beam scanned vertically to
methods are widespread employed to spectral data pre-proc- the metal strip contamination and the width of this contam-
essing [such as smoothing, standard normal variate (SNV), ination was smaller than the THz beam diameter (Figure
multiplicative scatter correction (MSC), Savitzky-Golay first 4a), the position during the one-dimensional scanning was
and second derivative] (Nakajima et al. 2007; Suhandy et al. difficult to be detected. For example, the peak of electric
2012) and multivariate analysis [e. g. principal component field decreased because of the interrupt by contamination
analysis (PCA), partial least squares (PLS) and support vec- (Figure 4b and c), however, the waveform changed gently
tor machine (SVM)] (Liu et al. 2018; Liu, Zhao, Wu, et al. (even an equal peak value as the typical spectra of fat parts),
2019). The accuracy and predictive ability of those analysis implying no contamination detected (Figure 4a).
are usually evaluated by the coefficients of determination The frozen pork loin meat was initially studied by
(R2), standard error (SE) for calibration and prediction and Hoshina et al. (2009). Due to the difference in absorbance
ratio of standard error of prediction to standard deviation values and refractive indices, the striated muscle and adipose
value (RPD) (Suhandy et al. 2012). A reasonable model pos- tissue were able to be clearly inspected by THz-TDS. The
sesses a high R2 (acceptable: 0.66–0.81; good: 0.82–0.90; absorbance of the striated muscle was 2.5–3.0 times higher
excellent: over 0.90) (Garc
ıa-Mart
ın 2015; Feng, Makino, than that of adipose tissue. Similar observation was obtained
Oshita, et al. 2018) with a low root mean square error where tissues had considerably higher absorption in THz
(RMSE) value along with a high RPD value (good model/ range than fat (Bowman et al. 2016). The various water
6 C.-H. FENG AND C. OTANI

Figure 3. THz technique applications in food field.

Note: Some of the figures are reproduced from Dobroiu, Otani, and Kawase 2006, Ok et al. (2014; 2018), Wang, Zhou, Huang, et al. (2019), and Jiang et al. (2019).

contents may be attributed to this phenomenon–the hydra- The reason for desiring chocolate may be probably not only
tion of a tissue greatly affected its absorption characteristics for its good taste, the enjoyable physiological effects from
(Bowman et al. 2016). the ingredients, but also may associate to some good experi-
The spatial distribution of adipose tissue in the striated ence or comfortable situation (Molinari and Callus 2012).
muscle was clearly displayed by two-dimensional map of From the health point of view, taking small-moderate
absorbance and the maximum peak position in the time- amounts of chocolate may have a number of health benefits
domain data (Hoshina et al. 2009). Similar findings were (Latif 2013). Some researchers stated that chocolate may
observed in the study of Bowman et al. (2016). The authors potentially possess anti-depressant effects and prevent cogni-
stated that the clear contrast between fat and muscle tissue tive dysfunction (Walcutt 2009; Latif 2013) while others did
were obtained in both THz time and frequency domain not find the relevant evidence (Scholey and Owen 2013;
images and in the reconstructed electrical properties. THz Veronese et al. 2019). Further investigation manifested that
imaging in reflection mode illustrated a higher resolution the consumption of chocolate enhanced positive mood,
and could be more appropriate for imaging fresh tissue especially when it was eaten mindfully (Meier et al. 2017).
(Bowman et al. 2016). All those may point out that chocolate exerts a crucial
impact on both economic development and people’s
daily life.
Chocolate
In the chocolate manufactory industry, chocolates are
From the prehistoric era, chocolate is known as “kakawa,” a susceptible to be contaminated. Therefore, a noninvasive
meaning of “Food of the Gods.” It is one of the most popu- and rapid prediction approach for monitoring chocolate to
lar craved food products in the world (Meier, Noll, and ensure good quality is highly demanded to meet today’s
Molokwu 2017; Veronese et al. 2019): approximately 7.3 fiercely competitive market. Detecting contamination like
million tons of retail chocolate confectionery were consumed metallic or nonmetallic contamination such as stone or glass
throughout the world in 2015/2016, with the expectation of splinter attracts great attention during chocolate production
up to 7.7 million tons by 2018/2019 (Veronese et al. 2019). (Redo-Sanchez et al. 2011). The stones may origin from
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 7

Table 1. Application of THz spectroscopy for detecting food products.

Chemometric
Product Applications THz range Key findings analysis Accuracy Reference
Meat Sausage foreign 0.10–3.50 Amplitude of fat piece (1 mm PCA, DA 98.3–100% Wang, Zhou, Huang,
body (aluminum thickness) was larger than et al. 2019
sheet) detection that with 2 mm.
The width of foreign body
smaller than diameter of
THz beam is hard
to identify
Frozen pork loin 0.35–2.30 Absorbance of the striated – – Hoshina et al. 2009
muscle was 2.5–3.0 times
higher than that of
adipose tissue
Spatial distribution for
adipose tissue and striated
muscle can be clearly
shown in dimensional map
Chocolate FB (stone, glass 0.05–2.50 The transmitted intensity was – – Jӧrdens et al. 2006
splinter, metallic lower when the
screw) detection contaminations were glass
and stone
Not suitable for chocolate
wrapped in aluminum foil
(for packaging)
More than one-time delay
and scanning speed of
0.55 m/s were
recommended
FB (metallic needle) 0.27–1.05 Detector made by a low-cost – – Schuster et al. 2011
silicon CMOS was capable
of detecting a metallic
needle buried in a
packaged chocolate bar
Melted 0.20–0.30 Continuous-wave sub-THz – – Ok et al. 2015
surface detection transmission imaging with
a polygonal mirror was
feasible to visualize interior
structure of the chocolate
bar with packaging.
Oil Types of vegetable 0.20–1.50 Time delays for sunflower – – Li 2010
oil identification seed, peanut, soybean, and
rapeseed oils were at
2.13 ps, 2.1 ps, 2.1ps and
2.1 ps
Inter-molecular vibrations
are clearly shown from 1.1
THz to 1.5 THz
Edible oil category 1.50–3.50 The total prediction accuracy GA-PLS-DA, PLS, 50–100% Yin, Tang, and
determination for GA-PLS-DA, PLS, iPLS, iPLS, biPLS Tong 2016b
biPLS were 100%, 50%,
54% and 88%, respectively.
Optimal spectral intervals
were 1.8–1.9THz, 2.5–2.6
THz, 3.4–3.5t THz.
Food and technical 0.05–3.00 Refractive index increased – – Karali
unas
oil identification with an increasing water et al. 2018
concentration
contaminated in oil.
Reason for non-uniform
dependence of optical
properties during
degradation was found
Transgenic 0.10–1.50 Correction for PLS-DA model PLS-DA, SPA-LDA 97% Liu et al. 2017
cottonseed oil can achieve to 97%
discrimination
Transgenic 0.10-1.50 The classification accuracy of PLS-WLDA; 96.53%, 97% Liu, Fan, Liu
camellia oil PLS-WLDA can reach SPA-WLDA et al. 2019
to 97%
AFB1 in soybean oil 0.10–1.50 1 mg/kg AFB1 in soybean oil (t-SNE BPNN, BPNN 90–100%; Rp > Liu, Zhao, Wu,
identification can be accuracy (90%) 0.9456; RMSEP < et al. 2019
detectable. 1.7433 mg/kg
Prediction accuracy
increased with an increase
of AFB1 concentration
Using t-SNE BPNN model:
(continued)
8 C.-H. FENG AND C. OTANI

Table 1. Continued.
Chemometric
Product Applications THz range Key findings analysis Accuracy Reference
Rp ¼ 0.9948; RMSEP ¼
0.7124 mg/kg;
Using BPNN model: Rp ¼
0.9456; RMSEP ¼ 1.7433
mg/kg
EVOO geographical 0.10–4.00 Different EVOO geographical PCA, GA-LS-SVM, >82.50% Liu et al. 2018
origin origins can be separated GA-BPNN, GA-RF
determination based on the absorbance
spectra range of 1.0–2.5
Accuracies for GA-LS-SVM,
GA-BPNN, and GA-RF
models were 96.25%,
86.25%, 82.50%,
respectively.
Crop products Embryo and 0.04–0.08 Embryonic and endosperm – – Li et al. 2018
endosperm region can be clearly
evolution in observed at 0.05 and 0.06
maize seed THz, respectively
The problem for applying
THz spectral images to
uneven samples can be
solved by double Gaussian
filter denoising and
another pretreatment
Starch content 3.00–13.50 No peaks at 5.0 and 7.8 THz 2nd D-LR R ¼ 0.98, RMSE Nakajima et al. 2019
evolution in were found in amorphous ¼ 4.7%
mung starch
bean seedlings Frequencies oat 9.0, 12.1,
13.1 THz may relate to
covalent modes, especially
for intensity at 9.0 THz
Relationship between
starch content and THz
spectra was quantified by
using 2nd D-LR model
FB (small stone, 0.10–3.00 The best reconstructed image – – Jiang, Ge, and
glass fragment, for detect FB was obtained Zhang 2019
wood chip and at frequency of 1.1 THz
metal screw) Only metal screw can be
detected in identifiable when it
wheat grain embedded in
and flour 10 mm depth
Moldy 0.20–1.60 Optimal frequencies (0.32, PCA- SVM, PCA- >86% Jiang et al. 2015
wheat detection 0.59, 0.87, 1.0, 1.29, and PLSR, PCA-BPNN
1.58 THz) were obtained
Overall prediction accuracy
for PCA-SVM, PCA-PLSR,
and PCA-BPNN using
optimal frequency were
95%, 92.5% and 86%,
respectively
THz imaging can visualize
the location when
mold occurred
Wheat quality 0.20–1.60 Satisfied prediction accuracy PCA-SVM, PCR, 95% Ge et al. 2014
identification (95%) was obtained under PLS, BPNN
PCA-SVM model
Wheat varieties 0.20–2.00 Optimal interval frequency iPLS, PLS Rc2: 0.982–0.992 Ge et al. 2015
identification range was obtained RMSEC:
between 0.787 and 0.900 0.573–1.472
THz
Prediction accuracy can
reach to R2 ¼ 0.992 and
RMSEC ¼ 0.9573 using
iPLS model.
Transgenic rice 0.00–5.00 Differences between the PCA-BPNN, RF-1st 66.67–96.67% Liu, Liu, Hu
identification transgenic and non- D, SNV et al. 2016
transgenic rice seeds
occurred in THz band
An accuracy classification
of 96.67% in the validation
group was achieved when
using RF model coupled
(continued)
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 9

