A Comprehensive Review of Non-Destructive Monitoring of Food Freshness and Safety Using NIR Spectroscopy and Biosensors: Challenges and Opportunities

Abstract

The demand for safe, high-quality, and minimally processed food has intensified interest in non-destructive analytical techniques capable of assessing freshness and safety in real time. Among these, near-infrared (NIR) spectroscopy and biosensors have emerged as leading technologies due to their rapid, reagent-free, and sample-preserving nature. NIR spectroscopy offers a holistic assessment of internal compositional changes, while biosensors provide specific and sensitive detection of biological and chemical contaminants. Recent advances in miniaturization, chemometrics, and deep learning have further enhanced their potential for inline and point-of-need applications across diverse food matrices, including meat, seafood, eggs, fruits, and vegetables. This review critically evaluates the operational principles, instrumentation, and current applications of NIR spectroscopy and biosensors in food freshness and safety monitoring. It also explores their integration, highlights practical challenges such as calibration transfer and regulatory hurdles, and outlines emerging innovations including hybrid sensing, Artificial Intelligence (AI) integration, and smart packaging. The scope of this review is to provide a comprehensive understanding of these technologies, and its objective is to inform future research and industrial deployment strategies that support sustainable, real-time food quality control. These techniques enable near real-time monitoring under laboratory and pilot-scale conditions, showing strong potential for industrial adaptation. The nature of these targets often determines the choice of transduction method.

1. Introduction

Ensuring the freshness and safety of food products is a global priority, with challenges spanning from postharvest handling to consumer purchase. The globalization of food supply chains, demand for minimally processed foods, and rising awareness of foodborne illnesses have heightened the need for rapid and reliable monitoring methods [1]. Traditional analytical approaches such as microbiological assays, chemical titrations, and chromatographic methods, while highly sensitive, are destructive, labor-intensive, and time-consuming, limiting their suitability for real-time or large-scale applications [2,3]. Non-destructive techniques offer fast, cost-effective, and sustainable alternatives by preserving sample integrity and enabling continuous monitoring. They allow early detection of spoilage and contamination, reducing waste and improving consumer safety [4]. Among these, Near-Infrared (NIR) spectroscopy and biosensors stand out as complementary approaches: NIR provides rapid, holistic evaluation of compositional changes such as moisture, protein, and lipid oxidation, while biosensors offer high specificity for hazards like pathogens, toxins, pesticides, and allergens through enzyme-, antibody-, or aptamer-based recognition systems [5,6,7,8,9].

Recent advances, including miniaturization, nanomaterial integration, and AI-driven chemometrics, have expanded the applicability of these systems, enabling portable, field-deployable devices and Internet of Things (IoT) linked platforms for predictive analytics and supply chain traceability [10,11]. Applications now span meat, seafood, eggs, cereals, fruits, and vegetables, with demonstrated potential for real-time monitoring across diverse processing and retail environments. The objective of this paper is to provide a comprehensive assessment of NIR spectroscopy and biosensor technologies for food freshness and safety monitoring. Specifically, it examines their working principles, instrumentation, and applications, while also evaluating challenges such as calibration transfer, sensor drift, manufacturing costs, and regulatory hurdles. Furthermore, it highlights emerging innovations such as hybrid NIR-biosensor integration, AI and cloud-based monitoring, blockchain-enabled traceability, and sustainable sensing materials. By synthesizing current progress and identifying key research and policy needs, this paper aims to inform strategies that will support the gradual change in these technologies from laboratory prototypes toward scalable, validated monitoring systems suitable for industrial implementation.

2. NIR Spectroscopy for Food Freshness and Quality

2.1. Principles of NIR Spectroscopy and Spectral Interpretation

Near-infrared (NIR) spectroscopy is a vibrational technique that measures molecular overtone and combination bands in the 780–2500 nm range. These absorptions arise from C-H, N-H, and O-H bonds, which are abundant in food constituents such as water, lipids, and proteins [12]. As a result, NIR provides a rapid, non-destructive assessment of compositional and structural changes associated with food quality and freshness. The resulting spectra are broad and overlapping, necessitating chemometric methods for interpretation. Pre-processing techniques such as standard normal variate (SNV), multiplicative scatter correction (MSC), and derivatives enhance signal clarity [13]. Chemometric methods, including Principal Component Analysis (PCA) and Successive Projection Algorithm (SPA), are applied for variable selection and noise reduction, while Partial Least Squares Regression (PLSR) and Support Vector Regression (SVR) enable quantitative prediction of contaminant levels. This workflow highlights FT-NIR’s capability for rapid, non-destructive process monitoring, allowing early detection of adulteration or degradation during production and storage. Such integration of spectroscopy and chemometrics offers a practical pathway for real-time quality and shelf-life assessment in food processing environments. Multivariate models like partial least squares regression (PLSR), principal component analysis (PCA), and, more recently, deep learning algorithms such as convolutional neural networks (CNNs) are employed to extract predictive information on parameters such as pH, moisture, total volatile basic nitrogen (TVB-N), and microbial load [14,15]. A study by Candeias et al. [16] demonstrated that wheat flour quality assessment using handheld NIR devices combined with chemometric pre-processing and PLS modeling can deliver accurate predictions for compositional parameters in blended flour, underscoring the practical utility of portable systems. Similarly, portable NIR devices have been applied to distinguish organic from conventional mango fruits, chips, and juices with high accuracy, demonstrating their versatility for both freshness monitoring and authenticity verification [17].

Recent applications further highlight the versatility of NIR: Zhang et al. successfully differentiated Radix Pueraria starch from different geographical origins using FT-NIR spectroscopy, demonstrating its power in authenticity and traceability studies [18]. Cheng et al. [19] combined hyperspectral imaging with convolutional neural networks to evaluate lipid oxidation in pork, showing the synergy of AI with NIR spectral data. Wang et al. [20] applied FT-NIR spectroscopy with Linear discriminant analysis via QR decomposition LDA/QR for rapid classification of green tea varieties, confirming its value in quality control. Likewise, Dai et al. [21] used NIR spectroscopy and chemometrics to monitor critical parameters during soybean meal fermentation, highlighting its role in process monitoring. NIR is particularly well-suited for bulk, in situ analysis, offering real-time insights with no sample destruction or reagents. However, it remains an indirect technique and depends on robust calibration models that account for sample variability and environmental factors [22]. Figure 1 shows the integration of Fourier Transform Near-Infrared (FT-NIR) spectroscopy with chemometric analysis for monitoring contamination and quality in edible oils. The process begins with peanut oil samples deliberately mixed with five pollutants—white mineral oil, diesel fuel, kerosene, lubricants, and engine oil—followed by homogenization through mechanical and ultrasonic vibration. Using FT-NIR in transmission mode (64 scans, 8 cm⁻¹ resolution, 4000–10,000 cm⁻¹ range), spectral data are acquired and transmitted for computational analysis [23].

Recent advancements in portable spectrometers and hyperspectral systems, when combined with chemometrics and AI, are further improving their applicability for on-site food safety monitoring. In this review, Near-Infrared (NIR) ranges from 780 to 2500 nm, while Visible–NIR (Vis–NIR) extends from 400 to 2500 nm, encompassing both visible and near-infrared regions. The visible range provides surface and color information, whereas NIR offers deeper penetration and captures C–H, O–H, and N–H vibrational features related to food composition. Accordingly, Vis–NIR systems are used for optical and surface assessments, while NIR-only platforms are suited for bulk compositional analysis. This clarification ensures consistent terminology and emphasizes how spectral range selection affects penetration depth and analyte specificity across food matrices.

2.2. Portable and Online NIR Systems

Traditional NIR instruments were bulky and limited to laboratory environments; however, advancements in optoelectronics and miniaturized light sources have enabled the development of portable and handheld NIR spectrometers suitable for in-field or inline applications. Such devices offer the flexibility of conducting rapid, non-invasive freshness evaluations directly in production, storage, or retail environments. Leon et al. used a handheld NIR spectrometer to classify Angus beef steaks by aging status, achieving over 90% accuracy, and to predict storage duration with strong reliability [24]. This demonstrated the suitability of handheld NIR for real-time meat traceability. Similarly, Yao et al. applied a portable NIR spectrometer combined with deep learning algorithms to monitor egg freshness, reporting prediction accuracies above 90%, which highlighted the added value of AI in improving calibration models [25]. In another seafood example, Vilkova et al. used infrared spectroscopy to monitor spoilage progression in rainbow trout during cold storage, confirming its utility as a rapid screening tool [26]. Recent studies further expand portable NIR applications. Qi et al. discriminated red jujube varieties using a portable NIR spectrometer with fuzzy improved linear discriminant analysis, showcasing its power for fruit classification in complex supply chains [27]. Zhang et al. [28] employed portable NIR with fuzzy uncorrelated discriminant transformation for rapid authentication of milk origin, enabling traceability in dairy systems. Shen et al. [29] demonstrated smartphone-controlled portable NIR for detecting fumonisin B1 and B2 in corn, achieving reliable toxin prediction under field conditions.

