Portable paper-based microfluidic device for rapid on-site screening of milk adulterants

Anushka, Aditya Bandopadhyay* and Prasanta Kumar Das
Department of Mechanical Engineering, Indian Institute of Technology Kharagpur, Kharagpur, West Bengal, India. E-mail: anushka@iitkgp.ac.in; aditya@mech.iitkgp.ac.in; pkd@mech.iitkgp.ac.in

Received 4th June 2025 , Accepted 18th August 2025

First published on 22nd August 2025


Abstract

Milk adulteration remains a significant public health concern in India, where conventional laboratory-based detection methods are often costly, time-consuming, and impractical for field use. This study introduces a novel paper-based microfluidic device designed for rapid, low-cost detection of multiple milk adulterants. The device comprises a 3D-printed strip holder and utilizes gravity-assisted capillary flow through porous paper, eliminating the need for hydrophobic barriers or external power sources. Its modular design allows for easy reuse of the holder while only replacing the paper strip for successive tests. The platform enables visual detection of common adulterants—including neutralizers, starch, hydrogen peroxide, urea, detergents, and boric acid—via reagent-specific colorimetric responses. The device meets the ASSURED criteria of World Health Organization for point-of-care diagnostics, offering a promising tool for decentralized milk quality monitoring and contributing to both consumer safety and improved supply chain transparency in the dairy industry. The device demonstrated a limit of detection (LOD) as low as 0.03% for urea and hydrogen peroxide, outperforming existing paper-based methods. The results were validated across five independent trials per condition, with high reproducibility and minimal cross-reactivity, confirming the diagnostic reliability of the platform.


1 Introduction

India is the largest producer of milk globally, with a diverse and extensive dairy farming sector.1–3 According to the National Dairy Development Board (NDDB), there are approximately 70 million dairy farms in India. These farms are predominantly small-scale, with an average of two to three milking animals each. The density of dairy farms is particularly high in a few states, which are major contributors to the national milk supply.2 The high density and large number of dairy farms underscore the importance of ensuring milk quality and safety. Given the widespread nature of dairy farming and the critical role milk plays in the Indian diet, there is a pressing need for effective detection devices to monitor and control milk adulteration.

Adulteration involves adding substances to increase volume or mask deficiencies, leading to serious health risks for consumers. The National Milk Safety and Quality Survey conducted by the Food Safety and Standards Authority of India (FSSAI) revealed that while large-scale adulteration is not widespread, instances of contamination and quality issues persist at a high level.4 The survey found that samples were adulterated with substances like hydrogen peroxide,5 detergents,6 urea,7 and neutralizers.8 This indicates a need for vigilant monitoring and stringent quality controls to ensure milk safety. The permissible limit of ingestion for these substances is outlined in the guidelines provided by the WHO.9 Consuming these toxins in excess of the permissible level can cause severe health issues like kidney failure, newborn mortality, gastrointestinal problems, diarrhea, and even cancer.10

To contextualize the chemical risks associated with milk adulteration, Table 1 summarizes each contaminant's purpose, typical concentration, and health implications.

Table 1 Summary of milk adulterants detected in this study, their intended use, typical concentrations, and associated health risks
Adulterant Purpose Typical conc. (%) Detrimental effects
Urea Boost SNF content 0.03–0.5 (ref. 4) Kidney damage, metabolic imbalance10
Starch Increase SNF, thickness 0.1–1.0 (ref. 4) Indigestion; risk for infants, diabetics9
Hydrogen peroxide Antibacterial preservative 0.03–0.1 (ref. 5) Oxidative stress, GI irritation11
Detergent Froth formation, emulsify fats 0.1–0.5 (ref. 6) GI irritation, liver toxicity11
Neutralizer Mask acidity, extend shelf life 0.05–0.2 (ref. 8) Alkalosis, pH imbalance10
Boric acid Prevent spoilage 0.1–0.2 (ref. 9) GI distress, kidney damage9


Given these widespread practices and their associated health risks, rapid and reliable detection of adulterants in milk becomes an urgent public health priority.

Despite their accuracy, traditional methods for milk adulteration detection suffer from drawbacks such as high cost, complex protocols, and dependency on sophisticated instrumentation and skilled personnel.12 Most existing methods are designed to target a single or narrow class of adulterants, limiting their scope for field-level diagnostics.

Various methods have been used for the detection of adulterants in milk such as liquid chromatography-tandem mass spectrophotometry,13–15 gas chromatography-mass spectrometry,16,17 high-performance liquid chromatography,18,19 matrix-assisted laser desorption/ionization mass spectrophotometry and nuclear magnetic resonance. Furthermore, enzyme-linked immunosorbent assays (ELISA) came out as a powerful alternative to the above methods.20–22

Silva et al.23 developed a smartphone-based system to evaluate milk quality by assessing parameters like refractive index and cryoscopic point, offering a portable yet indirect method for detecting dilution-based adulteration. Coitinho et al.24 employed FTIR spectroscopy to identify milk adulterants by analyzing absorption bands characteristic of various contaminants. Wang et al.25 developed a lateral flow nucleic acid assay (LFNAA) to authenticate yak milk, integrating isothermal recombinant polymerase amplification (RPA) with a lateral flow platform. This method targets species-specific genetic markers and produces visible results within 40 minutes, allowing field-level authentication without specialized instrumentation.

Lu et al.26 utilized near-infrared (NIR) spectroscopy combined with least squares support vector machines (LS-SVM) to rapidly detect melamine in milk powder. Feature extraction using partial least squares discriminant analysis (PLS-DA) improved classification accuracy, enabling 100% detection in both training and testing sets. Das et al.27 introduced an electrical impedance spectroscopy (EIS)-based biosensor for differentiating polar and non-polar adulterants. By analyzing variations in impedance and capacitance through a Pt/Teflon/SiO2/Si sensor stack, they developed a diagrammatic method to quantify adulteration. Nieuwoudt et al.28 demonstrated a portable Raman spectroscopic system integrated with fiber optics and hemispherical aluminum wells to enhance signal quality. This setup enabled rapid, preparation-free detection of sucrose, ammonium sulfate, melamine, and urea in milk, with sensitivity down to 50–1000 ppm (0.005–0.1%). In order to identify other potential contaminants in milk, He et al.29 integrated 2D correlation spectroscopy analysis with Fourier transform infrared techniques.

Although various lab-based methods are available for the accurate detection of adulterants, their use is limited for point-of-care testing and requires expensive, bulky instrumentation and highly trained personnel. This has led to the development of paper-based methods as the better alternative for the detection and quantification of adulterants. Anushka et al.30 reviewed the fabrication techniques and applications of various paper-based methods tailored for the analysis of milk contaminants. The study explores innovative techniques that eliminate the need for organic solvents, making the process more environmentally friendly. Chen et al.31 introduced wax valves designed to enhance distance-based analyte detection in paper microfluidic devices. By optimizing wax printing parameters such as line thickness and melting time, the study significantly improved the functionality of μPADs. The research demonstrates a linear relationship between wax line width and hydrophobic barrier formation, aiding in the efficient design of paper-based diagnostic tool. Fan et al.32 discussed the development of a novel milk carton integrated with paper-based microfluidics for rapid milk quality testing. The proposed design aims to provide an on-site, cost-effective solution for detecting contaminants in milk. The study highlights the feasibility and effectiveness of using paper-based sensors to ensure milk safety in real-time. A low-cost μPAD was created to identify milk adulterants like starches, urea, and sugars in the form of glucose and sucrose.33 The paper test card utilizes specialized chemical indicators that react specifically to common adulterants found in milk, such as water, urea, and starch. These reactions lead to discernible color changes or other visual cues, allowing for easy detection of adulteration.

