DOI: 10.3303/CET24112020 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Paper Received: 23 February 2024; Revised: 12 April 2024; Accepted: 14 June 2024 
Please cite this article as: Sanchez R., Alejo M., Escribano P., Arroyo P., Meléndez F., Lozano J., 2024, Evaluation of the Shelf Life of Fresh 
Fish Using an Electronic Nose, Chemical Engineering Transactions, 112, 115-120  DOI:10.3303/CET24112020 
  

 CHEMICAL ENGINEERING TRANSACTIONS  
 

VOL. 112, 2024 

A publication of 

 

The Italian Association 
of Chemical Engineering 
Online at www.cetjournal.it 

Guest Editors: Selena Sironi, Laura Capelli 

Copyright © 2024, AIDIC Servizi S.r.l. 
ISBN 979-12-81206-13-7; ISSN 2283-9216 

Evaluation of the Shelf Life of Fresh Fish Using an Electronic 

Nose 

Ramiro Sánchez a, María Alejo a, Patricia Escribano a, Patricia Arroyo b, Félix 

Meléndez b, Jesús Lozano*b 

a Center for Scientific and Technological Research of Extremadura (CICYTEX), 06007 Badajoz, Spain 

b Industrial Engineering School, University of Extremadura, 06006 Badajoz, Spain 

jesuslozano@unex.es 

The fishing industry faces significant challenges in supply chain management and fish quality assurance. 

Freshness, a crucial quality attribute, decreases over storage time. Traditional freshness control methods like 

sensory analysis, chromatography, or microbiological analyses, while effective, are slow and costly. Hence, the 

need for a tool like the electronic nose (E-nose) emerges, capable of predicting fish freshness by evaluating its 

odor using economical, quick, and portable equipment. This study aimed to classify the river fish Tinca tinca 

(tench) according to its freshness state and correlate its olfactory pattern with the presence of aerobic mesophilic 

microorganisms in the samples. The proliferation of aerobic mesophiles accelerates the sensory olfactory 

degradation of fish and is an indicator of quality. The E-nose used is a prototype developed by the University of 

Extremadura (UEx) with a matrix of 5 metal oxide semiconductor (MOS) gas sensors. Data were analyzed using 

Principal Component Analysis (PCA), where the first two components explained 93.4 % of the total variance, 

and Partial Least Squares Discriminant Analysis (PLSDA) was successful in classifying 95 % of the samples. 

Lastly, a Partial Least Squares (PLS) model was employed to relate the actual number of aerobic mesophilic 

bacteria to those predicted by the E-nose, achieving a determination coefficient of R2
Pred = 0.990. E-nose 

technology emerges as a promising alternative for enhancing the precision and efficiency in assessing fresh 

fish quality. 

Keywords: fresh fish, electronic nose, shelf life, quality assessment. 

1. Introduction 

In the fishing industry, ensuring quality is essential for protecting health and satisfying consumers. Tench is a 

freshwater fish, and interest in its cultivation has significantly increased due to its potential in aquaculture, taste, 

and high nutritional value. Tench production has grown in Europe, and studies on its intensive breeding have 

increased (Al Fatle et al., 2022). Fish freshness, a key quality indicator, is compromised by microbial activity (Li 

et al., 2019), which can cause unfavorable sensory changes like bad odors and tastes. The detection of aerobic 

mesophilic bacteria is a fish quality indicator, as their presence indicates contamination and deterioration, 

reflecting hygiene conditions during processing and storage, and therefore, the fish's freshness and shelf life 

(Peris and Escuder-Gilabert, 2009). 

Given the limitations of traditional analysis techniques for assessing fish freshness, emerging technologies like 

electronic noses (E-nose) offer a promising solution. Over the last two decades, a substantial number of studies 

have been conducted focusing on the application of E-nose technology to assess the freshness and stability of 

stored fish products. These studies have examined a series of volatile compounds such as sulfur and nitrogen 

compounds, which are released as a result of microbial activity and biochemical processes throughout the 

preservation period (Grassi et al., 2019). E-noses mimic the human sense of smell and allow for the identification 

of changes in the chemical composition of foods quickly and non-invasively (Wu et al., 2023). Integrating 

electronic noses with chemometric analysis to evaluate fish freshness based on olfactory profiles and the 

evaluation of bacterial load represents a significant advance. These methods provide a detailed perspective on 

115



quality, enabling precise discrimination between different states of fish freshness and correlating these data with 

levels of aerobic mesophilic bacteria (Grassi et al., 2019).  