Table 1. Continued.
Chemometric
Product Applications THz range Key findings analysis Accuracy Reference
with the first derivative
pretreatment.
Transgenic 0.20–2.50 The genetical modified cotton ADPSO, SVM 96.97% Liu and Fan, 2016
cotton seeds seeds were identified by
using ADPSO-SVM
Genetical modified 0.10–2.50 The genetical modified seeds DPLS, Grid Search- 96.15% Wei et al. 2020
soybean seeds can be accurately classified SVM, PCA-BPNN
classification by using Grid Search-SVM
combined with the
pretreatment of mean
center and iPLS
Genetical modified 0.10–4.00 Soybean seeds with LS-SVM, PCA- 66.67–88.33% Liu, Liu, Chen
soybean seeds glyphosate resistant, BPNN, SNV et al. 2016
identification hybrid descendants and
traditional non-
transformed could be
clearly distinguished using
LS-SVM and the
pretreatment of SNV with
the classification correction
of 88.33%
Fruit L-AA concentration 0.60–13.50 Identified eight important PLS; 1st D-PLS; Rp2: 0.916 Suhandy et al. 2012
determination frequencies: 1.62 THz, 1.97 RMSEP: 2.791 for
THz, 2.55 THz, 3.59 THz, 1st D-PLS
4.28 THz, 6.83 THz, 9.60
THz and 10.88 THz
Intermolecular bonding
force of L-AA related to a
lower frequency
L-AA concentration 0.60–13.50 Calibration model performed PLS, iPLS 0.900-0.965 Suhandy et al. 2013
determination well between 1.74 and
7.40 THz
Rc2: 0.965; RMSEC: 1.268
for using full spectra PLS
model;
Rcv2: 0.900 RMSECV: 2.099
for using interval
iPLS model
Mixture Food Matrices Food matrices (flour, 0.10-3.00 Absorption peaks of – – Baek, Lim, and
milk and melamine were at 2, 2.26 Chun 2014
chocolate and 2.6 THz
powder) Sensitivity was reduced in
adulteration flour and chocolate
powder when they
were covered
FB (maggot, 0.2 Whole sizes of maggot and – – Lee et al. 2012
crickets) cricket pieces were
detection identifiable.
Antibiotics residues 0.3–1.8 Spectral fingerprint of TCH PLS-DA; PLSR Rc: 0.9596–0.9933; Qin, Xie, and
identification can be clearly observed RMSEC: 1.94–4.71 Ying 2014
For all TCH using PLSR
model, Rc ¼ 0.9649; RPD
values ¼ 3.04
Amino acids mixture 0.3–1.9 Characteristic THz spectrum PLS; iPLS R2 > 0.9904; RMSEP Lu et al. 2016
for glutamine: at 1.71 and ¼ 0.39 ± 0.02
2.27 THz; for glutamic acid:
between 1.23 and 2.03
THz
iPLS model performed well
at frequency of
1.20–1.37 THz
Note: ADPSO: adaptive dynamic particle swarm optimization; AFB1: Aflatoxin B1; biPLS: backward interval PLS models; DA: discriminant analysis; 1st D-PLS:
Savitzky-Golay first derivative – partial least squares; 2nd D-LR: Savitzky-Golay second derivative–linear regression; DPLS: discriminant partial least squares;
EVOO: extra virgin olive oils; FB: Foreign body; GA-BPNN: genetic algorithm-back propagation neural network; GA-LS-SVM: genetic algorithm-least squares-sup-
port vector machine; GA-PLS-DA: genetic algorithm-partial least squares discriminant analysis; GA-RF: genetic algorithm-random forest; iPLS: interval PLS; L-AA:
L-ascorbic acid; PCA: principal component analysis; PLS: partial least squares; PLS-WLDA: partial least square-weighted linear discriminant analysis; RF: random
forest; SPA-LDA: successive projection arithmetic-linear discriminant analysis; SPA-WLDA: successive projection arithmetic-weighted linear discriminant analysis;
TCH: Tetracycline hydrochloride t-SNE BPNN: t-distributed stochastic neighbor embedding back propagation neural network

ingredients like nuts whilst glass may come from light bulbs employed to detect foreign body in chocolate (Koch 2007;
(Jӧrdens et al. 2006). Owing to the high fat with low mois- J€
ordens and Koch 2008; Schuster et al. 2011; Guillet et al.
ture content, the THz imaging has been intensively 2014). Contaminations such as stones, metallic screws or
10 C.-H. FENG AND C. OTANI

Figure 4. THz waveforms and frequency-domain amplitude spectra for beam scanning perpendicular to metal strip (a), horizon to metal strip (b) and using square
shape (c) (reproduced from Wang et al. 2019).

glass splinters in chocolate bar can be manifestly distin- the problems of melting and reshaping during handling
guished from desirable ingredients like nuts and raisins processing, especially in summer. As chocolates are com-
using THz technology (Jӧ rdens et al. 2006; Koch 2007). Jӧ monly packaged, a transmission technology like THz image
rdens et al. (2006) also mentioned that the different thick- is ideal for inspecting the quality before trade. A continu-
ness in chocolate bars might result in various signals and so ous-wave sub-THz transmission image was potentially
affected the foreign body detection. More than one-time employed to evaluate the quality of a chocolate bar (Ok
delay was thus recommended to address this problem (Jӧ et al. 2015). It can be clearly observed where the melted part
rdens et al. 2006; Gowen et al. 2012). Likewise, the feasibility of chocolate was located from the imaging without opening
of detecting a metallic needle buried in a packaged chocolate the packaging (Ok et al. 2015). Moreover, the inspection
bar using a low-cost THz detector in a 130 nm silicon com- speed was reported to be more rapid than THz-TDS imag-
plementary metal-oxide-semiconductor (CMOS) technology ing systems (Ok et al. 2015).
was investigated by Schuster et al. (2011). THz images taken
at 292 GHz (sub-THz) (0.292 THz) can clearly indicated the
needle position with a satisfactory spatial resolution Oil
(Schuster et al. 2011).
During chocolate transportation, cold chain–offers the Triglycerides are the main compounds of body fat in
refrigerated (includes chilling and freezing) preservation humans and other animals. Among them, a-linolenic acid
from post-harvest to home storage–is recommended to and linoleic acid are not synthesized in body and requires to
maintain the quality (Zhao et al. 2018). Due to the fluctu- be obtained from food such as edible oil (Liu et al. 2017).
ation of storage temperatures at initial storage, transporta- The values of acid, peroxide and so on are usually used to
tion, distribution and final retail shop, chocolates suffered evaluate the quality of an oil. The deterioration of vegetable
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 11

oils is usually through two main processes, i.e. oxidation high risk for human consumption (Liu, Zhao, Wu, et al.
and hydrolysis (Stark et al. 2011). The number and location 2019). The AFB1 level was limited to less than 2 mg/kg in
of the double bonds in triglycerides will be changed during the European Commission Regulation (EU) and less than 10
oxidation process. Meanwhile, peroxides and other oxygen- mg/kg for plant oil in China (excluding corn and peanut oil)
ated degradation products would be also generated, which whilst total aflatoxin level in grain by-product should be less
severely decreases the oil nutritional properties (Karali unas than 20 mg/kg in USA regulation (Liu, Zhao, Wu, et al.
et al. 2018). With regard to hydrolysis, the triglyceride mole- 2019). THz spectroscopy was capable of inspecting AFB1
cules break down with the reaction of water to diglycerides, even the level is 1 mg/kg in soybean oil, with over 90%
monoglycerides or glycerol, releasing free fatty acid (Garc
ıa- accuracy. With an increasing of AFB1 concentration, the
Mart
ın, Ruiz, Garc
ıa, et al. 2019). In order to maintain the prediction accuracy could reach up to 100% for 10 and 20
quality of oil, the concentration of free fatty acid should not mg/kg of AFB1 sample. THz spectroscopy with pretreatment
be over 1% in commercial edible oils. Four kinds of vege- greatly enhanced the robustness of the model: coefficient of
table oils were characterized by THz-TDS using transmis- prediction (Rp) can achieve to 0.9948 with RMSEP of 0.7124
sion mode in the frequency range from 0.2 to 1.5 THz (Li mg/kg when THz data was pretreated by t-distributed sto-
2010). The time delays of sunflower seed oil, peanut oil, soy- chastic neighbor embedding (t-SNE BPNN model) com-
bean oil, and rapeseed oil were reported to be 2.13 ps bined with back propagation neural network (BPNN) model,
(0.47 THz), 2.1 ps (0.48 THz), 2.1 ps (0.48 THz), and 2.1 ps in contrast with Rp of 0.9456 and RMSEP of 1.7433 mg/kg
(0.48 THz), respectively (Li 2010). Subsequently, the feasibil- for that without pretreatment using the same model (Liu,
ity of determining types of edible oil (peanut oil, corn oil, Zhao, Wu, et al. 2019).
pepper oil, sesame oil and sunflower seed oil) using THz In order to avoid fraud or mislabeling, the geographical
spectroscopy in the range of 1.5–3.5 THz was studied by origin (from Australia, Spain, Greece and Italy) of extra vir-
Yin, Tang, and Tong (2016b). The model developed by gen- gin olive oils (EVOO) were identified via THz spectroscopy
etic algorithm-partial least squares discriminant analysis coupled with chemometrics analysis (Liu et al. 2018).
(GA-PLS-DA) possessed a larger correlation coefficient of EVOOs from different areas possessed different fatty acid
prediction (Rp) with a smaller root mean square error of contents. For example, EVOO from Spain contained the
prediction (RMSEP). GA-PLS-DA model also showed a highest stearic acid (C18:0; 3.73 ± 0.06%) and oleic acid
higher classification accuracy than PLS (using full spectra), (C18:1; 78.55 ± 0.47%) contents whilst palmitic acid (C16:0;
interval PLS, and backward interval PLS models (Yin, Tang, 13.81 ± 0.12%), palmitoleic acid (C16:1; 1.28 ± 0.04%) and
and Tong 2016b). Pure, degraded oils and hydrocarbon linoleic acid (C18:2; 11.28 ± 0.13%) levels of EVOO from
chemicals were inspected using THz-TDS in the range of Italy showed the highest. Accordingly, the absorbance spec-
0.05 to 3.00 THz (Karali unas et al. 2018). Results showed tra of EVOOs from 1.0 to 2.5 THz were different and thus
that the spectra of refractive index (RI) and absorption coef- can be apparently discriminated their geographical origins.
ficient were significantly different when ester linkages pre- Further investigation demonstrated that EVOO samples ori-
sented in the both edible and technical oils (Karali unas et al. ginated from different zones can be accurately classified
2018). The oleic acid and water in vegetable oils resulted in (96.25% in prediction group) using least square (LS)-SVM
some changes in absorption coefficient and RI spectra. combined with GA using absorbance spectra abstract from
According THz technology, it is explicitly demonstrated that THz spectroscopy, indicating that THz tandem with chemo-
dynamics of inverse micelles and oil hydrolysis participated metrics can be a promising technique to rapidly and effi-
oil degradation and were accountable for non-uniform ciently discriminate the geographic origin of EVOOs (Liu
dependence of optical properties on extent of degradation et al. 2018).
(Karaliunas et al. 2018).
Edible oil was discriminated from transgenic cottonseed
Crop products
oil by using THz spectroscopy combined with PLS-DA
model with the correction up to 97% (Liu et al. 2017). Crop products are an important agricultural product serving
Likewise, the transgenic and non-transgenic camellia oil for our daily life but are vulnerable to resist insect damage,
were determined by THz spectroscopy coupled with succes- mycotic contamination or early germination (Wang, Sun,
sive projection arithmetic-weighted linear discriminant ana- and Pu 2017). The chemical composition and distribution in
lysis (SPA-WLDA) and partial least square-weighted linear crop products could offer a useful information for under-
discriminant analysis (PLS-WLDA) (Liu, Fan, Liu et al. standing physiological functions and basis for crop selection
2019). The model of PLS-WLDA elaborated a slightly higher and breeding (Li et al. 2018). It is thus of great interest to
classification accuracy (97%) than that of SPA-WLDA develop a rapid and noninvasive technique for evaluating
(96.53%), demonstrating the potential of THz spectroscopy quality and maintaining safety of the crop products. THz
combined with chemometric methods to classify the trans- spectroscopy elaborates this potential (Zhang et al. 2017).
genic organism. The spectral images for embryo and endosperm in maize
Aflatoxin B1 (AFB1), as a class I carcinogen according to seed were captured using THz time-domain spectral images
the International Agency for Research in Cancer (IARC, (Li et al. 2018). Different delay times were found to affect
1993), may widely exist in various agricultural food products the contour information of different tissues at different
like maize, wheat, rice and soybean during storage, posing a depths of the seed sample. To be specific, the endosperm
12 C.-H. FENG AND C. OTANI

