LWT - Food Science and Technology 87 (2018) 77e84

Contents lists available at ScienceDirect

LWT - Food Science and Technology

journal homepage: www.elsevier.com/locate/lwt

Evaluation of the freshness of fresh-cut green bell pepper (Capsicum

annuum var. grossum) using electronic nose
Hui-Zhi Chen a, Min Zhang a, b, *, Bhesh Bhandari c, Zhimei Guo d
a
State Key Laboratory of Food Science and Technology, Jiangnan University, 14122, Wuxi, Jiangsu, China
b
Jiangnan University (Yangzhou), Food Biotechnology Institute, Yangzhou, 225002, China
c
School of Agriculture and Food Sciences, University of Queensland, Brisbane, QLD, Australia
d
Delin Marine Equipment Corporation Limited, 214122, Wuxi, Jiangsu, China

articleinfo abstract

Article history: This study investigated a freshness assessment method utilizing electronic nose (e -nose) for fresh-cut
Received 16 April 2017 green bell pepper (Capsicum annuum var. grossum) stored at 7 ± 1 ○ C. Physicochemical results showed
Received in revised form
that the fresh-cut green bell peppers were fresh until 5 days and became spoiled at 7 days, particularly as
15 August 2017
evidenced by a surge in aerobic plate count and malondialdehyde content on subsequent days. The e -
Accepted 18 August 2017
nose data combined hierarchical cluster analysis (HCA) can preliminarily distinguish between fresh (days
Available online 19 August 2017
0, 1, 3 and 5) and spoiled (days 7 and 9) samples. Principal component analysis (PCA) result showed that
days 0 and 1 samples were mixed together in PCA plot, and the other different groups can be obtained
Keywords:
according to the different sampling days. Partial least squares (PLS) statistical model (R 2 ¼ 0.9783,
Fresh-cut vegetable
RMSE ¼ 0.3317) was used to correlate the e-nose data with the aerobic plate counts. The results sug-
Electronic nose
Freshness gested the promising possibility of e-nose system for monitoring freshness of fresh-cut green bell pepper.
Microbial spoilage © 2017 Elsevier Ltd. All rights reserved.
Shelf life

1. Introduction counts, texture, color, and sensory evaluation (Rodoni, Zaro,

Green bell pepper (Capsicum annuum var. grossum) is rich in
vitamins which is popular crops throughout the world for fresh
market consumption (Singh, Giri, & Kotwaliwale, 2014). A good
storage temperature for some peppers is 7e13 ○C, because some
cultivars are susceptible to chilling injury when stored below 7 ○C
(González-Aguilar, Ayala-Zavala, Ruiz-Cruz, Acedo-Félix, & Dıaz-
Cinco, 2004). Thanks to demands of convenience and healthiness,
sales of minimally processed vegetables with a fresh-like quality
are rapidly increasing. However, fresh-cut vegetables are highly
perishable and their quality deterioration usually occurs in a short
time. The storage life of fresh-cut pepper is limited by various
factors like different storage conditions, fresh-cut processes and
species which influence the freshness loss and spoilage pattern of
fresh-cut vegetables.
Traditional methods used to determine shelf life of fresh-cut
vegetable are based on chemical, microbiological, physical and
sensory indices, such as phenolic compounds, molds and bacterial
* Corresponding author. State Key Laboratory of Food Science and Technology,
Jiangnan University, 214122, Wuxi, China.
E-mail address: min@jiangnan.edu.cn (M. Zhang).
Hasperué, Concellón, & Vicente, 2015). Most traditional methods
are time-consuming and require skilled personnel. Moreover, gas
chromatography-mass spectrometry (GC-MS) has been applied to
analyze the volatile compounds of foods. However, it is neither
feasible to realize rapid diagnosis of volatile profiles.
Electronic nose (e-nose) is composed of a variety of gas sensors
that interact with odor molecules to generate electronic signals
called sensor responses. Then, the responses are collected by a
computer system and handled using multivariate data analysis
methods (Kiani, Minaei, & Ghasemi-Varnamkhasti, 2016a). Data
analysis methods of sensors’ response involve partial least squares
(PLS), cluster analysis (CA), principal component analysis (PCA),
linear discriminate analysis (LDA), functional discriminate analysis
(FDA), and so on which are considered as linear approaches, while
fuzzy logic, artificial neural network (ANN) and probabilistic neural
network (PNN) are based on nonlinear methods (Loutfi, Coradeschi,
Mani, Shankar, & Rayappan, 2015).
The use of e-nose is a promising method for the simple mea-
surement of maturity and other quality indicators of fruits and
vegetables. Application of e-nose to evaluate fruit ripening stage
during storage has been studied over the past years (Hernández
Gómez, Wang, Hu, & García Pereira, 2007), because e-nose