Several recent advances in paper-based analytical devices (μPADs) have explored hydrophobic patterning, multiplex detection, and digital quantification techniques. Rewatkar et al.34 fabricated μPADs using paraffin wax molding and laser patterning on A4 paper, enabling controlled microfluidic flow validated through dye leakage. Aksorn et al.35 created a multiplex assay using polylactic acid-based hydrophobic barriers for colorimetric detection of multiple sugars on chromatography paper.

Furthermore, Lima et al.36 developed a bio-active paper strip that detects hydrogen peroxide through guaiacol-peroxidase chemistry, producing a red tetraguaiacol product captured via digital photography. Govindarajalu et al.37 proposed a portable wax-patterned paper sensor for starch detection, relying on specific binding interactions. Guinati et al.38 reported a colorimetric μPAD with microfluidically connected test zones for simultaneous detection of urea, hydrogen peroxide, and pH, analyzed via RGB image analysis. Together, these studies underscore the promise of μPADs in decentralized diagnostics. However, many of these devices remain limited in scope—targeting single analytes, requiring precise handling, or lacking field robustness. The present work addresses these limitations by integrating multiplex detection, ambient operation, and ruggedized design, supporting consistent and reliable operation under field conditions.

Therefore, there is an urgent need for innovative solutions to address this issue. A milk adulteration detection device that is efficient, reliable, less expensive, portable, and easy to use can play a crucial role in ensuring that the milk reaching consumers is safe and of high quality. The device should offer high precision and reliability, irrespective of the surrounding conditions, and be capable of detecting even low concentrations of adulterants in milk. Such a device would not only benefit consumers by providing them with safer milk but also support dairy farmers and producers in maintaining high standards of quality. It would aid regulatory bodies in their efforts to monitor and control adulteration and ensure compliance with safety standards. By addressing the challenges of milk adulteration, this device has the potential to make a significant impact on public health and the dairy industry as a whole.

In response, we introduce a portable, paper-based microfluidic device engineered for the simultaneous detection of multiple milk adulterants. The device operates solely on natural capillary flow through porous paper substrates, eliminating the need for any external power source. It offers results comparable to laboratory-based methods while enabling point-of-care testing (POCT) in resource-limited settings. Notably, the configuration—including the inclined strip geometry and gravity-assisted flow—has been formally protected under Indian Patent No. 541444, titled “An easy method to detect milk adulteration”, filed at IIT Kharagpur in 2023. The device meets the ASSURED criteria (Affordable, Sensitive, Specific, User-friendly, Rapid and Robust, Equipment-free, and Deliverable to end-users) set by the World Health Organization 30, ensuring high relevance for decentralized diagnostics and public health surveillance.

2 Device design and working principle

The design features a platform comprising a set of interconnected test strips and a strip holder. The test strips are strategically inclined, enhancing capillary flow for effective fluid transport. To establish an effective fluid flow control within the devices, 3D printing technology was utilized to construct the strip holder. Furthermore, a sheet of grade 4 Whatman filter paper is cut in the required shape and placed in the 3D printed part. This sheet may be machine-punched or laser-cut. The development of a 3D-printed holder for test strips has streamlined the functionality and fabrication process of paper-based devices by eliminating the need for hydrophobic barriers. Traditionally, creating such barriers involved chemical treatments to alter the paper's surface properties or physical methods to block the pores. The device is robust and suitable for adulterant identification in liquid because no hydrophobic coatings are employed. Fabricating hydrophobic barriers in paper-based devices can introduce several technical drawbacks. Firstly, the process often takes more time due to the multiple steps required for chemical treatment, such as photolithography, inkjet etching, and chemical vapor deposition.39 This added complexity can also increase the cost of production. There are significant reliability concerns, such as the risk of leakage if the barriers are not perfectly formed, which can compromise the device's functionality. Wax-based barriers, while simple and cost-effective, are particularly prone to interference from external factors like high temperatures (above 60 °C), which can cause the wax to soften or melt, leading to device failure.40 Furthermore, the chemicals used for creating hydrophobic barriers can sometimes interfere with the assays being conducted, leading to inaccurate results.

The 3D device design ensures that the liquid flow rate is constant with that of an intact paper substrate. Even though several cuts are made in the paper sheet, the liquid flow rate remains the same as if the substrate were a complete sheet of paper. The reagents are preloaded into the detection zone of the paper and adulterants are identified using the colorimetric detection approach. A color intensity test has been used for quantitative analysis. The device can concurrently identify neutralizers, starch, hydrogen peroxide, urea, detergents, and boric acid in milk samples for the first time. Compared to traditional single strip-based adulterant detections, which require several experiments to identify a single adulterant, the simultaneous adulterant detection method is superior for milk samples. Adulterants added to milk are detected using a color intensity test by comparing their colors with a standard test strip. The device is capable of detecting multiple adulterants under ambient conditions. Detailed evaluation of detection sensitivity and comparison with existing methods is presented in the Results and discussion section. The device is particularly suitable for use in low-resource settings due to its straightforward fabrication, rapid analysis time, and robust design. By addressing both technological (ASSR) and user acceptance (UED) considerations, this device fully aligns with the ASSURED criteria outlined by WHO.30

One of the key advantages of the present devices is their superior limit of detection for various adulterants compared to the devices in the literature. The present device can identify trace amounts of adulterants with a higher sensitivity, providing users with more accurate and reliable results. The developed device is capable of detecting multiple milk adulterants with high sensitivity, achieving detection thresholds as low as 0.03% (v/v) under ambient conditions. Therefore, present adulterant detection device distinguish themselves from the other devices through their superior limit of detection.

3 Fabrication

The device comprises a simple platform featuring distinct functional elements: a reservoir positioned at one end to accommodate the milk sample and slots situated at the opposite end to hold the paper strips (as shown in Fig. 1). Between these endpoints, a set of test strips is securely mounted, each independently treated with reagents for the detection of common milk adulterants such as hydrogen peroxide, boric acid, urea, detergent, starch, and neutralizer.
image file: d5sd00090d-f1.tif
Fig. 1 Schematic representation of the capillary-driven device holder. (a) Isometric view. (b) Side view.

The holder is 3D-printed and designed with seven grooves that securely anchor the reagent-treated test strips in a parallel alignment, with the possibility to expand further if more tests are required. Each groove corresponds to a specific adulterant detection zone, allowing simultaneous colorimetric analysis across all strips. This structured layout ensures mechanical stability, prevents cross-strip reagent diffusion, and supports organized sample application. Milk is drawn to the other end of the reservoir due to capillarity aided by gravity.

