Trends in Food Science & Technology 156 (2025) 104872

Contents lists available at ScienceDirect

Trends in Food Science & Technology

journal homepage: [Link]/locate/tifs

Machine learning-enhanced electrochemical sensors for food safety:

Applications and perspectives
Wajeeha Pervaiz a , Muhammad Hussnain Afzal c, Niu Feng a , Xuewen Peng a , Yiping Chen a,b,*
a
College of Food Science and Technology, Huazhong Agricultural University, 1 Shizishan Street, Wuhan, China
b
Academy of Food Interdisciplinary Science, School of Food Science and Technology, Dalian Polytechnic University, Dalian, 116034, Liaoning, China
c
Hubei Key Laboratory of Material Chemistry and Service Failure, Key Laboratory of Material Chemistry for Energy Conversion and Storage (Ministry of Education),
Hubei Engineering Research Center for Biomaterials and Medical Protective Materials, School of Chemistry and Chemical Engineering, Huazhong University of Science
and Technology, 1037 Luoyu Rd, Wuhan, China

A R T I C L E I N F O A B S T R A C T

Keywords: Background: Food safety is a critical global concern that directly impacts human health and well-being. Elec­
Electrochemical sensors trochemical sensors have garnered considerable interest for detecting contaminants in food due to their sensi­
Machine learning tivity and selectivity; however, issues such as sensor instability and electrode fouling limit their effectiveness.
Food contaminants
The integration of machine learning (ML) into electrochemical sensing offers a transformative approach,
Food safety monitoring
enhancing sensor performance, stability, and data processing capabilities while enabling real-time monitoring.
Real-time monitoring
Scope and approach: This review succinctly explores the use of ML-enhanced electrochemical sensors specifically
for food safety applications. Initially, various ML algorithms applicable to electrochemical sensor technology for
food safety monitoring are discussed. The review then highlights the application of ML-enhanced sensors in
detecting food-related contaminants, such as pesticides, pharmaceutical residues, heavy metals, microorganisms,
artificial dyes, and phenolic compounds. Finally, it addresses the challenges and future prospects in advancing
electrochemical sensors for food safety, emphasizing the potential of appropriate ML algorithms to improve in-
situ food safety monitoring.
Key findings and conclusions: The integration of ML with electrochemical sensors improves their sensitivity,
selectivity, and stability, addressing issues like electrode fouling. ML algorithms such as support vector machines,
artificial neural networks, and random forests effectively detect food contaminants like pesticides, heavy metals,
and microorganisms. ML also enables real-time data processing for quick, accurate detection of trace-level
contaminants. However, challenges remain in sensor calibration, data reliability, and the need for high-
quality training datasets. Future research should focus on enhancing sensor robustness, refining ML models
for improved accuracy, and advancing the commercialization of ML-enhanced sensors for food safety monitoring.

1. Introduction contribute to widespread food safety concerns, making food contami­

nation a global issue. To address these challenges, it is essential to
Food contaminants pose critical threats to human health, impacting implement innovative technologies for real-time monitoring of food
food quality and safety. Heavy metals, pesticides, pathogens, and toxins contaminants, particularly through advanced sensing techniques, to
contaminate food sources directly or indirectly, entering the food chain effectively manage food safety risks (Bharti et al., 2024).
through plants, livestock and irrigation water, causing potential long- Various traditional approaches, including chromatographic and
term health issues such as kidney damage, neurological disorders and spectroscopic techniques are extensively applied for food contaminants
developmental problems (Afzal et al., 2025; Sweta & Singh, 2024). The detection. Chromatographic techniques, such as Gas chromatography
World Health Organization (WHO) estimates that foodborne illnesses (GC), High-performance liquid chromatography (HPLC) and Thin-layer
affect 600 million individuals each year worldwide, resulting an chromatography (TLC) (Chu & Zi, 2024; Okoye et al., 2024), are highly
approximately 33 million deaths (Du et al., 2022). Additionally, poor effective for separating complex mixtures and detecting trace level of
handling practices, contaminated water, and globalized food trade contaminants from food samples. Similarly, spectroscopic methods like

* Corresponding author. College of Food Science and Technology, Huazhong Agricultural University, 1 Shizishan Street, Wuhan, China.
E-mail address: chenyiping@[Link] (Y. Chen).

[Link]
Received 13 November 2024; Received in revised form 2 January 2025; Accepted 7 January 2025
Available online 9 January 2025
0924-2244/© 2025 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

Fourier-transform infrared (FT-IR) spectroscopy, Fluorescence spec­ food contaminants as shown in Fig. 1. First, a general overview of
troscopy (FLS), and UV–visible (UV–Vis) (Zein et al., 2023) are robust electrochemical sensors is provided followed by a discussion on various
tools for qualitative and quantitative analysis. They are characterized by ML algorithms and their advantages. Furthermore, the review highlights
extreme sensitivity, accuracy, and broad-spectrum detection of chemical the application of ML-enhanced electrochemical sensors for detecting
contaminants, which are essential features of these diagnostic methods different food contaminants, including pesticides, pharmaceutical resi­
in food safety monitoring. Other techniques, like Enzyme-linked dues, heavy metals, bacteria, food dyes, and phenolic compounds.
immunosorbent assay (ELISA) and Polymerase chain reaction (PCR) Additionally, it discusses the challenges and potential solutions for
(Weichman et al., 2024), are commonly used as well, offering reliable implementing ML in food safety sensors and outlines future prospects.
data particularly for biological contaminants such as pathogens or al­
lergens. These methods are highly accurate and are widely considered 2. ML-based electrochemical sensors for food safety
the benchmark for many foods safety tests.
Despite their numerous strengths, these conventional techniques are ML a key part of artificial intelligence, allows algorithms and soft­
encumbered by certain constraints that hinder their use for on-site food ware to predict outcomes from data without needing specific in­
safety monitoring. Although, Chromatographic and spectroscopic tech­ structions. It focuses on finding meaningful patterns in large datasets (a
niques are often time-consuming, require skilled personnel, and involve process called “learning”) using advanced computational methods. Once
expensive equipment, which can make them less practical for rapid, these patterns are learned, ML applies this knowledge to new data
real-time analysis (Shenashen et al., 2022; Yang et al., 2022). Moreover, making accurate predictions. When combined with electrochemical
the data generated from these methods can be complex and difficult to sensors ML revolutionizes data analysis by improving the ability to un­
interpret, requiring specialized knowledge and advanced training. derstand complex data. This powerful combination greatly boosts sensor
Additionally, many of these techniques are labor-intensive and not al­ performance and expands their practical applications.
ways suitable for large-scale or field-based applications (Caratti et al.,
2024; Patel et al., 2021). Therefore, there is a growing need for alter­ 2.1. Concept of machine learning
native, rapid, cost-effective, and efficient methods, such as sensor-based
technologies, which can overcome the limitations of traditional methods Since Frank Rosenblatt introduced the perceptron in 1957, ML and
and facilitate on-site, real-time food contaminant detection. artificial neural networks (ANN) have captured significant attention
Within this context, electrochemical sensors are a significant sub­ (Rosenblatt, 1958). By the late 1980s, neural networks became a central
group of chemical sensors that transform a chemical or physical prop­ focus of ML, thanks to breakthroughs in learning methods and network
erty of a system to an output signal proportional to the analyte architectures (Karhunen et al., 2015, pp. 125–142; Sarker, 2021). Over
concentration with the necessary attributes to effectively meet the de­ the last two decades, ML has evolved from a laboratory experiment into
mands of on-site food safety monitoring in terms of fast response, high a commercially viable technology (Yadav, 2020). It is now recognized as
sensitivity, good portability and low cost (Afzal et al., 2025; Jubeen a computational science dedicated to creating algorithms that can
et al., 2024). These advancements have made electrochemical sensors analyse and interpret patterns enabling machines to learn, reason, and
function in the identification of organic pollutants in groundwater make decisions much like the human brain (Mousavizadegan et al.,
(Coronado-Apodaca et al., 2023), the quantification of trace metals in 2023; Puthongkham et al., 2021). By feeding these models vast amounts
natural water bodies (Srivastava et al., 2024), the surveillance of car­ of data, they can make informed decisions and continually improve
cinogens in food (Manikandan et al., 2018; Munawar et al., 2020). The through error detection and correction (Aery & Ram, 2017). The goal of
adaptability of electrochemical sensors to a wide range of detection ML is not just to build predictive models but also to create systems that
scenarios is facilitated by the diversity of sensing modalities including can learn and evolve on their own. These systems improve their per­
amperometric, potentiometric, conductometric and impedimetric sen­ formance over time as they are exposed to more data (Lei et al., 2023).
sors each tailored to specific types of analytes and analytical re­ Based on the learning system, ML can be classified into four groups:
quirements (Oladipo et al., 2023; Sakthivel et al., 2024). However, the supervised learning, unsupervised learning, semi-supervised learning,
electrochemical sensors also have problems, such as drift, poor repeat­ and reinforcement learning (Sindayigaya & Dey, 2022).
ability, surface fouling, co-matrix effects and low selectivity. Therefore, In supervised learning algorithms are trained on labelled datasets,
scientific society is searching for breakthroughs to progress the effi­ where each example is paired with an output label. Supervised learning
ciency of electrochemical sensors in food safety applications. An intel­ contains classification and regression, where the aim is to predict the
ligent electrochemical sensor can predict analyte concentration outcome for new data based on the patterns learned from labelled data
according to a decision-making system. based on input features (Saravanan & Sujatha, 2018; Ullah et al., 2020).
In recent years, the integration of artificial intelligence (AI), partic­ Algorithms such as Support Vector Machines (SVM), Decision Trees
ularly machine learning (ML), has revolutionized electrochemical sen­ (DT), Artificial Neural Networks (ANN), Convolutional Neural Networks
sors, unlocking their full potential for food safety (Kim et al., 2020; Mou (CNNs) and Recurrent Neural Networks (RNNs), stochastic gradient
et al., 2023). ML processes complex data from conventional and widely descent algorithm (SGD), Linear Discriminant Analysis (LDA) and
used sensor methods with high precision providing detailed insights for Bayesian estimation, are frequently employed to classify food contami­
effective contamination detection. ML enables smart calibration, opti­ nants, making them valuable tools in electrochemical sensing applica­
mization and result analysis, transforming traditional sensors into tions (John Lu, 2010). On the other hand, unsupervised learning deals
intelligent devices capable of predicting analytes based on with data that lacks labelled responses. Unsupervised learning includes
pre-constructed models (Mostajabodavati et al., 2024). This enhances clustering methods, dimensionality reduction strategies, anomaly
sensor specificity, enabling more accurate identification of target sub­ detection, association rule learning algorithms etc. Clustering groups
stances and improving the detection of complex patterns in food similar data points together, while dimensionality reduction reduces the
matrices (Hassan et al., 2023). Additionally, ML boosts the tracking of number of features in the data (Usama et al., 2019). Algorithms like
various analytes and their interactions, allowing for real-time, multi-­ K-Means clustering and Principal Component Analysis (PCA) (Arellano
analyte monitoring in intricate food matrices (Li et al., 2023). The Vidal & Govan, 2024), can group similar sensor responses, helping re­
combination of ML and electrochemical transduction paves the way for searchers uncover hidden patterns and relationships among various
next-generation smart sensors, significantly advancing food safety analytes (Alghanmi et al., 2022).
monitoring (Liu et al., 2022; Xue et al., 2021; Zhou et al., 2022). Semi-supervised learning bridges the gap between supervised and
This review comprehensively explores the progress of ML-enhanced unsupervised methods by employing small amounts of labelled data
electrochemical sensors in improving their performance for monitoring alongside larger sets of unlabelled data. This approach is particularly

2
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

Fig. 1. Schematic representation of machine learning-enhanced electrochemical sensing for detecting food contaminants.