Beyond meat and seafood, Patel et al. explored a hybrid NIR and proton Nuclear magnetic resonance (NMR) approach for semi-hard cheese, enabling online monitoring of compositional changes during ripening [30]. Likewise, Bonazza et al. [31] demonstrated that portable NIR could effectively track oxidative stability in hempseed oil under varying storage conditions, supporting its use in shelf-life prediction for lipid-rich foods. As illustrated in Figure 2, portable and online NIR systems can be deployed across different stages of the food production pipeline. Figure 2A represents a schematic overview of a portable NIR setup, where handheld or benchtop devices are used for real-time quality control. These instruments often incorporate spectral acquisition modules (e.g., NIRQuest) that analyze raw food ingredients before processing. The results are fed into predictive models that support intelligent control systems, allowing for on-the-spot adjustments to maintain consistent quality. This is especially useful in on-site or decentralized processing environments, such as small-scale meat or dairy production facilities [32]. Figure 2B depicts an inline or online NIR system integrated directly into a continuous processing line. Here, NIR probes and sensors are embedded in equipment to enable process analytical technology (PAT) frameworks. These sensors continuously gather spectral data and feed it into advanced control models that perform real-time data analysis and predictive adjustments. This type of setup is particularly valuable for ensuring uniformity in large-scale food production, as it allows for closed-loop process control without interrupting the manufacturing flow [33]. Collectively, these studies illustrate that portable and online NIR systems, particularly when paired with chemometrics or AI, are transforming freshness evaluation from lab-based testing to flexible, real-time monitoring across diverse food matrices.

2.3. Chemometric Models and AI Integration

The overlapping nature of NIR spectra requires multivariate models to extract meaningful freshness and safety information. Classical approaches PCA for dimensionality reduction, PLSR for prediction, and classifiers such as LDA and SVM remain widely used [34]. Recent advances integrate AI to improve accuracy and adaptability. Zhao et al. Developed convolutional neural network–support vector regression CNN-SVR hybrids for pork freshness, achieving R² > 0.92 for pH and TVB-N [35]. Similarly, Tan et al. applied NIR with ensemble learning to classify fresh vs. refrigerated pork at >94% accuracy [36]. In plant-based foods, Gjonaj et al. [37] found artificial neural networks (ANNs) more accurate than PLS in detecting argan oil adulteration, while for poultry, Cai et al. [38] fused hyperspectral data with deep learning to predict chicken spoilage markers, outperforming chemometrics. More advanced AI-driven strategies are also emerging. Wang et al. [39] demonstrated that combining Markov Transition Fields with CNNs significantly improved NIR model performance for aflatoxin B1 (AFB1) prediction in maize, highlighting the potential of deep spatiotemporal representations in food safety applications.

Similarly, Guo et al. [40] applied a portable Vis-NIR system with Competitive adaptive reweighted sampling convolutional neural networks CARS-CNN to predict water core disorders in apples, showcasing how CNN-based variable selection enhances model robustness for fruit quality monitoring. In another study, Liang et al. employed Visible near-infrared vis-NIR hyperspectral imaging to non-destructively discriminate homochromatic foreign materials in cut tobacco, providing evidence for NIR’s utility in food authenticity verification when combined with advanced machine learning [41]. Beyond raw spectral data, multimodal fusion approaches have also been explored: Yolandani et al. compared predictive models of soy protein isolate bitterness using spectrofluorometric, chromatographic, and sensory data, emphasizing the growing role of integrated AI and chemometrics in food evaluation [42]. Overall, chemometrics remains foundational, but AI models offer greater robustness against batch variability, instrument transfer, and noise. Challenges persist in model interpretability and large data requirements, yet AI-driven NIR systems are increasingly positioned as scalable, real-time tools for food quality monitoring.

2.4. Applications Across Food Categories

NIR spectroscopy has been applied across diverse food matrices, demonstrating its adaptability for freshness and quality assessment in both animal- and plant-based systems. In eggs, Patil and Patil [43] used portable Vis-NIR (902–1810 nm) with PLS and PCA models to predict freshness from Haugh Units and storage duration, achieving R²c = 0.986 and 95% accuracy over 25 days of storage. Similarly, Wang et al. [44] employed a smartphone-connected NIR spectrometer for egg freshness prediction with ±2 days accuracy, highlighting its potential for consumer-level applications, as shown in Table 1. For shellfish, Ghidini et al. [45] applied portable NIR to live mussels, correlating spectral markers with nitrogen compounds and ATP-related metabolites. Their Orthogonal Partial Least-Squares Regression (OPLSR) model (R²p = 0.91) enabled non-invasive prediction of storage time. In poultry, Kim et al. [46] combined Vis-NIR with Partial Least Squares Discriminant Analysis (PLS-DA) to classify chicken breast samples by pH and drip loss with >95% accuracy, again demonstrating the power of chemometrics with portable systems (Table 1).

NIR has also been successfully employed in processed foods. Guo et al. estimated TVB-N in preserved eggs with SVR and PCA, achieving R²p = 0.91 for spoilage detection [47]. Expanding on processed applications, Deng et al. [48] enhanced FT-NIR with ensemble learning methods to detect mineral oil contamination in corn oil, highlighting the potential of NIR for food safety in edible oils. Plant-based foods provide further examples. In apples, phase-based reflectance spectroscopy tracked ripening and decay with minimal calibration [49]. Guo et al. [40] further demonstrated the non-destructive determination of edible quality and water core degree in apples using a portable Vis-NIR transmittance system combined with deep learning (CARS-CNN), underscoring the growing role of AI-enhanced NIR models in fruit quality monitoring. Beyond apples, Jiang et al. visualized the distribution of pesticide residues in mulberry leaves using NIR hyperspectral imaging, showing its applicability for fresh produce safety [50]. Broader reviews also support these findings; for example, Zareef et al. [51] summarized applications of non-linear algorithms coupled with NIR across diverse food matrices, confirming its versatility for both quality control and authenticity verification. NIR has also predicted rice moisture and amylose with R²c = 0.96 [52] and detected honey adulteration with 96% accuracy [53] as detailed in Table 1. Together, these examples underscore NIR’s broad adaptability across raw and processed food systems.

Table 1. Summary of NIR Spectroscopy applications across food categories.

Food Product/Matrix	Analytical Target(s)	NIR System and Wavelength Range	Acquisition Mode	Sample Preparation	Chemometric/AI Model	Sample Size/Replicates	Performance Metrics	Key Findings/Notes	Reference
Eggs	Haugh Unit, (Regression)	Portable Vis-NIR (902–1810 nm)	Reflectance	Whole eggs, unwashed	PLS, PCA, PLS-DA	180	R2c = 0.986, Accuracy = 95%	Portable NIR devices can provide rapid, non-destructive estimation of egg freshness comparable to lab measurements.	[43]
Eggs	Freshness (Regression)	Smartphone-connected NIR	Reflectance	Whole egg shell intact	ANN + Savitzky–Golay	120	R2 = 0.83, RMSE = 1.97 days	Consumer-level smartphone-linked NIR systems can reliably predict freshness within ±2 days.	[44]
Pork	pH, TVB-N	Vis–NIR (400–1000 nm)	Reflectance	Fresh meat slabs	CNN–SVR, CNN–PLSR	150	R2 > 0.92, RPD > 3.6	Hybrid CNN-based models improved prediction robustness, enabling inline industrial monitoring.	[35]
Mussels	Freshness Index, Viability (Regression)	Portable NIR (950–1650 nm)	Reflectance	Live mussels, shells intact	OPLSR	90	R2p = 0.91	Detected subtle spectral changes linked to spoilage without removing shells, suitable for aquaculture logistics.	[45]
Preserved Eggs	TVB-N	Lab-based NIR (1000–2500 nm)	Transmittance	Sliced samples	PCA, SVR	75	R2p = 0.91, RMSEP = 0.38	Adapted NIR to high-salt processed eggs for rapid spoilage screening.	[47]
Apples	Degradation Progress	Phase-based reflectance	Phase Reflectance (850–1700 nm)	Whole fruits		60	Low calibration requirement	Cost-efficient method requiring minimal calibration, ideal for field checks.	[49]
Chicken Meat	Drip Loss, pH	Vis–NIR	Reflectance	Fresh chicken breasts	PLS-DA	100	Accuracy > 95%	Industrial applicability for real-time poultry grading.	[46]
Salmon	Lipid oxidation, TVB-N	Hyperspectral NIR (900–1700 nm)	Reflectance (400–1000 nm)	Whole filets, skin on	PLSR, SVM	80	R2p = 0.94	Detected early spoilage and oxidation before visual signs.	[12]
Milk	Protein, fat content	FT-NIR (1000–2500 nm)	Transmittance	Homogenized	PLSR	200	R2 > 0.99	Lab-level compositional analysis in under 1 min.	[54]
Cheddar Cheese	Ripening stage	Portable NIR (950–1650 nm)	Reflectance	Sliced cheese	PCA + PLS-DA	50	Accuracy 92%	Discriminated maturity stages for optimized flavor profiles.	[55]
Coffee Beans	Moisture, defects	NIR (1100–2500 nm)	Reflectance	Whole beans	ANN	300	R2p = 0.97	Detected defects and optimized roasting profiles.	[56]
Wheat Flour	Moisture, protein	FT-NIR (1000–2500 nm)	Transmittance	Ground flour	PLSR	250	R2c = 0.98	Rapid quality grading for milling operations.	[57]
Peanuts	Aflatoxin contamination	NIR (900–1700 nm)	Reflectance	Whole kernels	PLS-DA	120	Accuracy > 90%	Early detection of contaminated lots before processing.	[58]
Beer	Alcohol %, turbidity	FT-NIR (1000–2500 nm)	Transmittance	Degassed beer	PLSR	60	R2 = 0.99	Accurate inline brewery QC.	[59]
Grapes	Sugar content (°Brix)	Portable NIR (900–1700 nm)	Reflectance	Intact grapes	PLSR	150	R2p = 0.97	Enabled selective harvesting based on ripeness.	[60]
Rice	Moisture, amylose	NIR (900–1700 nm)	Reflectance	Milled grains	PLSR	180	R2c = 0.96	Facilitated rapid classification for export quality compliance.	[52]
Honey	Adulteration detection	Portable NIR (950–1650 nm)	Transmittance	Liquid honey	PLS-DA	100	Accuracy = 96%	Detected multiple adulterants within 30 s.	[53]
Tomatoes	Lycopene, firmness	Hyperspectral NIR (900–1700 nm)	Reflectance	Intact fruits	PLSR, SVM	90	R2p = 0.95	Predicted ripeness and post-harvest storage potential.	[61]