This work focuses on the electronic nose's ability to differentiate between various degrees of tench fish 

freshness and associate these results with the presence of aerobic mesophilic bacteria, proposing a rapid and 

efficient method to improve quality control in the fishing industry. This study not only enhances the effectiveness 

of quality monitoring but also contributes to food safety by mitigating the risk of foodborne diseases and ensuring 

higher quality products for consumers.  

2. Materials and Methods 

2.1 Experimental Design 

The study was conducted with river tench samples acquired from a local market and refrigerated at a 

temperature of 2°C. Five samples were randomly selected for analysis, conducting aerobic mesophilic counts 

and E-nose measurements on days 0 (purchase day), 4, 12, and 18 post-purchase to encompass a 

representative range of the fish's shelf life. 

2.2 Microbiological Analysis 

The plate count method, as referenced in UNI EN ISO 4833-1: 2013, was used to determine the load of aerobic 

mesophiles. 10 g of fresh fish samples were homogenized in a peptone saline solution and serially diluted. 1 

mL of each dilution was inoculated onto Petri dishes with standard count agar and incubated at 30 °C for 72 h. 

Colony counts were expressed in logarithms of colony-forming units per gram (log CFU/g). 

2.3 Electronic Nose 

The E-nose device used is a prototype developed by the University of Extremadura (UEx) that communicates 

via Bluetooth with a mobile app. It is a compact, low-energy consumption portable device incorporating 5 metal 

oxide semiconductor (MOS) gas sensors from different manufacturers, providing broad selectivity. These 

sensors' distinctive feature is their ability to integrate both analog and digital electronic components on a hotplate 

with the detectors, all within a single integrated circuit. Additionally, the device is equipped with sensors for 

measuring other environmental parameters like temperature, relative humidity, and atmospheric pressure. 

Specific details of the sensors used are presented in Table 1. 

Table 1: Sensors included in the electronic nose prototype. 

Sensor Manufacturer Type Signals 

BME680 Bosch Sensortech Gmbh, 

Germany 
Metal-Oxide 

(MOS) 
Temperature, Relative Humidity, 

Pressure, Resistance value(Ω) 
SGP30 Sensirion AG, Switzerland MOS CO2, TVOCs 1, H2 (raw signal 2), 

Ethanol (raw signal) 
ZMOD4410 Renesas Electronic 

Corporation, Japan 
MOS Ethanol (raw signal), Resistance 

value(Ω), CO2, TVOCs, IAQ 3 
CCS811 ScioSense B.V., The 

Netherlands 
MOS CO2, TVOCs, Resistance value 

iAQ-Core ScioSense, Eindhoven, The 

Netherlands 
MOS CO2 (ppm), TVOCs (ppb), 

Resistance value (Ω) 

 

1. Total Volatile Organic Compounds. 
2. Signal derived from the sensor resistance. 
3. Air Quality Index. 

 

The device measures 118 x 82 x 22.5 mm operates on a +3.7 VDC rechargeable lithium battery and uses 

Bluetooth for communication with a mobile app dedicated to its control and configuration. It also includes a pump 

and a solenoid valve for extracting the headspace from the samples. A block diagram of the device's electronic 

board is shown in Figure 1. 

 

116



 

Figure 1. Block diagram of the electronic board of the prototype   

The measurement process with this device occurs in two stages: a 60-second desorption phase where sensors 

are only in contact with ambient air, serving as a reference and cleaning signal, followed by a 60-second 

adsorption or sampling phase where sensors interact with the sample's headspace. The solenoid valve, 

managed by the microcontroller, alternates between these two inputs of sample and air. Analyses were 

performed on 5 fish fillets. A 5 g sample was placed in a container with a lid that has one air inlet and another 

for aspirating the sample's headspace through the E-nose pump. The sample was placed in a thermostatic 

water bath at 25 ºC as illustrated in Figure 2. 

 

Figure 2. Measurements of the olfactory pattern of fish with E-nose 

2.4 Laboratory test 

To test the discrimination capability for gases indicative of fish spoilage due to microbial attack, such as 

ammonia gas, a laboratory test was conducted. The test consisted of preparing a solution of distilled water and 

another of 0.02 % ammonia. Ten samples were taken from each solution, and six measurements were made 

for each sample. Subsequently, the observations from the water and ammonia were plotted on a PCA graph. 