useful in areas where labelling data is resource-intensive, such as challenges. Although ANN is highly flexible and can address a wide
contaminant detection (Yan & Wang, 2022). Reinforcement learning range of problems in sensor applications, it requires large datasets and
takes a different approach, where agents learn to make decisions by runs the risk of overfitting or underfitting. Therefore, optimization
interacting with their environment and receiving feedback in the form of techniques are often applied to ensure accurate prediction models
rewards or penalties. Over time, this helps the agent optimize its (Kurucan et al., 2024). In conclusion, PCA helps simplify complex sensor
decision-making process. In recent years, reinforcement learning has data, while LDA, SVM, and ANN provide robust classification and pre­
emerged as a powerful technique for adaptive decision-making, finding dictive power in electrochemical sensing, each offering unique strengths
applications in areas that require real-time optimization and complex for different aspects of sensor data analysis.
problem-solving (Martín-Guerrero & Lamata, 2021). Additionally, reinforcement learning offers a unique approach by
Now a days, ML whether supervised, unsupervised, or allowing algorithms to learn from interactions with their environment
reinforcement-based has become an essential tool across various scien­ through a feedback-based system (Vázquez-Canteli & Nagy, 2019).
tific fields, offering versatile solutions that continue to shape the future While less commonly applied in electrochemical sensing, reinforcement
of technology. learning has the potential to optimize sensor operations by enabling
real-time adjustments to parameters based on performance feedback.
2.2. The role of machine learning in shaping future sensor technologies Furthermore, deep learning, a subset of ML, employs multi-layered
neural networks to extract complex features from data. Techniques
ML executes four main tasks: classification, regression, clustering, such as CNNs and RNNs can process intricate signals and temporal data
and dimensionality reduction. ML algorithms can deal with the problem from sensors, enhancing the detection capabilities of electrochemical
of classification, clustering, and regression (Han et al., 2022). PCA is a sensing technologies.
widely used unsupervised learning technique in electrochemical sensing The classification of ML and the selection of appropriate algorithms
primarily for reducing the dimensionality of sensor data while retaining are vital for advancing electrochemical sensing. Each classification
key discriminatory information. It helps remove redundancy from the serves a distinct purpose and is tailored to specific data types and ap­
dataset, making it easier to visualize general trends (Attouri et al., plications. The integration of these ML methods enhances the perfor­
2024). However, for more accurate classification and predictive analysis mance of electrochemical sensors (Kresge, 2020), improving their
of sensor data, PCA is often combined with supervised techniques like ability to detect and classify food contaminants effectively. As ML
LDA or ANN. Supervised learning, essential in electrochemical sensor techniques continue to evolve their application in electrochemical
applications involves training models using labelled sensor data to sensing is expected to expand (Ballard et al., 2021; da Silva et al., 2020),
predict outcomes. LDA, a popular method retains information about the leading to more sensitive, efficient, and robust monitoring solutions that
class of the analyte during data transformation and enhances class contribute to public health and food safety (Ayres et al., 2021). Details
separability while minimizing intraclass variation. This approach is used on the electrochemical sensing concept, advantages, and limitations of a
to create a linear classifier for analyte discrimination and its accuracy is set of ML algorithms can be found out in Table 1.
validated through cross-validation before applying the model to un­
known sensor data. In sensor array applications, LDA is commonly 2.3. Machine learning in data analysis
applied due to its effective classification capabilities (Hassan et al.,
2024). ML is a powerful set of algorithms designed to learn from large
However, SVM another supervised technique, is also advantageous datasets, identifying patterns for data analysis (Zhong et al., 2022). It
for both classification and regression tasks. SVM seeks to find the involves four key steps: data collection, pre-processing, model training
optimal decision boundary in sensor data, dividing the n-dimensional and validation, and evaluation (Yang et al., 2023; Zhou, 2022). The
space into distinct categories (Cervantes et al., 2020). It is particularly process begins with gathering extensive historical data from sources like
useful in electrochemical sensing as it requires less training data sensors, databases, and logs. In the pre-processing phase, the data is
compared to LDA, making it highly efficient for sensor-based applica­ cleaned, removing inconsistencies such as duplicates and missing values
tions. Both SVM and ANN are powerful tools for pattern recognition in to ensure high quality for analysis. Once pre-processed, the data is
electrochemical sensing. SVM utilizes kernel functions for nonlinear divided into three subsets: training (60%), validation (20%), and testing
mapping, making data linearly separable (Güven & Şimşir, 2020), while (20%), all drawn from the same distribution for consistency. The model
ANN uses multilayer connections to handle both linear and nonlinear training phase involves using the bulk of the data to train a ML

3
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

Table 1 Table 1 (continued )

Comparison of ML-Algorithms: Advantages, disadvantages, and applications in Algorithm Advantages Disadvantages Applications in
electrochemical sensors. Electrochemical
Algorithm Advantages Disadvantages Applications in Sensing
Electrochemical hierarchy of measurements,
Sensing relationships. grouping sensor
Artificial Excellent for Requires large Pattern readings based on
Neural complex non-linear dataset, prone to recognition and similarity.
Networks problems, learns overfitting and prediction of Gaussian Models complex, Computationally Detecting multiple
(ANN) from data and requires careful analyte Mixture overlapping data expensive, requires chemical species
generalizes for tuning of concentrations Models distributions, can careful in overlapping
novel situations parameters. like Glucose level (GMM) handle mixed data initialization to electrochemical
prediction. types. avoid poor results. signals, handling
Convolutional Minimal pre- Computationally Analysing sensor mixed chemical
Neural processing, high expensive, requires arrays, classifying data in sensor
Networks accuracy in image- large datasets. different chemical arrays.
(CNN) like data. compounds. Linear Simple to Assumes linear Refining
Support Vector Effective in high- Computationally Classifying Regression understand, relationships, predictions from
Machines dimensional expensive for large electrochemical implement and requires multiple sensors
(SVM) spaces, works well datasets, sensitive data (e.g., computationally regularization with correlated
with small to kernel selection identifying gases). efficient. parameter tuning. inputs, calibrating
datasets. electrochemical
Random Forest Reduces overfitting Requires more Predicting analyte sensors with
(RF) with bagging, computational concentrations. collinear data.
robust and resources, slower classification tasks LASSO Encourages Sensitive to the Selecting key
accurate with large in real-time with complex Regression sparsity (feature regularization features from
datasets. applications. sensor data. selection), useful parameter, may sensor data,
Principal Reduces data Only linear Reducing the with high- underperform with identifying
Component dimensionality, relationships complexity of dimensional data. correlated features. important
Analysis handles noisy captured, some loss sensor data, variables for
(PCA) datasets well. of interpretability. identifying key chemical
variables in detection.
electrochemical Least Squares Reduces prediction Prone to Reducing error in
measurements. Boosting error with overfitting with predicting analyte
Partial Least Handles Assumes linear Quantifying (LSBoost) boosting, handles large learners, concentrations,
Squares (PLS) multicollinear data relationships, analytes from classification and slower with large handling noisy
well, effective for sensitive to complex regression. datasets. electrochemical
high-dimensional outliers. electrochemical data in
datasets. signals, analysing classification
overlapping tasks.
signals. Recurrent Excellent for time- Suffers from Monitoring time-
Extreme High performance Parameter tuning Enhancing sensor Neural series data, vanishing dependent
Gradient and scalable, is complex, prone performance and Networks captures sequential gradients, changes in
Boosting handles missing to overfitting with classification, (RNN) dependencies. computationally electrochemical
(XGBoost) data well. improper prediction of gas expensive. sensor outputs,
regularization. concentrations. long-term tracking
Genetic Good for Computationally Optimizing sensor of chemical
Algorithm optimization tasks, intensive, requires configuration, concentrations.
(GA) adapts to various careful fitness selecting the best Long Short- Solves the High Continuous
datasets. function design. sensor Term vanishing gradient computational monitoring of
combination for Memory problem, excellent cost, requires large time-series
detecting specific (LSTM) for long-term datasets for electrochemical
analytes. dependencies optimal data, long-term
K-Nearest Simple to Sensitive to noisy Classifying gases performance. pollutant level
Neighbours implement, no data, or chemicals prediction using
(KNN) need for explicit computationally based on sensor sensor arrays.
training. expensive for large readings, T-distributed Excellent for Computationally Visualizing
datasets. grouping similar Stochastic visualizing high- expensive, not complex sensor
electrochemical Neighbor dimensional data, suitable for large data,
signals. Embedding separates clusters datasets. differentiating
Naive Bayes Fast and Assumes Probabilistic (t-SNE) effectively. clusters of sensor
computationally independence classification of readings for
efficient, works between features, compounds, various analytes.
well with small sensitive to categorizing Elastic Net Combines the Requires tuning of Analysing
datasets. outliers. chemical Regression advantages of regularization multicollinear
compounds in LASSO and ridge parameters, may electrochemical
sensor data. regression, reduces still be sensitive to data, selecting the
K-Means Easy to implement, Requires pre- Grouping similar multicollinearity highly correlated most important
Clustering fast convergence definition of sensor readings, issues. features. sensor inputs
for small datasets. cluster number, categorizing while refining
struggles with sensor outputs for predictions.
complex clusters. unknown Particle Swarm Effective for global Prone to local Optimizing sensor
compounds. Optimization optimization tasks, optima, configurations,
Hierarchical Does not require High Discovering (PSO) can handle non- performance improving sensor
Clustering pre-defined computational relationships convex search depends on calibration for
clusters, provides a cost, sensitive to between different spaces. parameter tuning more accurate
noisy data electrochemical measurements.
(continued on next page)

4
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

Table 1 (continued ) paves the way for proactive approaches in managing food safety chal­
Algorithm Advantages Disadvantages Applications in lenges. As these technologies continue to evolve, their integration
Electrochemical promises to deliver significant advancements in ensuring safe food for
Sensing consumers and protecting public health.
Sparse Partial Handles high- Complexity in Identifying
Least dimensional data, selecting sparsity patterns among 3.1. Electrochemical detection of pesticides
Squares-DA reduces data parameters, different chemical
(sPLS-DA) dimensionality and sensitive to data groups or classes,
In modern agriculture, pesticides play a crucial role in enhancing
performs variable scaling and pre- enhancing
selection processing. classification productivity by protecting crops from diseases and pests, while also
simultaneously. accuracy for high- supporting plant growth, reducing losses, and improving overall farm
dimensional yield and quality. A wide range of pesticides is employed, including
sensor data.
organophosphates (OPs) such as methyl parathion, methyl paraoxon,
and paraoxon, insecticides like dinotefuran, thiamethoxam, clothiani­
algorithm, fine-tuning it for optimal performance. The validation data is din, imidacloprid, and fenobucarb, herbicides such as diquat, meso­
then used to adjust the model, followed by testing to ensure real-world trione, 2,4-Dichlorophenoxyacetic acid and glyphosate, fungicides like
performance. After training, the model is deployed to make predictions, carbendazim and thiabendazole, lastly, carbamates including carbaryl,
classifications, or decisions, depending on the task at hand. After fenobucarb, and carbosulfan. These substances are designed not only to
deployment, models often require tuning to enhance performance. This manage pests but also to optimize crop growth, ensuring the consistent
can be done by incorporating more data, adjusting parameters, or using production of high-quality agricultural products (Tan et al., 2021).
different features (Ashmore et al., 2021; Lu & Martínez, 2011; Tanak However, increased use and mishandling of pesticides to sustain high
et al., 2022). Continuous evaluation of the model accuracy and effi­ food production worldwide have raised significant food safety concerns
ciency is crucial for ensuring reliable outcomes. due to the potential for harmful residues to remain in food products.
To evaluate ML models, different metrics are employed based on the Here is the main aim to comprehensively explore the detection of
type of task. For classification tasks, accuracy, precision, recall, F1 score, pesticides in agricultural food products, focusing on electrochemical
and area under the curve (AUC) are common metrics. For regression sensor techniques integrated with novel ML algorithms and models
tasks, metrics like the coefficient of determination (R2), mean absolute (Nehru, Chen, & Dong, 2024). Such advanced sensing approaches are
error (MAE), mean squared error (MSE), and root mean squared error urgently needed to control potential risks to public health, ensure food
(RMSE) are used. Clustering algorithms are assessed using metrics like safety and safeguard consumers from exposure to hazardous pesticide
the silhouette score, Davies-Bouldin index and adjusted Rand index residues. In this context, Xu et al. developed a simple, low cost,
(ARI), which measure the quality of clusters based on compactness, disposable and intelligent sensing platform laser-induced porous gra­
separation, and agreement between predicted and actual clusters. These phene (LIPG) electrode for detecting maleic hydrazide (MH) residues in
evaluation steps ensure the ML model generalizes well on new data and potatoes and peanuts via DPV ranging 0.9–101.9 μM. The LIPG electrode
aligns with specific objectives (Mahanti et al., 2024). created through a one-step laser process on polyimide film, offers flex­
ibility, a 3D porous structure, and excellent conductivity requiring no
3. Applications of ML-assisted electrochemical sensor in food polishing or cleaning before use like GCE. Among the ML models ANN,
safety RF, and LS-SVM, the LS-SVM achieved the best performance with an R2
of 0.9994 and an RPD of 42.44, as illustrated in Fig. 2a (Xu et al., 2021).
Food is essential for human health providing necessary nutrients and The LS-SVM model showed excellent prediction and recovery perfor­
energy, but it can also be a source of contamination through harmful mance, making it suitable for real-world applications. Combining elec­
microorganisms, toxins and chemicals like pesticides, heavy metals, and trochemical detection with ML is advancing, enabling the development
food dyes (Li et al., 2021). Contamination can occur at various stages of of portable sensors for food safety. In another study aimed at enhancing
production including harvesting, processing and distribution. The sensor sensitivity, Sheng et al. combined electrochemical sensing with
increasing global demand for food has led to intensified agricultural ML to detect MH residues in onion, rice, potato, and cotton leaves. They
practices exposing food to a wide range of pollutants (Zein et al., 2023). synthesized carboxyl-functionalized poly(3,4-ethylenedioxythiophene)
Moreover, food is often contaminated from the environment like (PEDOT-C4-COOH) and electro-deposited copper nanoparticles (CuNPs)
polluted water and crooked or inferior soil are responsible for this sit­ onto it, creating a CuNPs/PEDOT-C4-COOH/GCE sensor. This sensor
uation. These contaminants can cause serious health issues such as exhibited excellent stability and selectivity with a wide detection range
allergic reactions, hormonal imbalances and inflammation (Ayerdurai of 0.06–1000 μM, and low LOD of 10 nM. Various established regression
et al., 2023). Advanced sensors equipped with technologies like machine methods were compared to the ANN-based approach, which was found
learning can provide rapid, accurate detection, ensuring food safety and to outperform the traditional methods. The ANN model showed an R2 of
meeting regulatory standards across regions (Parihar et al., 2024). 0.9999 in training set and 0.9937 in testing with relative errors below
The applications of ML in the sensing of food contaminants are 0.5 with recoveries of 98.48–99.10% confirming the excellent perfor­
rapidly expanding, offering innovative solutions to some of the most mance and feasibility of ANN method (Sheng et al., 2019). Thus, the
pressing challenges in food safety (Ali & Khan, 2017). By employing electrochemical sensor combined with the ANN method offers a prom­
advanced algorithms and computational techniques, ML enhances the ising smart solution for digital output in agricultural analysis of MH.
capabilities of traditional sensing methods, enabling more accurate Pesticide residues in agricultural products, such as tea, present
detection and classification of foodborne hazards and contaminants. In serious health risks to consumers and have emerged as a major concern
food safety monitoring, ML models analyse complex datasets generated for the food safety. For this purpose, Tang et al. developed an electro­
by food sensors and testing equipment to identify contaminants at trace chemical sensor powered by ML for rapid detection of carbendazim
levels helping to inform regulatory standards and safety protocols. In the (CBZ) residue in tea. The sensor composite electrode consisting of cop­
food industry, ML aids in the real-time detection of contaminants like per oxide (CuO), reduced graphene oxide (rGO), and poly N-phenyl­
pathogens and adulterants ensuring food safety, quality and compliance glycine (NPAN), CuO/rGO/NPAN electrode achieved high sensitivity
with regulations (Castle et al., 2021; Gokulkumar et al., 2024). The and selectivity with a linear detection range of 0.5–100 μM and LOD of
synergy between ML and sensing technologies not only improves the 0.19 μM by (S/N = 3) even in the presence of other pesticides. ML
efficiency and effectiveness of food safety monitoring systems but also models, including SVM and RF were applied to analyse DPV data, with
SVM achieving 100% accuracy on the original dataset and exhibits