To achieve the desired geometry, the test strips are fabricated from Whatman filter paper using a CO2 laser system (10.6 μm wavelength, 60 W power), operated at 5% of maximum speed and 4% power specifically tailored for paper cutting. Extensive trials were conducted to optimize capillary flow and reagent interaction while ensuring the process remained free of contamination.

The present design leverages gravity to assist capillary flow, promoting efficient sample transport toward the reaction zones. This approach ensures smooth and consistent flow, facilitating timely and reliable test results. Additionally, a carefully considered structural design enhances fluid movement along the strip, contributing to faster sample progression and improved interaction with reagents. These strategic choices collectively boost the device's performance, offering greater reliability and ease of use.

3.1 Optimization of capillary pore size

Before finalizing the device, an extensive series of experiments was conducted to optimize various parameters critical to its performance. The selection of suitable materials for the test strips emerged as a pivotal aspect of the development process. After screening a diverse range of substrates, paper was identified as a particularly promising candidate due to its widespread availability, affordability, and favorable wicking characteristics. The porous nature of paper enables spontaneous fluid movement and uniform distribution of reagents without external actuation, ensuring consistent interaction between the sample and the reagents on each strip.

The investigation extended beyond material selection, delving into how pore size influences capillary flow. Experiments were performed using Whatman filter papers of grades 1, 4, and 41, selected for their distinct pore size distributions and flow properties. These grades were systematically evaluated for liquid flow rate, reagent retention, and compatibility with the intended device design (Table 2).

Table 2 Comparison of Whatman filter paper grades for device optimization
Grade Pore size Flow rate Remarks
G1 ∼11 μm Slow Good circularity; slower wicking not ideal for rapid testing
G4 20–25 μm Optimal Best balance of flow rate and retention; selected for final device
G41 >25 μm Very fast Rapid flow, but poor reagent control and uneven distribution


To visualize the extent of liquid spreading, dyed milk droplets were deposited onto each grade, and the final spread diameters were recorded post-absorption (Fig. 2), highlighting the influence of pore size on absorption and distribution characteristics.


image file: d5sd00090d-f2.tif
Fig. 2 Final spreading diameter of dyed milk droplets on Whatman filter papers (grade 1, grade 4, grade 41) after absorption, illustrating the effect of pore size on liquid spread.

To further quantify the spreading dynamics, droplet propagation was analyzed using the temporal evolution of the area-to-perimeter (A/P) ratio, serving as a geometric descriptor of front advancement and uniformity. As expected, an increase in pore size corresponded to enhanced wicking rates, consistent with classical capillary flow theory in porous media.

Grade 1, with a nominal pore size of approximately 11 μm, offers fine filtration capacity but restricts fluid flow, resulting in slower wicking dynamics. While beneficial for minimizing sample dispersion, the reduced flow rate can impede rapid reagent delivery across the substrate.

In contrast, grade 4, characterized by larger pores (20–25 μm), facilitates more efficient fluid transport, achieving a favorable balance between rapid wicking and adequate reagent distribution—making it highly suitable for microfluidic applications.

Grade 41, designed for coarse filtration, supports the fastest liquid flow owing to its larger pore architecture. However, the high flow rate can lead to less uniform reagent spread, potentially compromising consistency in reactions requiring even analyte distribution.

As illustrated in Fig. 3, an increase in pore size leads to a higher liquid flow rate due to reduced capillary resistance. Nevertheless, this accelerated flow may result in non-uniform reagent distribution. For applications requiring a balance between wicking efficiency and reagent uniformity, grade 4 provides an optimal compromise. Conversely, grade 1 favors controlled flow for tasks demanding minimal analyte dispersion, whereas grade 41 offers rapid transport at the expense of distribution uniformity.


image file: d5sd00090d-f3.tif
Fig. 3 Temporal evolution of area-to-perimeter (A/P) ratio and Hausdorff dimension during droplet spreading on different grades of filter paper (grade 1, grade 4, grade 41). A/P ratio (left axis) indicates geometric compactness; Hausdorff dimension (right axis) captures front complexity.

To further characterize the spreading behavior, the area-to-perimeter (A/P) ratio was complemented with Hausdorff dimension analysis, providing deeper insights into the evolving complexity of the spreading front. While the A/P ratio quantifies geometric compactness, the Hausdorff dimension captures deviations from idealized circular fronts, reflecting the intricacies of fluid propagation on porous media. A lower Hausdorff dimension indicates a smoother and more circular spreading front, while a higher value signifies greater irregularity and complexity in the droplet geometry. Notably, naive fitting based solely on geometric assumptions fails to account for the subtle deviations from circularity that naturally arise during droplet spreading. The observed trends indicate that more porous substrates, such as grade 4, exhibit a faster stabilization of the Hausdorff dimension, suggesting the formation of smoother, less complex fronts over time. This behavior implies a more homogeneous reaction environment, as the liquid uniformly permeates the substrate without following preferential pathways. Such uniform infiltration is critical for achieving consistent reagent distribution, thereby enhancing the reliability and reproducibility of detection outcomes.

Representative images of the actual device are shown in Fig. 4. Fig. 4(a) shows the assembled strip holder with dry reagent-loaded test strips before sample addition, while Fig. 4(b) captures the post-reaction colorimetric responses for six adulterants.


image file: d5sd00090d-f4.tif
Fig. 4 Capillary-driven device. (a) Device prior to sample addition. (b) Device after sample addition, showing colorimetric responses to different adulterants. Observed regions from top to bottom: yellow (urea detection); dark brown (starch); brownish-orange (hydrogen peroxide); rose red (neutralizer); blue (detergent); orange (boric acid); unreacted region (reference zone confirming no reagent interaction).

3.2 Design calibration of strip geometry

Determining the appropriate dimensions of the test strips—specifically their length, width, and spacing—was a critical step in achieving efficient capillary flow and reliable assay performance. These design choices significantly influenced fluid transport, reagent interaction, and manufacturability, while maintaining cost-effectiveness. Following a series of trials, the final geometry was calibrated to balance sensitivity, material usage, and ease of fabrication.

Initial prototypes employed strips with a width of 1 mm; however, this dimension proved insufficient to transport an adequate volume of milk to the reaction zones. The narrow width limited both flow rate and reagent interaction area, resulting in faint or inconsistent color changes. To address this, the strip width was increased to 2 mm, significantly improving sample transport, reaction uniformity, and colorimetric signal visibility. This width was finalized based on extensive testing as it offered the best compromise between material economy, capillary performance, and reagent spread.

The spacing between adjacent strips was optimized to balance two critical requirements: compactness of the overall device and prevention of reagent cross-talk. Strips placed too closely risked capillary flow bridging between adjacent strips, leading to false positives or mixed color responses. Conversely, excessive spacing increased the device footprint and disrupted uniform sample distribution. Spacing trials established a minimum safe distance that prevents lateral diffusion of reagents while preserving a compact form factor. This ensures that each test lane operates independently and provides reliable, isolated colorimetric outputs.