http://dx.doi.org/10.1016/j.lwt.2017.08.052
0023-6438/© 2017 Elsevier Ltd. All rights reserved.
78 H.-Z. Chen et al. / LWT - Food Science and Technology 87 (2018) 77e84
sensors are able to detect volatile aroma related to fruit ripening.
Metal oxide semiconductor (MOS) sensors, particularly SnO 2 sen-
sors, are able to detect ethylene which is heavily involved in
modulating the volatile emissions of fruits. Thus the MOS e-nose
can successfully distinguish two different stages of apricots matu-
rity after simulated shelf life storage (Defilippi et al., 2009). Lebrun,
Plotto, Goodner, Ducamp, and Baldwin (2008) used e-nose for
identifying volatiles of mango fruit in order to differentiate its
harvest maturity. Gómez, Wang, Hu, and Pereira (2008) found that
e-nose could assess tomato ripening stage during storage by PCA
and LDA, but showed poor prediction performance on fruit quality
factors like soluble solids content, pH and puncture force. More-
over, rapid diagnosis of microbial contamination of food products
by e-nose has been proposed. In a study of Gobbi et al. (2015), e-
nose with four MOS sensors was available for rapid diagnosis of
Enterobacteriaceae in vegetable soups. Results of Giovenzana,
Beghi, Buratti, Civelli, and Guidetti (2014)’s work demonstrated
that a commercial portable e-nose composed of 10 MOS sensors
was able to monitor freshness decay of fresh-cut Valerianella locusta
L. based on evolution of the aroma profile during storage.
Despite these encouraging examples, e-nose application for
freshness assessment of fresh-cut vegetables is still scarce. There-
fore, it will be beneficial to develop reliable methods to assess
freshness of fresh-cut vegetables and to establish evaluation
criteria of their quality. The objective of this study is to establish an
easy-to-use method based on e-nose for freshness assessment of
fresh-cut green bell pepper.
2. Materials and methods
2.1. Vegetable materials
Physiologically mature green bell peppers (Capsicum annuum
var. grossum) were purchased in located market, Wuxi, China.
Peppers without damage and defects were washed, cut in pieces of
2.2 ± 0.3 cm × 8.0 ± 0.5 cm, dipped in 0.1 mL/L sodium hypochlorite
solution for 5 min, washed again with water, and weighted within
150 ± 10 g. Then they were packaged in circular polyethylene
terephthalate (PET) trays and capsed (15.3 cm × 5.3 cm). Forty-
eight trays containing samples were stored at 7 ± 1 ○C.
2.2. Determinations
Samples were analyzed for traditional physicochemical indices
and e-nose responses to green bell pepper samples at different
storage periods (0, 1, 3, 5, 7 and 9 days). Eight trays with samples
were prepared for each point of determinations. Aerobic plate
count and e-nose test were conducted in at least eight repeats from
these eight trays and other traditional physicochemical experi-
ments were conducted in at least three repeats randomly from
three different trays.
2.2.1. Weight loss
Physiological loss in weight is leading to shriveling appearance
for fruits and vegetables. Weight loss was measured by a laboratory
level weighting scale (Precision Balance XPE303S, Mettler-Toledo
International Inc., Switzerland) at different storage period. The
weight loss can be expressed on a wet weight basis (g/100 g) by the
difference in initial and final weights of the sample (Singh et al.,
2014).
2.2.2. Chlorophyll content
Chlorophyll content is an important indicator for green vege-
tables, which is related to their color attribute. Determination of
chlorophyll content was adapted from Sgroppo and Pereyra (2009).
Four g of samples was homogenized with 100 mL of 0.8 L/L acetone
(mixing 4 L acetone and 1 L distilled water) and filtered through
filter paper. The absorbance of filtrate was determined using
spectrophotometer (UV2600, TECHCOMP, China) at 645 nm and
663 nm. Results are calculated using Eq. (1).
Chlorophyll content (mg/g) ¼ (20.21A645þ8.02A663) × V × N/
(W × 1000) (1)
Where A645 and A663 are the absorbance values at 645 nm and
663 nm, respectively. V is the volume of the extract solution (mL), N
is the dilution factor, W is the mass of the sample (g).
2.2.3. Aerobic plate count
Microbiological analysis is also one of important indexes for
judging the quality of fresh-cut vegetables. Microbiological exam-
ination was determined by aerobic plate count according to Na-
tional Standard of the People's Republic of China (GB 4789.2e2010).
Twenty-five g of samples were put in sterile stomacher bags and
homogenized for 2 min in 225 mL of 0.85 g/100 mL aseptic physi-
ological saline. A series of decimal dilutions were prepared and
spread over agar plates. The experiment was carried out in a clean
bench. Colonies on the plates were counted after incubation for
48 ± 2 h at 36 ± 1 ○C. The data were recorded as colony-forming
units (CFU) and expressed as log 10 CFU/g.
2.2.4. Malondialdehyde (MDA) content
MDA content is considered as an indicator to assess the degree
of plant oxidative stress. Five g of samples were homogenized with
50 mL of 10 g/100 mL trichloroacetic acid and centrifuged (2-16 PK,
Sigma Laboratory Centrifuges, Germany) for 10 min at 1700 g. Four
mL of the supernatant was mixed with equal volumes of 0.6 g/
100 mL 2-thiobarbituric acid (dissolved in 10 g/100 mL trichloro-
acetic acid). The solution was put into boiling water bath for 15 min
and then centrifuged for 10 min at 1700 g after rapidly cooling
down. The absorbance was measured at 450 nm, 532 nm and
600 nm (Hodges, DeLong, Forney, & Prange, 1999). MDA content
was calculated using Eq. (2).
MDA content (nmol/g) ¼ (6.45 × (A532—A600)—
0.56 × A450) × V1 × V3/(V2 × W) (2)
Where A450, A532 and A600 are the absorbance values at 450 nm,
532 nm and 600 nm, respectively. V1 is the total volume of the
solution obtained after reaction (mL), V2 is the volume of the
extract solution using for reaction (mL), V3 is the total volume of the
extract solution (mL), W is the mass of the sample (g).
2.2.5. Membrane permeability
Membrane permeability, an indicator of quantification of plant
cellular disruption, is also correlated to the quality of vegetables
and fruits during storage (Gonzalez, Anthon, & Barrett, 2010). The
samples were cut into 5 mm ×5 mm pieces and 10 g of samples
were rinsed with deionized water. After dehydrating the surface
water with filter paper, the sample were put into breaker con-
taining 50 mL of deionized water and kept in 30 ○C water bath for
1 h. Then the electric conductivity value (A 0) of deionized water
was measured using a conductivity meter (DS-11AT, Shanghai
Precision Science Instrument Co., Ltd, China). The electric conduc-
tivity value (A1) was determined again after heating and boiling the
water with samples for 15 min. Tests were carried out in three
repeats. Membrane permeability was calculated using Eq. (3)
(Meng, Zhang, & Adhikari, 2012).
Membrane permeability (%)¼(A0/A1) × 100% (3)
 H.-Z. Chen et al. / LWT - Food Science and Technology 87 (2018) 77e84 79