Table 1. Summary of NIR Spectroscopy applications across food categories.

Food Product/Matrix	Analytical Target(s)	NIR System and Wavelength Range	Acquisition Mode	Sample Preparation	Chemometric/AI Model	Sample Size/Replicates	Performance Metrics	Key Findings/Notes	Reference
Eggs	Haugh Unit, (Regression)	Portable Vis-NIR (902–1810 nm)	Reflectance	Whole eggs, unwashed	PLS, PCA, PLS-DA	180	R2c = 0.986, Accuracy = 95%	Portable NIR devices can provide rapid, non-destructive estimation of egg freshness comparable to lab measurements.	[43]
Eggs	Freshness (Regression)	Smartphone-connected NIR	Reflectance	Whole egg shell intact	ANN + Savitzky–Golay	120	R2 = 0.83, RMSE = 1.97 days	Consumer-level smartphone-linked NIR systems can reliably predict freshness within ±2 days.	[44]
Pork	pH, TVB-N	Vis–NIR (400–1000 nm)	Reflectance	Fresh meat slabs	CNN–SVR, CNN–PLSR	150	R2 > 0.92, RPD > 3.6	Hybrid CNN-based models improved prediction robustness, enabling inline industrial monitoring.	[35]
Mussels	Freshness Index, Viability (Regression)	Portable NIR (950–1650 nm)	Reflectance	Live mussels, shells intact	OPLSR	90	R2p = 0.91	Detected subtle spectral changes linked to spoilage without removing shells, suitable for aquaculture logistics.	[45]
Preserved Eggs	TVB-N	Lab-based NIR (1000–2500 nm)	Transmittance	Sliced samples	PCA, SVR	75	R2p = 0.91, RMSEP = 0.38	Adapted NIR to high-salt processed eggs for rapid spoilage screening.	[47]
Apples	Degradation Progress	Phase-based reflectance	Phase Reflectance (850–1700 nm)	Whole fruits		60	Low calibration requirement	Cost-efficient method requiring minimal calibration, ideal for field checks.	[49]
Chicken Meat	Drip Loss, pH	Vis–NIR	Reflectance	Fresh chicken breasts	PLS-DA	100	Accuracy > 95%	Industrial applicability for real-time poultry grading.	[46]
Salmon	Lipid oxidation, TVB-N	Hyperspectral NIR (900–1700 nm)	Reflectance (400–1000 nm)	Whole filets, skin on	PLSR, SVM	80	R2p = 0.94	Detected early spoilage and oxidation before visual signs.	[12]
Milk	Protein, fat content	FT-NIR (1000–2500 nm)	Transmittance	Homogenized	PLSR	200	R2 > 0.99	Lab-level compositional analysis in under 1 min.	[54]
Cheddar Cheese	Ripening stage	Portable NIR (950–1650 nm)	Reflectance	Sliced cheese	PCA + PLS-DA	50	Accuracy 92%	Discriminated maturity stages for optimized flavor profiles.	[55]
Coffee Beans	Moisture, defects	NIR (1100–2500 nm)	Reflectance	Whole beans	ANN	300	R2p = 0.97	Detected defects and optimized roasting profiles.	[56]
Wheat Flour	Moisture, protein	FT-NIR (1000–2500 nm)	Transmittance	Ground flour	PLSR	250	R2c = 0.98	Rapid quality grading for milling operations.	[57]
Peanuts	Aflatoxin contamination	NIR (900–1700 nm)	Reflectance	Whole kernels	PLS-DA	120	Accuracy > 90%	Early detection of contaminated lots before processing.	[58]
Beer	Alcohol %, turbidity	FT-NIR (1000–2500 nm)	Transmittance	Degassed beer	PLSR	60	R2 = 0.99	Accurate inline brewery QC.	[59]
Grapes	Sugar content (°Brix)	Portable NIR (900–1700 nm)	Reflectance	Intact grapes	PLSR	150	R2p = 0.97	Enabled selective harvesting based on ripeness.	[60]
Rice	Moisture, amylose	NIR (900–1700 nm)	Reflectance	Milled grains	PLSR	180	R2c = 0.96	Facilitated rapid classification for export quality compliance.	[52]
Honey	Adulteration detection	Portable NIR (950–1650 nm)	Transmittance	Liquid honey	PLS-DA	100	Accuracy = 96%	Detected multiple adulterants within 30 s.	[53]
Tomatoes	Lycopene, firmness	Hyperspectral NIR (900–1700 nm)	Reflectance	Intact fruits	PLSR, SVM	90	R2p = 0.95	Predicted ripeness and post-harvest storage potential.	[61]

3. Biosensor Applications in Food Safety and Spoilage Detection

3.1. Overview of Biosensor Mechanisms

Biosensors are integrated analytical devices that combine a biological recognition element with a physicochemical transducer to detect specific analytes with high sensitivity and specificity. Their application in food systems has expanded rapidly, driven by the demand for real-time, minimally invasive, and decentralized tools to safeguard quality and safety. The biological recognition element, commonly enzymes, antibodies, nucleic acids, aptamers, or whole cells, confers selectivity by binding to the target analyte, thereby initiating a biochemical interaction. This event is then converted by a transducer into a measurable signal. Biosensors have emerged as portable, sensitive, and low-cost tools for food safety monitoring. Figure 3 summarizes the major nucleic acid amplification techniques integrated into biosensing systems for food safety applications. Figure 3A–C show enzyme-free amplification methods such as Hybridization Chain Reaction (HCR) and Catalytic Hairpin Assembly (CHA), which rely on the self-assembly of DNA hairpins for signal enhancement. These isothermal reactions amplify nucleic acid targets without enzymes, improving stability and suitability for on-site food pathogen detection [62]. Loop-Mediated Isothermal Amplification (LAMP), shown in Figure 3B, uses multiple primers and a strand-displacing polymerase to rapidly amplify DNA sequences under constant temperature, enabling visual detection through color or fluorescence signals [63]. Figure 3D,E [64,65] show Salmonella or E. coli screening in food samples, while Rolling Circle Amplification (RCA) provides continuous amplification from circular templates, producing long single-stranded DNA with high sensitivity. Figure 3F,G demonstrates Exponential Amplification Reaction (EXPAR) and Nicking Enzyme Amplification Reaction (NEAR), both of which achieve ultra-fast amplification by recycling short DNA fragments through nicking and polymerase activity [66,67]. These methods enable near real-time detection and are increasingly coupled with colorimetric or electrochemical biosensors for food pathogen monitoring. These approaches enable rapid signal amls can enhance both recognition and signal transduction. For example, Gao et al. [68] reported the use of magnetic nanomaterials with dual-recognition strategies to improve sensitivity and enable multiplex detection of foodborne bacteria, while Li et al. [69] developed a DNA tetrahedral scaffold-based electrochemical biosensor for simultaneous detection of aflatoxin B1 and ochratoxin A, illustrating the potential of nucleic acid nanostructures for multi-analyte sensing.

