Procedia

Procedia Engineering 5 (2010) 1228–1231 Engineering

Procedia Engineering 00 (2009) 000–000
www.elsevier.com/locate/procedia
www.elsevier.com/locate/procedia

Proc. Eurosensors XXIV, September 5-8, 2010, Linz, Austria

Fish freshness decay measurement with a colorimetric artificial

olfactory system
F. Dinia, R. Paolesseb, D. Filippinic, A. D’Amicoa, I. Lundströmc, C. Di Natalea*
a
Department of Electronic Engineering, University of Rome Tor Vergata, Via Politecnico, 00133 Roma; Italy
b
Department of Chemical Science and Technology, University of Rome Tor Vergata, Via Ricerca Scientifica, 00133 Roma; Italy
c
Division of Applied Physics, Dept. Physics, Chemistry and Biology, Linköping University, Linköping, Sweden

Abstract

This paper reports about the application of an artificial olfactory system based on optical imaging technology. This arrangement
is formed by a distributed layer of chemical indicators illuminated by a computer screen and imaged by a digital camera. The
system has been applied to monitor the freshness decay in fish. The set of indicators is formed by porphyrinoids and acid-base
indicators, this combination provides an optimal capture of the process with some of the indicators sensitive to first stage, when
the product is still fresh, and others more sensitive to the last part of the freshness decay

2010 Published by Elsevier Ltd. Open access under CC BY-NC-ND license.

c 2009
©

Keywords: Type your keywords here, separated by semicolons ;

1. Introduction

In the last decades the development of artificial olfaction systems was mostly oriented to improve the application
target but the similarities between natural and artificial systems have been subject of little progresses. In natural
olfaction about 100,000 olfactory receptor neurons are found with only few hundreds kinds of receptors and for
many sensors technologies this situation is hardly reproducible. On the other hand, it was shown that this structure
can be easily imitated using optical indicators probed by image sensors [1]. Indeed, an image sensor can at the same
time measure the optical properties of a large sensitive area, and the area corresponding to a single image pixel may
correspond to an individual sensor. This concept has been recently utilized to define, considering the optical
properties of pixels, the second processing layer found in natural olfaction: the glomerular layer [2]. Furthermore,
the distributed sensing layer provides also the separation properties of the olfactory mucosa [3].
In this paper we present an application of such a system to the characterization of fish freshness decay. The
artificial olfactory system is composed by an almost continuous layer of porphyrinoids and acid-base indicators that

* Corresponding author. Tel.: +0-000-000-0000 ; fax: +0-000-000-0000 .

E-mail address: author@institute.xxx .

1877-7058 c 2010 Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
doi:10.1016/j.proeng.2010.09.334
 F. Dini et al. / Procedia Engineering 5 (2010) 1228–1231 1229

2 Author name / Procedia Engineering 00 (2010) 000–000

forms a sort of artificial epithelium. The layer is illuminated by a computer screen and imaged by a digital camera
that decomposes the sensing layer into the number of pixels defined by the image resolution.
The platform was tested to monitor the freshness decay of thawed hake fillets held at room temperature and for
several hours. During the process fish headspace is known to change both in quality and quantity [5]. Previous
investigations evidenced that the headspace composition is a result of the balance between the “fresh fish” odour
and the microbial spoilage produced compounds [4]. The most important chemicals involved in the fresh fish odour
are long-chain alcohols and carbonyls, bromophenols and N-cyclic compounds. Their concentration and the
presence of other compounds are rather typical of each species. On the other side, microbial spoilage produces
short-chain alcohols and carbonyls, amines, sulphur compounds, aromatic, N-cyclic, and acid compounds. The
concentrations of these chemicals are directly correlated to the degree of spoilage. Among these compounds amines
are considered as the typical markers for fish freshness detection. Since the use of a multeplicity of indicators
should allow for a proper detection of all the phases leading from a fresh to a stale product, fish freshness
classifications has been frequently used as a test for electronic noses since the very beginning [5].