The compact and lightweight design enhances portability, making the device ideal for field use in areas lacking laboratory infrastructure. Minimal setup requirements reduce costs and complexity compared to traditional methods. Affordable manufacturing and long shelf-life further improve accessibility for diverse applications. Each test strip is individually treated with specific reagents tailored to detect common milk adulterants such as hydrogen peroxide, boric acid, urea, detergent, starch, and neutralizers. Upon exposure to adulterated milk, the reagent-treated strips undergo color changes at designated reaction zones, providing clear visual feedback for rapid and reliable adulteration detection.

Liquid flows through the porous paper substrates via natural capillary flow, requiring no external power source and ensuring minimal manufacturing and operational costs. The device demonstrated robust reliability with a detection limit ranging from 0.02% to 0.5%, suitable for diverse detection scenarios. Overall, the design offers a versatile and cost-effective solution for milk adulteration detection, exhibiting significant advantages over traditional methods in terms of simplicity, reliability, and affordability.

4 Adulterants identification

Globally, milk quality is evaluated using organoleptic,41 chemical,42 physical,43 and microbiological44 criteria. Chemical tests are pivotal for detecting milk adulteration, ensuring final products meet international standards. Various adulterants can be identified using a paper-based microfluidic system via colorimetric techniques. Chemical assessment serves multiple purposes: acceptance or rejection decisions, farm-level hygiene monitoring, classification for payment and usage, and selection for culture preparation or specialized product manufacture.

For sample preparation, various adulterants were individually added to milk at known quantities. Stock solutions were prepared by dissolving solid adulterants (urea, starch, boric acid, sodium hydroxide) in distilled water to make 2% (w/v) solutions, while hydrogen peroxide (30% purity) was used directly. These solutions were volumetrically added to milk in concentrations ranging from 0.2% to 1.0% (v/v), simulating realistic contamination scenarios.

Milk samples were equilibrated at 26–28 °C before testing. Approximately 6 mL of milk was dispensed into the sample reservoir using a syringe or pipette, and the inclined geometry of the strip holder ensured even flow across all test zones.

Adulterants were detected using specific colorimetric reagents. Urea produced a yellow color with DMAB and TCA [Fig. 5(a)]; hydrogen peroxide yielded a black color via iodide-starch chemistry [Fig. 5(b)]; starch formed a blue-black complex with iodine [Fig. 5(c)]; detergent formed a blue complex with chloroform [Fig. 5(d)]; neutralizers caused a red hue with phenolphthalein [Fig. 5(e)]. Curcumin responded to boric acid with an orange shift, and detergents were visualized using methylene blue in the presence of chloroform. To ensure fast and visually distinct color development on paper, reagents were freshly prepared at working concentrations suitable for immobilization and rapid chromogenic reaction. Each reagent solution was spotted onto predefined zones of Whatman paper using a micropipette, independent of the milk volume (6 mL) introduced from the inlet reservoir. The strips were then air-dried at room temperature (25 °C) for 30 minutes in the absence of direct light. Chloroform-based reagents were dried for 10–12 minutes. After drying, the strips were sealed in foil-laminated zip-lock pouches with silica desiccant and stored at room temperature. The reagents remained visually responsive for up to three weeks. Full formulations and spotting protocols are detailed in Table S1 (SI).


image file: d5sd00090d-f5.tif
Fig. 5 Colorimetric responses of paper strips to different adulterants. (a) Urea; (b) hydrogen peroxide; (c) starch; (d) detergent; (e) neutralizer; (f) boric acid.

These reactions [Fig. 6] are rapid, visually distinct, and occur within seconds under ambient conditions.


image file: d5sd00090d-f6.tif
Fig. 6 Reaction mechanisms for visual detection of milk adulterants, illustrating chromophore formation upon reagent–analyte interaction.

To ensure specificity, each strip was challenged with all six adulterants at 1.0%. No colorimetric response was observed in non-target zones, confirming absence of cross-reactivity. Reproducibility tests further validated the platform's reliability for field deployment.

To ensure real-world applicability, all experiments were conducted using commercially available packaged milk commonly sold in India, including both cow and buffalo milk variants. To assess whether milk composition affects test accuracy, we performed recovery tests on four types of milk with varying fat contents: skimmed (0.1%), toned (1.5%), full cream (3.5%), and buffalo milk (6.5%). Each was spiked with known amounts of adulterants, and the platform's detection performance was assessed. Recovery values ranged from 91.8% to 107.4%, confirming that colorimetric detection remains robust and accurate across typical variations in fat and protein content.

4.1 Recovery analysis in toned and full cream milk

To validate analytical performance across typical milk matrices, spike-and-recovery experiments were performed using commercial toned and full cream milk. Four key adulterants (urea, starch, hydrogen peroxide, and detergent) were spiked into milk at known concentrations and tested using our colorimetric platform. The resulting color responses were compared across milk types, and no significant loss in color intensity or reaction delay was observed. Approximate recovery values, based on visual consistency and colorimetric response grading, ranged from 95.0% to 105.0%, confirming high reliability and minimal matrix interference. These values are summarized in Table S4 (SI) and are in agreement with previously reported trends for paper-based diagnostic devices.45–47

5 Incurred cost

Milk adulteration is a pervasive issue affecting milk quality and consumer health. To address this, a cost-effective detection device has been designed to enhance milk safety. A techno-economic analysis was performed to evaluate the feasibility of the device. The cost structure includes a single-use paper strip, which is replaced after each test and discarded according to biosafety protocols, while the holder is a one-time investment considered a fixed asset.

The cost analysis is segmented into two components: the consumable paper strip and the reusable strip holder. As summarized in Table 3, the paper strip loaded with reagents costs approximately 12 INR (0.15 USD). The 3D-printed holder, fabricated from PLA filament, costs about 8.5 INR (0.10 USD). The detection kit also includes a reusable dropper for sample application, priced at around 0.5 INR (0.006 USD) when procured in bulk. Thus, the total cost of the complete kit is 21 INR (0.26 USD), making it among the most affordable platforms available for simultaneous detection of multiple adulterants.

Table 3 Manufacturing cost analysis of the capillary-based device
Component Quantity Cost (INR)
3D-printed strip holder 1 8.5 (0.10 USD)
Whatman nitrocellulose filter paper 1 sheet 10 (0.12 USD)
Reagents 2 (0.024 USD)
Dropper 1 0.5 (0.006 USD)
Total cost 21 (0.26 USD)


The market potential for this device is considerable. With an estimated initial penetration of 10[thin space (1/6-em)]000 units, and a projected selling price of 50 INR (0.59 USD) per unit, the total revenue would be approximately 500[thin space (1/6-em)]000 INR (5900.88 USD). Although the gross profit per unit is about 29 INR (0.34 USD), approximately 70% of the surplus is allocated to operational expenses such as advertising, distribution, packaging, and sales. After accounting for these costs, a net profit margin of 30% is retained, resulting in an estimated net profit of 87[thin space (1/6-em)]000 INR (1026.75 USD) for the initial batch.

Beyond financial metrics, the device delivers substantial nonmonetary benefits: it improves public health by facilitating the detection of harmful adulterants, enhances consumer trust in dairy products, and supports regulatory compliance and enforcement.