region can be clearly revealed by reconstructed image at as the optimal (O) frequencies to input into SVM, PLSR,
18.0 ps (0.06 THz) while at 20.7 ps (0.05 THz) for embryonic and BPNN models. The best overall prediction accuracy was
region. After pretreatment (e.g. double Gaussian filter achieved by using O-PCA-SVM model (95%). Although
denoising, differential time slice, peak-peak in slice, edge overall prediction accuracy decreased slightly in comparison
detection extraction of contours) for THz optical informa- with the achieved one using the full frequencies (96.5%),
tion, the blurred visual problem of the images was solved approximate 96.49% frequencies was reduced from the full
and also showed the feasible application of THz imaging to spectra. Using optimal (feather) frequencies can not only
uneven samples (Li et al. 2018). accelerate the data analysis processing but also simplify the
Dynamical changes of starch content in germinating model and provide useful information for a simple cost-
mung bean seedlings during 7 days were studied using THz effective system. The output of the value from O-PCA-SVM
spectroscopy (Nakajima et al. 2019). Four peaks (9.0, 10.5, can reflect each pixel, therefore, an image can be made
12.2, and 13.1 THz) were shown to be the most suitable to according to the spatial positions of each pixel. Figure 5
estimate starch content. In addition, the identifiable peak at illustrates the wheat grains with four different mold stages
9.0 THz detected from seedlings may be due to a covalent and it can be clearly observed that the embryo [at regions
vibration of starch rather than being affected by the con- (1–3)] suffered from mold at the early stage (Figure 5b) and
stituent saccharides. Besides, the starch contents can be suc- the situation diffused to the whole interior structure at the
cessfully predicted using the peak intensity at 9.0 THz in the end of storage time (Figure 5d) (Jiang et al. 2015). Likewise,
second derivative spectra (R ¼ 0.98, RMSE ¼ 4.7%) wheat grain mixed with normal, worm-eaten, moldy and
(Nakajima et al. 2019) sprouting were identified using PCA-SVM model under
The foreign body (i.e. small stone, glass fragment, wood THz range of 0.2–1.6 THz, with prediction accuracy of 95%
chip and metal screw) embedded in wheat grain and flour (Ge et al. 2014).
was comprehensively investigated using THz imaging with The genetically modified and non-transgenic rice were
reflection mode (Jiang, Ge, and Zhang 2019). The foreign discriminated by THz spectroscopy imaging system in tan-
body became more apparent with an increase of frequency dem with chemometrics, with the classification accuracy of
(from 0.9 THz and 1.3 THz) while the most optimal recon- 96.67% in the validation set when using random forest with
structed image was obtained at frequency of 1.1 THz. The the first derivative pretreatment (Liu, Liu, Hu et al. 2016).
limitation of the system dynamic range may cause the for- Similarly, the genotype (glyphosate-resistant, hybrid
eign body undetectable at 1.3 THz (Jiang, Ge, and Zhang descendants and traditional non-transformed) soybean seeds
2019). Further study elucidated that only the metal screw could be noninvasively monitored by THz-TDS combined
can be clearly distinguished when the depth of foreign body with chemometrics, with the classification accuracy of
in wheat grain was increased to 10 mm. The absorption in 88.33% in the prediction group when using least squares-
the wheat layer and the scattering due to rough surface may support vector machines (LS-SVM) model with SNV pre-
attribute to the decrease of transmitted intensity when for- treatment (Liu, Liu, Chen et al. 2016). In the same way, two
eign body (small stone, glass fragment and wood chip) were hundred and twenty-five transgenic and non-transgenic soy-
embedded in a deeper location. As for metal screw, the con- beans were rapidly and efficiently identified by THz-TDS in
trast of refractive index with wheat grain is much higher a frequency range of 0.1–2.5 THz. With the pretreatment of
than glass, stone and wood chip (equivalent to wheat grain), mean center and interval partial least squares (iPLS), trans-
resulting in being detectable when buried in 10 mm (Jiang, genic seeds can be accurately (accuracy rate: 96.15%)
Ge, and Zhang 2019). The increased depth again led to the classified by THz using the model developed by Grid
obscuring shape of foreign bodies in flour. The cross-section Search-SVM (Wei et al. 2020). Likewise, three types of
images of wheat with foreign body were found to be con- transgenic cotton seeds (i.e. Xinqiu No. k638, Lumianyan
sisted with the corresponding THz waveforms. The No.36, and New luzhong No.6) were detected by THz. It
unknown depth of the material can be estimated by the fol- was difficult to distinguish the differences of those cotton
lowing equation: samples merely based on their absorption peaks. However, a
D ¼ DT  2e (1)where D is the depth of material, DT novel adaptive dynamic particle swarm optimization associ-
is the differences between the first peak and second peak, ated with SVM model enabled to identify those transgenic
and e is the speed of the light in vacuum. If the refractive cotton seeds, with the recognition rate up to 96.97% (Liu
index (n) is known, the depth of the material can be accur- and Fan 2016).
ately calculated using the Equation (2) (Jiang, Ge, and
Zhang 2019).
Fruit
D ¼ DTen=2 (2)
Jiang et al. (2015) distinguished the moldy wheat using With an important composition and being an important
THz imaging in the range of 0.2–1.6 THz in tandem with nutrient in fruit, the different concentrations (0–21%) of
multivariate classification. Spectra data was pretreated with vitamin C (L-ascorbic acid) was determined by using ATR-
PCA and the top four PCs explained 98.25% of the total THz (Suhandy et al. 2012). The highest coefficient of deter-
contribution. Six frequencies (0.32 THz, 0.59 THz, 0.87 THz, mination for prediction (Rp2, 0.916) with the lowest RMSEP
1.0 THz, 1.29 THz, and 1.58 THz), according to peaks and (2.791%) was obtained when the spectra was pre-processed
valleys of the loading weights of the top four PC, were used by Savitzky–Golay first derivative. Eight important
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 13

Figure 5. Wheat grains with four different mold statues captured by THz images: (a) normal; (b) slightly moldy; (c) moderately moldy; and (d) seriously moldy
(reproduced from Jiang et al. 2015).

frequencies (1.62 THz, 1.97 THz, 2.55 THz, 3.59 THz, mistrust, but also leads the relevant dairy company to go
4.28 THz, 6.83 THz, 9.60 THz and 10.88 THz) were selected bankrupt. The reason to add melamine (1,3,5-triazine-2,4,6-
according to the regression coefficients of the developed triamine) in milk powder is because of its high nitrogen
calibration model (Suhandy et al. 2012). The inter-molecular content (40% nitrogen by number of atoms), which can
vibration of solid L-ascorbic acid (L-AA) was reported at increase the total nitrogen concentration that is convention-
2.05 THz and 2.34 THz (Cao, Zhang, and Zhou 2008) while ally used as an indicator to evaluate the protein content
some peaks were observed at 1.97 and 3.59 THz (RIKEN (Baek, Lim, and Chun 2014). The different concentrations
THz Database), indicating that intermolecular bonding force (0–10 wt %) of melamine mixed in flour, milk, chocolate
of L-AA was more relevant to a lower frequency. A high powder, and high-density polyethylene (HDPE) were
correlation between THz spectra and L-AA concentration detected by using THz-TDS (Baek, Lim, and Chun 2014).
was obtained using PLSR with full spectrum (Suhandy et al. The absorption peaks of melamine mixed with different
2012). Subsequently, interval partial least squares (iPLS) was food matrices (flour, milk, chocolate) and HDPE were
employed to split the spectra into smaller equidistant 11 detected at 2, 2.26 and 2.6 THz. In addition, a strong linear
subintervals, the prediction ability was improved greatly regression relationship was observed between absorbance
with a decrease of square error prediction (SEP) from 2.146 and different concentrations in HDPE (R2 > 0.998) and
to 1.490. The performance of calibration model was again weight ratio (Wmelamine/Wfood matrix) (R2 > 0.967). THz
confirmed to be better at the lower frequency range
images were acquired at 2 THz where a strong and clear
(1.74–7.40 THz) (Suhandy et al. 2013). The absorbance at a
peak of melamine absorption was obtained. According to
higher L-AA concentration (21.5%) was lower than that at a
THz images, the concentration and location of melamine in
lower L-AA concentration (15.3%), which can be explained
food matrices and HDPE can be identifiable even they were
by the less free water amount and so less inter-molecular
wrapped with oriented polypropylene/polyethylene vinyl
vibration developing between water and L-AA (Suhandy
film (thickness: 40 lm) or roll paper/polyethylene film
et al. 2013).
The tomato quality was evaluated by THz wave and a (thickness: 80 lm). In relation to HDPE, the effect of cover-
decreased tendency in the THz reflectivity was found in the ing materials on sensitivity of THz image was little whilst a
damaged area, which may result from moisture loss (Ogawa reduction of sensitivity was exemplified in flour and choc-
et al. 2006). This is of practical use for identifying the fruit olate powder. The combined effects of particle size and
injury caused by pressing during handling or transportation. chemical composition of sample and wrapped materials may
THz-TDS with Gouy phase shift in the reflected waveform contribute to this phenomenon (Baek, Lim, and
was also successfully applied to differentiate grapes, cherries, Chun 2014).
blue berries, and plum from leaves and stems (Federici et al. Apart from adulteration, there were also some cases
2009). The authors stated that THz imaging may be a useful where foreign body was miscellaneous in the food matrices
method for predicting high water content fruits in a thin during harvest, producing, or other procedures. The ground
canopy (Federici et al. 2009). instant noodle powder contaminated by different sizes of
maggot (length: 8, 11, 15, 22 mm; thickness: 2, 2, 2.5, 3 mm)
and crickets (length: 35, 50 mm; thickness: 5.5, 7 mm) was
Mixture food matrices scrutinized by continuous wave (CW) THz imaging at
Mixture food matrices like milk powder, cereal powder, 0.2 THz and X-ray (Lee et al. 2012). The maggot pieces were
hamburgers, meatballs, patties are attractive targets for adul- invisible with X-ray imaging while the whole sizes of maggot
teration due to monetary profits. For instance, one of the and cricket pieces were detectable using CW THz imaging,
largest food safety events – adulteration of milk powder indicating that low-density foreign body hidden in the food
with melamine–shocked everyone in China. A lot of inno- matrices could be more effectively examined using CW THz
cent infants consumed this “poison milk power” and conse- imaging (Lee et al. 2012). The images captured by X-ray for
quently suffered from kidney stones and other larger crickets (length: 50 mm, thickness: 10 mm) embedded
complications. This appalling issue not only seriously in noodle power were much less clear, compared to those
infringers upon consumer rights and results in consumers’ taken by THz imaging (Lee et al. 2012).
14 C.-H. FENG AND C. OTANI