2.2.6. Electronic nose

The characteristic aroma of vegetables is important to their
sensory acceptance by consumers. Determination of flavor volatiles
using e-nose was adapted from Liu, Zhang, and Wang (2016). A
commercial e-nose instrument (iNose; Ruifen Trading Co.,
Shanghai, China) was used to obtain the information from the head
space of fresh-cut green bell pepper samples. This e-nose consisted
of 14 MOS gas sensors and its main applications is presented in
Table 1. This table shows current known or specified reaction (Qiu,
Wang, & Gao, 2015). At ambient temperature (25 ± 1 ○C), static
head space was generated before analysis in a 50 mL vial using 3.5 g
of each sample for 30 min. The gas path was cleaned by reference
air for 30 min before the experiment in order to normalize sensor
signals. The measurement phase lasted for 150 s, which was
enough for sensors to stabilize. Cleaning time (recovery time) for
sensors was 300 s. The gas was transferred to the sensors chamber Fig. 1. Changes in weight (-) and chlorophyll content (C) of fresh-cut green bell
at a flow rate of 1 L/min. pepper during storage at 7 ± 1 ○ C.
Values are the mean of three replicates and bars stand for standard deviation.

2.3. Statistical analysis

Data were analyzed by One-Way ANOVA using SPSS 16.0 (SPSS
Inc., Chicago, IL, USA). Means and standard deviation were calcu-
lated using Tukey's multiple range tests at the 5% significance level.
Hierarchical cluster analysis (HCA) and principal component anal-
ysis (PCA) were also carried out with the SPSS 16.0. Partial Least
Square (PLS) was conducted by OriginPro 9.1 (OriginLab Corpora-
tion, Northampton, MA, USA).
3. Results and discussion
3.1. Weight loss and chlorophyll content
There is a limit of wilting and shriveling appearance for fresh
vegetables and fruits. Generally, the weight loss should be less than
5% during the storage for whole bell peppers (González & Tiznado,
1993). In this work, physiological loss in weight significantly
increased at the end of the storage in Fig. 1. However, the samples
only underwent 0.60 ± 0.03% loss of initial weight, which was
below the acceptable limit of 2% reported by Manolopoulou,
Xanthopoulos, Douros, and Lambrinos (2010) for quality criteria
of green bell peppers. The reason for low weight loss may be owing
to the good barrier property of PET trays and caps for fresh-cut
peppers. Rodoni, Vicente, Azevedo, Concellón, and Cunha (2015)
also found that mass loss of fresh-cut green pepper strips was
less than 1% under 5 kPa O2 and 5 kPa CO2 controlled atmospheres
storage at 5 ○ C. Conesa, Verlinden, Artés-Hernández, Nicolaï, and
Artés (2007) also reported that the weight loss of fresh-cut bell
peppers under high O2 and low O2 levels stored up to 3 days at 7 ○C
was less than 0.4%.
Chlorophyll is the major pigment in green bell peppers, which
influences the color of peppers. Pigment evolution of pepper during
storage is related to its physiological state and other pre-harvest
and pre-treatment factors. As can be seen in Fig. 1, the chloro-
phyll content of peppers gradually decreased throughout storage.
The original chlorophyll content of sample was 0.72 ± 0.05 mg/g at
the beginning of storage and it went down to 0.44 ± 0.04 mg/g at
the end of storage. This is agreement with Sgroppo and Pereyra
(2009)’s results. The degradation of chlorophyll is caused by the
presence of oxygen and enzymes like chlorophyllase (Gong &
Mattheis, 2003). Chilling temperature reduces the enzymatic ac-
tivity and delays color change (Hedayati & Niakousari, 2015). Thus
high retention of green color was observed at 7 ± 1 ○ C in this study.
3.2. Aerobic plate count
Raw vegetables are easily contaminated by microorganisms
which might compromise the safety of the product. Sanitation is
critical to fresh-cut vegetables in keeping the microbial contami-
nation to a minimum (Alexopoulos et al., 2013). The aerobic
microbe counts of healthy raw vegetables are usually between
103 CFU/g and 107 CFU/g (Garg, Churey, & Splittstoesser, 1990).
Washing the whole and fresh-cut vegetables with sanitizer like

Table 1
Sensors used and their main applications in the electronic nose.