5
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

Fig. 2. (a) The experimental data with different ranges were employed for predicting results using different ML models, and the electrochemical sensor based on
LIPG electrode for the intelligent analysis of ultra-trace MH in agro-products was developed (Xu et al., 2021; Copyright © 2021, Royal Society of Chemistry). (b) The
fabrication of MoS2/MWCNTs porous nanohybrid network with oxidase-like characteristic as electrochemical nanozyme sensor coupled with machine learning for
intelligent analysis of carbendazim in tea and rice samples (Zhu et al., 2020; Copyright © 2020, Elsevier). (c) Machine learning for the fabrication of nanohybrid
electrochemical sensor based on AgNPs/MWCNTs/GO/GCE and voltametric determination of benomyl residues in agro-products (Xu et al., 2022; Copyright © 2022,
Institute of Physics (IOP) Science).

excellent test results, correctly identifying CBZ residues (Tang et al., compared with both traditional and SVM. LS-SVM achieving R2 of
2023). This electrode with ML offers a rapid, accurate method for 0.9836 for training and 0.9909 for testing, along with an RMSEC of
detecting CBZ residues, ensuring food safety and enhancing food qual­ 0.9979, MAE of 0.4350, and an RPD of 8.71 (Xu et al., 2022). This
ity. To increase the detection range for pesticides detection from tea, Xu innovative approach is a new guideline for the synthesis of sensing
et al. proposed a highly sensitive electrochemical sensor using a nano­ materials, intelligent sensing of target analytes and demonstrating
hybrid of multiwalled carbon nanotube (MWCNTs)/graphene oxide strong potential for real-world BN monitoring in agricultural products.
(GO), and dendritic silver nanoparticles (AgNPs) for detecting benomyl Meanwhile, Zhu et al. developed an advanced electrochemical
(BN) residues in tea and cucumber. The nanohybrid was synthesized by nanozyme sensor integrated with ML for detecting CBZ residues, a
electrodepositing AgNPs on an MWCNTs/GO substrate, forming an fungicide from tea and rice. The sensor is fabricated from a molybdenum
AgNPs/MWCNTs/GO by simple mixed strategy. The sensor showed disulfide (MoS2) and MWCNTs porous nanohybrid network, created by
excellent voltametric performance with a detection range of 0.2–122.2 ultrasonically dispersing MWCNTs into a MoS2 aqueous dispersion in a
μM, a Km of 27.58 μM, and a maximum current of 62.5 μA, as presented carboxymethyl cellulose (CMC) water solution, providing oxidase-like
in Fig. 2c. ML models, including SVM and LS-SVM, were used for BN properties, high functional surface area, low electron transfer resis­
intelligent concentration analysis compared with the traditional linear tance, excellent electrode stability and high stability. CMC-MWCNTs/
regression method, while LS-SVM shows a more superior performance as MoS2 achieved a detection range of 0.04–100 μM with LOD of 7.4 nM, as

6
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

illustrated in Fig. 2b. The use of ML through ANN for the intelligent 3.2. Electrochemical detection of pharmaceutical residues
analysis of CBZ was compared with traditional regression analysis. DPV
data processed using an ANN model yielded R2 values > 0.997, an RMSE Pharmaceutical residues and antibiotics, are increasingly found in
below 0.2 and an RPD of 20.781 (Zhu et al., 2020). Results, indicates the environment due to their widespread use in healthcare, agriculture,
that ANN outperforms traditional regression models (i.e., squaring, and animal farming (Majdinasab et al., 2020; Wang et al., 2022). These
logarithm, polynomial, exponential, etc) with accuracy by effectively residues enter ecosystems through improper disposal and excretion,
predicting the CBZ concentration and demonstrated as potential tool for contaminating water, soil, and even food products like meat, fish, and
CBZ detection. milk. The presence of antibiotic residues poses significant health risks i.
To summarize above discussion, the integration of electrochemical e., allergy, hepatotoxicity, nephropathy, reproductive disorders
sensors with ML techniques offers a powerful and innovative solution for including the development of antibiotic resistance, making the need for
detecting pesticide residues in agricultural products. These approaches advanced detection technologies critical for ensuring food safety and
enhance the sensitivity, accuracy, and efficiency of monitoring systems, environmental protection (Adane et al., 2024).
ensuring food safety and supporting sustainable agricultural practices. Acyclovir (ACY), an antiviral medication, is primarily used to treat
The promising results from various studies demonstrate the potential for herpes simplex and zoster infections. While it offers significant thera­
real-world applications, offering a smarter, more reliable system for peutic benefits, it is crucial to closely monitor its use due to the potential
food safety monitoring. for adverse effects such as neurotoxicity, urticaria, malaise, and neph­
rotoxicity. To address this need, Bux et al. proposed a ML-based elec­
trochemical sensor using hydrothermally synthesized nickel manganese

Fig. 3. (a) Schematic diagram depicting the simple hydrothermal strategy adopted to synthesize NiMnO3 flower-like structures and their utilization as catalytic
surfaces for direct oxidative detection of ACV (Bux et al., 2024; Copyright © 2024, Elsevier). (b) A ML strategy to assist design of AgNPs decorated BP nanocomposites
as nanozyme sensor for intelligent analysis of clenbuterol in animal-derived agro-products (Xiong et al., 2023; Copyright © 2023, the Institute of Physics (IOP)
Science). (c) The fabrication and application of ANN for sensor (Lu et al., 2022; Copyright © 2022, Elsevier). (d) Schematic representation of machine
learning-enabled wireless microneedle array integrated screen-printed electrode-based electrochemical point-of-care device for lidocaine detection (Kadian et al.,
2024; Copyright © 2024, Elsevier).

7
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

oxide (NiMnO3) with flower like morphology provided abundant active novel ML-enabled electrochemical point-of-care device for rapid and
sites an electroactive surface area (EASA) of 0.0952 cm2 for detecting onsite detection of lidocaine. The fabricated device based on wireless 3D
ACV in water. The detection was validated across both high (15–75 μM) microneedle array integrated with GCEs for electrochemical sensing, as
and low (0.1–1.0 μM) concentration ranges using DPV and amperom­ presented in Fig. 3d featuring ultra-sharp microneedle arrays with a
etry, achieving a LOD of 1.59 nM and a sensitivity of 1.039 μA μM− 1 fluid reservoir to collect interstitial fluid and under optimal conditions
cm2, owing selective oxidation potential, excellent selectivity towards sensor exhibited a linear detection range of 1–120 μM, with LOD of 0.13
ACY, even in the presence of common interferents, confirming its suit­ μM and a sensitivity of 0.83 μA/μM. A univariate linear regression model
ability for practical applications as shown in Fig. 3a. Among ML models was applied on experimental data, achieved an R2 value of 0.98 and an
tested, ANN outperformed others like RF and XGBoost, achieving an R2 RMSE of 0.032, demonstrating good accuracy in predicting concentra­
of 0.9836 with superior error minimization (Bux et al., 2024). This study tions of analyte (Kadian et al., 2024). The developed ML model and user
highlights the sensor high sensitivity for ACY detection, potential of interface can be effectively used for onsite low detection of analyte,
ANN in predicting electrocatalytic sensor behaviour confirming its offering onsite digital readout of lidocaine concentration.
suitability for practical applications. In conclusion, the integration of ML with electrochemical sensors
In another study, Fatima et al. fabricated a ML-optimized electro­ offers a powerful and efficient approach for detecting pharmaceutical
chemical sensor as AgNPs infused MWCNTs for the sensitive detection of residues with exceptional sensitivity and accuracy. This combination
ACY. AgNPs were evenly deposited on the surface of MWCNTs using enables real-time, on-site analysis, ensuring precise monitoring across
polydopamine (PDA) functionalization. Ag/PDA-MWCNTs-based sensor food safety and provide public health applications. With enhanced
shows ACY electrooxidation achieved with an impressive LOD of 2.7408 predictive capabilities, these technologies present a scalable, reliable
nM, high sensitivity of 2.8796 μA nM⁻1, and a linear range of 5–290 nM. solution for safeguarding health and promoting sustainability. The
The sensor demonstrated high selectivity, even in the presence of performance of various ML-based sensors for pharmaceutical residue
interfering substances like lamivudine and zanamivir. Integration with detection is summarized in Table 2.
GBR model achieved the highest R2 of 0.95, indicating a strong linear
correlation between the actual and predicted voltametric current. The 3.3. Electrochemical detection of heavy metal ions and anions
sensor also exhibited excellent stability retaining 83% efficacy after 10
days, with high reproducibility (RSD = 0.33%), affirming the model Heavy metals with high atomic weight and density, between 3.5 and
robustness, predictive capabilities and its effectiveness in generalizing 7 g/cm3. The most well-known heavy metals are iron (Fe), zinc (Zn),
across diverse scenarios and setting a new benchmark for antiviral drug arsenic (As), copper (Cu), nickel (Ni), cadmium (Cd), mercury (Hg), lead
monitoring (Fatima et al., 2024). (Pb), tin (Sn), etc. Though some are essential in trace amounts for bio­
On the other hand, Clenbuterol (CLB), an adrenergic receptor agonist logical functions, such as Fe2+, Zn2+, manganese (Mn2+), and cobalt
used for asthma treatment, is often illegally added to animal feed to Co2+, but may become toxic at higher concentrations. Some heavy metal
promote lean meat production. Prolonged consumption of CLB ions (HMIs) like lead Pb2+, mercury Hg2+, arsenic As2+, and cadmium
contaminated meat can cause symptoms like muscle tremors, palpita­ Cd2+ are toxic in nature, and their intake through food has become an
tions, and dizziness in humans. To detect CLB residues from pork meat increasing global ecological and public health concern (Baassiri et al.,
and pig serum, Xiong et al. designed a ML-assisted electrochemical 2024). These metals enter the environment through industrial processes,
sensor with unique structural characteristic of black phosphorene (BP), mining, waste disposal, and agricultural practices, leading to contami­
Nafion (Nf), MWCNTs and AgNPs. To increase stability and electro nation of water and soil. Heavy metals may cause nausea, diarrhea,
catalytic capacity of BP, AgNPs decorated on BP surface and hybridize vomiting, or adverse reactions for low-level or short-duration exposure,
with MWCNTs containing Nf. exhibiting excellent performance with a once they have invaded the human body through ingestion (drinking or
detection range of 0.05–30.0 μM, and a LOD of 3.2 nM. A BP-ANN model eating), inhalation, or skin contact (Sulthana et al., 2024). Advanced
with GA to optimize sensor performance for CLB detection. The current detection techniques are essential for managing heavy metal pollution
response with BP-ANN-GA boosted by 9.02%. LS-SVM implemented to and reducing its impact. Continuous efforts have been made to trace
establish quantitative analysis model for CLB and achieved an R2 of heavy metals and anions in food and water.
0.9977, RMSE of 0.0303 and MAE of 0.0225 with average recoveries Leads ions (Pb2+) that seriously affect children growth and mental
from real sample between 98.66 to 101.67%, as illustrated in Fig. 3b development and the maximum level of lead in drinking water set by the
(Xiong et al., 2023). These results indicate that the electrochemical World Health Organization was 0.01 mg/L. So, the determination of
sensor, combined with ML for intelligent analysis, is both feasible and Pb2+ in aquatic systems was required with high efficiency. Recently, Su
practical for detecting CLB in animal-derived food products. et al. designed ML based electrochemical sensor for detecting Pb2⁺ in
Furthermore, Niclosamide (NA) and lidocaine contain an amide water, utilizing tea residue as raw material to produce nano-biochar
group, which is crucial to their function. Niclosamide treats parasitic (NBC). NBC was in situ reshaped on the electrode surface to obtain
infections, while lidocaine is used as a local anaesthetic and antiar­ electro reduced NBC (ErNBC) via CV with enhanced material properties,
rhythmic. Despite sharing the amide group, they have different mech­ including size reduction, improved electrical conductivity, increased
anisms and uses. Recent studies have focused on the development of response current for Pb2⁺ detection with a detection range of 80–1200
advanced detection technologies for both drugs. For instance, Lu et al. μg/L and LOD of 0.49 μg/L, as depicted in Fig. 4a. ML specifically RF
focused on a ML-assisted electrochemical sensor and supercapacitor model optimized sensor performance achieving an R2 of 0.998 MSE of
based on carbonized metal-organic frameworks (C-ZIF-67) for detecting 0.0013 and accurately predict Pb2+ concentration, showing superiority
NA, as illustrated in Fig. 3c. The C-ZIF-67, fabricated through pyrolysis, over the other two models (linear regression and K-nearest neighbor)
showed good conductivity and stability, high effective surface area, (Su et al., 2024). Due to the limited data samples in this experiment, the
improved catalytic performance (196.6-fold) over bare GCE. The sensor integrated ML model could not fully capitalize on its potential but pro­
achieved a detection range of 1 nM–9 μM with a low LOD of 0.3 nM. An vide satisfactory recovery results. To overcome this challenge, Yao et al.
ANN model enhanced prediction accuracy, achieving an R2 of 0.9954, developed simple, efficient and advanced ML based electrochemical
RMSE of 0.5524, and MAE of 0.1262 (Lu et al., 2022). The ML model via sensor combined with BiFeO3/MXene composite for detecting Pb2+ in
ANN offer theoretical support for the electrochemical application of ML water. The interconnected material composite possessed high conduc­
in both supercapacitors and sensors by reducing data reading errors. It tivity, abundant active sites and high accumulation of Pb2+ on sensor
also lays the foundation for creating self-storage energy systems for surface offering a wide linear detection range of 0.0001–3000 μg/L, LOD
future agro-food safety monitoring in outdoor environments. of 0.0001 μg/L and high reproducibility. OED-ML method optimized
Similarly, Kadian et al. developed a rapid, sensitive, user-friendly, sensor performance and greatly simplified experimental procedure