Transduction modalities vary widely. Optical biosensors, such as surface-enhanced Raman scattering (SERS), offer label-free and non-destructive detection capabilities. Guo et al. [70] highlighted SERS-based applications for monitoring pesticide residues and spoilage in fruits and vegetables; meanwhile, Zou et al. [71] reviewed the integration of quantum dots as advanced nanomaterials for optical biosensing in food quality assessment. Complementary to optical methods, electrochemical sensors provide high sensitivity and compatibility with portable formats. Dong et al. [72] emphasized photoelectrochemical biosensors as a hybrid platform, combining the selectivity of photoactive materials with electrochemical readouts for pathogen detection. Further, the coupling of immunosensors and aptasensors with magnetic nanoparticles, as described by Gao et al. [73], enhances both capture efficiency and signal transduction in mycotoxin monitoring. These examples collectively demonstrate the versatility of biosensors across recognition chemistries and signal modalities, underscoring their central role in advancing rapid, real-time, and automated food safety assessment.

3.2. Detection of Pathogens, Pesticides, and Spoilage Metabolites

One of the most critical applications of biosensors in the food industry is the detection of pathogenic microorganisms such as Salmonella, E. coli, and Listeria monocytogenes. Figure 4 shows the conventional workflow used for isolating and identifying microbial contaminants from food matrices. The process typically begins with Figure 4A, sample collection, where rinsing, swabbing, or homogenization is performed on diverse food products such as meat, dairy, and baked goods to extract potential pathogens. Next, pathogen isolation, as shown in Figure 4B, involves centrifugation, filtration, and enrichment culture, often enhanced by magnetic separation techniques for selective bacterial capture. The final stage, Figure 4C pathogen detection, integrates molecular biology assays (e.g., PCR amplification), immunological testing (including lateral flow and ELISA), and biochemical identification using selective culture media [74]. Gao et al. [68] demonstrated how magnetic nanomaterial-assisted biosensors using milk and poultry wash water samples can significantly improve sensitivity and allow multiplex pathogen detection. Similarly, Zhu et al. [75] developed a colorimetric biosensor capable of simultaneously detecting ochratoxin A and aflatoxin B1 in maize, ensuring analytical compatibility with complex sample matrices. Similarly, biosensors targeting pesticide residues were tested in fruit homogenates and surface rinsates, confirming matrix-adjusted performance. Beyond pathogens and toxins, biosensors also offer effective solutions for pesticide detection. Figure 5 entails different electrochemical and immunoassay-based biosensing strategies for detecting foodborne pathogens. Figure 5A shows an immunomagnetic–electrochemical platform for Salmonella detection. Magnetic nanoparticles coated with antibodies capture target cells, which are enzymatically labeled and detected electrochemically on a modified glassy carbon electrode (m-GEC), enabling rapid and specific pathogen identification [76]. Figure 5B illustrates a sandwich-type electrochemical immunoassay on a gold electrode. Anti-Salmonella antibodies conjugated with alkaline phosphatase catalyze a redox reaction that produces a measurable electrochemical signal, enhancing detection sensitivity [77] while Figure 5C depicts a dual detection workflow combining electrochemical immunoassay (ELIME) and real-time PCR for Salmonella in ready-to-eat vegetables [78]. The electrochemical assay targets surface antigens, while PCR verifies genetic material, improving result reliability in complex food matrices. Finally, Figure 5D presents a multiplex electrochemical immunosensor for simultaneous detection of E. coli, Salmonella, and Campylobacter. Different metal sulfide nanocrystals (CdS, CuS, PbS) serve as electrochemical tracers, allowing distinct signals for each pathogen via anodic stripping voltammetry [79].

Enzyme-driven electrochemical biosensors have been highlighted for their versatility in tracking diverse chemical contaminants, with nanomaterial integration improving signal transduction [9]. At the same time, smart packaging approaches have gained attention; Yang et al. reported bilayer pH-sensitive colorimetric films that not only monitor crucian carp spoilage but also provide a visual indication of volatile metabolite accumulation [80]. Spoilage detection is particularly important for perishable foods such as fruits, fish, and meat. Guo et al. used electronic nose technology to identify spoilage-causing fungi and predict the spoilage degree of apples [81], while earlier reviews have emphasized the utility of electronic noses for volatile metabolite-based pathogen detection [82]. Collectively, these works illustrate the adaptability of biosensors in monitoring pathogens, pesticides, and spoilage metabolites, reinforcing their role as integral tools for modern food safety assurance.

3.3. Optical and Electrochemical Transduction Strategies

Optical and electrochemical methods are the most widely applied biosensing strategies in food safety. Optical biosensors detect changes in absorbance, fluorescence, or refractive index, providing real-time and non-destructive analysis. Meng et al. engineered signal transduction between Cadmium telluride (CdTe) and Cadmium selenide (CdSe) quantum dots to design a photoelectrochemical immunoassay, showing the potential of nanomaterial-based optical sensors for enhanced food safety testing [83]. Similarly, Dong et al. reviewed advances in photoelectrochemical biosensors for pathogen detection, highlighting the role of photoactive materials in improving sensitivity and adaptability [72]. As shown in Figure 6, nanomaterial-assisted photoelectrochemical biosensors utilize light-induced electron transfer between semiconductors, metals, and carbon-based materials to enhance sensitivity and specificity. Figure 6A shows a heterojunction PEC biosensor composed of Bi₂WO₆ (BWO-1), Ag₂S, and Au nanoparticles on an indium tin oxide (ITO) electrode. Upon light irradiation, electrons transfer through the conduction band (CB) cascade while holes migrate oppositely in the valence band (VB), promoting efficient charge separation. This enhances photocurrent response and reduces recombination, improving sensitivity and stability [84]. Figure 6B presents an immunomagnetic PEC sensor for Staphylococcus aureus detection using ZnS-CdS semiconductor coupling. After bacterial incubation and magnetic separation, Cys-mediated charge transfer occurs between the ZnS/CdS interface under illumination, amplifying the photocurrent for rapid, selective detection in complex matrices [85]. Figure 6C displays an NIR-responsive MoS₂/rGO composite electrode, where near-infrared excitation induces electron transfer from MoS₂ to reduced graphene oxide (rGO). This heterostructure enhances conductivity, broadens absorption into the NIR region, and supports deep-penetration analysis suitable for turbid food samples [86] and Figure 6D shows a UCNP-based hybrid PEC biosensor integrating upconversion nanoparticles (UCNPs@SiO₂@Ag–C₃N₄) with antimicrobial peptide Magainin I for E. coli O157:H7 detection. NIR irradiation activates UCNPs, driving photoinduced charge transfer and generating a strong PEC response for highly specific, low LOD [87]. These systems demonstrate the synergy between optical excitation and electrochemical readouts, enabling improved detection of pathogens and contaminants in food safety monitoring. Multiplex sensing has also been advanced through optical-electrochemical platforms for simultaneous mycotoxin detection [88], with quantum dot systems offering strong portability [9]. Electrochemical biosensors, in contrast, translate recognition events into electrical signals via amperometric, potentiometric, or impedimetric modes. Zhang et al. introduced a structure-switching aptasensor for mercury in dairy products, illustrating high specificity [89], and Zhai et al. applied pH-sensitive electrochemical films in smart packaging to monitor spoilage [90]. Their simplicity, low cost, and miniaturization potential make electrochemical sensors particularly suited for field deployment. Together, these complementary strategies balance non-destructive optical sensing with robust and scalable electrochemical detection, forming the backbone of next-generation smart food monitoring.

3.4. Miniaturized and Integrated Biosensing Platforms

Advances in microfabrication, printed electronics, and wireless technologies are driving the development of compact biosensing platforms for real-time food monitoring. Miniaturized systems are increasingly embedded in smart packaging, handheld detectors, and wearable devices, enabling continuous quality assessment during storage and distribution. Zhai et al. pioneered edible, pH-sensitive films with electrochemical writing [90], while Fan et al. introduced visual intelligent labels capable of tracking meat freshness in real time, underscoring the university’s contributions to smart packaging innovations [91]. Beyond freshness indicators, miniaturized biosensors are also being adapted for chemical hazard detection. For instance, Wang et al. developed a tunable multiplexed fluorescence biosensing platform capable of simultaneously detecting multiple pesticides such as paraquat and carbendazim, highlighting the feasibility of compact, multi-residue monitoring systems in food safety [92]. Similarly, Wu et al. introduced smartphone-integrated AuAg nanocluster membranes for the multivariate fluorescent sensing of Hg²⁺, Cu²⁺, and amino acids, offering portability while maintaining high sensitivity [93]. These approaches indicate promising portability and sensitivity for laboratory trials; however, full-scale deployment requires additional testing for stability and calibration consistency. Beyond fluorescence, miniaturized Raman-based microfluidic biosensors have demonstrated strong applicability. Jayan et al. reported a Raman spectroscopy-integrated microfluidic platform capable of rapid pathogen detection in food, offering a scalable and smartphone-compatible solution for real-time monitoring [94].