Sensor No. General description Reference

S1 Ammonia, used as sensor for aromatic compounds Benzene

S2 Reacts on sulphur compounds, H2S 1 × 10— 4 g/L. Otherwise sensitive to many terpenes and sulphur organic H2S
compounds, which are important for smell, limonene, pyrazine
S3 Mainly hydrogen, selectively, (breath gases) H2
S4 Organic acid esters and terpenes, aromatic compounds, less polar compounds Propane
S5 Detects alcohol's, partially aromatic compounds, broad range CO
S6 Sensitive to methane (environment) ca. 1 × 10—2 g/L. Broad range CH4
S7 Aromatics compounds, sulphur organic compounds H2S
S8 Very sensitive, broad range sensitivity, react on nitrogen oxides NO2
S9 Aliphatic hydrocarbons, aromatic compounds Propane
S10 Hydrocarbons CH4
S11 Aromatic compounds Toluene
S12 Alcohol, organic solvents C2H5OH
S13 Alkenes, aromatic compounds, less polar compounds Propane
S14 Reacts on high concentrations >0.1 g/L, sometime very selective (methane) CH4
80 H.-Z. Chen et al. / LWT - Food Science and Technology 87 (2018) 77e84
chlorine is necessary to decrease the microbial counts (González-
Aguilar, Ayala-Zavala, Ruiz-Cruz, Acedo-Félix, & Dıaz-Cinco,
2004). The evolution of aerobic microorganism counts of pepper
samples in this research is presented in Fig. 2. The low levels of
aerobic plate count at the beginning of storage reflected an excel-
lent quality of raw materials. Then, aerobic plate counts gradually
increased during the storage. Some studies of microbiological
analysis on ready-to-use vegetables reported that 6.0 log 10 CFU/g of
aerobic plate count is preferred to be tolerance limit in fresh-cut
vegetables, and when the total aerobic plate count is more than 8
log10 CFU/g, this is a sign of vegetable spoilage (Gómez-López,
Ragaert, Jeyachchandran, Debevere, & Devlieghere, 2008). In this
work, 5.0 log10 CFU/g of aerobic plate counts was considered as a
tolerance limit to evaluate microbial quality of ready-to-eat or
ready-to-cook fresh-cut green bell peppers. Samples reached this
acceptable limit at days 7 in this study. In a study of Ranjitha,
Sudhakar Rao, Shivashankara, and Roy (2015), the microbial
counts of fresh-cut green bell pepper (Capsicum annuum L.) main-
tained an acceptable level within 6 days during storage at 8 ○C.
3.3. Malondialdehyde content and membrane permeability
Malondialdehyde (MDA) is a secondary end product of poly-
unsaturated fatty acid oxidation of cell membrane, which is caused
by accumulation of reactive oxygen species (Xing et al., 2010). As
shown in Fig. 3, both the MDA content and membrane permeability
increased over the 9 d-storage time. A continuous increase in MDA
contents was also observed by Chen, Hu, Wang, Hu, and Cui (2016)
in fresh-cut green peppers during the whole storage period. In this
work, the original MDA content of sample was 0.74 ± 0.08 nmoL/g
at the beginning of storage and it rose to 4.61 ± 0.41 nmoL/g at the
end of storage. A surge in MDA content occurred at the 7-d storage
time with the value of 4.36 ± 0.53 nmoL/g. This result was in
accordance with aerobic plate count. It indicated that the pepper
samples become spoiled at 7 days of storage.
Membrane permeability is expressed as leakage of ions in tis-
sues. The higher its value, the worse the cell membrane integrity.
The deterioration of membrane permeability is also related to lipid
peroxidation and accumulation of active oxygen species (Duan
et al., 2009). In Fig. 3, the membrane permeability of the sample
was 18.48 ± 1.32% at the beginning of storage and rose to
27.97 ± 2.80% and 31.30 ± 0.91% respectively at 7 and 9 days of
Fig. 2. Changes in aerobic plate count of fresh-cut green bell pepper during storage at
7 ± 1 ○ C.
Values are the mean of eight replicates and bars stand for standard deviation.
Fig. 3. Changes in malondialdehyde content (-) and membrane permeability (C) of
fresh-cut green bell pepper during storage at 7 ± 1 ○C.
Values are the mean of three replicates and bars stand for standard deviation.
storage. An increment in membrane permeability of fresh-cut
green peppers with increasing storage time was also observed by
Raymond, Zhang, Karangwa, and Chesereka (2013) who found the
original and final permeabilities of about 10% and 41.42%, respec-
tively under cold storage at 5 ○C for 15 days. Meng et al. (2012) also
reported similar changes in MDA content and membrane perme-
ability of fresh-cut green peppers during storage at 4 ○C.
According to above-mentioned quality indicators, the fresh-cut
green bell peppers remained fresh until 7 days at 7 ± 1 ○ C and
considered being spoil beyond this period.
3.4. Electronic nose
3.4.1. Electronic nose response to pepper samples
Many volatile compounds are produced from secondary me-
tabolites with various biosynthetic pathways in tissues of vegeta-
bles (Berna, 2010). The flavors of fresh-cut green bell peppers at