8
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

Table 2
ML-based electrochemical sensors for the detection of food contaminants.
Sensing material ML Real sample Analyte Concentration LOD Accuracy R2 value Ref.
algorithm Range (%)

Pesticides
Laser-induced porous LS-SVM Peanut, potato Maleic hydrazide 0.9–101.9 μM 0.1 μM – 0.9994 Xu et al.
graphene (LIPG) electrode (2021)
CuO/rGO/NPAN SVM Tea Carbendazim 0.5–100 μM 0.19 μM 100 – Tang et al.
(2023)
AgNPs/MWCNTs/GO LS-SVM Tea, cucumber Benomyl 0.2–122.2 μM 0.02 μg/ – 0.9836 Xu et al.
mL (2022)
MoS2/MWCNTs ANN Tea, rice Carbendazim 0.04–100 μM 7.4 nM – 0.997 Zhu et al.
(2020)
CuNPs/PEDOT-C4-COOH ANN Onion, rice, Maleic hydrazide 0.06–1000 μM 10 nM 99 – Sheng et al.
potato, cotton (2019)
leaves
Pharmaceutical residue
NiMnO3 ANN Water Acyclovir 0.1–1.0 μM 1.59 nM – 0.9836 Bux et al.
(2024)
AuNPs/CHI/GCE CatBoost Water Ciprofloxacin 0.0–1.25 μM 0.049 μM – 0.9985 Ahmed et al.
(CB) (2024)
AgNPs/MWCNT GBR Wastewater Acyclovir 5–290 nM 2.7408 83 0.95 Fatima et al.
nM (2024)
BP-AgNPs/MWCNTs-Nf LS-SVM Pork Clenbuterol 0.05–30.0 μM 3.2 nM – 0.9977 Xiong et al.
(2023)
C-ZIF-67 ANN Water, drug Niclosamide 1 nM-9 μM 0.3 nM – 0.9954 Lu et al.
samples (2022)
Graphene/SPCE RMSE Interstitial fluid Lidocaine 1–120 μM 0.13 μM – 0.9817 Kadian et al.
(2024)
Heavy metals
Nafion/ErNBC/GCE RF Water Pb2+ 80–1200 μg/L 0.49 μg/L – 0.998 Su et al.
(2024)
Nickel hexacyanoferrate CNN Seawater Cs+
1 ppb to 1000 ppm 10 ppb 100 ​ Lee et al.
(NiHCF) (2024)
2+
BiFeO3/Ti3C2 BPANN Lake water Pb 0.0001–3000 μg/L 0.0001 99.5 – Yao et al.
μg/L (2023)
Ag2O-BiOBr Naïve Bayes Water Ni, Cu – – 93.2 0.98 Kailasam et al.
model (2024)
PEDOT-C@Cu-NPs XGBR Pickle NO−2 5–580 μM 3.91 μM 95.3 – Abbas et al.
model (2024)
Organic dyes
AgNPs/SSN CNN Food Sudan red 0.1–20 μM 10 nM 97 – Sun et al.
(2024b)
AgNPs-Co3O4/fCNTs ANN Fanta, pineapple Sunset yellow 0.06–60 μM 0.8 nM 100 0.99 Balram et al.
juice (2023)
Graphite intercalation ANN Wastewater Methyl orange – – 98 0.909 Ganthavee
compound (GIC) et al. (2024)
Phenolic compounds
CoWO4/AC ANN River water Tetrabromobisphenol A 0.1–1.0 μM 0.024 μM – 0.9659 Jawaid et al.
(2024)
Fe-MOF/rGO ANN Water Penta-chlorophenol 50 mM-200 μM 75.63 nM – 0.99 Meskher et al.
(2022)
Zn-Co MOF/Ti3C2/Fe3O4- ANN Silage Mycophenolic acid 3.21 × 10− 8- 2.1 × – 0.96 Ge et al.
MGO 8.9856 × 10− 6 mol/ 10− 8 mol/ (2022)
L L
MIL-101(Cr)–NH2 CNN Coffee powder, Caffeic acid 0.5–100 μM 0.011 μM 100 – Kavya et al.
red wine (2024)
Bacteria
Gallium-indium (eGaIn) MLP-NN Food, Escherichia coli (EC) 102–109 CFU/mL – 97 0.99 Aliev et al.
alloy environment (2024)
sample
3-Aminophenylboronic acid RF Seawater, milk Escherichia coli 1.0 × 101–106 CFU/ – – 0.998 Wang et al.
(3-APBA) monomer mL (2023)
– – – Staphylococcus aureus – – – 0.994 –
(SA)
– – – Vibrio parahaemolyticus – – – 0.996 –
(VP)
Adulterants
Multi-electrode system ELM model Alcohol, water Liquor adulteration – – 90 >0.97 Sun et al.
(2024a)
Graphene composite CNN beef, lamb, pork, Meat adulteration – – 99 – Fu (2024)
chicken
Silicon chip SSDL Goat milk Milk adulteration – – >94 – Xue and Zhao
(2024)
Others
MWCNT/AuNPs ANN Stimorol Xylitol 9.9 × 10− 6-2.9 × 9.8 × – 0.9992 Uwaya et al.
10− 5 pM 10− 6 pM (2022)
Fc-GO-MWCNT ANN Pork Palmitic acid 1 nM-1 mM 0.48 nM – 0.995 Zhang et al.
(2024)

9
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

Fig. 4. (a) Schematic diagram of Preparation of NBC and in situ electrochemical reshaping of NBC to prepare ErNBC-based sensor for selective detection of Pb2+ (Su
et al., 2024; Copyright © 2024, Elsevier). (b) Schematic illustration showing road map adopted for the synthesis of PEDOT-C@Cu-NPs (Abbas et al., 2024; Copyright
© 2024, Elsevier). (c) Machine learning strategy-incorporated BiFeO3/Ti3C2 MXene electrochemical sensor for simple, rapid and highly sensitive detection of Pb2+
(Yao et al., 2023; Copyright © 2023, Elsevier).

achieving 99.5% accuracy and an MSE of 0.00002 with recovery rates of Nitrite (NO−2 ) a common nitrogen-containing compound, is widely
98.79–101.3% as shown in Fig. 4c (Yao et al., 2023). The prepared used as an additive and preservative in the food, drug, and helps prevent
sensor demonstrated excellent performance, high sensitivity, excellent bacterial growth in processed meats. NO−2 is also extensively used as a
selectivity and ML optimization system offers new insights for advancing fertilizer in agriculture, leading to significant residual NO−2 levels in
heavy metal ions detection sensors. water and soil. Higher concentration in these environments may in­
In addition, Kailasam et al. fabricated a nano sensor coupled with crease concentrations in vegetables and can be produced through
electrochemical transducer for analyzing trace nickel and copper metal ammonia oxidation or nitrate reduction, NO−2 is toxic and classified as a
ions (Ni2+, Cu2+) in aqueous media. They proposed a nanocomposite of hazardous substance, Abbas et al. developed a sensitive electrochemical
silver oxide-bismuth oxy bromide coated with Nf (Ag2O-BiOBr) and sensor through green approach using poly(3,4-ethylenedioxythiophene)
sensor showed excellent sensitivity and stability. DPV study indicated PEDOT functionalized carbon matrix suspended copper nanoparticles
increasing current with Ni2+ and Cu2+ different concentration provided (C@Cu-NPs). The nanocomposite (PEDOT-C@Cu-NPs) electrode has
a linear response with a correlation coefficient of 98%. Further, ML al­ shown excellent electro-oxidation capability towards NO−2 due to highly
gorithms such as Naïve Bayes, ANN, SVM and DT are employed for Ni2+ exposed Cu and high π-electron density as illustrated in Fig. 4b. The
and with Naïve Bayes achieving the highest accuracy of 93.2% found sensor showed high sensitivity, a linear detection range of 5–580 μM and
better among other algorithms with an R2 of 0.98 (Kailasam et al., a LOD of 3.91 μM with excellent selectivity against interference yielding
2024). The proposed sensor using DPV techniques combined with ML recovery rates of 94.28–96.72%, and the sensor maintained 95.3% of its
models, ensures superior detection of Ni2+ and Cu2+ in aqueous solu­ response after 10 days. ML models including XGBR, RF and DTR opti­
tions, offering higher detection accuracy providing a robust and mized sensor performance. ML optimized designed sensor has shown
affordable solution for water quality assessment. exceptional electrocatalytic activity in terms of sensitivity, selectivity,