Integration with smartphones and cloud-based platforms is further enhancing scalability. Zhai et al. demonstrated a smartphone-coupled Raman microfluidic platform for rapid detection of foodborne pathogens, providing a compact solution for real-time microbial diagnostics. Additionally, Wang et al. [95] explored volatile-based sensing using the behavioral responses of Drosophila species to plant essential oils, a concept that parallels the development of smart packaging sensors for spoilage-associated volatiles. Global advances complement these efforts, for example, Naik et al. [96] demonstrated near-field communication (NFC)-enabled wireless gas sensors for spoilage monitoring, and Peddareddigari et al. [97] proposed IoT and AI frameworks for decentralized food safety control. Nanotechnology-enabled sensors that replace bulky spectrometers with low-cost photodiode systems have further advanced real-time Vis-NIR quality assessment [98]. Figure 7 illustrates that biosensors are increasingly integrated into smart traceability systems that simultaneously address multiple food safety challenges. These platforms combine analyte-specific bioreceptors with miniaturized transducers and amplifiers, enabling rapid detection of contaminants such as food pathogens, allergens, pesticides, heavy metals, mycotoxins, veterinary drugs, and illegal additives. The versatility and portability of such systems underscore their potential as sensitive, low-cost, and user-friendly tools for real-time food safety monitoring. Integration with smartphones and cloud-based platforms is extending these systems toward scalable, intelligent monitoring [99]. Although issues of sensor stability, calibration, and multiplexing remain, the convergence of biosensor innovations with global digital tools highlights the future of real-time, intelligent food safety management.

4. Synergistic Use of NIR Spectroscopy and Biosensors

4.1. Complementary Strengths and Integration Rationale

Near-Infrared (NIR) spectroscopy and biosensors provide complementary advantages in food quality monitoring. NIR excels in rapid, non-destructive assessment of overall composition and structural changes, while biosensors target specific hazards such as pathogens, toxins, and residues with high sensitivity. When combined, they create a dual-layered system: NIR delivers broad, fast screening of degradation, and biosensors provide molecular-level confirmation of contamination. For example, Wang et al. [100] assessed egg freshness using transmission–reflection NIR spectroscopy with multivariate models, demonstrating its ability to rapidly track spoilage in perishable foods. Similarly, Fan et al. [101] applied NIR to monitor physicochemical and fermentation changes in Zhenjiang vinegar, highlighting its effectiveness in process control. In contrast, biosensors provide molecular precision. Zhang et al. developed an electrochemical aptasensor for trace mercury in dairy, showing their strength in contaminant-specific detection [89].

The integration of these methods enhances reliability: NIR delivers a quick overview of food quality, while biosensors confirm specific hazards. This combined approach has been further supported by studies, such as Tahir et al. on vibrational spectroscopy in food analysis [102], and Guo et al. on NIR-based apple quality evaluation [103], both underscoring the adaptability of optical methods for food matrices. Further, Shen et al. demonstrated chemometric-assisted NIR for flavonoid detection in goji berries, underscoring how broad-spectrum and targeted sensing can be merged for comprehensive food analysis [104]. These efforts suggest that integrating NIR spectroscopy with biosensors creates a multi-layered, rapid sensing framework by NIR, followed by biosensor-based molecular verification, paving the way for more reliable, real-time food safety and quality monitoring.

4.2. Hybrid Sensing Systems and Real-Time Monitoring

The development of hybrid sensing systems integrating NIR spectroscopy and biosensors has enabled simultaneous monitoring of compositional and microbial changes in food. These systems combine the broad, non-destructive screening ability of NIR with the molecular precision of biosensors, making them powerful tools for quality and safety assurance. Figure 8 shows that aptamer-based biosensors already provide multiplex, real-sample monitoring of mycotoxins such as Ochratoxin A OTA, Zearalenone ZEN, and AFB1. Figure 8A shows an aptamer–magnetic bead–gold nanorod assay for detecting ochratoxin A (OTA). Biotin-labeled OTA aptamers bind selectively to the toxin, while silver deposition amplifies the optical signal [105]. The resulting color change correlates with OTA concentration, enabling both visual and quantitative analysis. Figure 8B depicts a metal–organic framework (MOF)-based aptasensor for zearalenone (ZEN) detection. The UiO-66–hemin complex catalyzes a peroxidase-like reaction that oxidizes TMB (3,3′,5,5′-tetramethylbenzidine), producing a colorimetric readout proportional to toxin levels [106], while Figure 8C shows an aptamer–silver nanoparticle (AgNP) colorimetric assay for aflatoxin B1 (AFB1). In Figure 8A, the process starts with Step 1 where biotin-labeled OTA aptamers bind specifically to ochratoxin A (OTA) in the sample. In Step 2, streptavidin-coated magnetic beads (SA-MBs) are introduced, which bind to the biotin-labeled aptamers, allowing the aptamer-OTA complex to be captured and separated using a magnet. Step 3 involves the deposition of silver nanoparticles (Ag) onto the complex, which enhances the signal for detection. Finally, in Step 4, alkaline phosphatase (AAP) reacts with a substrate, producing a color change that increases with the concentration of OTA, providing a visual indicator of its presence in the sample. [106]. Structural changes in the aptamer upon toxin binding prevent aggregation of AgNPs, resulting in visible color variation from yellow to red-brown, which can be quantified spectrophotometrically [107]. Figure 8D highlights a MnO₂ nanozyme-based biosensor for OTA detection. MnO₂ nanoflowers mimic enzymatic catalysis, converting TMB to its oxidized form (oxTMB). Aptamer–toxin interactions inhibit this catalytic process, leading to a measurable decrease in color intensity for rapid on-site toxin assessment in food samples like wheat flour and red wine [108]. For instance, Wu et al. developed a visible colorimetric sensor array combined with chemometric algorithms for real-time potato quality assessment, showing how colorimetric and NIR features can be jointly exploited [109]. Similarly, Guo et al. demonstrated a multi-sensor fusion framework combining NIR data with biosensor-driven deep learning models for real-time monitoring and spoilage prediction in apples [110]. Other innovations have focused on integrating functional films with biosensing elements to support freshness monitoring in packaging. Zhang et al. reported the use of Co-MOF-derived sodium carboxymethyl cellulose films for in situ monitoring of apple freshness, enabling both visual and instrumental quality tracking [111]. In another study, Zhang et al. developed a pH-indicator film from dragon fruit peel pectin and cassava starch for real-time pork freshness evaluation, highlighting the role of natural materials in hybrid sensing [112].

These hybrid platforms are increasingly being adapted for near real-time monitoring in controlled environments, showing progress toward eventual industrial integration. For example, Wang et al. applied in situ and real-time monitoring of enzymatic processes using NIR spectrometry combined with biosensor-based pre-treatment strategies [113]. Similarly, Zhang et al. constructed real-time monitoring systems for soy sauce fermentation using miniature fiber NIR spectrometers, advancing their application in complex fermentation systems [114]. From Figure 9, we can see the optical principles and performance optimization of visible and near-infrared (Vis–NIR) surface plasmon resonance (SPR) biosensing systems used for food quality monitoring. Figure 9A–C show the fundamental distinction between visible and near-infrared sensing ranges. Visible light (400–800 nm) offers strong surface sensitivity, while near-infrared light (1000–1600 nm) enhances penetration depth, making it suitable for detecting analytes within turbid food matrices. Figure 9C indicates the relationship between wavelength and spectral sensitivity, showing that NIR wavelengths exhibit higher sensitivity (up to ~100 μm/RIU) due to reduced scattering and deeper light–matter interaction and Figure 9D,E present the simulated electromagnetic field distributions under different illumination conditions, indicating stronger evanescent field confinement in NIR regions compared to visible regions—critical for detecting trace biomolecular interactions at sensor surfaces. Figure 9F–I highlight both simulated and experimental optimization results. The reflectance spectra shift systematically with changes in incident angle (Figure 9F,H), confirming the wavelength-dependent resonance condition. Figure 9G,I show the sensitivity-wavelength relationship and the Figure of Merit (FOM) curves, validating that NIR-based configurations yield superior spectral sensitivity and detection resolution compared to visible counterparts [115]. These studies tell us that hybrid sensing systems not only enhance detection accuracy but also facilitate early warning systems for quality degradation and microbial contamination, paving the way for integration into digitalized smart food safety frameworks. Most reported NIR–biosensor systems are complementary rather than fully integrated and, in many cases, NIR spectroscopy is used for rapid, non-destructive compositional screening, while biosensors provide molecular-level verification for specific contaminants or spoilage markers. Only a few studies demonstrate true hybrid operation, where both modalities function simultaneously and share synchronized data outputs. Current examples mostly remain at the beginner or laboratory validation stage, with ongoing efforts focused on achieving real-time, in situ performance in industrial settings. This clarification ensures accurate representation of current technological maturity while highlighting the progress toward fully integrated hybrid monitoring platforms.