regular intervals were measured by e-nose during storage. Average
responding values of 14 e-nose sensors to each sample are pre-
sented in Fig. 4. It can be clearly observed that the responding
signals of sensors significantly increased at 7 days except the sen-
sors S3 and S14, which indicated that changes in responding values
of these sensors were significantly different between fresh and
spoiled samples. It is in concordance with different intermolecular
interactions between different MOS sensors and the volatile com-
pounds produced during pepper spoilage. It is noteworthy that
responding values of sensors S1, S2, S5, S8 and S11 were over 10 at 9
days, which had more contributions than others on the evaluating
flavor intensities in peppers samples. Details and main applications
of MOS gas sensor array are shown in Table 1. The results indicated
that the off-flavors of the sample have changed during the storage
period and the flavor quality deteriorated at the end of storage.
Particularly, emissions of alcohols, ketones and aldehydes, esters
and terpenes, sulfides, etc. continuously increased. This is in
agreement with Ranjitha, Rao, Shivashankara, and Roy (2015)’s
results, i.e., volatile analysis of fresh-cut green bell peppers by GC-
MS showed a simultaneous increase of ketones and aldehydes,
esters, sesquiterpenoids, furans and pyrazines during 9 days stor-
age period at 8 ○C.
3.4.2. Hierarchical cluster analysis
The responding signals of 14 e-nose sensors were analyzed us-
ing hierarchical cluster analysis (HCA) to attempt to separate data
into specific groups based on similarity or distances between
 H.-Z. Chen et al. / LWT - Food Science and Technology 87 (2018) 77e84
Fig. 4. Radar chart of electronic nose data of fresh-cut green bell peppers obtained at
different storage periods (-: Days 0; C: Days 1; :: Days 3; ;: Days 5; >: Days 7;
9: Days 9).
Values are the mean of eight replicates and bars stand for standard deviation.
Fig. 5. Hierarchical cluster analysis (HCA) dendrogram performed with electronic nose
date of fresh-cut green bell peppers obtained at different storage periods. (For inter-
pretation of the references to colour in this figure legend, the reader is referred to the
web version of this article.)
81
Fig. 6. Principal component analysis plot performed with electronic nose date of fresh-
cut green bell peppers obtained at different storage periods (⅄ : Days 0; × : Days 1;
þ: Days 3; △ : Days 5; B: Days 7; ,: Days 9).
different observations (Xu, Yu, Liu, & Zhang, 2016). HCA results can
be presented as a dendrogram, in which distances between ob-
servations are determined similarity in different observations
(Haddi et al., 2013). In this work, cluster method of between-groups
linkage with squared Euclidean distance measure was used to
analyze cluster membership. Agglomerative method was adopted
for plotting dendrogram, which means that each observation starts
in its own cluster and mergences are performed recursively as one
moves up the hierarchy (Arana, Ibanez, & Torre, 2016).
As shown in Fig. 5, HCA can successfully differentiate samples
into four groups (i.e., Days 0 and 1, Days 3 and 5, Days 7, Days 9).
They were also divided into two groups at a distance of 8. Samples
stored at days 0, 1, 3, and 5 formed the first group, while the other
group included samples stored at days 7 and 9. As mentioned
before, the fresh-cut green bell peppers were fresh until 5 days and
became spoiled at 7 days in this study according to the obtained
results of traditional quality indicators. Therefore, e-nose combined
Fig. 7. PCA loading plot performed with electronic nose date of fresh-cut green bell
peppers obtained at different storage periods.
82 H.-Z. Chen et al. / LWT - Food Science and Technology 87 (2018) 77e84
Table 2
Principal component 1 (PC1) and Principal component 2 (PC2) mean scores obtained at different storage periods based on data s et of electronic nose.
Scores Storage time
Days 0 Days 1
PC1 —1.01 ± 0.10a —1.04 ± 0.05a
PC2 0.34 ± 0.81b —0.86 ± 0.45a
Values are mean ± standard deviation (n ¼ 8).
Different letters (a, b, c, d and e) within the same row indicate a significant difference (p < 0.05).
HCA can be utilized to preliminarily classify fresh and spoiled
groups for fresh-cut green bell peppers.
3.4.3. Principal components analysis
The responding values of e-nose were also analyzed using
principal component analysis (PCA) to determine if there were
differences in volatile patterns from samples at different storage
periods. PCA is an effective dimension reduction method which can
maintain the most information of original data and allocate each
observation to a particular group (Zhang, Tian, & Pei, 2014).
The PCA results are presented in Fig. 6. The two-dimensional
PCA plots of different cultivars represent the difference among
samples at different storage periods. The principal components