10
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

repeatability, and reproducibility when it was subjected to NO−2 solution their safety is a growing concern (Li et al., 2020).
(Abbas et al., 2024). The developed electrode can successfully apply for Synthetic dyes are extensively used worldwide although they insti­
the real-time monitoring of NO−2 from spiked pickled vegetable extract gate adverse environmental and health issues at higher concentrations.
thus implying the suitability of the designed electrode for NO−2 moni­ A reliable and ultrasensitive electrochemical sensor for nanomolar
toring from vegetables. determination of common food colorant Sudan Red I and sunset yellow.
The reviewed studies demonstrate the promising capabilities of ML- Electrochemical sensors offer a sensitive and rapid method for detecting
assisted electrochemical sensors for detecting harmful substances with these dyes, while ML algorithms enhance the analysis and classification
high sensitivity and accuracy. By integrating ML models like RF and of complex sensor data (Kaya et al., 2021; Yu et al., 2022). Recently, Sun
ANN, sensor performance is significantly enhanced, ensuring better et al. presented a novel approach for food safety by developing a rapid
precision and reduced errors. While challenges in sensor stability and and precise method for detecting Sudan Red I by combining electro­
interference remain, future advancements should focus on improving chemical sensors with ML. Using a modified stainless-steel needle (SSN)
robustness, portability, and real-time monitoring for large-scale appli­ electrode enhanced with AgNPs investigated three distinct CNN models
cations. The summarized performance of various ML-based sensors for Inception V1, ResNet-50, and SqueezeNet V1.1. The AgNPs/SSN elec­
heavy metal detection is presented in Table 2. trode displays electrochemical activity towards Sudan Red I achieving a
linear detection range of 0.1–20 μM and 10 nM detection limit. Notably,
3.4. Electrochemical detection of organic dyes the Inception V1 and ResNet-50 networks exhibit superior predictive
capability compared to SqueezeNet V1.1 with 97% accuracy showcasing
Organic dyes are synthetic compounds, used in textile and food in­ their potential to solve complex relationships within electrochemical
data (Sun et al., 2024b). The incorporation of ML not only enhanced the
dustries, are known for their vibrant colours and chemical stability,
posing environmental and health risks. Textile dyes, such as azo dyes, precision of detection system but also emphasized the value of inter­
disciplinary collaboration in tackling the intricate challenges of food
contribute significantly to water pollution due to their resistance to
degradation, while food dyes are subject to strict safety regulations to safety.
Meanwhile, Balram et al. developed an advanced electrochemical
prevent adverse health effects. Food dyes are synthetic or natural colour
additives used to enhance the appearance of food and beverages. They sensor for detecting sunset yellow utilizing a ternary nanocomposite of
AgNPs doped with cobalt oxide (Co3O4) embedded in fCNTs and applied
can improve visual appeal, help indicate flavour (Shaikshavali et al.,
2020), and compensate for colour loss during processing or storage, but on SPCE surface (AgNPs-Co₃O₄/fCNTs/SPCE) from beverages like Fanta

Fig. 5. (a) Schematic diagram sowing AgNPs-Co3O4/fCNTs and detection of commonly used synthetic dye sunset yellow FCF (Balram et al., 2023; Copyright © 2023,
Elsevier). (b) Schematic illustration depicting the formation of CoWO4 composite with activated carbon and their utilization as an electrochemical platform for
oxidative detection of TBPA pollutant in PBS buffer (Jawaid et al., 2024; Copyright © 2024, Elsevier). (c) Schematic Au/MIL-101(Cr)–NH2 composite synthesis
diagram for CA detection and collaborative AI and computer vision workflow (Kavya et al., 2024; Copyright © 2024, Elsevier). (d) The platform consists of an
electrochemical cell associated with Keithley 6430 device and a connected computer with a neural network for analysing the results (Aliev et al., 2024; Copyright ©
2024, Elsevier). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

11
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

and pineapple juice. This sensor achieved a linear detection range of showcases ML potential in analyzing electrochemical data and high­
0.06–60 μM with LOD of 0.8 nM and high sensitivity of 8.79 μA μM− 1 lights how ML optimizes sensor performance for efficient pollutant
cm− 2. The proposed sunset yellow sensor achieves one of the lowest detection and dynamic monitoring.
detection limits reported to date. The performance of sensor integrated Similarly, Meskher et al. fabricated an electrochemical sensor using a
with ANN model achieved best performance with R2 of 0.99 indicating reduced rGO and metal-organic framework (MOF) nanocomposite for
excellent correlation between the true response and predicted response. detecting penta-chlorophenol (5-CP). This novel nanocomposite rGO/
The MSE of the developed ANN model for intelligent sensing of sunset MOF demonstrated a linear range of 50 μM–200 μM, sensitivity of 3.4
yellow is just 0.29 with a maximum error of 1.32 indicating excellent nA nM⁻1 with LOD and LOQ of 75.63 nM and 254.54 nM respectively.
performance of sensor, as displayed in Fig. 5a (Balram et al., 2023). The main parameters CV, EIS, SWV of sensor exhibited optimal perfor­
Based on extensive experimental results, we conclude that the developed mance. An ANN model was developed to analyse experimental data and
electrochemical sensor, using AgNPs-Co3O4/fCNTs/SPCE efficiently achieved R2 of 0.99, RMSE of 9.4744 × 10− 12 with high accuracy in
detects sunset yellow. This study demonstrates the effectiveness of predicting 5-CP concentrations (Meskher et al., 2022). These promising
nanocomposite sensors with ML for rapid and reliable food safety results of the sensor support the development of a low-cost 5-CP sensing
monitoring. system for in-field screening of aquatic environments, enabling efficient
On the other hand, textile, printing, and dyeing industries are among detection of hazards substances. Meanwhile, Ge et al. introduced a
the largest sources of dye wastewater, producing approximately 0.7 portable ML-based electrochemical nano sensor for detecting myco­
million metric tons of chemical dyes annually, which accounts for phenolic acid (MPA) in variable-pH silage environment. This sensor
17–20% of global water pollution. The presented study by Ganthavee used Zn-Co MOF integrated with Ti3C2 MXene and Fe3O4-magnetic
et al. investigated the electrochemical treatment of methyl orange (MO) graphene oxide (MGO) for enhanced electrocatalytic performance. The
dye in textile wastewater using graphite intercalation compound (GIC) Zn-Co MOF/MXene/Fe3O4-MGO nanohybrid composite offers a linear
particle electrodes. The process achieved 98% MO removal within 30 range 3.21 × 10− 8-8.9856 × 10− 6 mol/L with LOD of 2.1 × 10− 8 mol/L
min at a current density of 15 mA/cm2, energy consumption of 3.62 and good recovery rates between 104 and 107% across various samples.
kWh/kg, 79.53% of current efficiency and a half-life of 4.66 min. GIC An ANN model predicts MPA concentrations based on pH input,
particles enhanced electrooxidation and hydroxyl radical generation achieving R2 of 0.96, RMSE of 0.5883 and RPD of 5.0223 indicating
boosting treatment efficiency. ML models including ANN, SVM, and RF excellent accuracy (Ge et al., 2022). This rapid portable solution excels
that optimized the sensor performance with ANN achieving the highest in monitoring MPA in complex conditions reinforcing the role of elec­
predictive efficiency R2 of 0.909 with the lowest uncertainty and vari­ trochemical sensors in sensitive and accurate detection through ML
ability level. Monte Carlo simulations validated the model by confirming integration.
ANN robustness (Ganthavee et al., 2024). The Future studies should Moreover, Kavya et al., developed a ML based electrochemical sensor
focus on evaluating other robust classifiers to optimize electricity costs using a composite of 1D gold nanofibers decorated 2D amine-terminated
in industrial operations and will help identify the gap between pollutant chromium metal–organic framework (MIL-101(Cr)–NH2) composite for
generation and discharge sources, ultimately improving the efficacy and the determination of caffeic acid (CA) in complex matrices like coffee
expanding the applicability of optimized models. The performances of powder and red wine. The gold nanofibers with a diameter of 12 nm
various ML-based electrochemical sensors for organic dye detection are enhanced the structural integrity and performance of the framework.
summarized in Table 2. The sensor exhibited a detection range of 0.5–100 μM, 0.011 μM LOD
and sensitivity of 2.53 μA/μM/cm2 with recovery rates of 96–102% in
3.5. Electrochemical detection of phenolic compounds coffee and 99–101% in wine. Experimental CV and DPV data aligned
with ML predictions highlighted the ML role in optimizing sensor per­
Phenolic compounds, or phenols, are a captivating class of chemicals formance and accurately interpreting complex data, as depicted in
characterized by one or more hydroxyl groups (-OH) attached to an Fig. 5c (Kavya et al., 2024). This study demonstrates the potential of
aromatic hydrocarbon structure. These versatile substances play crucial combining advanced materials with computational techniques for
roles in various industries, particularly in the production of plastics and high-performance electrochemical sensing of CA, surpassing traditional
resins, while also presenting significant environmental and health methods.
challenges. A notable member of this group is tetrabromobisphenol A In concluding remarks, the integration of electrochemical sensors
(TBBPA), a brominated derivative of bisphenol A (BPA). TBBPA features with ML presents a highly effective solution for detecting phenolic
multiple bromine atoms that enhance its flame-retardant properties, compounds, significantly enhancing both sensitivity and accuracy.
making it widely used in electronics and other products. However, its However, further advancements are necessary to improve the robustness
extensive application has led to its detection as a common contaminant of these sensors for long-term field applications and expand their
in indoor environments and across various environmental compart­ detection capabilities to a broader range of contaminants. Future
ments, including water, soil, and air. This dual nature of serving in­ research should focus on optimizing sensor materials and refining ML
dustrial needs while posing potential risks underscores the complexity algorithms to enable real-time, adaptive monitoring, ensuring efficient
and significance of phenolic compounds in our modern world. Some and reliable pollutant detection across diverse food matrices.
examples are ML-assisted sensors that have been applied for the detec­
tion of POPs content in water, as shown in Table 2. 3.6. Electrochemical detection of bacteria
Recently, Jawaid et al. introduced an innovative and simple method
combining cobalt tungstate (CoWO4) with activated carbon (AC) elec­ Bacteria are ubiquitous both in the environment and in the human
trochemical sensor integrated with ML to detect TBBPA in river water. body putting us in contact with countless different species daily.
Incorporating AC with CoWO4 enhances ESA by 1.8-fold and increase of Fortunately, most of these are not harmful to humans with many even
CV-based oxidation current by 2.0-fold with improving conductivity and forming a beneficial relationship with their hosts. However pathogenic
surface properties. DPV optimization confirmed robust TBPA oxidation strains that lead to human disease do exist and pose a serious threat to
with a detection limit of 0.024 μM in PBS (0.1 M, pH 5.0). The DPV public health especially considering the generally small numbers of
sensing interpreted using ML algorithms and confirmed the high bacteria required for infectivity and the ever-increasing rates of anti­
detection accuracy of the developed sensor. Among the models tested, microbial resistance in these infections. The most common bacterial-
ANN achieved the highest R2 of 0.9659 and RMSE of 0.062, with DFT derived diseases are foodborne illnesses, urinary tract infections, sexu­
confirming strong interactions between TBBPA and CoWO4, as shown in ally transmitted infections, and healthcare-associated infections (Furst
Fig. 5b (Jawaid et al., 2024). Integrating ANN with DPV sensing & Francis, 2018).

12
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

Electrochemical methods with ML are very advantageous in rapid 1DCNN. The method analysed beef, lamb, pork, chicken, and lard
field detection due to their fast signal acquisition, easy portability and adulterants in raw and cooked forms using binary mixtures with 10–50%
cheap equipment. Escherichia coli (E. coli) is one of the most common lard. Electrochemical fingerprints captured by disposable SPCEs and
pathogens that can cause diarrhea and spreads among both humans and processed with DPV and splited into 60% training, 20% validation, and
animals. Aliev et al. introduced an advanced electrochemical platform 20% testing sets. The 1DCNN achieved 95.2–98.1% accuracy with
for rapid detection of E. coli, integrating a gallium-indium (eGaIn) alloy sensitivity and specificity exceeding 95% detecting adulterants as low as
interface with a hydrogel system to preserve cells and enhance reactions, 10%, even in cooked samples. Compared to FTIR-PCA and UV-PLS, this
as depicted in Fig. 5d. MLP model was employed trained on 200 CV at approach offers better accuracy, faster processing, and minimal sample
bacterial concentrations from 102 to 109 CFU/mL and achieved an preparation (Fu, 2024). This electrochemical-1DCNN fusion provides a
impressive detection accuracy of 97% with an R2 of 0.99. (Aliev et al., cost-effective, automated solution for food authentication, ensuring
2024). This platform reduced detection time to under 30 min faster than compliance with dietary restrictions and enabling efficient on-site
the traditional 24–48 h. It provided a linear response across various testing.
concentrations and was validated with environmental samples, high­ In another study, Xue et al. presented a novel method for detecting
lighting its potential for fast, accurate, and scalable pathogen moni­ goat milk adulteration with cow milk using a SSDL combined with
toring in food safety and biomedical diagnostics. voltametric microsensors. A compact ten-electrode array captured vol­
Moreover, Wang et al. developed a cell-imprinted electrochemical tametric fingerprints and distinguished pure and adulterated milk
impedance sensor combined with ML for qualitative and quantitative samples with 2–5% precision. Autoencoder-based feature extraction and
detection of Escherichia coli (EC), Staphylococcus aureus (SA), and Vibrio an MLP regression network processed the data, trained on 88 labelled
parahaemolyticus (VP). The sensor was created by electrochemical samples and unlabelled data across 0–100% adulteration levels. This
polymerization using 3-aminophenylboronic acid as a monomer. A RF approach improved accuracy by 4.3% compared to supervised models,
algorithm was optimized to classify bacteria, enabling qualitative achieving an accuracy of >94%. It detected adulteration levels as low as
identification and semi-quantitative detection across a concentration 1%, with under 2% error for blends up to 90% cow milk (Xue & Zhao,
range of 1.0 × 101-1.0 × 106 CFU/mL. The approach demonstrated high 2024). This solution provides rapid method for on-site milk authenti­
sensitivity and accuracy for both source and real samples including cation, ensuring scalability across production chains and robust per­
seawater and milk offering a fast, affordable and reliable solution for formance in diverse environment. SSDL enables the transition from
field-based bacterial screening and diagnosis (Wang et al., 2023). controlled lab settings to real-world applications while maintaining
However, the proposed sensor currently supports qualitative and predictive accuracy and adaptability, addressing environmental vari­
quantitative detection of unknown samples containing one of the three ability and enhancing user confidence. Future research should focus on
pathogenic bacteria. It is not yet capable of measuring mixture of sam­ refining these models for broader applications, potentially transforming
ples containing all three pathogenic bacteria simultaneously. Further quality assurance practices across the food industry.
studies are needed to enable the simultaneous detection of multiple
bacteria. 3.8. Others