4.3. Real-World Applications and Hybrid Systems

The real-world application of integrated NIR-biosensor platforms has gained momentum as the food industry seeks faster and more reliable tools for quality and safety monitoring. Perishable supply chains, especially for fresh fruits, vegetables, and seafood, are particularly vulnerable to losses during postharvest handling, transportation, and retail display. Traditional inspection methods are often too slow to prevent spoilage-related waste, making hybrid systems increasingly valuable. In these systems, NIR spectroscopy provides rapid, non-destructive insights into bulk compositional attributes such as soluble solids, firmness, and moisture, while biosensors offer molecular-level detection of specific contaminants or spoilage metabolites. For instance, Sun et al. developed a hybrid surface-enhanced Raman scattering (SERS)-active platform using gold nano-dendrites decorated with silver nanoparticles, enabling ultrasensitive detection of food contaminants [116]. Similarly, Gan et al. reported a ratiometric fluorescent biosensor based on metal–organic frameworks for the ultrasensitive detection of acrylamide in foods, highlighting the precision of biosensors in targeting molecular hazards often invisible to broad-spectrum NIR techniques [117].

As summarized in

Table 2. Hybrid NIR-biosensor applications in food quality and safety monitoring.

Target/Analyte	Biosensor Type	Recognition Element	Signal Transduction	Sensitivity/LOD	Application Example	Reference
Salmonella spp., allergens (peanut protein), bacterial toxins	Electrochemical	Enzymes, antibodies, aptamers	Amperometric, impedimetric, potentiometric	Varies; 1–10 CFU/mL for pathogens, ng/mL for allergens	Rapid (<30 min) screening across multiple contaminants in dairy, meat, and produce	[118,119]
Pesticide residues (organophosphates, carbamates)	Optical/Electrochemical	Enzymes (AChE), aptamers	Fluorescence quenching, electrochemical inhibition	ppt levels for chlorpyrifos, malathion	Real-time produce safety checks at farms and markets	[125,126]
Foodborne pathogens (E. coli, Listeria, Salmonella)	Flexible electrochemical biosensors	Antibodies, aptamers, phage proteins	Electrochemical impedance, voltammetry	Sub-femtomolar; <10 CFU/mL	On-site testing using flexible substrates. Spiked milk and chicken samples were validated	[127,128]
Food safety monitoring (multi-target)	Microfluidic biosensors	Antibodies, aptamers, MIPs	Electrochemical + optical	fM to nM; <5 μL volume	Dairy, cereal food matrices, real-time field monitoring, multiplexed detection	[129,130]
Pathogens, toxins, heavy metals	Optical biosensors	Antibodies, aptamers, enzymes, DNA probes	SPR, FRET, colorimetry	pM–nM	Multi-target screening in processed foods, beverages, produce	[131,132]
Ochratoxin A (mycotoxin)	Graphene FET aptasensor	ssDNA aptamer	Field-effect modulation	LOD ≈ 1.4 pM, 10 s response	Wine safety monitoring with reusable channels	[132]
Smart traceability systems	Integrated NIR + IoT + Blockchain		Optical NIR + digital ledger		Real-time traceability of food origin, freshness, handling history	[133]
Heavy metals (Pb2+, Hg2+, Cd2+)	Portable paper-based electrochemical	DNAzyme/aptamer	Electrochemical stripping voltammetry	Low ppb	Seafood and rice screening with smartphone readout	[134,135]
Food allergens (gluten, peanut, shellfish)	Lateral flow optical biosensor	Antibody or aptamer	Colorimetric (AuNP)	ng/mL	Consumer self-testing and restaurant verification.	[136,137]
Antimicrobial residues (β-lactams, tetracyclines)	Multiplex electrochemical biosensor	Enzyme inhibition assays	Differential pulse voltammetry	ng/mL–pg/mL	Dairy industry compliance monitoring	[138]
Microplastics in seafood	Optical biosensor	MIP	Fluorescence quenching	0.5 μg/L	Rapid detection in fish and shellfish	[134]
Acrylamide in baked products	Electrochemical	Enzyme (asparaginase)	Amperometric	0.02 μg/g	Monitoring contaminants in bread, biscuits	[139]
Norovirus in fresh produce	Microfluidic optical biosensor	DNA aptamer	SPR	20 genome copies/mL	Detection in lettuce wash water	[140]
Biogenic amines in fish	Wearable electrochemical patch	Enzyme (DAO)	Potentiometric	0.1 mg/L	Real-time freshness monitoring in packaging	[141]
Cyanotoxins in freshwater fish	Electrochemical aptasensor	DNA aptamer	Square wave voltammetry	0.8 ng/L	On-site aquaculture monitoring	[142]
Mycotoxins in cereals (multi-class)	Multiplex optical biosensor	Aptamers	FRET	0.2–1 ng/g	Simultaneous maize/wheat mycotoxin screening	[143,144]
pH and spoilage gases in packaged foods	Hybrid NIR–electrochemical smart label	pH-sensitive dye, conductive polymer	Optical + electrochemical		Smart packaging shelf-life tracking for meat/dairy	[145]

, hybrid approaches are particularly effective in real-world food monitoring. In poultry, antibody-functionalized gold nanoparticle biosensors have achieved highly sensitive colorimetric detection of Salmonella Typhimurium, with potential for on-site food safety screening [118]. Complementary to this, Konstantinou et al. demonstrated the use of B.EL.D^TM portable biosensor technology for detecting Salmonella spp. in meat, showing practical adaptability in industrial food environments [119]. Similarly, portable Vis-NIR modules combined with biosensor models have been used to monitor pesticide residues in cucumbers, while freshness-indicating smart labels for pork and fish combine pH-responsive films with spectroscopic readouts for rapid spoilage detection. These platforms demonstrate versatility in detecting bacteria such as E. coli and Salmonella, as well as heavy metals, providing a low-cost and rapid alternative to conventional methods like PCR. These biosensors combine bioreceptors with transducers and signal processors, enabling portable, low-cost, and sensitive tools for real-time food safety analysis and crop monitoring. Zhang et al. further emphasized the integration of micro/nano-bubble technology with hybrid sensing platforms to improve food preservation and extend shelf life [120]. The overview of Figure 10 demonstrates a series of electrochemical immunosensors developed for the rapid and sensitive detection of Salmonella. Figure 10A shows the fabrication of a gold electrode modified with cysteamine and glutaraldehyde to immobilize anti-Salmonella immunoglobulins (IgC). The immobilized antibodies provide specific biorecognition sites, allowing stable electron transfer and improved signal reproducibility for pathogen detection [121] and Figure 10B presents a magnetic bead-based impedance biosensor, where Salmonella cells are captured using antibody-functionalized magnetic beads (MBs) and glucose oxidase (GOx) conjugates. The enzymatic conversion of glucose to gluconic acid generates measurable impedance changes that correlate with bacterial concentration [122], whereas Figure 10C depicts a chronoamperometric immunosensor employing antibody-functionalized transducer wires (T-DW). Upon antigen–antibody binding with S. typhimurium, a distinct increase in current signal is observed, enabling quantitative detection through chronoamperometric analysis [123], and finally, Figure 10D demonstrates a nanocomposite-based immunosensor utilizing a glassy carbon electrode coated with carbon nanotube–poly(amidoamine)–chitosan (CNT–PAMAM–Chi) and gold nanoparticles (AuNPs). This nanostructured surface enhances electron transfer and antibody immobilization, yielding high sensitivity and selectivity for Salmonella detection in complex food samples [124]. To improve comparability and transparency, performance indicators summarized in Table 2 have been standardized and explicitly defined. Coefficient of determination (R²) reflects model fit; root mean square error (RMSE) represents prediction deviation; ratio of performance to deviation (RPD) indicates calibration robustness; the root mean square error of prediction (RMSEP) is used to assess the predictive capacity of calibration models; and classification accuracy denotes categorical predictive success. This ensures that NIR, biosensor, and hybrid systems can be compared on a consistent analytical basis across studies and food matrices.