(PCs) PC1 and PC2 represented 96.22% and 2.26% of the total vari-
ance, respectively. The cumulative contribution rate of first two PCs
accounted for 98.48%, which represented the largest fraction of
overall variability in the observations. Several groups can be ob-
tained in Fig. 6. The data of samples stored at days 0 and 1 were
mixed in a group, whereas the other samples stored at days (3, 5, 7,
9) clustered in four separate groups, with the distance from sam-
ples stored at day 0 being longer as storage time increased.
Considering the loading plot (Fig. 7) showing the relationship
between the e-nose variables (Benedetti, Buratti, Spinardi,
Mannino, & Mignani, 2008), the S11, S8, S2 and S5 had the higher
influence in the patter file. Moreover, the result of ANOVA analysis
of PC1 and PC2 scores (Table 2) showed that both PC1 and PC2
scores of fresh samples at days (0, 1, 3, 5) had significantly different
with spoiled samples' at days (7, 9). According to the results of HCA
and PCA, the e-nose can be used to discriminate between fresh and
spoiled samples by measuring sensors’ response of the volatile
compounds.
Fig. 8. Variables important plot of Partial least squares statistical model based on data
set of electronic nose.
Days 3 Days 5 Days 7 Days 9
—0.38 ± 0.07b —0.01 ± 0.07c 0.66 ± 0.15d 1.77 ± 0.06e
—0.73 ± 0.65a 1.30 ± 0.34bc 0.75 ± 0.34c —0.80 ± 0.55a
3.4.4. Partial least squares statistical model
Partial least squares (PLS) can be implemented to handle
collinear data and reduce the number of required calibration ob-
servations (Sanaeifar, Mohtasebi, Ghasemi-Varnamkhasti, &
Ahmadi, 2016). PLS statistical model is a fixed linear regression
model with the type of y¼a*x þ b. x is the original data, a means
linear combinations called orthogonal latent components. PLS re-
duces the size of variables by extracting orthogonal latent compo-
nents from the original data (Kiani, Minaei, & Ghasemi-
Varnamkhasti, 2016b). The determination coefficient (R2), root
mean squared error (RMSE) or residual sum of squares were
calculated to assess the predictive ability of the models.
Variable Importance in Projection (VIP) tool in PLS statistical
models was developed to analyze the contribution of the various
sensors. Fig. 8 shows the PLS VIP score of the 1e14 sensors. As can
be seen in Fig. 8, all sensors had contribution to the PLS models.
However, the first two sensors which had higher contribution were
S11 and S8, followed by S2, S5, S1 and S4 successively. This result
was similar to Fig. 7. This may be related to the high response values
of these sensors. Other sensors showed a poor contribution. S3 had
the lowest contribution, followed by S14. The reason is that S3 and
S4 are sensitive to hydrogen and volatile gas produced in cooking,
thus are not related to volatile compounds released by raw pepper
samples during storage.
The PLS statistical model was implemented to predict aerobic
plate counts from the MOS array data. Experimental values versus
the values predicted by the PLS models for aerobic plate counts are
presented in Fig. 9. A simple linear model with a fitting line
(y 0.973*x
¼ 0.074,
þ R2 0.9783,¼ RMSE 0.3317)¼and fitting pa-
rameters were obtained. Slope of the fitting line (0.973) and
intercept with the y axis (0.074) are related to prediction accuracy
and the R2¼(0.9783) is associated with precision of the PLS model.
Fig. 9. Experimental aerobic plate count versus the values predicted by Partial least
squares statistical model (dashed lines) based on data set of electronic nose. The solid
line represents ideal behavior.
 H.-Z. Chen et al. / LWT - Food Science and Technology 87 (2018) 77e84
The residual sum of squares of the fitting line was 5.02. Ideally, the
predicted values should lie along the diagonal line, which repre-
sents that the predicted and measured values are the same.
Therefore, the PLS model is much better as both the slope of the
fitting line and R2 are close to 1. PLS statistical studies showed a
good correlation of the e-nose data with aerobic plate counts of
pepper samples. Considering that the microbial count is an
important index to evaluate fresh-cut vegetable, the results of the
PLS model showed the potential application of e-nose in assessing
the freshness of fresh-cut green bell pepper in cold storage.
Overall, the above multivariate analysis results indicate that the
e-nose system combined with HCA, PCA and PLS can be trained for
effectively differentiating between freshness and spoilage of fresh-
cut green bell pepper and monitoring the aerobic plate counts
during the cold temperature storage.
4. Conclusions
An analytical method using e-nose for evaluating freshness of
fresh-cut green bell pepper was investigated in this work. Hierar-
chical cluster analysis (HCA) of e-nose data showed good separa-