2.5 Statistical Analysis 

The data collected by the E-nose were subjected to Principal Component Analysis (PCA) to reduce data 

dimensionality and explore possible sample groupings according to their freshness. Additionally, Partial Least 

Squares Discriminant Analysis (PLSDA) was performed to classify samples based on storage days. For 

correlating freshness with microbial load, a Partial Least Squares (PLS) model was used. All statistical analyses 

were conducted using Matlab R2023a Software version 9.14.0.2206163 (The Mathworks Inc., Natick, MA, USA) 

with the PLS_Toolbox 9.1 (Eigenvector Research Inc., Wenatchee, WA, USA). 

117



3. Results and Discussion 

3.1 Microbiological Results 

The microbiological analysis (Table 2) revealed a significant increase in the load of aerobic mesophiles over the 

storage period. 

 

Table 2: Aerobic mesophilic values during storage 

 Day Zero Day 4 Day 12 Day 18 

Aerobic 

Mesophiles 

(log UFC/g) 

3.60 3.30 7.00 9.44 

 

Initially, on day zero, the bacterial load was 3.60 log CFU/g, slightly decreasing to 3.30 log CFU/g by day 4, 

which could be attributed to sampling variability. However, by day 12, the load increased to 7.00 log CFU/g, and 

by day 18, it reached 9.44 log CFU/g, indicating significant bacterial proliferation and a reduction in fish 

freshness. 

3.2 Electronic Nose Results 

PCA analysis of the data obtained by the electronic nose showed clear discrimination between samples from 

different storage days. In Figure 3, the principal components PC1 and PC2 explained 77.21% and 16.17% of 

the total variance, respectively, demonstrating a good separation between fresh samples and those stored for 

longer periods. 

 

 

Figure 3. Grouping of observations with E-nose performed on different storage days 

PLSDA provided precise classification of samples according to storage days, with a correct prediction 

percentage of 95 % (Table 3). To construct the classes in the PLSDA model, fish samples were collected at four 

different time points: Day 0 (fresh), Day 4, Day 12 and Day 18. Each time point was considered a different class. 

The E-nose data were preprocessed using standard normalization techniques and PLSDA was used to classify 

samples based on storage days. To evaluate the accuracy of the model, internal cross-validation was performed 

using the "leave-one-out" technique. This high success rate indicates the effectiveness of the electronic nose in 

conjunction with chemometric analysis in classifying fish based on its freshness. 

Table 3: Classification Prediction by PLSDA Model 

 Real Class 

Predicted Day Zero Day 4 Day 12 Day 18 

Day Zero 4 0 0 0 

Day 4 1 5 0 0 

Day 12 0 0 5 0 

Day 18 0 0 0 5 

118



Finally, the results of the PLS prediction model are depicted in Figure 4. This graph shows the relationship 

between the actual microbial load values of aerobic mesophiles and those predicted by the electronic nose for 

samples at different storage times. 

 

Figure 4. PLS Graph of Correlation between Actual and E-nose Predicted Values of Aerobic Mesophiles  

 

The orange dots represent validation data, and the blue dots are calibration data. The overall trend indicated by 

the data is that the electronic nose is capable of accurately predicting the microbial load in samples based on 

data collected at different storage times. The high R² value in both data sets (calibration and validation) and the 

low values of RMSEC and RMSECV suggest that the electronic nose could be an effective tool for monitoring 

the microbial load of aerobic mesophiles in these samples. 

The findings of this study affirm the hypothesis that the electronic nose, in conjunction with chemometric analysis 

tools, offers an effective and efficient alternative for fish quality control. The correlation between microbial load 

and olfactory profiles detected by the electronic nose highlights the significance of these microorganisms in 

determining fish freshness. These results align with recent scientific literature, such as Al-Hooti et al. (2024), 

which explored the efficacy and reliability of an E-nose in detecting microbial decomposition in fresh sardines, 

comparing its readings with total bacterial count and other indicators under varying storage conditions. 

Furthermore, the success of an electronic nose is closely linked to proper sensor selection and data analysis 

system design. Recent studies in the field of electronic noses have emphasized the importance of feature 

selection and sensor array optimization for optimal device performance (Borowik et al., 2020). In this study, the 

selection of MOS sensors and the use of PCA, PLSDA, and PLS were shown to be suitable for the task, but 

ongoing optimization of these components is crucial for enhancing accuracy and the technology's applicability 

in different contexts. 