2. Experimental

The artificial epithelium was formed by an almost continuous layer of

Figure 1. Image acquired by the camera of
the artificial epithelium. Sensitive
indicators are labelled according to the
following list: 1: FeTPP 2: SiPC, 3: Methyl
orange, 4: ZnTPP, 5: Bromocresol purple,
6: Biladiene, 7: TPC, 8: Nile, 9: Red
Cresol, 10: MnTPP.
chemical indicators. The indicators used in this experiments were three
porphyrins, a silicon-phthalocyanine, a corrole, biladiene, and four acid-
base inidicators known to cover with different color different pH ranges.
Indicators were nile blue, methyl orange, bromocreol purple, and red cresol;
porphyrins were tetraphenylporphyrin (TPP) complexed with Mn, Fe, and
Zn.
Indicators were spotted in a PVC continuous membrane uniformly
coating a 25 mm diameter plastic substrate. The appearance of the substrate
is visible in fig. 1.
The artificial epithelium was placed in a cell with gas-tight optical
windows and equipped with tube connectors for gas delivery. The
epithelium was illuminated through the transparent window with a LCD
computer monitor (Philips 1704S) and imaged, in transmission with a digital
camera (Philips SPC900NC). A diagram of the experimental arrangement is
shown in fig. 2. Changes during gas exposure were probed by a sequence of
the three pure colors: red (255 0 0), green (0 255 0), and blue (0 0 255).
Such an experimental setup is dubbed as computer screen photoassisted
technology (CSPT ) [6].
The headspace of a hake fillet was continuously injected for five hours in
the sensors cell.

3. Results and discussions

Figure 2. Sketch of the CSPT
Fig. 3 shows the differential camera signals in the three channels (red,
green, and blue) measured under red, green and blue screen illumination.
The signals from each spot were subtracted from the background and the
first image corresponding to the start of the experiments was subtracted in
order to put into evidence only the differences due to the exposure of fish
headspace. Signals were further processed eliminating the mean value and
scaling all the signals to unitary variance. This procedure, known as
1230 F. Dini et al. / Procedia Engineering 5 (2010) 1228–1231

Author name / Procedia Engineering 00 (2010) 000–000 3

autoscaling, helps in eliminating the dominance of signals of larger magnitude.

From Fig. 3 it is evident that indicators differently behave at the beginning of the process, as the the spoilage
progress the headspace becomes so rich that all indicators provide the same value. It is worth to mention the
behavior of bromocresol purple that undergoes a sudden change of intensity at a point of evolution where likely the
amount of amines exceed a certain threshold.


!  
! 

!
! 

  
"



"



"


  

  

  













        
        
  
Figure 4. autoscaled indicator signals plotted verus the the logairthm of time. Logarithmic scale has been chosen to put in evidence the
change occuring in the first stage of the process.

 

 






 6
16


 

7 3
15 97 120
101
102
9499117
112
118
113
119
116
109
111
100
103 125
124
121
107122
128
126
123
114 135
134
131
133
129136
517
30 73
78
75
72
70 8186
87
88
82
83
84
79
7680
85
92
95
89110
105
90
9193115
104
96
106
108
98 127
132
130
137
138
140
139
142
141
66
6874 143
145
144







4
28 56 69
64
63
71 77 147





 188
22 9
12 52
53
61
55
5154 65
57
58
5962
60 67 146
148
155
152
149
153
158
151
157
163
 48
49
46 150
127 35
37
32
33
31
36
43
41
44
42
39
40
38 4750
45 156
162
154
160
159203
170
164
168
166
173
174198
167
196
171
179
169
187
200
194
204
175
172
176
178
189
186
197
193
214
165
161190
181
183
177
202
180
215
184
239
208
188
182
199
222
229
240
191
195
223
235
228
234
243 201
207
205
230
192
217
209
231
210
237
213
233
247
10
20
14 116
19
13 2
23
28
27
7
3
4
8
21 30
26
25
29
22
5 34 249
252
277
260
263
264
269
255
241
266
253
250
270
251 206
212
216
185
224
232
226
219
236
221
227
242
218
220
244
225
238
259
276
211
245
254
262
273
272
256
257
248
268
275
274
267246
25
21
23
29 18
12
20
9
17
15
1016
14 24 261
271
258
265
 11
13
19
2




24







26












  



 

      



  





Figure 5. Scores plot of the first two principal components. The Figure 6. Biplot of the first two principal components. All
abrupt change along the second principal component occurs at indicator have little differences excepts four cases.
about 150 minutes after the experiment start.