3.7. Electrochemical detection of adulterants Xylitol is a low-calorie sugar alcohol used as a sweetener widely used
in food, pharmaceutical, and dental products for its anti-cariogenic and
Food adulteration involves deliberate contamination or degradation antibacterial properties requires precise detection methods due to po­
of food products by substitution, omission or addition of harmful sub­ tential gastrointestinal side effects from excessive consumption. Mean­
stances thereby affecting food quality and safety. This issue is not only a while, Uwaya et al. developed a novel electrochemical sensor by
health hazard but also a significant economic burden with the WHO combining MWCNTs and AuNPs on GCE for highly sensitive xylitol
estimating that foodborne illnesses lead to approximately 420,000 detection as shown in Fig. 6a. The sensor achieved a LOD of 9.8 × 10− 6
deaths annually including treatment costs of around $100 billion pM, a 45-fold increase in sensitivity, and a linear range of 9.9 × 10− 6 to
(Thiruvengadam et al., 2024). Traditional detection methods often lack 2.9 × 10− 5 pM. Integration with ANN further enhanced predictive ac­
the precision required to detect subtle adulterants that are harmful. curacy, achieving an R2 of 0.9992. In addition, DFT results revealed that
The applications of electrochemical sensor with ML algorithms for the hydroxyl group on the second carbon as an active site for xylitol
enhancing and detecting food adulteration mainly focused on speci­ electrooxidation on AuNPs/MWCNTs/GCE, supported by monte carlo
ficity, sensitivity, and practical utility in the development of resilient and molecular dynamics simulations confirmed the stable interactions.
food safety systems. In this context, Sun et al. study presented a novel These features make the sensor an effective tool for detecting xylitol in
method for detecting liquor adulteration by combining multi-frequency consumer products (Uwaya et al., 2022). This new strategy exhibits a
pulse voltammetry with ML models. Electrochemical sensors using great potential for rapid detection of xylitol in food and dental products.
platinum, gold, nickel, chromium, tungsten, and silver electrodes eval­ Additionally, palmitic acid (PA), essential for physiological safety
uated six liquor samples (L-50 to L-100) at 1 Hz, 10 Hz, and 100 Hz, with and food quality, was detected using a surface molecularly imprinted
platinum at 10 Hz and gold at 100 Hz offering the highest differentiation polymer (MIP) electrochemical sensor. Zhang et al. introduced an
accuracy. The ELM model developed for qualitative differentiation of innovative MIP-based electrochemical sensor on an MXene surface using
liquors of varying purities achieved 90% accuracy with 36 hidden nodes, dopamine (DA) as the functional monomer. To enhance the electroactive
while the ABC-LS-SVM model used for quantitative analysis and pre­ area and conductivity the electrode was modified with AuNPs and a
dicted the liquor purity which outperformed others models with R2 > ferrocene-graphene oxide-multiwalled carbon nanotube (Fc-GO-
0.92 and RMSE <0.03. This is due to the robust cooperative mechanism MWCNT) composite. The sensor demonstrated a detection range of 1
of ABC, which offers quick convergence, resilience, and strong global nM-1mM) with LOD of 0.48 nM and LOQ of 1.61 nM. Data analysis was
optimization performance allowing for more precise parameter identi­ improved using an BP-ANN model delivered the most effective results
fication. Sixty samples of Moutai liquor adulterated with industrial R2 of 0.995, as shown in Fig. 6b (Zhang et al., 2024). This sensor was
alcohol and water were analysed, reducing data points from 222,000 to successfully tested on pork samples, confirming its practicality for
120 per electrode for faster processing (Sun et al., 2024a). This method real-world use. As the first MIP-based electrochemical sensor for PA, it
provides a fast, cost-effective, and accurate solution for liquor quality shows great potential for food quality and safety monitoring.
control, with broad potential in the food and beverage industry.
On the other hand, Fu et al. developed a rapid and precise method for
detecting meat adulteration by integrating electrochemical sensing with

13
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

Fig. 6. (a) Illustration of the fabrication of GCE/MWCNT/AuNPs and performance of ANN model (Uwaya et al., 2022; Copyright © 2022, Elsevier). (b) Preparation
of MIP-MXene, MIP-MXene/Fc-GO-MWCNT/AuNPs/SPE and Regression analysis of RBF-ANN (A) and BP-ANN (B) with ΔI as input and PA concentration as output
(Zhang et al., 2024; Copyright © 2024, Elsevier).

4. Challenges also limited, as traditional food safety testing methods are slow and
resource-intensive. Large-scale implementation of electrochemical sen­
Integrating ML with electrochemical sensors offers great potential for sors in the food industry is hindered by the incompatibility of these
improving food safety monitoring, but several challenges need to be sensors with existing monitoring systems, requiring significant adjust­
addressed for successful real-world applications. A primary issue is data ments to infrastructure. Furthermore, the changes in environmental
quality, as ML models require large, diverse, and accurate datasets, factors such as temperature and humidity, which can impact the per­
which are difficult to obtain in food safety frameworks due to the formance of the sensor and result in inaccurate or unreliable readings.
complex nature of food matrices and the presence of multiple contam­ This problem is exacerbated by the complex chemical interactions in
inants. Electrochemical data is often noisy and overlapping, creating food, which may interfere with the sensor readings. To overcome these
difficulties in distinguishing target molecules, particularly with real- challenges, adaptive algorithms that can adapt to changing environ­
world sensor exposure variability. Additionally, accuracy in ML ments and improve the robustness of models are crucial. Ethical con­
models decreases in varied environments due to the systemic drift of the cerns, including data privacy and algorithmic bias, also need to be
sensor and changing of system components. Addressing these issues addressed as these technologies collect sensitive information in food
requires robust data pre-processing, advanced noise-filtering algo­ safety monitoring.
rithms, and the development of models that can adapt to different However, to ensure the effective application of electrochemical
conditions. sensors for food safety, these challenges must be addressed through
Regulatory challenges further complicate the use of these technolo­ improvements in data processing, calibration of the sensors, and ML
gies in food safety. Food safety regulations differ by region, requiring techniques. In particular, by intending to bring these technologies to a
sensors to meet specific standards such as those set by the U.S. Food and reliable operation, real-time monitoring of contaminants can be per­
Drug Administration (FDA) or European Food Safety Authority (EFSA), formed to meet food safety standards worldwide.
which delays widespread adoption. The scalability of these sensors is

14
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

5. Conclusion Acknowledgments

Rapid industrialization, urbanization, and evolving environmental The authors are grateful for the financial support from National
challenges have increased the demand for reliable food safety moni­ Natural Science Foundation of China (32172293), Key R&D Program of
toring solutions. Electrochemical sensors have emerged as valuable al­ Hubei Province (2022BBA0057), and the Open Project of Key Labora­
ternatives to traditional laboratory methods, offering high sensitivity, tory of Detection Technology of Focus Chemical Hazards in Animal-
selectivity, and miniaturization potential. However, maintaining derived Food for State Market Regulation (NO. KF-202201).
reproducibility and stability is essential to meet the stringent re­
quirements of real-world food safety monitoring. Integrating ML with Data availability
electrochemical sensors has driven progress by enabling advanced data
analysis and improving sensor performance. This review provides a Data will be made available on request.
comprehensive overview of ML-powered sensors for detecting contam­
inants in food, with various ML algorithms enhancing detection accu­ References
racy, reliability, and data handling. Applications of ML-based
electrochemical sensors are highlighted for identifying key food con­ Abbas, W., Zafar, F., Abou Taleb, M. F., Ameen, M., Sami, A., Mazhar, M. E., Akhtar, N.,
Fazal, M. W., Ibrahim, M. M., & El-Bahy, Z. M. (2024). Machine learning trained poly
taminants, including pesticides, heavy metals, pharmaceuticals, bacteria (3, 4-ethylenedioxythiophene) functionalized carbon matrix suspended Cu
etc. Additionally, the review discusses challenges, future research op­ nanoparticles for precise monitoring of nitrite from pickled vegetables. Food
portunities, and the need for continuous innovation to improve sensor Chemistry, 460, Article 140395.
Adane, W. D., Chandravanshi, B. S., & Tessema, M. (2024). A novel electrochemical
durability and usability, paving the way for scalable solutions that meet sensor for the detection of metronidazole residues in food samples. Chemosphere,
global food safety standards and promote smarter, sustainable moni­ 359, Article 142279.
toring systems. Aery, M., & Ram, C. (2017). A review on machine learning: Trends and future prospects.
Research Cell: International Journal of Engineering Sciences, 25(63019), 89–96.
Afzal, M. H., Pervaiz, W., Asif, M., Huang, Z., Dai, J., Xu, Y., … Li, G. (2025). Engineering
6. Future perspectives MXene for electrochemical environmental pollutant sensing. Environmental Science:
Nano. [Link]
Ahmed, Y., Akter, T., Prima, M., Dutta, K. R., Mukut, S., Ahsan, M., Rahman, M. M., &
ML coupled with electrochemical sensors will revolutionize food
Hyder, M. M. Z. (2024). Advanced ciprofloxacin quantification: A machine learning
safety monitoring because these tools provide rapid, on-site detection of and metaheuristic approach using ultrasensitive chitosan-gold nanoparticle based
contaminants. Novel sensor materials such as graphene, MXene, and bio- electrochemical sensor. Journal of Environmental Chemical Engineering, Article
inspired materials will be key to improving the sensitivity, mechanical 115094.
Alghanmi, N., Alotaibi, R., & Buhari, S. M. (2022). Machine learning approaches for
properties, and durability of the electrochemical sensors in the anomaly detection in IoT: An overview and future research directions. Wireless
emerging future, which will facilitate the detection of broad-spectrum Personal Communications, 122(3), 2309–2324.
contaminants with improved sensitivity in different samples. In partic­ Ali, H., & Khan, E. (2017). Environmental chemistry in the twenty-first century.
Environmental Chemistry Letters, 15(2), 329–346.
ular, the introduction of more sophisticated machine learning tech­ Aliev, T. A., Lavrentev, F. V., Dyakonov, A. V., Diveev, D. A., Shilovskikh, V. V., &
niques, including deep learning and transfer learning, will enable the Skorb, E. V. (2024). Electrochemical platform for detecting Escherichia coli bacteria
sensors to process complex, high-dimensional datasets, generated by a using machine learning methods. Biosensors and Bioelectronics, 259, Article 116377.
Arellano Vidal, C. L., & Govan, J. E. (2024). Machine learning techniques for improving
wide range of food-like matrices, in real-time. It would improve the nanosensors in agroenvironmental applications. Agronomy, 14(2), 341.
detection of the contaminant by displaying a greater degree of accuracy Ashmore, R., Calinescu, R., & Paterson, C. (2021). Assuring the machine learning
and reliability despite the inherent variability of food products and lifecycle: Desiderata, methods, and challenges. ACM Computing Surveys, 54(5), 1–39.
Attouri, K., Mansouri, M., Hajji, M., Kouadri, A., Bouzrara, K., & Nounou, H. (2024).
environmental fluctuations. Real-time fault detection scheme for industrial chemical Tennessee eastman process.
Moreover, the scalability of relatively inexpensive sensors, will­ 2024 10th international conference on control, decision and information technologies
enable researchers to incorporate these devices into global food safety (CoDIT).
Ayerdurai, V., Cieplak, M., & Kutner, W. (2023). Molecularly imprinted polymer-based
infrastructures to support rapid implementation of real-time monitoring
electrochemical sensors for food contaminants determination. TrAC, Trends in
systems into multiple industries. However, this will facilitate the timely Analytical Chemistry, 158, Article 116830.
and potent detection of contaminants and adherence to evolving stan­ Ayres, L. B., Gomez, F. J., Linton, J. R., Silva, M. F., & Garcia, C. D. (2021). Taking the
dards of compliance. Also, Edge computing will help ML models to leap between analytical chemistry and artificial intelligence: A tutorial review.
Analytica Chimica Acta, 1161, Article 338403.
process data locally, which can result in decreased latency and faster Baassiri, K., Antonios, D., & Milane, A. (2024). Heavy metal contamination in Lebanese
decision-making. As systems become increasingly capable of self- lettuce: Quantification and health risk assessment. Journal of Food Composition and
direction, they will perform not only contaminate detection but also Analysis, Article 106871.
Ballard, Z., Brown, C., Madni, A. M., & Ozcan, A. (2021). Machine learning and
adjust to transitioning conditions, provide used insights around the computation-enabled intelligent sensor design. Nature Machine Intelligence, 3(7),
likely risks, and facilitate real-time action in food safety governance. 556–565.
The evolution of ML-driven electrochemical sensors will rely on Balram, D., Lian, K.-Y., Sebastian, N., Alharthi, S. S., Al-Saidi, H. M., Yadav, V. K.,
Kumar, D., & Kumar, V. (2023). A novel ternary nanocomposite based
cross-field cooperation between food scientists, data scientists, chemists, electrochemical sensor coupled with regularized neural network for nanomolar
and ethicists to tackle the burning questions around data privacy, al­ detection of sunset yellow FCF. Journal of Alloys and Compounds, 968, Article
gorithm transparency, and bias mitigation. If managed prudently and 171934.
Bharti, A., Jain, U., & Chauhan, N. (2024). Progressive analytical techniques utilized for
harnessed through sustained technological evolution, these sensors the detection of contaminants attributed to food safety and security. Talanta Open,
can profoundly revolutionize food safety monitoring, facilitating safer 100368.
food products, compliance with international food safety criteria, and Bux, N., Hussain, S., Küçükdeniz, T., Soomro, R. A., Mersal, G. A., Karakuş, S.,
Alhadhrami, A., & El-Bahy, Z. M. (2024). Intelligent machine learning enabled
ultimately bolstering public health and consumer protection.
sensor for acyclovir using NiMnO3 flower-like electrocatalyst. Materials Science and
Engineering B, 309, Article 117668.
Data availability Caratti, A., Squara, S., Bicchi, C., Liberto, E., Vincenti, M., Reichenbach, S. E., Tao, Q.,
Geschwender, D., Alladio, E., & Cordero, C. (2024). Boosting comprehensive two-
dimensional chromatography with artificial intelligence: Application to food-omics.
The data used to support the findings of this review are available TrAC, Trends in Analytical Chemistry, Article 117669.
from the corresponding author upon request. Castle, L. M., Schuh, D. A., Reynolds, E. E., & Furst, A. L. (2021). Electrochemical sensors
to detect bacterial foodborne pathogens. ACS Sensors, 6(5), 1717–1730.
Cervantes, J., Garcia-Lamont, F., Rodríguez-Mazahua, L., & Lopez, A. (2020).
Declaration of Competing Interest A comprehensive survey on support vector machine classification: Applications,
challenges and trends. Neurocomputing, 408, 189–215.
The authors report no declarations of interest