tion between fresh and spoiled samples and principal component
analysis (PCA) of e-nose data partially distinguished fresh-cut green
bell pepper samples under different storage days. Besides, a sig-
nificant correlation (R2 ¼ 0.9783, RMSE ¼ 0.3317) between the
sensors responses and aerobic plate count was obtained by PLS
statistical analyses method. These results indicated the feasibility of
easy-to-use e-nose for freshness and spoilage determination of
fresh-cut green bell pepper. A further investigation of e-nose is
needed to determine fresh, medium, and spoiled stages of fresh-cut
green bell peppers and evaluate freshness of other fresh-cut
vegetables.
Acknowledge
This work was financially supported by Jiangsu Province (China)
Key Project in Agriculture (Contract No. BE2015310217), Jiangsu
Province (China) “Collaborative innovation Center for Food Safety
and Quality Control” Industry Development Program, Jiangsu
Province(China) infrastructure Project(Contract No.BM2014051)
which have enabled us to carry out this study.
References
Alexopoulos, A., Plessas, S., Ceciu, S., Lazar, V., Mantzourani, I., Voidarou, C., et al.
(2013). Evaluation of ozone efficacy on the reduction of microbial population of
fresh cut lettuce (Lactuca sativa) and green bell pepper (Capsicum annuum).
Food Control, 30(2), 491e496.
Arana, I., Ibanez, F. C., & Torre, P. (2016). Monitoring the sensory quality of canned
white asparagus through cluster analysis. Journal of the Science of Food and
Agriculture, 96(7), 2391e2399.
Benedetti, S., Buratti, S., Spinardi, A., Mannino, S., & Mignani, I. (2008). Electronic
nose as a non-destructive tool to characterise peach cultivars and to monitor
their ripening stage during shelf-life. Postharvest Biology and Technology, 47(2),
181e188.
Berna, A. (2010). Metal oxide sensors for electronic noses and their application to
food analysis. Sensors (Basel), 10(4), 3882e3910.
Chen, J., Hu, Y., Wang, J., Hu, H., & Cui, H. (2016). Combined effect of ozone treat-
ment and modified atmosphere packaging on antioxidant defense system of fresh-
cut green peppers. Journal of Food Processing and Preservation, 40(5),
1145e1150.
Conesa, A., Verlinden, B. E., Artés-Hernández, F., Nicolaï, B., & Artés, F. (2007).
Respiration rates of fresh-cut bell peppers under supertamospheric and low
oxygen with or without high carbon dioxide. Postharvest Biology and Technology,
45(1), 81e88.
Defilippi, B. G., Juan, W. S., Valdés, H., Moya-León, M. A., Infante, R., & Campos-
Vargas, R. (2009). The aroma development during storage of Castlebrite apricots
as evaluated by gas chromatography, electronic nose, and sensory analysis.
Postharvest Biology and Technology, 51(2), 212e219.
Duan, X., Su, X., Shi, J., Yi, C., Sun, J., Li, Y., et al. (2009). Effect of low and high
oxygen-controlled atmospheres on enzymatic browning of litchi fruit. Journal of
83
Food Biochemistry, 33(4), 572e586.
Garg, N., Churey, J. J., & Splittstoesser, D. F. (1990). Effect of processing conditions on
the microflora of fresh-cut vegetables. Journal of Food Protection, 53(8),
701e703.
GB 4789.2. (2010). National Standard of the People's Republic of China: National food
safety standard, Food microbiological examination: Aerobic plate count.
Giovenzana, V., Beghi, R., Buratti, S., Civelli, R., & Guidetti, R. (2014). Monitoring of
fresh-cut Valerianella locusta Laterr. shelf life by electronic nose and VIS-NIR
spectroscopy. Talanta, 120, 368e375.
Gobbi, E., Falasconi, M., Zambotti, G., Sberveglieri, V., Pulvirenti, A., & Sberveglieri, G.
(2015). Rapid diagnosis of Enterobacteriaceae in vegetable soups by a metal
oxide sensor based electronic nose. Sensors and Actuators B: Chemical, 207,
1104e1113.
Gómez-López, V. M., Ragaert, P., Jeyachchandran, V., Debevere, J., & Devlieghere, F.
(2008). Shelf-life of minimally processed lettuce and cabbage treated with
gaseous chlorine dioxide and cysteine. International Journal of Food Microbi-
ology, 121(1), 74e83.
Gómez, A. H., Wang, J., Hu, G., & Pereira, A. G. (2008). Monitoring storage shelf life of
tomato using electronic nose technique. Journal of Food Engineering, 85(4),
625e631.
Gong, Y., & Mattheis, J. P. (2003). Effect of ethylene and 1-methylcyclopropene on
chlorophyll catabolism of broccoli florets. Plant Growth Regulation, 40(1),
33e38.
González-Aguilar, G. A., Ayala-Zavala, J. F., Ruiz-Cruz, S., Acedo-Félix, E., & Dıaz-
Cinco, M. E. (2004). Effect of temperature and modified atmosphere packaging
on overall quality of fresh-cut bell peppers. LWT - Food Science and Technology,
37(8), 817e826.
Gonzalez, M. E., Anthon, G. E., & Barrett, D. M. (2010). Onion cells after high pressure
and thermal processing: Comparison of membrane integrity changes using
different analytical methods and impact on tissue texture. Journal of Food Sci-
ence, 75(7), E426eE432.
González, G., & Tiznado, M. (1993). Postharvest physiology of bell peppers stored in