In order to prevent and control disease of livestock, anti- chemical- or biological-based techniques, THz spectros-
biotics are always used as a type of feed additive to improve copy is environmentally friendly, toxic-free, noninvasive,
animal growth. However, an overuse of the antibiotics will and time-saving (Wang, Sun, and Pu 2017). Owing to
lead to unsafe residue levels in various milk or meat prod- its low energy (1–10 meV) and nonionizing property, it
ucts, which is highly concerned by consumers. The antibiot- is safe for both biomolecule samples and operators
ics [Tetracycline hydrochloride (TCH), oxytetracycline (Qin, Xie, and Ying 2014; Wang, Sun, and Pu 2017).
hydrochloride (OTCH), doxycycline hydrochloride (DTCH), 2. Fingerprint features: THz spectroscopy is ideal for ana-
and chlortetracycline hydrochloride (CTCH)] residues in lyzing molecular vibrations, rotations and vibrational
milk powder were assessed by using THz-TDS within the energy levels since those physical phenomena of biomo-
0.3–1.8 THz range (Qin, Xie, and Ying 2014). The spectral lecules lie in the THz frequency range, revealing useful
fingerprints of TCH can be clearly discriminated while information such as protein conformational changes,
OTCH possessed very weak absorption peak. It was found inter-molecular interactions, hydrogen bonding stretches
that the absorption peak was little different from their own and Van der Waals forces in many chemical or bio-
fingerprints when they were mixed with milk powder, which logical materials (Yang et al. 2016). To this end, it is
may be due to the destruction of the crystalline structure very different from spectroscopic techniques in the
and the interactions between food and antibiotics. The near-infrared range (e.g. NIR spectroscopy, hyperspec-
absorbances of TCH and OTCH were trendy to increase tral imaging) that concentrates the intra-molecular
with the increasing concentration while there were no such vibration modes arising from specific molecular bond-
correlations in relation to DTCH and CTCH (concentration ing like O-H and C-H bonds (Suhandy et al. 2013).
below 10%) (Qin, Xie, and Ying 2014). Infant milk powder 3. Transmission capability: THz waves exert very weak
mixed with different concentrations of TCH, except for impact on nonpolar material. Therefore, THz spectros-
OTCH and DTCH groups, can be clearly classified by using copy provides an increased objective and effective
PLS-DA model (Qin, Xie, and Ying 2014). When powder method to evaluate packaged samples as materials like
was mixed with both TCH and DTCH in concentrations paper, vinyl, plastics, textiles, ceramics and semiconduc-
from 5% to 15%, each concentration could be identified tors are transparent to THz waves (Mathanker et al.
with high precision and the position of absorption peaks 2013). It is very useful to detect foreign body in pack-
slightly shifted to higher frequencies with the increased con- aged food products without opening the package.
centration. The vibrational modes caused by change of par- Several publications have addressed the successful appli-
ticle shape after mixing may take into account for this cation of THz spectroscopy and imaging to determine
phenomenon (Qin, Xie, and Ying 2014). the quality of packed chocolate (Koch 2007; J€ ordens
The mixtures of L-glutamic acid and L-glutamine in yel- and Koch 2008; Schuster et al. 2011; Redo-Sanchez
low foxtail millet powder were evaluated by combination of et al. 2013; Guillet et al. 2014) and sausages (Wang,
THz-TDS with chemometrics (Lu et al. 2016). The charac- Zhou, Huang, et al. 2019). With the combination of
teristic THz spectrum for glutamine was at 1.71 and chemometrics analysis with THz imaging, the early ger-
2.27 THz whilst between 1.23 and 2.03 THz for glutamic
mination, mycotic contamination, harmful residues of
acid. Again, the sensitivity of amino acids in sample powder
the crop seed and the interior nutritional composition
was weaken, which could attribute to the THz absorption by
evolution of seeds during different storage days can be
other components (e.g. saccharides, cellulose, proteins, etc.)
clearly observed and predicted (Chen et al. 2015; Li
(Lu et al. 2016). The concentrations of amino acids mixed
et al. 2018; Wang, Wang, Zhao, et al. 2018). Moreover,
in yellow foxtail millet samples were quantified by using
the nutritional components in the mixture food matrix
partial least squares (PLS) and intervals partial least squares
can be quantified as well (Lu et al. 2016; Wang, Wang,
(iPLS; interval was set as 9). iPLS model performed well at
Zhao, et al. 2018; Du, Zhang, and Zhang 2019).
frequency of 1.20–1.37 THz, with the correlation coefficient
4. Adulteration detection: THz spectroscopy can detect
of 0.9904 and 0.9906 for glutamine and glutamic acid,
products’ adulteration efficiently based on the finger-
respectively (Lu et al. 2016).
print of each mixed chemical composition with high
accuracy (Baek, Lim, and Chun 2014; Liu 2016). Apart
Advantages and challenge of THz spectro- from the aforementioned transgenic oil, THz spectros-
scopic technology copy associated with chemometrics methods is pio-
Over the past few years, THz spectroscopy and imaging neered and extensive for differentiating the transgenic
technology, as a promising and emerging technology, offer a soybean seeds (Liu, Liu, Chen, et al. 2016), rice and its
new strategy for evaluating food and ensuring food safety. seeds (Liu, Liu, Hu et al. 2016). All these publications
The advantages of the THz spectroscopy are listed emphasize THz spectroscopic imaging technology as an
as follows: attractive and unique technology for fast analysis in
online, off-line or at-line food inspections that can be
1. Low energy: In contrast to traditional analytical meth- hardly judged by the naked eyes.
ods such as high performance liquid chromatography 5. Comparable little scatter effect: as the wavelength of the
(HPLC), polymerase chain reaction (PCR), enzyme- THz spectral band is longer than conventional optical
linked immunosorbent assay (ELISA) and other and near infrared spectra, it will not be easily affected
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 15

by scattering when detecting the biological tissue sample important parameters in evaluating imaging processing.
with THz radiation. However, the scattering caused by The resolution of THz far-field imaging was lower than
particles in food matrices cannot be ignored because the that of THz near-field image. The pulsed THz imaging
sensitivity will decrease (Baek, Lim, and Chun 2014; Lu was reported approximately 25 times slower than that
et al. 2016). Dobroiu, Otani, and Kawase (2006) also using continuous wave radiation, which limits its wide-
mentioned that the grain size dominated the scattering spread. Therefore, different applications corresponding
properties of powders. If the grain size is equivalent to to different imaging systems should be considered. For
or larger than the wavelengths, the effect caused by example, it is recommended to using CW THz imaging
scattering is significantly (Dobroiu, Otani, and Kawase when many items with presumably similar properties
2006). Coating plasma and metamaterial are recom- are going to be screened. By contrast, when the aim is
mended to reduce the scattering effects (Ji et al. 2018). to identify unknown objects such as concealed explo-
6. Unique water absorption property: the absorption coef- sives, pulsed THz imaging may be more suitable (Baxter
ficient of water was reported around 200 cm1 at 1 and Guglietta 2011).
THz, almost 105 times higher than that in the visible
region (Qin, Ying and Xie, 2013). Therefore, this high
absorptive property can be utilized for accurately moni- Conclusions
toring and analyzing moisture in agri-food products. An overview of some recent developments and highlighted
applications of THz spectroscopy and imaging in food is
In spite of the above advantages, the application of THz addressed in this review. Different types of food products,
spectroscopy in agri-food industry is still in its early stage. such as meat, chocolate, oil, crop products, fruit and mix-
There are also a number of inherent limitations, which have ture food matrices have been extensively discussed.
so far become the challenges for its comprehensively appli- Although previous reviews have highlighted the development
cation in the foodstuffs. of THz spectroscopy and applications of THz in food, agri-
culture, chemistry, astronomy, modern web technology and
1. Limitation for high moisture content products: as THz biology, a few reviews have addressed evaluating the safety
signals are greatly attenuated by water, which limits its and quality of agricultural and food products using THz
application mainly to dry food samples. The alternative spectroscopy coupled with chemometrics since 2017. This
method to overcome this limitation is to employ the review provides the latest information on the application of
attenuated total reflectance (ATR) method rather than THz spectroscopy to agri-food productions and THz spec-
transmission method (Jepsen et al. 2011; Suhandy et al. troscopy in tandem with chemometrics. Although THz spec-
2012). In order to eliminate the influence caused by troscopic imaging technology is an attractive and sound
vapor, the THz system is either flushed with nitrogen technique that can rapidly determine the safety and quality
gas or in vacuum conditions, which may not be prac- of various food commodities, especially for the packaged
tical in a large food processing scenario. foodstuffs, without destructing the food structure, some
2. Penetration limitation: Although THz radiation can inherent limitations in terms of its extensive application,
penetrate the packaged samples, the penetration depth complex data handle, and initial investment are still needed
depends on several factors such as the properties of the to be solved.
food material, thickness of the sample, power of the From the information available in literature, THz spec-
THz source, wavelength of THz radiation and the angle troscopy and imaging technologies coupled with chemomet-
at the food sample (Qin, Ying, and Xie 2013). The rics have been demonstrated to efficiently mine the spectral
thickness of a high-water content product greater than data. Relationship between spectrum data and measured
1 mm was reported to be not suitable for THz spectros- parameters was clearly explained via multivariate analysis
copy. For instance, the spectra for thin slice of skin and with a high accuracy prediction. In the future, more robust
lean tissue samples were very weak. As stated above, the models need to be developed to qualitatively and quantita-
foreign bodies in the flour were not easily identified tively detect food composition. Future studies focused on
when they were hidden in a deeper place (above establishing a library for the wide group of compounds rele-
10 mm) (Jiang, Ge, and Zhang 2019). vant to food safety are also demanded for food process
3. Cost limitation: the high cost of installation also frus- monitoring and control.
trates THz extensive application. Economically, it costs As a rapid, nondestructive, innovative and promising
around 119,000-U.S. dollar for a commercial equipment, technique, THz spectroscopy and imaging technologies pos-
which may not be affordable for a small-scale industry. sess the potential to be applied to food field. More research
Developing a cost-effective and portable THz spectrom- is needed to elucidate the quality evolution during storage
eter with a high sensitivity may be a direction for the using THz technique.
future research tendency.
4. Scanning speed and spatial resolution: THz imaging,
which provides spectral information of interior pack- Acknowledgements
aged materials, is also one of the important THz tech- The authors would also like to thank Professor Juan Francisco Garc
ıa-
nologies. Spatial resolution and speed are the most Mart
ın from Universidad de Sevilla for thorough proofreading of the
16 C.-H. FENG AND C. OTANI