In summary, the milk adulteration detection device demonstrates strong economic viability with a substantial profit margin. Its production cost is justified by the potential revenue and societal benefits. The device offers an accessible and reliable means of detecting adulterants, thereby protecting vulnerable populations, including children and the elderly, who are particularly susceptible to milk-borne health risks. Moreover, the device promotes transparency and accountability within the dairy sector, fostering consumer confidence.

The broader societal benefits include reduced healthcare costs associated with contaminated milk consumption. The use of biodegradable PLA filament for 3D printing aligns with sustainable practices, addressing the growing demand for eco-friendly solutions. Collectively, these factors underscore the device's contribution to advancing food safety standards, public health, and environmental sustainability.

6 Results and discussion

Upon sample introduction, gravity-assisted capillary flow ensures consistent transport of milk along the test strips, delivering the sample to reagent-loaded detection zones without requiring external pumps. The reagents undergo specific colorimetric reactions with their target adulterants, producing time-evolving intensity changes that form the basis for detection.

To investigate the dynamic colorimetric behavior of the proposed detection system, grayscale intensity at the reaction zones was recorded as a function of time for six different adulterants, each tested across five concentrations (0.2% to 1.0%). These intensity–time plots capture the full transient response of the colorimetric reaction, from initial contact with the reagent to the final saturation stage. The rise in intensity corresponds to the generation of chromophoric species as the adulterant reacts with its respective reagent. Initially, this reaction is rapid due to high reactant availability, leading to a steep increase in color intensity. As the reaction proceeds, the system approaches chemical equilibrium and the rate of chromophore formation slows down, eventually leading to saturation.

Fig. 7 presents time-resolved grayscale intensity profiles for each adulterant across five concentrations (0.2% to 1.0%). These curves capture the transient dynamics of chromophore formation following reagent–analyte interaction in the detection zones. The initial steepness of the curves reflects the rate of color development, while the saturation plateau corresponds to the maximum chromophore concentration achieved under the given conditions.


image file: d5sd00090d-f7.tif
Fig. 7 Variation in color intensity over time for different concentrations of six common milk adulterants. (a) Boric acid, (b) detergent, (c) hydrogen peroxide, (d) neutralizer, (e) starch, (f) urea.

Urea and detergent exhibit the fastest responses, with higher concentrations reaching saturation within 30 seconds. This rapid color development is attributed to the strong reactivity of the DMAB–TCA and methylene blue–chloroform systems, respectively. Neutralizer also demonstrates prompt signal formation, stabilizing around 40–50 seconds due to its acid–base interaction with phenolphthalein. Hydrogen peroxide shows a sustained rise in intensity, particularly at lower concentrations, consistent with its gradual redox-based chromophore formation.

In contrast, starch and boric acid produce slower responses. Starch displays a steady increase in intensity, reflecting moderate interaction with iodine–potassium iodide, while boric acid shows delayed and relatively weak chromophore generation. This is consistent with the slow formation of the boric acid–curcumin complex, which exhibits both lower signal amplitude and delayed saturation. These differences in temporal response provide a foundation for kinetic modeling in subsequent analysis.

Together, these plots demonstrate that the rate of color development varies significantly across adulterants and concentrations. This variation is crucial for both qualitative and semi-quantitative analysis, as faster kinetics correspond to quicker detection times while well-separated intensity curves improve discrimination between concentrations.

6.1 Estimation of time constant

To quantitatively analyze the rate at which each colorimetric reaction proceeds, a pseudo-first-order kinetic model was applied to the intensity–time data. This was achieved by transforming the raw intensity values y(t) using the relation:
image file: d5sd00090d-t1.tif
Here, y represents the saturation intensity for a given concentration, and τ is the kinetic time constant. This transformation linearizes the reaction curve, allowing the extraction of τ from the slope of the resulting plot, which equals −1/τ. A smaller τ indicates a faster reaction, while a larger τ signifies slower kinetics.

Urea and detergent display the steepest slopes, reflecting rapid reaction kinetics and short τ values in the range of 13–20 s. This is consistent with the quick saturation observed in the raw intensity plots (Fig. 8). Neutralizer also shows relatively steep slopes, with moderate τ values of 30–50 s. In contrast, boric acid and starch exhibit flatter lines and longer time constants, consistent with slower chromophore formation. The closer slopes in boric acid suggest weak concentration dependence and limited kinetic resolution between concentrations.


image file: d5sd00090d-f8.tif
Fig. 8 Linearized kinetic plots of ln(1 − y/y) versus time for various concentrations (0.2–1.0%) of different adulterants: (a) boric acid, (b) detergent, (c) hydrogen peroxide, (d) neutralizer, (e) starch, and (f) urea.

The linearity observed across concentrations supports the pseudo-first-order assumption for most systems. Slight deviations at low concentrations for hydrogen peroxide and detergent are attributed to delayed onset of reaction or lower initial reaction rates. This kinetic modeling complements the earlier intensity–time analysis by shifting focus from equilibrium endpoints to transient dynamics. While intensity plots describe overall chromophore accumulation, the log-transformed representation isolates reaction speed—helping identify the fastest-responding adulterants and guiding time optimization for field deployment. From the slopes of these lines, the extracted τ values are plotted against concentration in Fig. 9, enabling direct comparison of kinetic speed across all analytes.


image file: d5sd00090d-f9.tif
Fig. 9 Kinetic time constant (τ) and saturation intensity (y) for six adulterants across different concentrations (0.2–1.0%).

6.2 Kinetics-based comparison across adulterants

To further understand the kinetic behavior of each adulterant, the extracted time constants (τ) and saturation intensities (y) were plotted as a function of concentration, as shown in Fig. 9. These plots reveal two key trends for τ: (1) for most adulterants, τ decreases monotonically with increasing concentration, and (2) at sufficiently low concentrations, τ values tend to plateau, indicating a kinetic detection limit.

The monotonic decrease in τ reflects faster reaction kinetics at higher concentrations, due to greater reactant availability. For instance, urea and detergent show significant drops in τ from approximately 48 s to 13–17 s between 0.2% and 1.0%. Neutralizer and starch display a more gradual decline, while boric acid maintains moderate τ values across all concentrations.

Interestingly, for some adulterants such as hydrogen peroxide, the τ values begin to converge between 0.2% and 0.4%, suggesting a limit beyond which kinetic differentiation is not easily discernible. This convergence aligns with the experimentally determined LODs, reinforcing that kinetic analysis can serve as an auxiliary metric for confirming detection thresholds.

In addition to kinetic behavior, the saturation intensity (y) trends provide complementary information about the extent of the reaction. For all adulterants, y increases with concentration, consistent with enhanced signal generation at higher analyte levels. Notably, the rate of increase in y varies across different adulterants, reflecting differences in their reaction efficiency and signal saturation characteristics. Thus, y serves as a valuable steady-state metric that, when combined with τ, offers a more comprehensive understanding of adulterant detection performance.