To ensure reliability and industrial relevance, validation environments for hybrid NIR–biosensor systems were clearly specified. Gan et al. [117] validated a ratiometric fluorescent metal–organic framework (MOF) biosensor for acrylamide detection using spiked biscuit and fried potato extracts, confirming that the limit of detection (LOD) and recovery rates aligned with regulatory thresholds in real food matrices. In contrast, studies that conducted measurements solely in buffer-based media were categorized as conceptual or pre-validation demonstrations. This clarification ensures that analytical performance indicators—such as sensitivity, repeatability, and selectivity—reflect true food-matrix conditions rather than simplified laboratory environments. These applications, among others listed in Table 2, confirm that hybrid NIR-biosensor systems not only enhance detection accuracy but also enable automated decision-making, supporting early interventions such as cold-chain optimization, targeted recalls, and better segregation of produce for export or processing.

4.4. Potential for Blockchain and IoT-Enabled Traceability

The integration of hybrid NIR–biosensor systems with Internet of Things (IoT) infrastructure and blockchain-based traceability frameworks is advancing from concept to practical application in food quality assurance. In these systems, sensor outputs—both spectral and biochemical—are transmitted wirelessly to cloud databases, where blockchain secures the data within a tamper-resistant ledger [146]. Each data entry becomes a verifiable point in the product’s digital history, accessible to producers, distributors, retailers, and even consumers. This integration creates a closed-loop feedback system: real-time measurements from hybrid devices can be linked to logistics operations under controlled conditions, though large-scale deployment requires more robust communication and data harmonization frameworks. Recent deployments demonstrate that linking hybrid sensing platforms with IoT and blockchain can enhance traceability efficiency and data reliability across food supply chains. For example, IoT-linked NIR modules have been integrated into agricultural logistics networks for real-time monitoring of apple quality and spoilage dynamics, showing measurable gains in predictive accuracy when coupled with deep learning analytics [110]. Similarly, cellulose nanocrystal-based smart packaging compatible with IoT-enabled biosensors has been developed for on-package freshness monitoring and digital record keeping, supporting end-to-end transparency [147]. The current challenge lies in scaling these systems beyond pilot-level integration by standardizing data formats, improving communication stability, and ensuring interoperability between spectroscopic and biochemical nodes. By transitioning from conceptual demonstrations to field-tested implementations, hybrid NIR–biosensor networks are positioned to support real-time, secure, and traceable food quality management within connected supply chains.

5. Challenges in Commercial Deployment

5.1. Calibration Transfer and Environmental Variability

One of the main obstacles in commercializing hybrid NIR-biosensor platforms is maintaining calibration accuracy when instruments or environments change. Small shifts in light source stability, detector sensitivity, or optical geometry can distort NIR spectra, while biosensors are sensitive to variations in temperature, pH, and food matrix composition. Fan et al. demonstrated that apple quality prediction by NIR suffered from biological variability [148], while Xing et al. showed that fermentation processes introduced spectral shifts complicating quantitative analysis [149]. Environmental factors add further complexity. In supply chains, devices may be used in warehouses, transport containers, or open markets, where humidity, vibration, and ambient light can reduce performance. Tian et al. further demonstrated that Vis-NIR hyperspectral imaging for purple sweet potato quality assessment required careful calibration to handle variability in moisture and anthocyanin distribution, underscoring the difficulties of transferability across food matrices [150]. Other studies confirm these challenges across different matrices: NIR-based antioxidant assessments in honey [151], multi-platform saffron authentication using High-Performance Liquid Chromatography (HPLC), Gas chromatography Mass Spectroscopy (GC/MS), and NIRS [152], and Roselle origin verification via NIR and NMR [153]. To improve robustness, researchers recommend calibration transfer algorithms, adaptive machine learning, and reference standards. Without such strategies, large-scale deployment risks inconsistent results, undermining both compliance and user trust. These challenges highlight the need for further inter-laboratory validation and robust calibration transfer methods before scaling hybrid systems into IoT frameworks.

5.2. Sensor Drift, Fouling, and Shelf-Life Issues

Hybrid sensing systems face gradual performance deterioration during long-term use. In NIR devices, optical drift caused by lamp decay, detector aging, or window contamination can shift baseline readings and bias predictive models. Biosensors are more vulnerable: enzymes lose catalytic activity through denaturation, antibodies undergo conformational changes, and aptamers degrade in complex food matrices. Biofouling adds another layer of difficulty, as proteins, fats, or microbial films accumulate on sensor surfaces, blocking active sites and reducing sensitivity. To mitigate these issues, strategies such as protective nano coatings, self-cleaning microfluidic flushing, disposable cartridges, and on-board recalibration have been explored. However, these solutions introduce trade-offs in cost, waste generation, and maintenance complexity. Effective deployment requires balancing sensor stability with economic feasibility, as performance drift and fouling directly increase operational costs and may compromise regulatory acceptance.

Recent studies highlight these challenges and solutions: Shoaib et al. emphasized aptamer instability in food biosensors [154]; Zhu et al. demonstrated matrix fouling issues in mycotoxin detection [75]; and Qu et al. reviewed nanotechnology-driven enzyme biosensors as a path to enhanced stability [9]. Similarly, Xu et al. developed an impedimetric aptasensor based on highly porous gold that addressed biofouling and stability concerns in complex food matrices [155], while Zhang et al. reported optical drift in heavy metal detection sensors [156]. More recently, Chen et al. advanced portable NIR-fluorescent sensing systems that incorporated recalibration protocols to counteract drift [157]. Additional contributions reinforce these findings; Hassan et al. employed SERS-based nanostructures for rapid contaminant detection, showing that surface modifications can mitigate fouling [158]. Lin et al. demonstrated how multi-sensor arrays enhance robustness in rice freshness monitoring, reducing the impact of drift [159]. Studies such as Golly et al. on grape preservation and Anandkumar et al. on toxic element accumulation further underscore the importance of stability and recalibration in food safety monitoring [160,161]. In Figure 11, multifunctional biosensors integrating PEC, SERS, and colorimetric modalities provide redundant signal channels that cross-validate measurements. This design minimizes the impact of biofouling or drift in any single modality, enhancing stability and extending operational shelf-life in real-world food monitoring applications. Figure 11A presents a FePor-TPA-based dual-mode PEC–colorimetric biosensor integrating glucose oxidase (GOx) and AuNPs for Staphylococcus aureus detection. This system not only identifies microbial contamination but also correlates enzymatic oxidation by-products (e.g., H₂O₂) with freshness indicators such as oxidative spoilage. The visible color change and photocurrent variation allow rapid on-site estimation of microbial load, a critical determinant of food shelf life [162] and Figure 11B illustrates a CM–MXene hybrid PEC–SERS platform that detects heavy metal ions (Pb²⁺) and pathogens through DNAzyme-assisted signal generation. The “on–off” PEC and SERS responses provide dual confirmation of spoilage or contamination events, offering a robust strategy for predicting the remaining shelf life of stored products by continuously sensing biochemical and environmental changes [163]. Collectively, these advances highlight that while sensor drift and fouling remain critical challenges, integrated nanotechnology and calibration strategies are paving the way toward more durable and reliable systems.

5.3. Economic and Manufacturing Constraints

The commercial deployment of hybrid NIR-biosensor systems faces not only technical but also economic and manufacturing barriers. High-performance NIR components such as photodiodes, diffraction gratings, and filters are costly due to their precision engineering and stability requirements. Similarly, biosensors using biological recognition elements (e.g., enzymes, antibodies, aptamers) demand stringent production conditions to ensure reproducibility and reliability. This dual complexity often translates into elevated manufacturing costs, creating significant barriers for small- and medium-sized enterprises (SMEs) operating on limited budgets [154]. Additionally, field applications demand robust housings, environmental sealing, and impact resistance to protect sensitive optical and biochemical parts during handling and transportation. Economies of scale could lower per-unit costs, but widespread adoption is typically delayed without strong business incentives [164]. Large-scale manufacturing requires high initial investment, while early adopters often face financial risk before sufficient market penetration is achieved. To overcome these constraints, modular designs have been proposed where biosensor cartridges can be interchanged across standardized NIR platforms, thus reducing upgrade costs. Multi-analyte biosensing platforms, capable of detecting multiple hazards simultaneously, also present a cost-efficient solution, as they replace the need for several single-function devices [165]. Recent advances in nanotechnology-driven biosensors have further highlighted scalable production routes that balance sensitivity, durability, and affordability, improving prospects for commercialization [9,166]. While hybrid NIR-biosensor systems hold strong potential for food safety and quality monitoring, manufacturing scalability and affordability remain key bottlenecks. Addressing these requires a combination of modular designs, multi-analyte integration, and advanced materials engineering. Continued collaboration between academia and industry will be essential to translate laboratory innovations into commercially viable platforms.