manuscript. The authors also appreciate Schmuttenmaer research group Conference on Bioinformatics and Biomedical Engineering, Shanghai,
for sharing their figures and images and would also like to thank the 47–9.
anonymous reviewers for their constructive comments. Cheng, J. H., and D.-W. Sun. 2015. Rapid and non-invasive detection
of fish microbial spoilage by visible and near infrared hyperspectral
imaging and multivariate analysis. LWT - Food Science and
Funding Technology 62 (2):1060–8. doi: 10.1016/j.lwt.2015.01.021.
Cheng, W. W., D.-W. Sun, and J. H. Cheng. 2016. Pork biogenic amine
Chao-Hui Feng wishes to thank for the financial support of her index (BAI) determination based on chemometric analysis of hyper-
research work under Special Postdoctoral Researcher Program at spectral imaging data. LWT- Food Science and Technology 73:13–9.
Riken, “Exploratory Collaboration Seed” in FY2020 “Collaboration doi: 10.1016/j.lwt.2016.05.031.
Seed Fund” at Riken, and JSPS Grant-in-Aid for Early-Career Scientists Cheng, L. N., D.-W. Sun, Z. W. Zhu, and Z. H. Zhang. 2017. Effects of
(No. 20K15477). high pressure freezing (HPF) on denaturation of natural actomyosin
extracted from prawn (Metapenaeus ensis). Food Chemistry 229:
252–9. doi: 10.1016/j.foodchem.2017.02.048.
Chen, T., Z. Li, X. H. Yin, F. R. Hu, and C. Hu. 2016. Discrimination
ORCID of genetically modified sugar beets based on terahertz spectroscopy.
Spectrochimica Acta. Part A, Molecular and Biomolecular
Chao-Hui Feng http://orcid.org/0000-0001-8372-196X Spectroscopy 153:586–90. doi: 10.1016/j.saa.2015.09.028.
Chiko Otani http://orcid.org/0000-0002-9406-2602 Chen, Z. W., Z. Y. Zhang, R. H. Zhu, Y. H. Xiang, Y. P. Yang, and
P. B. Harrington. 2015. Application of terahertz time-domain spec-
troscopy combined with chemometrics to quantitative analysis of
References imidacloprid in rice samples. Journal of Quantitative Spectroscopy
and Radiative Transfer 167:1–9. doi: 10.1016/j.jqsrt.2015.07.018.
Abdul-Munaim, M. A., M. Reuter, O. M. Abdulmunem, J. C. Balzer, Cheville, R. A., and D. Grischkowsky. 1995. Far-infrared terahertz
M. Koch, and D. G. Watson. 2016. Using terahertz time-domain time-domain spectroscopy of flames. Optics Letters 20 (15):1646–8.
spectroscopy to discriminate among water contamination levels in doi: 10.1364/ol.20.001646.
diesel engine oil. Transactions of the ASABE 59:795–801. Deng, X. Y., L. Y. Li, M. Enomoto, and Y. Kawano. 2019.
Ahmad, M. H., M. Nache, S. Waffenschmidt, and B. Hizmann. 2016. Continuously frequency-tuneable plasmonic structures for terahertz
Characterization of farinographic kneading process for different bio-sensing and spectroscopy. Scientific Reports 9 (1):3498doi: 10.
types of wheat flours using fluorescence spectroscopy and chemo- 1038/s41598-019-39015-6.
metrics. Food Control 66:44–52. Dhillon, S. S., M. S. Vitiello, E. H. Linfield, A. G. Davies, M. C.
Ajito, K. 2015. Terahertz spectroscopy for pharmaceutical and biomed- Hoffmann, J. Booske, C. Paoloni, M. Gensch, P. Weightman, G. P.
ical applications. IEEE Transactions on Terahertz Science and Williams, et al. 2017. The 2017 terahertz science and technology
Technology 5:1140–5. roadmap. Journal of Physics D: Applied Physics 50 (4):043001. doi:
Ando, Y., S. Hagiwara, H. Nabetani, I. Sotome, T. Okunishi, H. 10.1088/1361-6463/50/4/043001.
Okadome, T. Orikasa, and A. Tagawa. 2019. Improvements of dry- Dobroiu, A., C. Otani, and K. Kawase. 2006. Terahertz-wave sources
ing rate and structural quality of microwave-vacuum dried carrot by and imaging applications. Measurement Science and Technology 17
freeze-thaw pretreatment. LWT 100:294–9. doi: 10.1016/j.lwt.2018. (11):R161–R174. doi: 10.1088/0957-0233/17/11/R01.
10.064. Du, C.-M., X. Zhang, and Z. Y. Zhang. 2019. Quantitative analysis of
Arikawa, T., M. Nagai, and K. Tanaka. 2008. Characterizing hydration ternary isomer mixtures of saccharide by terahertz time domain
state in solution using terahertz time-domain attenuated total reflec- spectroscopy combined with chemometrics. Vibrational Spectroscopy
tion spectroscopy. Chemical Physics Letters 457 (1–3):12–7. doi: 10. 100:64–70. doi: 10.1016/j.vibspec.2018.11.003.
1016/j.cplett.2008.03.062. Elmas, N. K., Ş. F. N. Arslan, G. Akin, A. Kenar, H.-G. Janssen, and I.
Ariyoshi, S., C. Otani, A. Dobroiu, H. Sato, K. Kawase, H. M. Shimizu, Yilmaz. 2019. Synchronous fluorescence spectroscopy combined
T. Taino, and H. Matsuo. 2006. Terahertz imaging with a direct with chemometrics for rapid assessment of cold-pressed grape seed
detector based on superconducting tunnel junctions. Applied Physics oil adulteration: Qualitative and quantitative study. Talanta 196:
Letters 88 (20):203503–1. 3. doi: 10.1063/1.2204842. 22–31. doi: 10.1016/j.talanta.2018.12.026.
Baek, S. H., H. B. Lim, and H. S. Chun. 2014. Detection of melamine Federici, J. F., R. L. Wample, D. Rodriguez, and S. Mukherjee. 2009.
in foods using terahertz time-domain spectroscopy. Journal of Application of terahertz Gouy phase shift from curved surfaces for
estimation of crop yield. Applied Optics 48 (7):1382–8. doi: 10.1364/
Agricultural and Food Chemistry 62 (24):5403–7. doi: 10.1021/
ao.48.001382.
jf501170z.
Feng, C. H., L. Drummond, and D.-W. Sun. 2014. Modeling the
Baxter, J. B., and G. W. Guglietta. 2011. Terahertz spectroscopy.
growth parameters of lactic acid bacteria and total viable count in
Analytical Chemistry 83 (12):4342–68. doi: 10.1021/ac200907z.
vacuum-packaged Irish cooked sausages cooled by different meth-
Bilbao-Sainz, C., A. Sinrod, M. J. Powell-Palm, L. Dao, G. Takeoka, T.
ods. International Journal of Food Science & Technology 49 (12):
Williams, D. Wood, G. Ukpai, J. Aruda, D. F. Bridges, et al. 2019.
2659–67. doi: 10.1111/ijfs.12603.
Preservation of sweet cherry by isochoric (constant volume) freez- Feng, C. H., L. Drummond, D.-W. Sun, and Z. H. Zhang. 2014.
ing. Innovative Food Science & Emerging Technologies 52:108–15. Evaluation of natural hog casings modified by surfactant solutions
doi: 10.1016/j.ifset.2018.10.016. combined with lactic acid by response surface methodology. LWT -
Bowman, T., M. El-Shenawee, and L. K. Campbell. 2016. Terahertz Food Science and Technology 58 (2):427–38. doi: 10.1016/j.lwt.2014.
transmission vs reflection imaging and model-based characterization 03.012.
for excised breast carcinomas. Biomedical Optics Express 7 (9): Feng, C. H., L. Drummond, Z. H. Zhang, and D.-W. Sun. 2013. Effects
3756–83. doi: 10.1364/BOE.7.003756. of processing parameters on immersion vacuum cooling time and
Brucherseifer, M., M. Nagel, P. Haring Bolivar, H. Kurz, A. Bosserhoff, physico-chemical properties of pork hams. Meat Science 95 (2):
and R. B€ uttner. 2000. Label-free probing of binding state of DNA by 425–32. doi: 10.1016/j.meatsci.2013.04.057.
time-domain terahertz sensing. Applied Physics Letters 77 (24): Feng, C. H., L. Drummond, Z. H. Zhang, and D.-W. Sun. 2014.
4049–51. doi: 10.1063/1.1332415. Evaluation of innovative immersion vacuum cooling with different
Cao, B.-H., G.-X. Zhang, and Z. K. Zhou. 2008. Far-infrared vibrational pressure reduction rates and agitation for cooked sausages stuffed in
spectra of L-ascorbic acid investigated by terahertz time domain natural or artificial casing. LWT - Food Science and Technology 59
spectroscopy. In The Proceedings of the Second International (1):77–85. doi: 10.1016/j.lwt.2014.04.035.
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 17

Feng, C.-H., J. F. Garc
ıa Mart
ın, C. Li, B.-L. Liu, X.-Y. Song, Q.-L. Garc
ıa-Mart
ın, J. F. 2015. Optical path length and wavelength selection
Dong, W. Wang, and Y. Yang. 2016. Evaluation of physicochemical using Vis/NIR spectroscopy for olive oil’s free acidity determination.
properties and microbial attributes of cooked sausages stuffed in International Journal of Food Science and Technology 50:1461–7.
casing modified by surfactants and lactic acid after immersion vac-

Garc
ıa-Mart
ın, J. F., F. J. Al
es-Alvarez, M. d C. L
opez-Barrera, I.
uum cooling and long-term storage. International Journal of Food

Mart
ın-Dom
ınguez, and P. Alvarez-Mateos. 2019. Cetane number
Science & Technology 51 (10):2270–9. doi: 10.1111/ijfs.13224. prediction of waste cooking oil-derived biodiesel prior to transesteri-
Feng, C. H., and C. Li. 2015. Immersion vacuum-cooling as a novel fication reaction using near infrared spectroscopy. Fuel 240:10–5.
technique for cooling meat products: Research advances and current doi: 10.1016/j.fuel.2018.11.142.
state-of-the art. Comprehensive Reviews in Food Science and Food Garc
ıa-Mart
ın, J. F., J. C. Ruiz, M. T. Garc
ıa, C. H. Feng, and P.
Safety 14 (6):785–95. doi: 10.1111/1541-4337.12157.

Alvarez-Mateos. 2019. Esterification of free fatty acids with glycerol
Feng, C.-H., C. Li, J. F. Garc
ıa-Mart
ın, P. K. Malakar, Y. Yan, Y.-W. within the biodiesel production framework. Processes 7 (11):832–10.
Liu, W. Wang, Y.-T. Liu, and Y. Yang. 2017. Physical properties doi: 10.3390/pr7110832.
and volatile composition changes of cooked sausages stuffed in a Garc
ıa-Mu~ noz, E., K. A. Abdalmalak, G. Santamar
ıa, A. Rivera-Lavado,
new casing formulation based in surfactants and lactic acid during D. Segovia-Vargas, P. Castillo-Aran
ıbar, F. Van Dijk, T. Nagatsuma,
long-term storage. Journal of Food Science 82 (3):594–604. doi: 10. E. R. Brown, R. C. Guzman, et al. 2019. Photonic-based integrated
1111/1750-3841.13641. sources and antenna arrays for broadband wireless links in terahertz
Feng, C. H., Y. W. Liu, Y. Makino, J. F. Garc
ıa-Mart
ın, and E. communications. Semiconductor Science and Technology 34 (5):
Cummins. 2017. Evaluation of modified casings and chitosan-PVA 054001., doi: 10.1088/1361-6641/aaf8f2.
packaging on the physicochemical properties of cooked Sichuan sau- Ge, H. Y., Y. Y. Jiang, F. Y. Lian, Y. Zhang, and S. H. Xia. 2015.
sages during long term storage. International Journal of Food Science Characterization of wheat varieties using terahertz time-domain
& Technology 52 (8):1777–88. doi: 10.1111/ijfs.13451. spectroscopy. Sensors (Basel, Switzerland) 15 (6):12560–72. doi: 10.
Feng, C. H., and Y. Makino. 2020. Colour analysis in sausages stuffed 3390/s150612560.
in modified casings with different storage days using hyperspectral Ge, H., Y. Jiang, Z. Xu, F. Lian, Y. Zhang, and S. Xia. 2014.
imaging – A feasibility study. Food Control: 107047. https://doi.org/ Identification of wheat quality using THz spectrum. Optics Express
10.1016/j.foodcont.2019.107047. 22 (10):12533–44. doi: 10.1364/OE.22.012533.
Feng, C. H., Y. Makino, S. Oshita, and J. F. Garc
ıa-Mart
ın. 2018. Gowen, A. A., O. C. O’Sullivan, and C. P. Donnell. 2012. Terahertz
Hyperspectral imaging and multispectral imaging as the novel tech- time domain spectroscopy and imaging: Emerging techniques for
niques for detecting defects in raw and processed meat products: food process monitoring and quality control. Trends in Food Science
Current state-of-the-art research advances. Food Control 84:165–76.
& Technology 25 (1):40–6. doi: 10.1016/j.tifs.2011.12.006.
doi: 10.1016/j.foodcont.2017.07.013.
Guillet, J. P., B. Recur, L. Frederique, B. Bousquet, L. Canioni, I.
Feng, C. H., Y. Makino, M. Yoshimura, and F. J. Rodr
ıguez-Pulido.
Manek-H€ onninger, P. Desbarats, and P. Mounaix. 2014. Review of
2018a. Estimation of adenosine triphosphate content in ready-to-eat
terahertz tomography techniques. Journal of Infrared, Millimeter,
sausages with different storage days, using hyperspectral imaging
and Terahertz Waves 35 (4):382–411. doi: 10.1007/s10762-014-0057-
coupled with R statistics. Food Chemistry 264:419–26. doi: 10.1016/j.
0.
foodchem.2018.05.029.
Haddad, J. E., B. Bousquet, L. Canioni, and P. Mounaix. 2013. Review
Feng, C. H., Y. Makino, M. Yoshimura, and F. J. Rodr
ıguez-Pulido.
in terahertz spectral analysis. TRAC Trends in Analytical Chemistry
2018b. Real-time prediction of pre-cooked Japanese sausage color
44:98–105. doi: 10.1016/j.trac.2012.11.009.
with different storage days using hyperspectral imaging. Journal of
Han, P. Y., M. Tani, M. Usami, S. Kono, R. Kersting, and X. C. Zhang.
the Science of Food and Agriculture 98 (7):2564–72. doi: 10.1002/jsfa.
2001. A direct comparison between terahertz time-domain spectros-
8746.
Feng, C. H., Y. Makino, M. Yoshimura, D. C. Thuyet, and J. F. Garc
ıa- copy and far-infrared Fourier transform spectroscopy. Journal of
Mart
ın. 2018. Hyperspectral imaging in tandem with R statistics and Applied Physics 89 (4):2357–9. doi: 10.1063/1.1343522.
image processing for detection and Visualization of pH in Japanese He, M., A. K. Azad, S. Ye, and W. Zhang. 2006. Far-infrared signature
Big Sausages under different storage conditions. Journal of Food of animal tissues characterized by terahertz time-domain spectros-
Science 83 (2):358–66. doi: 10.1111/1750-3841.14024. copy. Optics Communications 259 (1):389–92. doi: 10.1016/j.optcom.
Feng, C. H., and D.-W. Sun. 2014. Optimization of immersion vacuum 2005.08.029.
cooling operation and quality of Irish cooked sausages by using Heshmat, B., G. M. Andrews, O. A. Naranjo-Montoya, E. Castro-
response surface methodology. International Journal of Food Science Camus, D. Ciceri, A. R. Sanchez, A. Allanore, A. A. Kmetz, S. L.
& Technology 49 (8):1850–8. doi: 10.1111/ijfs.12494. Eichmann, M. E. Poitzsch, et al. 2017. Terahertz scattering and
Feng, C. H., D.-W. Sun, J. F. Garc
ıa-Mart
ın, and Z. H. Zhang. 2013. water absorption for porosimetry. Optics Express 25 (22):27370–82.,
Effects of different cooling methods on shelf-life of cooked jumbo doi: 10.1364/OE.25.027370.
plain sausages. LWT - Food Science and Technology 54 (2):426–33. Hillger, P., J. Grzyb, R. Jain, and U. Pfeiffer. 2019. Terahertz imaging
doi: 10.1016/j.lwt.2013.05.033. and sensing applications with silicon-based technologies. IEEE
Feng, C. H., W. Wang, Y. Makino, J. F. Garc
ıa-Mart
ın, P. Alvarez- Transactions on Terahertz Science and Technology 9 (1):1–19. doi:
Mateos, and X.-Y. Song. 2019. Evaluation of storage time and tem- 10.1109/TTHZ.2018.2884852.
perature on physicochemical properties of immersion vacuum Hirori, H.,. K. Yamashita, M. Nagai, and K. Tanaka. 2004. Attenuated
cooled sausages stuffed in the innovative casings modified by surfac- total reflection spectroscopy in time domain using terahertz coher-
tants and lactic acid. Journal of Food Engineering 257:34–43. doi: 10. ent pulses. Japanese Journal of Applied Physics 43 (No. 10A):
1016/j.jfoodeng.2019.03.023. L1287–L1289. doi: 10.1143/JJAP.43.L1287.
Fukasawa, T., T. Sato, J. Watanabe, Y. Hama, W. Kunz, and R. Ho, I. C., X. Y. Guo, and X. C. Zhang. 2010. Design and performance
Buchner. 2005. Relation between dielectric and low-frequency of reflective terahertz air-biased-coherent-detection for time-domain
Raman spectra of hydrogen-bond liquids. Physical Review Letters 95 spectroscopy. Optics Express 18 (3):2872–83. doi: 10.1364/OE.18.
(19):197802–4. doi: 10.1103/PhysRevLett.95.197802. 002872.
Fukunaga, K., Y. Ogawa, S. I. Hayashi, and I. Hosako. 2007. Terahertz Hoshina, H., A. Hayashi, N. Miyoshi, F. Miyamaru, and C. Otani.
spectroscopy for art conservation. IEICE Electronics Express 4 (8): 2009. Terahertz pulsed imaging of frozen biological tissues. Applied
258–63. doi: 10.1587/elex.4.258. Physics Letters 94 (12):123901. doi: 10.1063/1.3106616.
Garc
ıa-Mart
ın, J. F., M. d C. L