From a practical point of view, urea and detergent emerge as the most rapidly detected adulterants, with low τ values even at minimal concentrations, while also achieving high saturation intensities. In contrast, boric acid and starch exhibit slower detection kinetics and comparatively lower y values, requiring longer reaction times and higher concentrations for reliable quantification.

6.3 Chemical basis of color formation

At the molecular level, when an adulterant is present in milk, it reacts with the reagent to form chromophores—molecules that absorb light in the visible range and impart distinct colors. The nature of chromophore formation depends on the specific reaction chemistry: redox processes (e.g., hydrogen peroxide), acid–base indicators (e.g., neutralizers), or complex formation (e.g., boric acid with curcumin). The more chromophores produced, the stronger the color response.

This formation process is typically concentration-dependent. Higher concentrations of adulterants result in more reagent–analyte interactions, producing deeper colors due to greater chromophore density. Conversely, lower concentrations yield fewer chromophores and hence lighter color intensities.

This trend aligns with Beer–Lambert's law:

A = ε·c·l
where A is absorbance, ε is molar absorptivity, c is analyte concentration, and l is the optical path length. Although this relationship is classically used in spectrophotometry, a similar proportionality holds between grayscale intensity and concentration in image-based systems.

The time-dependent behavior of color formation can be understood through classical chemical kinetics. For instance:

• Zero-order kinetics: A = A0kt

• First-order kinetics: A = A0·ekt
These simplified forms explain whether reactions are limited by reagent depletion, concentration, or diffusion in the porous substrate.

The time dependence of color formation follows pseudo-first-order kinetics, where the rate of chromophore accumulation decreases exponentially as reagent is consumed. This supports the use of log-transformed kinetic models (ln(1 − y/y)) shown in Fig. 8, where the slope reflects the inverse kinetic time constant (τ). To support the choice of the pseudo-first-order kinetic model, comparative fits using zero-order and diffusion-controlled models for representative adulterants (boric acid, detergent, and urea) are included in the SI (Fig. S1).

Once the saturation color was reached, the intensity remained stable for at least 60 minutes, with less than 2% variation across replicates. This robustness indicates that immediate image capture is not required, supporting delayed analysis in field settings.

6.4 Quantitative detection and calibration curves

Following the evaluation of time-dependent color intensity behavior, this section focuses on establishing a quantitative correlation between grayscale intensity and known adulterant concentrations. For each adulterant, calibration curves were generated by measuring grayscale intensity at saturation. Five concentrations (0.2%, 0.4%, 0.6%, 0.8%, and 1.0% v/v) were tested in quintuplicate. To enable uniform interpretation, all g L−1 concentrations for solid adulterants were converted to v/v as described in the adulterants identification section.

The mean values were plotted against concentration, and standard deviations were shown as error bars. Fig. 10 presents the resulting calibration curves. Each adulterant demonstrates a strong linear relationship between concentration and grayscale intensity, with coefficients of determination (R2) exceeding 0.97, confirming the analytical reliability of the device.


image file: d5sd00090d-f10.tif
Fig. 10 Calibration curves for six milk adulterants. Normalized grayscale intensity vs. concentration with error bars representing standard deviation (n = 5). Linear regression lines and R2 values are annotated. The subfigures correspond to different adulterants: (a) boric acid; (b) starch; (c) hydrogen peroxide; (d) detergent; (e) neutralizer; (f) urea.

Table 4 summarizes the regression parameters obtained. Neutralizer exhibited the highest slope, indicating the greatest detection sensitivity, while starch showed relatively lower sensitivity due to weaker chromophore development.

Table 4 Summary of regression parameters for the calibration curves of all six adulterants
Adulterant Regression equation Slope R2
Urea I = 62.28x + 37.87 62.28 0.99946
Starch I = 50.64x + 148.87 50.64 0.98000
Hydrogen peroxide I = 51.71x + 194.87 51.71 0.98300
Detergent I = 70.06x + 34.40 70.06 0.97300
Neutralizer I = 102.38x + 115.19 102.38 0.99900
Boric acid I = 35.00x + 80.40 35.00 0.9976


Table 5 Comparison of limit of detection (LOD) values for various adulterants in the present work and literature
Adulterant LOD (literature) LOD (present work) LOD (g L−1)
Hydrogen peroxide 0.1% (ref. 48) 0.03% 0.099
Neutralizer 0.13% 0.017
Starch 0.1% (ref. 48) 0.10% 0.02
Detergent 0.2% (ref. 48) 0.10% 0.02
Boric acid 0.20% 0.04
Urea 0.05%,48 0.07% (ref. 49) 0.03% 0.06


While linear calibration curves establish the device's semi-quantitative detection capability, the limit of detection (LOD)—the lowest concentration producing a reliable response—was also determined for each adulterant and compared with literature reports.

Taken together with the calibration slopes, these findings confirm that τ not only reflects the temporal resolution of the detection system but also complements equilibrium-based intensity measurements, providing mechanistically relevant validation of detection performance.

Based on the extracted τ values, the fastest-responding adulterants were urea and detergent, achieving saturation within 15–20 seconds at higher concentrations. Hydrogen peroxide demonstrated rapid but slightly slower kinetics, while boric acid and starch consistently exhibited higher τ values. Neutralizer showed intermediate kinetics with moderate τ values. This aligns with the temporal behavior observed in intensity and kinetic plots.

Neutralizer exhibits the highest calibration slope (102.38), indicating the most sensitive intensity change per unit concentration. Detergent and urea follow with steep slopes (70.06 and 62.28, respectively), reflecting rapid kinetics and strong color development. Hydrogen peroxide and starch show moderate slopes, while boric acid has the lowest sensitivity.

6.5 Limit of detection

The experimental limit of detection (LOD) was determined for each adulterant based on the minimum concentration at which a perceptible and statistically significant color change was observed across multiple replicates. A series of milk samples were prepared with increasing concentrations, starting from 0.01%. The LOD was defined as the lowest concentration where a reproducible color change appeared.

For urea and hydrogen peroxide, no response was observed below 0.03%, establishing the LOD at this concentration. Given the health risks associated with trace levels of adulterants, these LOD values are particularly significant (Table 5).

Compared to previous reports, the present device demonstrates lower LOD values for several adulterants, particularly hydrogen peroxide and urea, due to optimized geometry, reagent retention, and enhanced optical contrast.

To benchmark the performance and practicality of our device, we compared it with two previously reported paper-based platforms: a wax valve-based μPAD developed by Chen et al.31 and a multiplex sugar detection platform by Aksorn et al.35 A detailed comparison is presented in Table 6.