Fig. 4 shows the scores plot of the principal component analysis of the thirty signals shown in Fig. 3. The
behaviour is non linear, changes tend to be smaller as the time increases, and rather surprisingly, at the end of the
process the data points in the scores plot tend toward the fresh fish state. Actually a similar behaviour has been
observed before with other electronic noses, and this situation has been recently shown to be likely due to the
contemporaneous decrease and increase of different classes of chemicals [7]. The contribution of different indicators
and light illumination to Fig. 4 can be adequately studied by the biplot. This is a plot where both the scores (data
points) and the loadings (direction of the original axis) are simultaneously presented. Fig. 5 shows the biplot of the
first two principal components. Most of the indicators lies together along the first principal component, this accounts
of the correlation between signals driven by the fact that the process under study is characterized by a constant
increase of the total amount of volatile compounds. Nonetheless few indicators can be identified as bearer of
 F. Dini et al. / Procedia Engineering 5 (2010) 1228–1231 1231

4 Author name / Procedia Engineering 00 (2010) 000–000

additional information, and they are methyl orange under red illumination (label 3), biladiene under green and blue
illuminations (labels 16 and 26), and the Zinc Tetraphenylporphyrin under green illumination (label 14). It is worth
to note that these indicators are displaced towards the second principal component where the “curling” effect of the
data points takes place.

4. Conclusions

In this paper we demonstrated that an array of different chemical indicators can adequately monitor all the
processes occurring in fish during the transition from fresh to stale conditions. Colorimetric detection of fish
freshness has been studied in the past by several authors that focused their attention to the detection of amines with a
pH sensitive layer [8]. In this case, the detection is limitated to the stale conditions while this results suggest arrays
of proper indicator can give a full account of fish state identifying also the changes occurring in the first phase of
freshness decay.

Reference

[1] Dickinson T, White J, Kauer K, Walt D, A chemical detectings ystem based on a cross-reactive optical sensor array, Nature, 1996; 382:697–
700.
[2] Di Natale C, Martinelli E, Paolesse R, D’Amico A, Filippini D, Lundstrom I An experimental biomimetic platform for artificial olfaction,
PLoS ONE, 2008; 3:e3139
[3] Dini F, Filippini D, Paolesse R, D’Amico A, Lundström I, Di Natale C, Polymers with embedded chemical indicators as an artificial olfactory
mucosa Analyst 2010, 135: 1245-1252.
[4] Di Natale C, Olafsdottir, G. Electronic noses and electronic tongues in “Fishery products: quality, safety, and authenticity” in (Rebein H, and
Oehlenschlager J Eds.). Wiley-Blackwell, 2009
[5] Olafsson R, Martinsdottir E, Olafsdottir G, Sigfusson T, Gardner J, Monitoring of fish freshness using tin dioxide sensors, in (Gardner J,
Bartlett P Eds.) “Sensors and Sensory Systems for an Electronic Nose”, Kluwer Academic Publishers, 1992.
[6] Filippini D, Svensson S, I. Lundström I, Computer screen as a programmable light source for visible absorption characterization of
(bio)chemical assays, Chem. Commun. 2003 2:240–241.
[7] Di Natale C, Chemical sensors, in ( Nollet L, Toldrà F. Eds.) “Handbook of seafood and seafood product analysis”, CRC Press, 2010
[8] Pacquit A, Lau K, McLaughlin H, Frisby J, Quilty B, Diamond D, Development of a volatile amine sensor for the monitoring of fish spoilage,
Talanta, 2006; 69: 515–520.