opez Barrera, M. Torres Garc
ıa, Q.-A. Hu, B. B., and M. C. Nuss. 1995. Imaging with terahertz waves. Optics


Zhang, and P. Alvarez Mateos. 2019. Determination of the acidity of Letters 20 (16):1716–8. doi: 10.1364/ol.20.001716.
waste cooking oils by near infrared spectroscopy. Processes 7 (5): Huang, H.,. L. Liu, and M. O. Ngadi. 2017. Assessment of intramuscu-
304–7. doi: 10.3390/pr7050304. lar fat content of pork using NIR hyperspectral images of rib end.
18 C.-H. FENG AND C. OTANI

Journal of Food Engineering 193:29–41. doi: 10.1016/j.jfoodeng.2016. wood. Wood Science and Technology 32 (6):421–427. doi: 10.1007/
07.005. BF00702799.
Huang, H. C., Q. Liu, L. G. Zhu, Y. Zou, Z. H. Li, and Z. R. Li. 2019. Krumbholz, N., K. Gerlach, F. Rutz, M. Koch, R. Piesiewicz, T. K€ urner,
Dual-prism based terahertz time-domain attenuated total reflection and D. Mittleman. 2006. Omnidirectional terahertz mirrors: A key
spectroscopy and its application to characterise the hydration state element for future terahertz communication systems. Applied Physics
of L-threonine in solution. Optics Communications 437:133–8. doi: Letters 88 (20):202905. doi: 10.1063/1.2205727.
10.1016/j.optcom.2018.12.057. Latif, R. 2013. Chocolate/cocoa and human health: A review. The
International Agency for Research on Caner (IARC). 1993. Monographs Netherlands Journal of Medicine 71 (2):63–68.
on the evaluation of carcinogenic risks of human. Lyon, France: Leahy-Hoppa, M. R., M. J. Fitch, and R. Osiander. 2009. Terahertz
IARC Press. spectroscopy techniques for explosives detection. Analytical and
Jepsen, P. U., D. G. Cooke, and M. Koch. 2011. Terahertz spectroscopy Bioanalytical Chemistry 395 (2):247–257. doi: 10.1007/s00216-009-
and imaging–Modern techniques and applications. Laser & 2803-z.
Photonics Reviews 5 (1):124–66. doi: 10.1002/lpor.201000011. Lee, Y.-K., S.-W. Choi, S.-T. Han, D.-H. Woo, and H. S. Chun. 2012.
Ji, J., J. Jiang, J. Chen, F. Du, and P. Huang. 2018. Scattering reduction Detection of foreign bodies in foods using continuous wave tera-
of perfectly electric conductive cylinder by coating plasma and meta- hertz imaging. Journal of Food Protection 75 (1):179–183. doi: 10.
material. Optik 161:98–105. doi: 10.1016/j.ijleo.2018.02.033. 4315/0362-028X.JFP-11-181.
Jiang, Y. Y., H. Y. Ge, F. Y. Lian, Y. Zhang, and S. D. Xia. 2015. Li, J. S. 2010. Optical parameters of vegetable oil studied by terahertz
Discrimination of moldy wheat using terahertz imaging combined time-domain spectroscopy. Applied Spectroscopy 64:231–234.
with multivariate classification. RSC Advances 5 (114):93979–86. doi: Liaskos, C., S. Nie, A. Tsioliaridou, A. Pitsillides, S. Ioannidis, and I.
10.1039/C5RA15377H. Akyildiz. 2019. A novel communication paradigm for high capacity
Jiang, Y. Y., H. Y. Ge, and Y. Zhang. 2019. Detection of foreign bodies and security via programmable indoor wireless environments in
in grain with terahertz reflection imaging. Optik 181:1130–8. doi: 10. next generation wireless systems. Ad Hoc Networks 87:1–16. doi: 10.
1016/j.ijleo.2018.12.066. 1016/j.adhoc.2018.11.001.
Jiang, H. Z., S.-C. Yoon, H. Zhuang, W. Wang, Y. F. Li, and Y. Yang. Li, Y., X. Chen, F. Hu, D. Li, H. Teng, Q. Rong, W. Zhang, J. Han,
2019. Integration of spectral and textural features of visible and and H. Liang. 2019. Four resonators based high sensitive terahertz
near-infrared hyperspectral imaging for differentiating between nor- metamaterial biosensor used for measuring concentration of protein.
mal and white striping broiler breast meat. Spectrochimica Acta. Journal of Physics D: Applied Physics 52 (9):095105. doi: 10.1088/
Part A, Molecular and Biomolecular Spectroscopy 213:118–26. doi: 1361-6463/aaf7e9.
10.1016/j.saa.2019.01.052. Liu, J. 2017. Terahertz spectroscopy and chemometric tools for rapid
J€
ordens, C., and M. Koch. 2008. Detection of foreign bodies in choc- identification of adulterated dairy product. Optical and Quantum
olate with pulsed terahertz spectroscopy. Optical Engineering 47 (3): Electronics 49 (1):1–8. doi: 10.1007/s11082-016-0848-8.
037003. doi: 10.1117/1.2896597. Liu, J. J., and L. L. Fan. 2016. A novel ADPSO-SVM combined with
Jӧ rdens, C., F. Rutz, and M. Koch. 2006.. Quality assurance of choc- terahertz spectroscopy for recognition of transgenic organisms.
olate products with terahertz imaging. Paper presented at the Optik 127 (4):1957–1961. doi: 10.1016/j.ijleo.2015.11.124.
European conference on non-destructive testing. Liu, J. J., L. L. Fan, Y. M. Liu, L. L. Mao, and J. Q. Kan. 2019.
Kamruzzaman, M., Y. Makino, and S. Oshita. 2016a. Hyperspectral Application of terahertz spectroscopy and chemometrics for discrim-
imaging for real-time monitoring of water holding capacity in red ination of transgenic camellia oil. Spectrochimica Acta. Part A,
meat. LWT - Food Science and Technology 66:685–691. doi: 10.1016/ Molecular and Biomolecular Spectroscopy 206:165–169. doi: 10.1016/
j.lwt.2015.11.021. j.saa.2018.08.005.
Kamruzzaman, M., Y. Makino, and S. Oshita. 2016b. Parsimonious Liu, J. J., and J. Q. Kan. 2018. Recognition of genetically modified
model development for real-time monitoring of moisture in red product based on affinity propagation clustering and terahertz spec-
meat using hyperspectral imaging. Food Chemistry 196:1084–1091. troscopy. Spectrochimica Acta. Part A, Molecular and Biomolecular
doi: 10.1016/j.foodchem.2015.10.051. Spectroscopy 194:14–20. doi: 10.1016/j.saa.2017.12.074.
Kamruzzaman, M., Y. Makino, and S. Oshita. 2016c. Online monitor- Liu, J. J., and Z. Li. 2014. The terahertz spectrum detection of trans-
ing of red meat color using hyperspectral imaging. Meat Science genic food. Optik 125 (23):6867–6869. doi: 10.1016/j.ijleo.2014.08.
116:110–117. doi: 10.1016/j.meatsci.2016.02.004. 114.
Karacaglar, N. N. Y., T. Bulat, I. H. Boyaci, and A. Topcu. 2019. Liu, W., C. H. Liu, F. Chen, J. B. Yang, and L. Zheng. 2016.
Raman spectroscopy coupled with chemometric methods for the dis- Discrimination of transgenic soybean seeds by terahertz spectros-
crimination of foreign fats and oils in cream and yogurt. Journal of copy. Scientific Reports 6:35799doi: 10.1038/srep35799.
Food and Drug Analysis 27 (1):101–110. doi: 10.1016/j.jfda.2018.06. Liu, W., C. H. Liu, X. H. Hu, J. B. Yang, and L. Zheng. 2016.
008. Application of terahertz spectroscopy imaging for discrimination of
Karali unas, M., K. E. Nasser, A. Urbanowicz, I. Kasalynas, D. transgenic rice seeds with chemometrics. Food Chemistry 210:
Brazinskien_e, S. Asadauskas, and G. Valusis. 2018. Non-destructive 415–421. doi: 10.1016/j.foodchem.2016.04.117.
inspection of food and technical oils by terahertz spectroscopy. Liu, Y., H. Liu, M. Q. Tang, J. Q. Huang, W. Liu, J. Y. Dong, … Y.
Scientific Reports 8 (1):18025. doi: 10.1038/s41598-018-36151-3. Zhang. 2019. The medical application of terahertz technology in
Kawase, K. 2004. Terahertz imaging for drug detection and large-scale non-invasive detection of cells and tissues: Opportunities and chal-
integrated circuit inspection. Optics and Photonics News 15 (10): lenges. RSC Advances 9:9354.
34–39. doi: 10.1364/OPN.15.10.000034. Liu, W., C. H. Liu, J. J. Yu, Y. Zhang, J. Li, Y. Chen, and L. Zheng.
Kawase, K.,. Y. Ogawa, Y. Watanabe, and H. Inoue. 2003. Non-destruc- 2018. Discrimination of geographical origin of extra virgin olive oils
tive terahertz imaging of illicit drugs using spectral fingerprints. using terahertz spectroscopy combined with chemometrics. Food
Optics Express 11 (20):2549–2554. doi: 10.1364/oe.11.002549. Chemistry 251:86–92. doi: 10.1016/j.foodchem.2018.01.081.
Kiwa, T., M. Tonouchi, M. Yamashita, and K. Kawase. 2003. Laser ter- Liu, J. J., L. L. Mao, J. F. Ku, H. J. Peng, Z. F. Lao, D. Chen, and B. H.
ahertz-emission microscope for inspecting electrical faults in inte- Yang. 2017. Using terahertz spectroscopy to identify transgenic cot-
grated circuits. Optics Letters 28 (21):2058–2060. doi: 10.1364/ol.28. tonseed oil according to physicochemical quality parameters. Optik
002058. 142:483–488. doi: 10.1016/j.ijleo.2017.05.103.
Koch, M. 2007. Terahertz technology: A land to be discovered. Optics Liu, W., P. G. Zhao, C. S. Wu, C. H. Liu, J. B. Yang, and L. Zheng.
and Photonics News 18 (3):20–25. doi: 10.1364/OPN.18.3.000020. 2019. Rapid determination of aflatoxin B1 concentration in soybean
Koch, M., S. Hunsche, P. Schumacher, M. C. Nuss, J. Feldmann, and J. oil using terahertz spectroscopy with chemometric methods. Food
Fromm. 1998. THz-imaging: A new method for density mapping of Chemistry 293:213–219. doi: 10.1016/j.foodchem.2019.04.081.
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 19