Table 6 Comparison of the present paper-based microfluidic device with selected μPAD platforms reported by Chen et al. (2019)31 and Aksorn et al. (2020)35
Parameter Present work Chen et al. (2019)31 Aksorn et al. (2020)35
Target analytes 6 adulterants (urea, starch, boric acid, NaOH, peroxide, detergent) Glucose, potassium iodate Urea, starch, boric acid, detergent, formalin
Multiplexing capability Yes (6 in parallel) No Yes (3 analytes)
Detection method Color intensity (visual and RGB analysis) Distance-based color front RGB-based smartphone quantification
LOD range 0.2–0.3% (w/v) 0.05–0.5 mM 0.3–0.5% (w/v)
Time to result <2 minutes 5–8 minutes 3–6 minutes
Sample volume 6 mL 40 μL 2 mL
Reagent volume (per test) 8–12 μL ∼3–5 μL 5–10 μL
Fabrication Manual spotting + 3D holder Wax printing with valve pattern Inkjet-printed assay pads
User equipment None (visual) or mobile camera (optional) Ruler or mobile phone Mobile app
Shelf life/storage 3 weeks at 4–8 °C in foil pouch 2–3 weeks with desiccant Not reported
Estimated cost/test INR 3.5–4.0 (USD ∼0.05) USD 0.10–0.15 USD 0.12–0.15
Meets WHO ASSURED? Yes Partial Partial


6.6 Sensitivity analysis based on kinetic response

To validate the LODs, sensitivity analysis was performed by evaluating the sensor response at 0.01% and 0.02%, below the established LODs. Fig. 11 presents the time-resolved intensity profiles for these sub-LOD concentrations.
image file: d5sd00090d-f11.tif
Fig. 11 Time-resolved intensity profiles for milk samples adulterated with six common substances at sub-LOD concentrations of 0.01% and 0.02% with error bars indicating standard deviation. (a) Boric acid, (b) detergent, (c) hydrogen peroxide, (d) neutralizer, (e) starch, (f) urea.

The lack of distinguishable intensity evolution at these concentrations confirms that the device exhibits no false-positive behavior. The kinetic approach—monitoring signal evolution overtime—adds an additional layer of validation to the robustness of the detection platform.

6.7 Recovery tests in commercial milk samples

To ensure real-world applicability and matrix tolerance, we validated the performance of the device using commercially available milk samples with different fat contents—specifically, toned milk and full cream milk. Each sample was spiked with known concentrations of four representative adulterants (urea, starch, hydrogen peroxide, and detergent), and recovery was calculated based on the ratio of detected to spiked concentration.

The results (Table S2 in the SI) show that the recoveries across both milk types ranged between 95.0% and 105.0%, confirming high accuracy and minimal matrix interference. These findings further validate the robustness and reliability of the device across real-world milk compositions.

Conclusions

This study introduces a low-cost, field-deployable paper-based microfluidic platform capable of detecting multiple milk adulterants with high sensitivity and rapid response. The device operates on gravity-assisted capillary flow without requiring hydrophobic patterning or external power sources. It supports simultaneous detection of six common adulterants—urea, starch, hydrogen peroxide, detergent, neutralizers, and boric acid.

Kinetic and calibration-based analyses confirmed the device's quantitative performance, with reaction speed modeled using pseudo-first-order kinetics and saturation intensities yielding R2 values exceeding 0.97. Experimentally determined LODs for urea and hydrogen peroxide were as low as 0.03%, outperforming existing methods.

The device fabrication—including laser-cut Whatman grade 4 paper and optimized strip dimensions—ensures reproducible fluid transport and chromophore formation. Its inclined channel configuration stabilizes flow and color development even under field conditions. In alignment with WHO's ASSURED framework for point-of-care diagnostics, the developed device is affordable (cost <$0.05 per test), user-friendly, and equipment-free, requiring no skilled operation or electronic reader. It demonstrates high sensitivity and specificity for multiple milk adulterants, with clear visual outputs emerging within 30–60 seconds. These responses remained stable across milk types and environmental conditions, confirming its robustness. The strips are compact, foil-sealed, and deliverable in field settings, with a storage life of 2–3 weeks. A point-wise summary of the ASSURED criteria and how our device fulfills them is provided in Table S3 (SI).

Overall, the proposed device offers a low-cost, rapid, and field-deployable strategy for simultaneous detection of multiple milk adulterants, with strong alignment to WHO's ASSURED criteria. While the platform demonstrates robust performance, its current implementation has a few practical constraints. The shelf life of pre-loaded strips is approximately 2–3 weeks, and reagent response may vary slightly under prolonged exposure to extreme temperatures. Additionally, manual reagent spotting—although reproducible under lab conditions—could benefit from automation in future scaled-up fabrication.

Intellectual property disclaimer

The design and configuration of the device described in this work are formally protected under Indian Patent No. 541444, titled “An easy method to detect milk adulteration”, filed by the authors at IIT Kharagpur. The present manuscript reports the scientific validation and performance evaluation of the patented device. Publication of this work does not conflict with the patent, and the authors retain the right to disseminate the research findings.

Conflicts of interest

There are no conflicts of interest to declare.

Data availability

Supplementary information is available: The SI file includes Table S1: Reagent formulations, concentrations, and application protocols for colorimetric detection of adulterants. Fig. S1: Comparative kinetic fitting of colorimetric data using different models (zero-order, pseudo-first-order, and diffusion-controlled). Table S2: Recovery performance of the paper-based device across commercial milk types spiked with adulterants. Table S3: Summary of how the device meets the WHO ASSURED criteria for point-of-care diagnostics. See DOI: https://doi.org/10.1039/D5SD00090D.

All data supporting the findings of this study are available within the article and its SI file.