low density polyethylene bags. LWT - Food Science and Technology, 26(5),
450e455.
Haddi, Z., Alami, H., El Bari, N., Tounsi, M., Barhoumi, H., Maaref, A., et al. (2013).
Electronic nose and tongue combination for improved classification of Moroc-
can virgin olive oil profiles. Food Research International, 54(2), 1488e1498.
Hedayati, S., & Niakousari, M. (2015). Effect of coatings of silver nanoparticles and
gum Arabic on physicochemical and microbial properties of green bell pepper
(Capsicum annuum). Journal of Food Processing and Preservation, 39(6),
2001e2007.
Hernández Gómez, A., Wang, J., Hu, G., & García Pereira, A. (2007). Discrimination of
storage shelf-life for mandarin by electronic nose technique. LWT - Food Science
and Technology, 40(4), 681e689.
Hodges, D. M., DeLong, J. M., Forney, C. F., & Prange, R. K. (1999). Improving the
thiobarbituric acid-reactive-substances assay for estimating lipid peroxidation
in plant tissues containing anthocyanin and other interfering compounds.
Planta, 207(4), 604e611.
Kiani, S., Minaei, S., & Ghasemi-Varnamkhasti, M. (2016a). Application of electronic
nose systems for assessing quality of medicinal and aromatic plant products: A review.
Journal of Applied Research on Medicinal and Aromatic Plants, 3(1), 1e9. Kiani, S.,
Minaei, S., & Ghasemi-Varnamkhasti, M. (2016b). A portable electronic nose as an
expert system for aroma-based classification of saffron. Chemo-
metrics and Intelligent Laboratory Systems, 156, 148e156.
Lebrun, M., Plotto, A., Goodner, K., Ducamp, M.-N., & Baldwin, E. (2008). Discrimi-
nation of mango fruit maturity by volatiles using the electronic nose and gas
chromatography. Postharvest Biology and Technology, 48(1), 122e131.
Liu, Z., Zhang, M., & Wang, Y. (2016). Drying of restructured chips made from the old
stalks of Asparagus officinalis: Impact of different drying methods. Journal of the
Science of Food and Agriculture, 96(8), 2815e2824.
Loutfi, A., Coradeschi, S., Mani, G. K., Shankar, P., & Rayappan, J. B. B. (2015). Elec-
tronic noses for food quality: A review. Journal of Food Engineering, 144, 103e111.
Manolopoulou, H., Xanthopoulos, G., Douros, N., & Lambrinos, G. (2010). Modified
atmosphere packaging storage of green bell peppers: Quality criteria. Bio-
systems Engineering, 106(4), 535e543.
Meng, X., Zhang, M., & Adhikari, B. (2012). Extending shelf-life of fresh-cut green
peppers using pressurized argon treatment. Postharvest Biology and Technology,
71, 13e20.
Qiu, S., Wang, J., & Gao, L. (2015). Qualification and quantisation of processed
strawberry juice based on electronic nose and tongue. LWT - Food Science and
Technology, 60(1), 115e123.
Ranjitha, K., Sudhakar Rao, D. V., Shivashankara, K. S., & Roy, T. K. (2015). Effect of
pretreatments and modified atmosphere packaging on the shelf life and quality
of fresh- cut green bell pepper. Journal of Food Science and Technology, 52(12),
7872e7882.
Raymond, L. V., Zhang, M., Karangwa, E., & Chesereka, M. J. (2013). Mixed noble gas
effect on cut green peppers. International Agrophysics, 27(1), 75e79.
Rodoni, L., Vicente, A., Azevedo, S., Concellón, A., & Cunha, L. M. (2015). Quality
retention of fresh-cut pepper as affected by atmosphere gas composition and
ripening stage. LWT - Food Science and Technology, 60(1), 109e114.
Rodoni, L. M., Zaro, M. J., Hasperué, J. H., Concellón, A., & Vicente, A. R. (2015). UV-C
treatments extend the shelf life of fresh-cut peppers by delaying pectin solu-
bilization and inducing local accumulation of phenolics. LWT - Food Science and
Technology, 63(1), 408e414.
Sanaeifar, A., Mohtasebi, S. S., Ghasemi-Varnamkhasti, M., & Ahmadi, H. (2016).
84 H.-Z. Chen et al. / LWT - Food Science and Technology 87 (2018) 77e84

Application of MOS based electronic nose for the prediction of banana quality
properties. Measurement, 82, 105e114.
Sgroppo, S. C., & Pereyra, M. V. (2009). Using mild heat treatment to improve the
bioactive related compounds on fresh-cut green bell peppers. International
Journal of Food Science & Technology, 44(9), 1793e1801.
Singh, R., Giri, S. K., & Kotwaliwale, N. (2014). Shelf-life enhancement of green bell
pepper (Capsicum annuum L.) under active modified atmosphere storage. Food
Packaging and Shelf Life, 1(2), 101e112.
Xing, Y., Li, X., Xu, Q., Jiang, Y., Yun, J., & Li, W. (2010). Effects of chitosan-based
coating and modified atmosphere packaging (MAP) on browning and shelf
life of fresh-cut lotus root (Nelumbo nucifera Gaerth). Innovative Food Science &
Emerging Technologies, 11(4), 684e689.
Xu, L., Yu, X., Liu, L., & Zhang, R. (2016). A novel method for qualitative analysis of
edible oil oxidation using an electronic nose. Food Chemistry, 202, 229e235.
Zhang, L., Tian, F., & Pei, G. (2014). A novel sensor selection using pattern recog-
nition in electronic nose. Measurement, 54, 31e39.