3.3 Laboratory Test Results 

The PCA graph depicted in Figure 5 demonstrates two distinct data sets, one representing water and the other 

a 0.02 % ammonia solution. 

 
Figure 5. Graph of discrimination between distilled water and 0.02% ammonia solution 

119



The pronounced separation of ammonia, a marker for fish freshness, even at such low concentration, highlights 

the sensors' efficiency in discriminating these sample types. A Neural Network Analysis (ANN) classification 

model was applied (Table 4), yielding a 100 % success rate in identifying the samples. 

 

Table 4: Classification Prediction by ANN Model 

 Real Class 

Predicted Water Ammonia 

Water 60 0 

Ammonia 0 60 

4. Conclusions 

This study demonstrated the E-nose's capability, along with chemometric analyses, to effectively classify fish 

freshness based on its olfactory profile and correlate these results with the load of aerobic mesophilic bacteria. 

The findings suggest that the electronic nose can serve as a valuable tool for quality control in the fishing 

industry, offering a rapid, accurate, and non-destructive alternative to conventional methods based on 

microbiological analysis. The established correlation between the olfactory profiles detected by the electronic 

nose and the microbial load highlights the relevance of this technological approach for assessing fish freshness 

and, consequently, its quality. This study should be completed with an increase in the number of samples 

analysed to establish more robust models and confirm their potential use. This work paves the way for future 

research to further optimize sensor selection and analysis algorithms, expanding the electronic nose's 

applicability to a wider range of food products and storage conditions. It is also imperative to continue exploring 

the integration of these technologies into real-time monitoring systems, thus enhancing the food industry's ability 

to ensure the safety and quality of its products. In conclusion, the electronic nose represents a significant 

advancement in fish quality control, facilitating more informed and rapid decisions that can improve supply chain 

management and meet consumer demands for high-quality products. 

Acknowledgments 

The authors would like to thank the Scientific and Technological Research Center of Extremadura (Spain) for 

providing their facilities and materials for the conduct of the tests. Additionally, gratitude is extended to the 

Research Group on Perception and Intelligent Systems of the University of Extremadura (GR21045) for their 

support in carrying out this work. 

References 

Al Fatle, F. A., Meleg, E. E., Sallai, Z., Szabó, G., Várkonyi, E., Urbányi, B., ... & Lehoczky, I., 2022, Genetic 

Structure and Diversity of Native Tench (Tinca tinca L. 1758) Populations in Hungary—Establishment of 

Basic Knowledge Base for a Breeding Program. Diversity, 14(5), 336. 

Al-Hooti, H. S., Al-Bulushi, I. M., Al-Attabi, Z. H., Rahman, M. S., Al-Subhi, L. K., & Al-Habsi, N. A., 2024, 

Efficiency of Electronic Nose in Detecting the Microbial Spoilage of Fresh Sardines (Sardinella longiceps). 

Foods, 13(3), 428. 

Borowik, P., Adamowicz, L., Tarakowski, R., Siwek, K., & Grzywacz, T., 2020, Odor detection using an e-nose 

with a reduced sensor array. Sensors, 20(12), 3542. 

Grassi, S., Benedetti, S., Opizzio, M., di Nardo, E., & Buratti, S., 2019, Meat and fish freshness assessment by 

a portable and simplified electronic nose system (Mastersense). Sensors, 19(14), 3225. 

ISO 4833-2: Microbiology of the Food Chain-Horizontal Method for the Enumeration of Microorganisms: Part 1: 

Colony Count at 30 C by the Pour Plate Technique, 2013, Available online: 

https://www.iso.org/standard/53728.html (accessed on 09.09.2024). 

Li, P., Ren, Z., Shao, K., Tan, H., & Niu, Z., 2019, Research on distinguishing fish meal quality using different 

characteristic parameters based on electronic nose technology. Sensors, 19(9), 2146. 

Peris, M., & Escuder-Gilabert, L., 2009, A 21st century technique for food control: electronic noses. PubMed. 

Recuperado de https://pubmed.ncbi.nlm.nih.gov/ 

Wu, K., Debliquy, M., & Zhang, C., 2023, Metal–oxide–semiconductor resistive gas sensors for fish freshness 

detection. Comprehensive Reviews in Food Science and Food Safety, 22(2), 913-945. 

120


	30sanchez.pdf
	Evaluation of the Shelf Life of Fresh Fish Using an Electronic Nose