Liu, H.-B., H. Zhong, N. Karpowicz, Y. Q. Chen, and X.-C. Zhang. Nunes, K. M., M. V. O. Andrade, M. R. Almeida, C. Fantini, and
2007. Terahertz spectroscopy and imaging for defense and security M. M. Sena. 2019. Raman spectroscopy and discriminant analysis
applications. Proceedings of the IEEE 95 (8):1514–1527. doi: 10.1109/ applied to the detection of frauds in bovine meat by the addition of
JPROC.2007.898903. salts and carrageenan. Microchemical Journal 147:582–589. doi: 10.
Li, H., J. Z. Wu, C. L. Liu, X. R. Sun, and L. Yu. 2018. Study on pre- 1016/j.microc.2019.03.076.
treatment methods of terahertz time domain spectral image for Ogawa, Y., S. Hayashi, N. Kondo, K. Ninomiya, C. Otani, and K.
maize seeds. IFAC-PapersOnLine 51:206–210. Kawase. 2006. Feasibility on the quality evaluation of agricultural.
Lu, S., X. Zhang, Z. Zhang, Y. Yang, and Y. Xiang. 2016. Quantitative products with terahertz electromagnetic wave. American Society of
measurements of binary amino acids mixtures in yellow foxtail mil- Agricultural and Biological Engineering. In Annual International
let by terahertz time domain spectroscopy. Food Chemistry 211: Meeting Oregon Convention Center. St. Joseph, MI: ASABE.
494–501. doi: 10.1016/j.foodchem.2016.05.079. Ok, G., H. J. Kim, H. S. Chun, and S.-W. Choi. 2014. Foreign-body
Luna, A. S., A. P. Silva, C. S. Silva, I. C. A. Lima, and J. S. Gois. 2019. detection in dry food using continuous sub-terahertz wave imaging.
Chemometric methods for classification of clonal varieties of green Food Control 42:284–289. doi: 10.1016/j.foodcont.2014.02.021.
coffee using Raman spectroscopy and direct sample analysis. Journal Ok, G., K. Park, H. S. Chun, H. J. Chang, N. Lee, and S.-W. Choi.
of Food Composition and Analysis 76:44–50. doi: 10.1016/j.jfca.2018. 2015. High-performance sub-terahertz transmission imaging system
12.001. for food inspection. Biomedical Optics Express 6 (5):1929–1941. doi:
Markelz, A. G., J. R. Knab, J. Y. Chen, and Y. He. 2007. Protein 10.1364/BOE.6.001929.
dunamical transition in terahertz dielectric response. Chemical Ok, G., K. Park, M.-C. Lim, H.-J. Jang, and S.-W. Choi. 2018. 140-GHz
Physics Letters 442 (4-6):413–417. doi: 10.1016/j.cplett.2007.05.080. subwavelength transmission imaging for foreign body inspection in
Markelz, A. G., A. Roitberg, and E. J. Heilweil. 2000. Pulsed terahertz food products. Journal of Food Engineering 221:124–131. doi: 10.
spectroscopy of DNA, bovine serum albumin and collagen between 1016/j.jfoodeng.2017.10.011.
0.1 and 2.0 THz. Chemical Physics Letters 320 (1-2):42–48. doi: 10. Ok, G., H. J. Shin, M.-C. Lim, and S.-W. Choi. 2019. Large-scan-area
1016/S0009-2614(00)00227-X. sub-terahertz imaging system for nondestructive food quality inspec-
Mathanker, S. K., P. R. Weckler, and N. Wang. 2013. Terahertz (THz) tion. Food Control 96:383–389. doi: 10.1016/j.foodcont.2018.09.035.
applications in food and agriculture: A review. Transations of the Pal, S. K., J. Peon, B. Bagchi, and A. H. Zewail. 2002. Biological water:
ASABE 56:1–12. Femtosecond dynamics of macromolecular hydration. The Journal of
McIntosh, A. I., B. Yang, S. M. Goldup, M. Watkinson, and R. Physical Chemistry B 106 (48):12376–12395. doi: 10.1021/jp0213506.
Donnan. 2012. Terahertz spectroscopy: A powerful new tool for the Pu, Y. Y., and D.-W. Sun. 2015. Vis-NIR hyperspectral imaging in vis-
chemical sciences? Chemical Society Reviews 41 (6):2072–2082. doi: ualizing moisture distribution of mango slices during microwave-
10.1039/c1cs15277g. vacuum drying. Food Chemistry 188:271–278. doi: 10.1016/j.food-
Meier, B. P., S. W. Noll, and O. J. Molokwu. 2017. The sweet life: The chem.2015.04.120.
effect of mindful chocolate consumption on mood. Appetite 108: Qin, B. Y., Z. Li, T. Chen, and Y. Chen. 2017. Identification of genetic-
21–27. doi: 10.1016/j.appet.2016.09.018. ally modified cotton seeds by terahertz spectroscopy with MPGA-
Mittleman, D. M., M. Gupta, R. Neelamani, R. G. Baraniuk, J. V. SVM. Optik 142:576–582. doi: 10.1016/j.ijleo.2017.06.030.
Rudd, and M. Koch. 1999. Recent advances in terahertz imaging. Qin, J. Y., L. J. Xie, and Y. B. Ying. 2014. Feasibility of terahertz time-
Applied Physics B: Lasers and Optics 68 (6):1085–1094. doi: 10.1007/ domain spectroscopy to detect tetracyclines hydrochloride in infant
s003400050750. milk powder. Analytical Chemistry 86 (23):11750–11757. doi: 10.
Molinari, E., and E. Callus. 2012. Psychological drivers of chocolate 1021/ac503212q.
consumption. In Chocolate and Health, eds. A. Conti, R. Paoletti, A. Qin, J. Y., L. J. Xie, and Y. B. Ying. 2015. Determination of tetracycline
Poli, and F. Visioli, 137–46. Milano: Springer. hydrochloride by terahertz spectroscopy with PLSR model. Food
Monago-Mara~ na, O., M. Guzm
an-Becerra, A. M. Pe~ na, and T. Chemistry 170:415–422. doi: 10.1016/j.foodchem.2014.08.050.
Galeano-D
ıaz. 2018. Determination of pungency in spicy food by Qin, J. Y., L. J. Xie, and Y. B. Ying. 2017. Rapid analysis of tetracycline
means of excitation-emission fluorescence coupled with second- hydrochloride solution by attenuated total reflection terahertz time-
order chemometric calibration. Journal of Food Composition and domain spectroscopy. Food Chemistry 224:262–269. doi: 10.1016/j.
Analysis 67:10–18. doi: 10.1016/j.jfca.2017.12.031. foodchem.2016.12.064.
Nagai, N., and R. Fukasawa. 2004. Abnormal dispersion of polymer Qin, J. Y., Y. B. Ying, and L. J. Xie. 2013. The detection of agricultural
films in the THz frequency region. Chemical Physics Letters 388 (4- products and food using terahertz spectroscopy: A review. Applied
6):479–482. doi: 10.1016/j.cplett.2004.03.044. Spectroscopy Reviews 48 (6):439–457. doi: 10.1080/05704928.2012.
Nagai, M., H. Yada, T. Arikawa, and K. Tanaka. 2006. Terahertz time- 745418.
domain attenuated total reflection spectroscopy in water and bio- Redo-Sanchez, A., N. Laman, B. Schulkin, and T. Tongue. 2013.
logical solution. International Journal of Infrared and Millimeter Review of terahertz technology readiness assessment and applica-
Waves 27 (4):505–515. doi: 10.1007/s10762-006-9098-3. tions. Journal of Infrared, Millimeter, and Terahertz Waves 34 (9):
Nakajima, S., H. Hoshina, M. Yamashita, C. Otani, and N. Miyoshi. 500–518. doi: 10.1007/s10762-013-9998-y.
2007. Terahertz imaging diagnostics of cancer tissues with a chemo- Redo-Sanchez, A., G. Salvatella, R. Galceran, E. Rold
os, J.-A. Garc
ıa-
metrics technique. Applied Physics Letters 90 (4):041102–1. doi: 10. Reguero, M. Castellari, and J. Tejada. 2011. Assessment of terahertz
1063/1.2433035. spectroscopy to detect antibiotic residues in food and feed matrices.
Nakajima, S., K. Shiraga, T. Suzuki, N. Kondo, and Y. Ogawa. 2019. The Analyst 136 (8):1733–1738. doi: 10.1039/c0an01016b.
Quantification of starch content in germinating mung bean seed- Ren, A., A. Zahid, D. Fan, X. Yang, M. A. Imran, A. Alomainy, and
lings by terahertz spectroscopy. Food Chemistry 294:203–208. doi: Q. H. Abbasi. 2019. State-of-the-art in terahertz sensing for food
10.1016/j.foodchem.2019.05.065. and water security – A comprehensive review. Trends in Food
Notake, T., R. Endo, K. Fukunaga, I. Hosako, C. Otani, and H. Science & Technology 85:241–251. doi: 10.1016/j.tifs.2019.01.019.
Minamide. 2014. State-of-the art database of terahertz spectroscopy Rogalski, A., and F. Sizov. 2011. Terahertz detectors and focal plane
based on modern web technology. IEEE Transactions on Terahertz arrays. Opto-Electronics Review 19 (3):346–404. doi: 10.2478/s11772-
Science and Technology 4 (1):110–115. doi: 10.1109/TTHZ.2013. 011-0033-3.
2284862. Schmuttenmaer Research Group. 2019a. Time-resolved THz spectros-
Notake, T., K. Kamata, T. Iyoda, C. Otani, and H. Minamide. 2019. copy. Accessed December 16, 2019. https://thz.yale.edu/technique/
Expression of various polarization effects by using Spirulina-tem- time-resolved-thz-spectroscopy
plated metal lcoils at the terahertz frequency region. Japanese Schmuttenmaer Research Group. 2019b. THz emission spectroscopy.
Journal of Applied Physics 58 (3):032007. doi: 10.7567/1347-4065/ Accessed December 16, 2019. https://thz.yale.edu/technique/thz-
aafca6. emission-spectroscopy
20 C.-H. FENG AND C. OTANI