15
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

Chu, Z.-Y., & Zi, X.-J. (2024). Network toxicology and molecular docking for the toxicity Lee, J., Baek, K., Jeong, H., Doh, S., Kim, K., & Cho, K. H. (2024). Revolutionizing cesium
analysis of food contaminants: A case of aflatoxin B1. Food and Chemical Toxicology, monitoring in seawater through electrochemical voltammetry and machine learning.
188, Article 114687. Journal of Hazardous Materials, Article 136558.
Coronado-Apodaca, K. G., Rodríguez-De Luna, S., Araújo, R., Oyervides-Muñoz, M. A., Lei, L., Pang, R., Han, Z., Wu, D., Xie, B., & Su, Y. (2023). Current applications and future
González-Meza, G. M., Parra-Arroyo, L., Sosa-Hernandez, J. E., Iqbal, H. M., & Parra- impact of machine learning in emerging contaminants: A review. Critical Reviews in
Saldivar, R. (2023). Occurrence, transport, and detection techniques of emerging Environmental Science and Technology, 53(20), 1817–1835.
pollutants in groundwater. MethodsX, 102160. Li, C., Li, C., Yu, H., Cheng, Y., Xie, Y., Yao, W., Guo, Y., & Qian, H. (2021). Chemical
da Silva, G. S., de Oliveira, L. P., Costa, G. F., Giordano, G. F., Nicoliche, C. Y., da food contaminants during food processing: Sources and control. Critical Reviews in
Silva, A. A., Khan, L. U., da Silva, G. H., Gobbi, A. L., & Silveira, J. V. (2020). Food Science and Nutrition, 61(9), 1545–1555.
Ordinary microfluidic electrodes combined with bulk nanoprobe produce Li, H., Sheng, W., Haruna, S. A., Bei, Q., Wei, W., Hassan, M. M., & Chen, Q. (2023).
multidimensional electric double-layer capacitances towards metal ion recognition. Recent progress in photoelectrochemical sensors to quantify pesticides in foods:
Sensors and Actuators, B: Chemical, 305, Article 127482. Theory, photoactive substrate selection, recognition elements and applications.
Du, Q., Rao, R., Bi, F., Yang, Y., Zhang, W., Yang, Y., Liu, N., & Zhang, X. (2022). TrAC, Trends in Analytical Chemistry, 164, Article 117108.
Preparation of modified zirconium-based metal-organic frameworks (Zr-MOFs) Li, G., Wu, J., Jin, H., Xia, Y., Liu, J., He, Q., & Chen, D. (2020). Titania/electro-reduced
supported metals and recent application in environment: A review and perspectives. graphene oxide nanohybrid as an efficient electrochemical sensor for the
Surfaces and Interfaces, 28, Article 101647. determination of allura red. Nanomaterials, 10(2), 307.
Fatima, M., Hanif, S., Elsharkawy, E. R., Zafar, F., Zulfiqar, A., Khan, M. A., Akhtar, N., Liu, Z., Shurin, G. V., Bian, L., White, D. L., Shurin, M. R., & Star, A. (2022). A carbon
Fareed, Z., El-Bahy, Z. M., & Shafiq, Z. (2024). Design and fabrication of machine nanotube sensor array for the label-free discrimination of live and dead cells with
learning trained silver nanoparticles-infused multi-walled carbon nanotube-based machine learning. Analytical Chemistry, 94(8), 3565–3573.
sensor for antiviral drug monitoring. Microchemical Journal, Article 110921. Lu, X., Liu, P., Bisetty, K., Cai, Y., Duan, X., Wen, Y., Zhu, Y., Rao, L., Xu, Q., & Xu, J.
Fu, J. (2024). Research on meat adulteration based on one-dimensional convolutional (2022). An emerging machine learning strategy for electrochemical sensor and
neural network (1DCNN) combined with electrochemical technology. Journal of supercapacitor using carbonized metal–organic framework. Journal of
Food Measurement and Characterization, 1–9. Electroanalytical Chemistry, 920, Article 116634.
Furst, A. L., & Francis, M. B. (2018). Impedance-based detection of bacteria. Chemical Lu, J. J., & Martínez, L. (2011). International journal of computational intelligence
Reviews, 119(1), 700–726. systems. Mathware and Soft Computing: Magazine of the European Society for Fuzzy
Güven, I., & Şimşir, F. (2020). Demand forecasting with color parameter in retail apparel Logic Technology, 18(1), 55-55.
industry using artificial neural networks (ANN) and support vector machines (SVM) Mahanti, N. K., Shivashankar, S., Chhetri, K. B., Kumar, A., Rao, B. B., Aravind, J., &
methods. Computers & Industrial Engineering, 147, Article 106678. Swami, D. (2024). Enhancing food authentication through E-nose and E-tongue
Ganthavee, V., Fernando, M. M., & Trzcinski, A. P. (2024). Monte Carlo simulation, technologies: Current trends and future directions. Trends in Food Science &
artificial intelligence and machine learning-based modelling and optimization of Technology, Article 104574.
three-dimensional electrochemical treatment of xenobiotic dye wastewater. Majdinasab, M., Mishra, R. K., Tang, X., & Marty, J. L. (2020). Detection of antibiotics in
Environmental Processes, 11(3), 1–31. food: New achievements in the development of biosensors. TrAC, Trends in Analytical
Ge, Y., Camarada, M. B., Liu, P., Qu, M., Wen, Y., Xu, L., Liang, H., Liu, E., Zhang, X., & Chemistry, 127, Article 115883.
Hao, W. (2022). A portable smart detection and electrocatalytic mechanism of Manikandan, V. S., Adhikari, B., & Chen, A. J. A. (2018). Nanomaterial based
mycophenolic acid: A machine learning-based electrochemical nanosensor to adapt electrochemical sensors for the safety and quality control of food and beverages. 143
variable-pH silage microenvironment. Sensors Actuators B: Chemical, 372, Article (19), 4537–4554.
132627. Martín-Guerrero, J. D., & Lamata, L. (2021). Reinforcement learning and physics. Applied
Gokulkumar, K., Huang, S.-J., Kogularasu, S., Aljuwayid, A. M., Maheshwaran, S., & Sciences, 11(18), 8589.
Govindasamy, M. (2024). Enhanced electrochemical detection of tartrazine in Meskher, H., Achi, F., Ha, S., Berregui, B., Babanini, F., & Belkhalfa, H. (2022). Sensitive
beverages and liquid soap via nickel phosphide-adorned functionalized carbon rGO/MOF based electrochemical sensor for penta-chlorophenol detection: A novel
nanofibers. Journal of the Taiwan Institute of Chemical Engineers, 157, Article 105420. artificial neural network (ANN) application. Sensors Diagnostics, 1(5), 1032–1043.
Han, J., Pei, J., & Tong, H. (2022). Data mining: Concepts and techniques. Morgan Mostajabodavati, S., Mousavizadegan, M., Hosseini, M., Mohammadimasoudi, M., &
kaufmann. Mohammadi, J. (2024). Machine learning-assisted liquid crystal-based aptasensor
Hassan, M. M., Yi, X., Sayada, J., Zareef, M., Shoaib, M., Chen, X., Li, H., & Chen, Q. for the specific detection of whole-cell Escherichia coli in water and food. Food
(2024). Progress of machine learning-based biosensors for the monitoring of food Chemistry, 448, Article 139113.
safety: A review. Biosensors and Bioelectronics, Article 116782. Mou, J., Ding, J., & Qin, W. (2023). Deep learning-enhanced potentiometric aptasensing
Hassan, M. M., Yi, X., Zareef, M., Li, H., & Chen, Q. (2023). Recent advancement of with magneto-controlled sensors. Angewandte Chemie, 135(3), Article e202210513.
optical, electrochemical, and photoelectrochemical transducer-based microfluidic Mousavizadegan, M., Firoozbakhtian, A., Hosseini, M., & Ju, H. (2023). Machine learning
devices for pesticide and mycotoxins in food and water. Trends in Food Science & in analytical chemistry: From synthesis of nanostructures to their applications in
Technology, Article 104230. luminescence sensing. TrAC, Trends in Analytical Chemistry, Article 117216.
Jawaid, S., Sharma, B. P., Tumrani, S. H., Abbas, Z., Soomro, R. A., Karakuş, S., … Munawar, H., Mankar, J. S., Sharma, M. D., Garcia-Cruz, A., Fernandes, L. A. L.,
Nafady, A. (2024). Enhanced electrochemical oxidation and machine learning- Peacock, M., & Krupadam, R. J. (2020). Highly selective electrochemical nanofilm
assisted sensing of tetrabromobisphenol A using activated carbon facilitated CoWO4 sensor for detection of carcinogenic PAHs in environmental samples. Talanta, 219,
heterostructures. Materials Science and Engineering B, 308, Article 117546. Article 121273.
John Lu, Z. (2010). The elements of statistical learning: Data mining, inference, and Nehru, R., Chen, C. W., & Dong, C.-D. (2024). A review of smart electrochemical devices
prediction. Oxford University Press. for pesticide detection in agricultural food and runoff contaminants. Science of The
Jubeen, F., Batool, A., Naz, I., Sehar, S., Sadia, H., Hayat, A., & Kazi, M. (2024). Total Environment, Article 173360.
Mycotoxins detection in food using advanced, sensitive and robust electrochemical Okoye, C. O., Jiang, H., Nazar, M., Tan, X.-L., & Jiang, J. (2024). Redefining modern food
platform of sensors: A review. Sensors and Actuators A: Physical, Article 115045. analysis: Significance of omics analytical techniques integration, chemometrics and
Kadian, S., Sahoo, S. S., Kumari, P., & Narayan, R. J. (2024). Machine learning enabled bioinformatics. TrAC, Trends in Analytical Chemistry, Article 117706.
onsite electrochemical detection of lidocaine using a microneedle array integrated Oladipo, A. A., Oskouei, S. D., & Gazi, M. (2023). Metal-organic framework-based
screen printed electrode. Electrochimica Acta, 475, Article 143664. nanomaterials as opto-electrochemical sensors for the detection of antibiotics and
Kailasam, V., Sankararajan, R., Kailasam, M., & Suseela, S. B. (2024). Machine learning hormones: A review. Beilstein Journal of Nanotechnology, 14(1), 631–673.
assisted metal oxide-bismuth oxy halide nanocomposite for electrochemical sensing Parihar, A., Sharma, P., Choudhary, N. K., Khan, R., & Mostafavi, E. (2024). Internet-of-
of heavy metals in aqueous media. Crystal Research and Technology, 59(5), Article Things-integrated molecularly imprinted polymer-based electrochemical nano-
2300173. sensors for pesticide detection in the environment and food products. Environmental
Karhunen, J., Raiko, T., & Cho, K. (2015). Unsupervised deep learning: A short review. Pollution, Article 124029.
Advances in independent component analysis learning machines. Patel, S., Jamunkar, R., Sinha, D., Patle, T. K., Kant, T., Dewangan, K., & Shrivas, K.
Kavya, K., Kumar, R. S., Kumar, R. R., Ramesh, S., Yang, W., Kakani, V., & Haldorai, Y. (2021). Recent development in nanomaterials fabricated paper-based colorimetric
(2024). Fabrication of 1D/2D Au nanofiber/MIL-101 (Cr)–NH2 composite for and fluorescent sensors: A review. Trends in Environmental Analytical Chemistry, 31,
selective electrochemical detection of caffeic acid: Predicting sensor performance by Article e00136.
machine learning and investigating the porosity using AI and computer vision-based Puthongkham, P., Wirojsaengthong, S., & Suea-Ngam, A. (2021). Machine learning and
image analysis. Microchemical Journal, 200, Article 110490. chemometrics for electrochemical sensors: Moving forward to the future of
Kaya, S. I., Cetinkaya, A., Ozkan, S. A. F., & Toxicology, C. (2021). Latest advances on the analytical chemistry. Analyst, 146(21), 6351–6364.
nanomaterials-based electrochemical analysis of azo toxic dyes Sunset Yellow and Rosenblatt, F. (1958). The perceptron: A probabilistic model for information storage and
Tartrazine in food samples. Food and Chemical Toxicology, 156, Article 112524. organization in the brain. Psychological Review, 65(6), 386.
Kim, H., Park, S., Jeong, I. G., Song, S. H., Jeong, Y., Kim, C. S., & Lee, K. H. (2020). Sakthivel, K., Balasubramanian, S., Chang-Chien, G.-P., Wang, S.-F., Billey, W.,
Noninvasive precision screening of prostate cancer by urinary multimarker sensor Platero, J., Soundappan, T., & Sekhar, P. (2024). Editors’ choice—review—advances
and artificial intelligence analysis. ACS Nano, 15(3), 4054–4065. in electrochemical sensors: Improving food safety, quality, and traceability. ECS
Kresge, L. (2020). Data and algorithms in the workplace: A primer on new technologies. Sensors Plus, 3(2), Article 020605.
eScholarship, University of California. Saravanan, R., & Sujatha, P. (2018). A state of art techniques on machine learning
Kurucan, M., Özbaltan, M., Yetgin, Z., & Alkaya, A. (2024). Applications of artificial algorithms: A perspective of supervised learning approaches in data classification.
neural network based battery management systems: A literature review. Renewable 2018 Second international conference on intelligent computing and control systems
Sustainable Energy Reviews, 192, Article 114262. (ICICCS).
Sarker, I. H. (2021). Deep learning: A comprehensive overview on techniques, taxonomy,
applications and research directions. SN computer science, 2(6), 420.