References

  1. National Dairy Development Board, Annual Report 2023–24, 2024, https://www.nddb.coop/sites/default/files/pdfs/NDDB_AR_2023_24_Eng.pdf, Accessed: 2025-05-24 Search PubMed.
  2. USDA Foreign Agricultural Service, Dairy and Products Annual – 2017, 2017, https://apps.fas.usda.gov/newgainapi/api/report/downloadreportbyfilename?filename=Dairy+and+Products+Annual_New+Delhi_India_10-13-2017.pdf, Accessed: 2025-05-24 Search PubMed.
  3. Food and Agriculture Organization, Agro-industries Characterization and Appraisal: Dairy in India, 2007, https://www.fao.org/4/ap299e/ap299e.pdf, Accessed: 2025-05-24 Search PubMed.
  4. Food Safety and Standards Authority of India, National Milk Safety and Quality Survey 2018, 2019, https://fssai.gov.in/upload/uploadfiles/files/Report_Milk_Survey_NMQS_Final_18_10_2019.pdf, Accessed: 2025-05-24 Search PubMed.
  5. P. Singh and N. Gandhi, Food Rev. Int., 2015, 31, 236–261 CrossRef CAS.
  6. M. K. Sukumaran and H. Singuluri, Indian J. Dairy Sci., 2015, 68(2), 190–192 Search PubMed.
  7. S. Kandpal, A. Srivastava and K. Negi, Indian J. Community Health, 2012, 24, 188–192 Search PubMed.
  8. K. Sharma and M. Paradakar, Food Secur., 2010, 2, 97–107 CrossRef.
  9. K. Girma, Z. Tilahun and D. Haimanot, Review on milk safety with emphasis on its public health, World J. Dairy Food Sci., 2014, 9(2), 166–183 Search PubMed.
  10. T. K. Thorning, A. Raben, T. Tholstrup, S. S. Soedamah-Muthu, I. Givens and A. Astrup, Food Nutr. Res., 2016, 60, 32527 CrossRef.
  11. Y. T. Hsu and C. H. Kao, Plant Soil, 2007, 298, 231–241 CrossRef CAS.
  12. S. Das, B. Goswami and K. Biswas, Sens. Lett., 2016, 14, 4–18 CrossRef.
  13. D. M. Luykx, J. H. Cordewener, P. Ferranti, R. Frankhuizen, M. G. Bremer, H. Hooijerink and A. H. America, J. Chromatogr. A, 2007, 1164, 189–197 CrossRef CAS.
  14. J. Yang, N. Zheng, H. Soyeurt, Y. Yang and J. Wang, Food Sci. Nutr., 2019, 7, 56–64 CrossRef CAS PubMed.
  15. J. H. Cordewener, D. M. Luykx, R. Frankhuizen, M. G. Bremer, H. Hooijerink and A. H. America, J. Sep. Sci., 2009, 32, 1216–1223 CrossRef CAS PubMed.
  16. X. Dai, X. Fang, F. Su, M. Yang, H. Li, J. Zhou and R. Xu, J. Chromatogr., B, 2010, 878, 1634–1638 CrossRef CAS.
  17. P. Scano, A. Murgia, F. M. Pirisi and P. Caboni, J. Dairy Sci., 2014, 97, 6057–6066 CrossRef CAS.
  18. A. S. Ivanova, A. D. Merkuleva, S. V. Andreev and K. A. Sakharov, Food Chem., 2019, 283, 431–436 CrossRef CAS.
  19. R.-K. Chen, L.-W. Chang, Y.-Y. Chung, M.-H. Lee and Y.-C. Ling, Rapid Commun. Mass Spectrom., 2004, 18, 1167–1171 CrossRef CAS PubMed.
  20. L. Sanchez, M. D. Perez, P. Puyol, M. Calvo, G. Brett and G. E. Vegarud, Determination of vegetal proteins in milk powder by enzyme-linked immunosorbent assay: Interlaboratory study, J. AOAC Int., 2002, 85, 1390–1397 CrossRef CAS.
  21. L. Asensio, I. Gonzalez, T. Garcia and R. Martin, Food Control, 2008, 19, 1–8 CrossRef CAS.
  22. H. Benli and E. Barutcu, Anim. Biosci., 2021, 34, 1995 CrossRef CAS PubMed.
  23. A. F. S. Silva and F. R. Rocha, Food Control, 2020, 115, 107299 CrossRef CAS.
  24. T. B. Coitinho, L. D. Cassoli, P. H. R. Cerqueira, H. K. da Silva, J. B. Coitinho and P. F. Machado, J. Food Sci. Technol., 2017, 54, 2394–2402 CrossRef CAS.
  25. Z. Wang, T. Li, W. Yu, L. Qiao, S. Yang and A. Chen, LWT--Food Sci. Technol., 2020, 122, 109038 CrossRef CAS.
  26. C. Lu, B. Xiang, G. Hao, J. Xu, Z. Wang and C. Chen, J. Near Infrared Spectrosc., 2009, 17, 59–67 CrossRef CAS.
  27. C. Das, B. N. Chowdhury, S. Chakraborty, S. Sikdar, R. Saha, A. Mukherjee, A. Karmakar and S. Chattopadhyay, LWT--Food Sci. Technol., 2021, 136, 110347 CrossRef CAS.
  28. M. K. Nieuwoudt, S. E. Holroyd, C. M. McGoverin, M. C. Simpson and D. E. Williams, Appl. Spectrosc., 2017, 71, 308–312 CrossRef CAS PubMed.
  29. B. He, R. Liu, R. Yang and K. Xu, Optical Diagnostics and Sensing X: Toward Point-of-Care Diagnostics, 2010, pp. 160–168 Search PubMed.
  30. Anushka, A. Bandopadhyay and P. K. Das, Eur. Phys. J.: Spec. Top., 2023, 232, 781–815 CAS.
  31. C. Chen, L. Zhao, H. Zhang, X. Shen, Y. Zhu and H. Chen, Anal. Chem., 2019, 91, 5169–5175 CrossRef CAS PubMed.
  32. Y. Fan, H. Wang, S. Liu, B. Zhang and Y. Zhang, J. Food Saf., 2018, 38, e12548 CrossRef.
  33. J. L. Luther, V. H. de Frahan and M. Lieberman, Anal. Methods, 2017, 9, 5674–5683 RSC.
  34. P. Rewatkar and S. Goel, ECS J. Solid State Sci. Technol., 2020, 9, 115025 CrossRef CAS.
  35. J. Aksorn and S. Teepoo, Talanta, 2020, 207, 120302 CrossRef CAS PubMed.
  36. L. S. Lima, E. L. Rossini, L. Pezza and H. R. Pezza, Spectrochim. Acta, Part A, 2020, 227, 117774 CrossRef CAS.
  37. A. K. Govindarajalu, M. Ponnuchamy, B. Sivasamy, M. V. Prabhu and A. Kapoor, Bull. Mater. Sci., 2019, 42, 1–6 CrossRef CAS.
  38. B. G. Guinati, L. R. Sousa, K. A. Oliveira and W. K. Coltro, Anal. Methods, 2021, 13, 5383–5390 RSC.
  39. T. Lam, J. P. Devadhasan, R. Howse and J. Kim, Sci. Rep., 2017, 7, 1188 CrossRef.
  40. H. J. Chun, Y. M. Park, Y. D. Han, Y. H. Jang and H. C. Yoon, BioChip J., 2014, 8(3), 218–226 CrossRef CAS.
  41. S. Clark, M. Costello, M. Drake and F. Bodyfelt, The Sensory Evaluation of Dairy Products, Springer, 2009, vol. 571 Search PubMed.
  42. S. Sultana, A. Hosen, A. Afsana, E. Pehan, A. Islam, N. Sultana and M. Hassan, J. Food Qual. Hazards Control, 2024, 11, 214–222 CAS.
  43. P. S. Deshmukh, S. D. Deshmukh and S. A. Deshmukh, Indian J. Dairy Sci., 2021, 74, 374–378 Search PubMed.
  44. J. B. Ndahetuye, K. Artursson, R. Bage, A. Ingabire, C. Karege, J. Djangwani, A.-K. Nyman, M. P. Ongol, M. Tukei and Y. Persson, J. Dairy Sci., 2020, 103, 9730–9739 CrossRef CAS.
  45. S. Patari and P. Mahapatra, Sci. Rep., 2022, 12, 13465 CrossRef PubMed.
  46. S. Patari and P. Mahapatra, Food Chem., 2023, 421, 136122 Search PubMed.
  47. M. Ahsan, R. Singh and A. Sharma, Sens. Imaging, 2025, 32, 1–12 Search PubMed.
  48. S. Patari, P. Datta and P. S. Mahapatra, Sci. Rep., 2022, 12, 13657 CrossRef CAS.
  49. K. M. Khan, H. Krishna, S. K. Majumder and P. K. Gupta, Food Anal. Methods, 2015, 8, 93–102 CrossRef.

This journal is © The Royal Society of Chemistry 2025
Click here to see how this site uses Cookies. View our privacy policy here.