Scholey, A., and L. Owen. 2013. Effects of chocolate on cognitive func- Walcutt, D. L. 2009. Chocolate and mood disorders. PsychCentral
tion and mood: A systematic review. Nutrition Reviews 71 (10): Accessed July 8, 2018. https://psychcentral.com/blog/chocolate-and-
665–681. doi: 10.1111/nure.12065. mood-disorders/.
Schuster, F., D. Coquillat, H. Videlier, M. Sakowicz, F. Teppe, L. Wallace, V. P., E. MacPherson, J. A. Zeitler, and C. Reid. 2008. Three-
Dussopt, B. Giffard, T. Skotnicki, and W. Knap. 2011. Broadband dimensional imaging of optically opaque materials using nonioniz-
terahertz imaging with highly sensitive silicon CMOS detectors. ing terahertz radiation. Journal of the Optical Society of America A,
Optics Express 19 (8):7827–7832. doi: 10.1364/OE.19.007827. Optics, Image Science, and Vision 25 (12):3120–3133. doi: 10.1364/
Shen, Y.-C. 2011. Terahertz pulsed spectroscopy and imaging for josaa.25.003120.
pharmaceutical applications: A review. International Journal of Walther, M., B. M. Fischer, and P. Uhd Jepsen. 2003. Noncovalent
Pharmaceutics 417 (1-2):48–60. doi: 10.1016/j.ijpharm.2011.01.012. intermolecular forces in polycrystalline and amorphous saccharides
Shin, H. J., S.-W. Choi, and G. Ok. 2018. Qualitative identification of food in the far infrared. Chemical Physics 288 (2-3):261–268. doi: 10.
materials by complex refractive index mapping in the terahertz range. 1016/S0301-0104(03)00031-4.
Food Chemistry 245:282–288. doi: 10.1016/j.foodchem.2017.10.056. Wang, J., C.-L. Law, P. K. Nema, J.-H. Zhao, Z.-L. Liu, L.-Z. Deng, Z.-
Shiraga, K., Y. Ogawa, and N. Kondo. 2016. Hydrogen bond network J. Gao, and H.-W. Xiao. 2018. Pulsed vacuum drying enhances dry-
of water around protein investigated with terahertz and infrared ing kinetics and quality of lemon slices. Journal of Food Engineering
spectroscopy. Biophysical Journal 111 (12):2629–2641. doi: 10.1016/j. 224:129–138. doi: 10.1016/j.jfoodeng.2018.01.002.
bpj.2016.11.011. Wang, K. Q., D.-W. Sun, and H. B. Pu. 2017. Emerging non-destruc-
Siripatrawan, U., and Y. Makino. 2018. Simultaneous assessment of tive terahertz spectroscopic imaging technique: Principle and appli-
various quality attributes and shelf life of packaged bratwurst using cations in the agri-food industry. Trends in Food Science &
hyperspectral imaging. Meat Science 146:26–33. doi: 10.1016/j. Technology 67:93–105. doi: 10.1016/j.tifs.2017.06.001.
meatsci.2018.06.024. Wang, Y. M., Q. Wang, Z. S. Zhao, A. F. Liu, Y. Tian, and J. Y. Qin.
Siripatrawan, U., Y. Makino, S. Kawagoe, and S. Oshita. 2011. Rapid 2018. Rapid qualitative and quantitative analysis of chlortetracycline
detection of Escherichia coli contamination in packaged fresh spin- hydrochloride and tetracycline hydrochloride in environmental sam-
ach using hyperspectral imaging. Talanta 85 (1):276–281. doi: 10. ples based on terahertz frequency-domain spectroscopy. Talanta
1016/j.talanta.2011.03.061. 190:284–291. doi: 10.1016/j.talanta.2018.08.008.
Sotome, I., M. Takenaka, S. Koseki, Y. Ogasawara, Y. Nadachi, H. Wang, T., H.-L. Wu, W. Long, J. Y. Hu, L. Cheng, A.-Q. Chen, and
Okadome, and S. Isobe. 2009. Blanching of potato with superheated R.-Q. Yu. 2019. Rapid identification and quantification of cheaper
steam and hot water spray. LWT - Food Science and Technology 42 vegetable oil adulteration in camellia oil by using excitation-emis-
(6):1035–1040. doi: 10.1016/j.lwt.2009.02.001. sion matrix fluorescence spectroscopy combined with chemometrics.
Stark, M. S., J. J. Wilkinson, J. R. Smith, A. Alfadhl, and B. A. Food Chemistry 293:348–357. doi: 10.1016/j.foodchem.2019.04.109.
Wang, C., R.-Y. Zhou, Y.-X. Huang, L. J. Xie, and Y. B. Ying. 2019.
Pochopien. 2011. Autooxidation of branched alkanes in the liquid
Terahertz spectroscopic imaging with discriminant analysis for
phase. Industrial & Engineering Chemistry Research 50 (2):817–823.
detecting foreign materials among sausages. Food Control 97:
doi: 10.1021/ie101695g.
100–104. doi: 10.1016/j.foodcont.2018.10.024.
Suhandy, D., T. Suzuki, Y. Ogawa, N. Kondo, H. Naito, T. Ishihara, Y.
Wei, X., W. Q. Zheng, S. P. Zhu, S. L. Zhou, W. Wu, and Z. Y. Xie.
Takemoto, and W. Liu. 2012. A quantitative study for determination
2020. Application of terahertz spectrum and interval partial least
of glucose concentration using attenuated total reflectance terahertz
squares method in the identification of genetically modified soy-
(ATR-THz) spectroscopy. Engineering in Agriculture, Environment
beans. Spectrochimica Acta Part A: Molecular and Biomolecular
and Food 5 (3):90–95. doi: 10.1016/S1881-8366(12)80020-6.
Spectroscopy 238:118453. doi: 10.1016/j.saa.2020.118453.
Suhandy, D., M. Yulia, Y. Ogawa, and N. Kondo. 2013. Prediction of
Wietzke, S., F. Rutz, C. Jӧrdens, N. Krumbholz, N. Vieweg, C. Jansen,
L-Ascorbic acid using FTIR-ATR terahertz spectroscopy combined
… M. Koch. 2008. Applications of terahertz spectroscopy in the
with interval partial least squares (iPLS) regression. Engineering in plastics industry. Proceedings of SPIE, vol. 6840, 1–9. 68400V-1-9.
Agriculture, Environment and Food 6 (3):111–117. doi: 10.1016/ Xiao, H., Bai, -W. J.-W. Xie, L. Sun, D.-W.. and Gao Z.-J. 2015. Thin-
S1881-8366(13)80020-1. layer air impingement drying enhances drying rate of American gin-
Sun, W., X. Wang, and Y. Zhang. 2012. A method to monitor the oil seng (Panax quinquefolium L.) slices with quality attributes consid-
pollution in water with reflective pulsed terahertz tomography. ered. Food and Bioproducts Processing 94:581–591. doi: 10.1016/j.fbp.
Optik 123 (21):1980–1984. doi: 10.1016/j.ijleo.2011.10.002. 2014.08.008.
Suzuki, T., Y. Ogawa, and N. Kondo. 2011. Characterization of pesti- Xu, W., L. Xie, J. Zhu, W. Wang, Z. Ye, Y. Ma, C.-Y. Tsai, S. Chen,
cide residue, cis-Permethrin by terahertz spectroscopy. Engineering and Y. Ying. 2017. Terahertz sensing of chlorpyrifos-methyl using
in Agriculture, Environment and Food 4 (4):90–94. doi: 10.1016/ metamaterials. Food Chemistry 218:330–334. doi: 10.1016/j.food-
S1881-8366(11)80007-8. chem.2016.09.032.
Taday, P. F., I. V. Bradley, D. D. Arnone, and M. Pepper. 2003. Using ter- Yada, H., M. Nagai, and K. Tanaka. 2008. Origin of the fast relaxation
ahertz pulse spectroscopy to study the crystalline structure of a drug: A component of water and heavy water revealed by terahertz time-
case study of the polymorphs of ranitidine hydrochloride. Journal of domain attenuated total reflection spectroscopy. Chemical Physics
Pharmaceutical Sciences 92 (4):831–838. doi: 10.1002/jps.10358. Letters 464 (4-6):166–170. doi: 10.1016/j.cplett.2008.09.015.
Teixeira, A. M., and C. Sousa. 2019. A review on the application of Yada, H., M. Nagai, and K. Tanaka. 2009. The intermolecular stretch-
vibrational spectroscopy to the chemistry of nuts. Food Chemistry ing vibration mode in water isotopes investigate with broadband ter-
277:713–724. doi: 10.1016/j.foodchem.2018.11.030. ahertz time-domain spectroscopy. Chemical Physics Letters 473 (4-6):
Tonouchi, M. 2007. Cutting-edge terahertz technology. Nature 279–283. doi: 10.1016/j.cplett.2009.03.075.
Photonics 1 (2):97–105. doi: 10.1038/nphoton.2007.3. Yamazaki, S., M. Harata, T. Idehara, K. Konagaya, G. Yokoyama, H.
Tonouchi, M., M. Yamashita, and M. Hangyo. 2000. Terahertz radi- Hoshina, and Y. Ogawa. 2018. Actin polymerization is activated by
ation imaging of supercurrent distribution in vortex-penetrated terahertz irradiation. Scientific Reports 8 (1):9990. doi: 10.1038/
YBa2Cu3O7d thin film strips. Journal of Applied Physics 87 (10): s41598-018-28245-9.
7366–7375. doi: 10.1063/1.372995. Yang, Q., D.-W. Sun, and W. W. Cheng. 2017. Development of simpli-
Veronese, N., J. Demurtas, S. Celotto, M. G. Caruso, S. Maggi, F. fied models for non-destructive hyperspectral imaging monitoring of
Bolzetta, J. Firth, L. Smith, P. Schofield, A. Koyanagi, et al. 2019. Is TVB-N contents in cured meat during drying process. Journal of
chocolate consumption associated with health outcomes? An Food Engineering 192:53–60. doi: 10.1016/j.jfoodeng.2016.07.015.
umbrella review of systematic reviews and meta-analyses. Clinical Yang, X., X. Zhao, K. Yang, Y. P. Liu, Y. Liu, W. L. Fu, and Y. Luo. 2016.
Nutrition (Edinburgh, Scotland) 38 (3):1101–1108. doi: 10.1016/j. Biomedical applications of terahertz spectroscopy and imaging. Trends
clnu.2018.05.019. in Biotechnology 34 (10):810–824. doi: 10.1016/j.tibtech.2016.04.008.
 CRITICAL REVIEWS IN FOOD SCIENCE AND NUTRITION 21

Ye, H. B., Y. H. Zhang, and W. Z. Shen. 2006. Carrier transport and Zhan, H. L., J. F. Xi, K. Zhao, R. Bao, and L. Z. Xiao. 2016. A spectral-
optical properties in GaAs far-infrared/terahertz mirror structures. mathematical strategy for the identification of edible and swill-
Thin Solid Films. 514 (1-2):310–315. doi: 10.1016/j.tsf.2006.03.009. cooked dirty oils using terahertz spectroscopy. Food Control 67:
Yin, M., S. Tang, and M. Tong. 2016a. The application of terahertz 114–118. doi: 10.1016/j.foodcont.2016.02.043.
spectroscopy to liquid petrochemicals detection: A review. Applied Zhang, H., Z. Li, T. Chen, and B. Y. Qin. 2017. Quantitative determin-
Spectroscopy Reviews 51 (5):379–396. doi: 10.1080/05704928.2016. ation of auramine O by terahertz spectroscopy with 2DCOS-
PLSR model. Spectrochimica Acta. Part A, Molecular and
1141291.
Biomolecular Spectroscopy 184:335–341. doi: 10.1016/j.saa.2017.05.
Yin, M., S. Tang, and M. Tong. 2016b. Identification of edible oils
017.
using terahertz spectroscopy combined with genetic algorithm and
Zhang, H., X. Xu, F. Fan, and S. Chang. 2015. Study on a tunable nar-
partial least squares discriminant analysis. Analytical Methods 8 (13): row-band filter based on magnetic defects in photonic crystal in the
2794–2798. doi: 10.1039/C6AY00259E. terahertz region. Optical Engineering 54 (4):047104. doi: 10.1117/1.
Zhai, H. L., B. Q. Li, J. Chen, X. Wang, M. L. Xu, J. J. Liu, and S. H. OE.54.4.047104.
Lu. 2018. Chemical image moments and their applications. TRAC Zhao, H. X., S. Liu, C. Q. Tian, G. Yan, and D. Wang. 2018. An over-
Trends in Analytical Chemistry 103:119–125. doi: 10.1016/j.trac.2018. view of current status of cold chain in China. International Journal
03.017. of Refrigeration 88:483–495. doi: 10.1016/j.ijrefrig.2018.02.024.