16
W. Pervaiz et al. Trends in Food Science & Technology 156 (2025) 104872

Shaikshavali, P., Reddy, T. M., Lakshmi Narayana, A., Hussain, O., Venkataprasad, G., & qualitative and quantitative analysis of three bacteria. Sensors Actuators B: Chemical,
Venu Gopal, T. (2020). A powerful electrochemical sensor based on Fe3O4 384, Article 133672.
nanoparticles-multiwalled carbon nanotubes hybrid for the effective monitoring of Wang, X., Xie, Y., & Lin, L. (2022). Recent development of microfluidic biosensors for the
sunset yellow in soft drinks. Journal of Food Measurement and Characterization, 14, analysis of antibiotic residues. TrAC, Trends in Analytical Chemistry, 157, Article
3319–3332. 116797.
Shenashen, M. A., Emran, M. Y., El Sabagh, A., Selim, M. M., Elmarakbi, A., & El- Weichman, K., Miller, K. G., Malaca, B., Mori, W. B., Pierce, J., Ramsey, D., Vieira, J.,
Safty, S. A. (2022). Progress in sensory devices of pesticides, pathogens, coronavirus, Vranic, M., & Palastro, J. P. (2024). Analytic pulse technique for computational
and chemical additives and hazards in food assessment: Food safety concerns. electromagnetics. Computer Physics Communications, 298, Article 109096.
Progress in Materials Science, 124, Article 100866. Xiong, Y., Wu, R., Xu, L., Zhong, Y., Ge, Y., Wen, Y., Yao, H., Zhou, W., & Ai, S. (2023).
Sheng, Y., Qian, W., Huang, J., Wu, B., Yang, J., Xue, T., Ge, Y., & Wen, Y. (2019). Machine learning for design of phosphorene nanozyme sensor and its intelligent
Electrochemical detection combined with machine learning for intelligent sensing of analysis of clenbuterol in animal-derived agro-products. Journal of the
maleic hydrazide by using carboxylated PEDOT modified with copper nanoparticles. Electrochemical Society, 170(4), Article 047505.
Microchimica Acta, 186, 1–12. Xu, L., Wu, R., Zhu, X., Wang, X., Geng, X., Xiong, Y., Chen, T., Wen, Y., & Ai, S. (2021).
Sindayigaya, L., & Dey, A. (2022). Machine learning algorithms: A review. Information Intelligent analysis of maleic hydrazide using a simple electrochemical sensor
Systems Journal, 11(8), 1127–1133. coupled with machine learning. Analytical Methods, 13(39), 4662–4673.
Srivastava, R., Singh, Y., White, J. C., & Dhankher, O. P. (2024). Mitigating toxic metals Xu, L., Xiong, Y., Wu, R., Geng, X., Li, M., Yao, H., Wang, X., Wen, Y., & Ai, S. (2022). An
contamination in foods: Bridging knowledge gaps for addressing food safety. Trends emerging machine learning strategy for the fabrication of nanozyme sensor and
in Food Science & Technology, Article 104725. voltametric determination of benomyl in agro-products. Journal of the
Su, Z., Wang, J., Hu, S., Cheng, Y., Yang, Y., Zhou, S., … Yang, L. (2024). In-situ Electrochemical Society, 169(4), Article 047506.
reshaping nano-biochar on electrode surface for machine learning assisted selective Xue, Y., Ji, W., Jiang, Y., Yu, P., & Mao, L. (2021). Deep learning for voltammetric
sensing of Pb2+ in real water samples. Applied Surface Science, 665, Article 160294. sensing in a living animal brain. Angewandte Chemie International Edition, 60(44),
Sulthana, S. F., Iqbal, U. M., Suseela, S. B., Anbazhagan, R., Chinthaginjala, R., 23777–23783.
Chitathuru, D., … Kim, T. H. (2024). Electrochemical sensors for heavy metal ion Xue, D., & Zhao, H. (2024). Machine learning and electrochemistry techniques for
detection in aqueous medium: A systematic review. ACS Omega, 9(24), detecting adulteration of goat milk with cow milk. Journal of Food Measurement and
25493–25512. Characterization, 1–8.
Sun, X., Liu, F., & Xue, X. (2024a). Identification of liquor adulteration based on machine Yadav, D. (2020). Machine learning: Trends, perspective, and prospects. Science of The
learning and electrochemical sensor. Journal of Food Measurement and Total Environment, 349, 255–260.
Characterization, 1–8. Yan, J., & Wang, X. (2022). Unsupervised and semi-supervised learning: The next frontier
Sun, X., Liu, F., & Xue, X. (2024b). Machine learning combined with electrochemical in machine learning for plant systems biology. The Plant Journal, 111(6), 1527–1538.
sensor for rapid detection of Sudan Red I in food. Journal of Food Measurement and Yang, D., Ngoc, K. M., Shin, I., & Hwang, M. (2023). DPReLU: Dynamic parametric
Characterization, 18(1), 95–104. rectified linear unit and its proper weight initialization method. International Journal
Sweta, & Singh, B. (2024). A review on heavy metal and metalloid contamination of of Computational Intelligence Systems, 16(1), 11.
vegetables: Addressing the global safe food security concern. International Journal of Yang, Z., Zhang, W., Yin, Y., Fang, W., & Xue, H. (2022). Metal-organic framework-based
Environmental Analytical Chemistry, 104(16), 4762–4783. sensors for the detection of toxins and foodborne pathogens. Food Control, 133,
Tan, J., Wang, T., Li, Y., Xu, S., Chen, S., & Hao, H. (2021). Review on functionalized Article 108684.
magnetic nanoparticles for the pretreatment of organophosphorus pesticides. Green Yao, H., Wu, R., Zou, J., Liu, J., Peng, G., Wang, X., … Lu, L. (2023). A machine learning
Processing and Synthesis, 10(1), 485–498. strategy-incorporated BiFeO3/Ti3C2 MXene electrochemical platform for simple,
Tanak, A. S., Sardesai, A., Muthukumar, S., & Prasad, S. (2022). Simultaneous detection rapid detection of Pb2+ with high sensitivity. Chemosphere, 340, Article 139728.
of sepsis host response biomarkers in whole blood using electrochemical biosensor. Yu, L., Chang, J., Zhuang, X., Li, H., Hou, T., & Li, F. (2022). Two-dimensional cobalt-
Bioengineering Translational Medicine, 7(3), Article e10310. doped Ti3C2 MXene nanozyme-mediated homogeneous electrochemical strategy for
Tang, M., Guo, J., & Shen, Z. (2023). Rapid detection of carbendazim residue in tea by pesticides assay based on in situ generation of electroactive substances. Analytical
machine learning assisted electrochemical sensor. Journal of Food Measurement and Chemistry, 94(8), 3669–3676.
Characterization, 17(6), 6363–6369. Zein, M. I. H., Hardianto, A., Irkham, I., Zakiyyah, S. N., Devi, M. J., Manan, N. S. A.,
Thiruvengadam, M., Kim, J.-T., Kim, W.-R., Kim, J.-Y., Jung, B.-S., Choi, H.-J., Chi, H.-Y., Ibrahim, A. U., & Hartati, Y. W. (2023). Recent development of electrochemical and
Govindasamy, R., & Kim, S.-H. (2024). Safeguarding public health: Advanced optical aptasensors for detection of antibiotics in food monitoring applications.
detection of food adulteration using nanoparticle-based sensors. Critical Reviews in Journal of Food Composition and Analysis, 124, Article 105644.
Analytical Chemistry, 1–21. Zhang, H., Luo, B., Liu, K., Wang, C., Hou, P., Zhao, C., & Li, A. (2024). Development of
Ullah, Z., Al-Turjman, F., Mostarda, L., & Gagliardi, R. (2020). Applications of artificial surface molecular-imprinted electrochemical sensor for palmitic acid with machine
intelligence and machine learning in smart cities. Computer Communications, 154, learning assistance. Talanta, 275, Article 126124.
313–323. Zhong, Z., Xu, M., Rodriguez, M. A., Xu, C., & Buyya, R. (2022). Machine learning-based
Usama, M., Qadir, J., Raza, A., Arif, H., Yau, K.-L. A., Elkhatib, Y., Hussain, A., & Al- orchestration of containers: A taxonomy and future directions. ACM Computing
Fuqaha, A. (2019). Unsupervised machine learning for networking: Techniques, Surveys, 54(10s), 1–35.
applications and research challenges. IEEE Access, 7, 65579–65615. Zhou, Z.-H. (2022). Open-environment machine learning. National Science Review, 9(8),
Uwaya, G. E., Sagrado, S., & Bisetty, K. (2022). Smart electrochemical sensing of xylitol Article nwac123.
using a combined machine learning and simulation approach. Talanta Open, 6, Zhou, Z., Wang, L., Wang, J., Liu, C., Xu, T., & Zhang, X. (2022). Machine learning with
Article 100144. neural networks to enhance selectivity of nonenzymatic electrochemical biosensors
Vázquez-Canteli, J. R., & Nagy, Z. (2019). Reinforcement learning for demand response: in multianalyte mixtures. ACS Applied Materials & Interfaces, 14(47), 52684–52690.
A review of algorithms and modeling techniques. Applied Energy, 235, 1072–1089. Zhu, X., Liu, P., Ge, Y., Wu, R., Xue, T., Sheng, Y., … Wen, Y. (2020). MoS2/MWCNTs
Wang, C., Hao, T., Wang, Z., Lin, H., Wei, W., Hu, Y., Wang, S., Shi, X., & Guo, Z. (2023). porous nanohybrid network with oxidase-like characteristic as electrochemical
Machine learning-assisted cell-imprinted electrochemical impedance sensor for nanozyme sensor coupled with machine learning for intelligent analysis of
carbendazim. Journal of Electroanalytical Chemistry, 862, Article 113940.

17