5.4. Regulatory Hurdles and Market Acceptance

Regulatory approval of hybrid NIR-biosensor systems is challenging because they integrate two distinct domains: physical-chemical measurement (NIR spectroscopy) and biochemical detection (biosensors). Agencies often require separate validations for each modality, including equivalence to gold-standard methods, stability over the intended shelf life, and reproducibility under real-world conditions. In the EU, hybrid devices for official food control must meet both analytical instrumentation standards and diagnostic regulations, while in the U.S., pathogen- or toxin-detecting biosensors may fall under FDA or USDA oversight depending on commodity type [167,168]. Beyond regulatory clearance, market acceptance depends on stakeholder trust and perceived ease of use. Food producers, distributors, and regulators must believe that hybrid systems provide reliable, actionable insights without adding complexity. Field trials, transparent validation reports, and partnerships with industry stakeholders are key to building confidence. Moreover, user-friendly interfaces and automated data interpretation tools can lower technical barriers, facilitating adoption in resource-limited settings. Recent studies provide examples of how validation across diverse contexts is critical. Wang et al. showed that pulsed light can induce structural and functional changes in horseradish peroxidase, illustrating the need for regulatory evaluation of system stability under non-traditional treatments [169]. In the postharvest sector, Raynaldo et al. demonstrated the potential of Wickerhamomyces anomalous to control black spot disease in cherry tomatoes, but also highlighted mechanistic uncertainties that must be resolved before regulatory approval [170]. Packaging innovations raise similar concerns: Lin et al. developed nanofibers containing chrysanthemum essential oil for beef packaging, presenting safety and migration issues that regulators must scrutinize before adoption [171]. At a broader food systems level, Navarro-Hortal et al. emphasized that functional foods such as berries, curcumin, and olive oil require clear molecular-level validation before claims can be accepted by authorities [172]. Without these trust-building measures, even highly advanced hybrid systems risk limited penetration in the global food monitoring market [164,173,174].

6. Future Directions and Innovation Potential

6.1. Advances in Wearables, Handhelds, and Wireless Platforms

The shift toward portable and wearable hybrid systems reflects the growing demand for real-time, point-of-use food quality assessments. Handheld NIR-biosensor devices are increasingly deployed in farms, fisheries, and retail outlets, enabling inspectors to test products directly in the field. Wearable systems, such as sensor-integrated gloves or wristbands, provide continuous monitoring during handling and logistics. Recent progress in miniaturized photonics and low-power biosensor electronics has allowed both sensing modalities to be combined in compact, battery-operated platforms. Smartphone integration further strengthens these systems by harnessing computational power, GPS tagging, and wireless connectivity for real-time, contextualized reporting. These advances enable distributed monitoring networks where many devices feed into unified dashboards, enhancing transparency and accelerating corrective actions across supply chains [175,176]. Researchers have contributed significantly to this field, particularly in miniaturized spectroscopic sensing and real-time food monitoring. Examples include applications of NIR and FT-NIR spectroscopy for tea quality assessment [165], portable spectrometers for egg protein and food quality analysis [177], hybrid approaches combining NIRS with chemometrics for saffron authentication [152], and Vis-NIR applications for rice freshness evaluation [159]. These works highlight the department’s leadership in advancing hybrid sensing toward handheld, wearable, and wireless systems.

6.2. AI and Cloud-Based Food Monitoring Systems

Hybrid NIR–biosensor systems generate large, multi-modal datasets demanding advanced analytics. Deep learning models excel here, merging continuous spectral data with biosensor signals to produce unified freshness or safety indices that evolve with incoming data. These AI frameworks learn across seasons, geographies, and commodity types, enhancing adaptability and accuracy over static chemometric models [178]. Cloud integration elevates this capability; aggregating data across distributed devices enables predictive intelligence at scale. Automated shelf-life forecasts, spoilage alerts, and compliance dashboards can trigger actions in near-real time, such as rerouting perishable goods before spoilage occurs. AI-fed, cloud-connected systems thus transform hybrid setups from passive sensors to dynamic agents in quality management [179,180]. Supportive advances in hardware from electrochemical biosensors suited for rapid in-field measurement [181] to cutting-edge non-destructive spectroscopic methods [60] provide robust data streams for AI deployment. Moreover, NIR-based classification tools and IoT-enabled risk models highlight growing feasibility for scalable, real-time decision-making architectures [180,182].

6.3. Green and Sustainable Sensing Materials

The future of hybrid sensing will be shaped not only by performance and connectivity but also by sustainability. Researchers are increasingly adopting biodegradable polymers and cellulose-based substrates as flexible, eco-friendly alternatives to traditional plastics [183,184]. Paper-based and nanocellulose platforms offer low-cost, disposable, and recyclable options for food quality monitoring while minimizing environmental impact [185,186]. For biosensors, plant-derived recognition elements and green synthesis methods are being explored to replace environmentally taxing reagents. Recent studies have highlighted bio-based solutions that align with circular economy principles. For instance, Otero et al. emphasized the revalorization of olive oil processing by-products through green extraction strategies, showing how agro-industrial waste can be converted into sensing materials [187]. Similarly, Zhou et al. demonstrated how deep eutectic solvent processing can serve as a green, water-saving method for preparing biodegradable sensing substrates [188]. From a nanomaterial perspective, Wang et al. reported the design of flexible, superhydrophobic biomass carbon aerogels from corn bracts, underscoring the potential of agricultural residues as high-performance, sustainable sensor scaffolds [189]. In addition, Ariza-Calahorra et al. developed Pickering emulsions stabilized by biopolymers to enhance the bio-accessibility of curcumin, showcasing how food-derived and biodegradable systems can integrate into sustainable sensing and delivery platforms [190]. Such bio-based nanosensors align with global sustainability goals by using renewable and biodegradable components [191]. Circular economy principles are also being integrated into NIR device construction, ensuring that optical and electronic components can be recovered and reused at the end of service life [192]. Designing devices for modular disassembly further reduces waste, since only worn components need replacement rather than discarding entire units. These environmentally conscious innovations not only enhance the market appeal of hybrid sensing systems but also ensure compliance with tightening sustainability regulations and growing consumer expectations for greener technologies.

6.4. Research Needs and Industry–Policy Recommendations

For hybrid NIR-biosensor systems to transition from laboratory prototypes to mainstream industry tools, targeted research and coordinated policy support are essential. A key research priority is the development of robust, transferable calibration models capable of functioning across devices, environments, and food categories, supported by large-scale, real-world datasets that capture seasonal, geographic, and processing variability [156,193]. Long-term stability studies of biosensor components, especially those using biological recognition elements such as enzymes, antibodies, or aptamers, are also required to ensure durability under diverse operational conditions [154]. Another critical need is the creation of integrated data fusion frameworks that can merge NIR spectral data with biosensor outputs into unified freshness and safety indicators. Open-access spectral–biosensor databases and standardized validation protocols would accelerate algorithm development and allow meaningful cross-platform comparisons [9,152]. On the policy front, harmonized regulatory pathways for multimodal sensing technologies could streamline approvals, avoiding separate validations for physical-chemical and biochemical modalities. Incentives such as subsidies, tax credits, and inclusion in official monitoring programs would encourage broader adoption, especially in high-risk food sectors [173,174]. Coordinated industry-policy initiatives should also ensure adoption extends beyond high-value exports to domestic markets. This includes pilot programs for small and medium-sized enterprises, operator training, and public–private partnerships to implement hybrid sensing at critical control points, ultimately strengthening food safety and reducing waste across supply chains.

7. Conclusions

The rapid, non-destructive assessment of food freshness and safety is increasingly critical in an era of globalized supply chains, consumer demand for minimally processed products, and heightened food safety regulations. This review shows that NIR spectroscopy and biosensor technologies each offer unique yet complementary advantages: NIR provides broad, compositional insights into quality changes, while biosensors deliver targeted, high-specificity detection of safety hazards. Advances in portable instrumentation, chemometrics, AI integration, and IoT connectivity have brought both technologies closer to real-time, inline deployment across diverse food matrices, including meat, seafood, eggs, fruits, and vegetables. The combination of these sensing platforms into hybrid systems holds significant promise, enabling simultaneous monitoring of freshness and contamination while supporting emerging concepts like blockchain-enabled traceability and predictive shelf-life modeling. Nevertheless, practical challenges, including calibration transfer, sensor drift, economic constraints, and complex regulatory pathways, remain significant barriers to widespread adoption.

Overcoming these hurdles will require targeted research into robust calibration models, stable and sustainable sensing materials, and harmonized validation frameworks, alongside industry policy collaborations that incentivize adoption. Looking ahead, hybrid NIR-biosensor systems integrated with AI-driven analytics, cloud-based monitoring, and sustainable materials are well-positioned to transform food quality control from reactive inspection to proactive, predictive management. By addressing current technical and operational limitations while fostering cross-sector cooperation, these technologies can play a central role in reducing food waste, improving safety, and enhancing consumer trust in an increasingly complex global food system. While hybrid NIR–biosensor systems show strong laboratory performance, industrial translation remains limited by calibration robustness, sensor drift, and environmental variability. Continued validation under real-world conditions, supported by cross-sector collaborations, will be essential to realize reliable, near real-time applications in food safety and quality monitoring.