Nondestructive Classification of Yellow and Orange Colored Yolk Eggs using Hyperspectral Imaging Combined with PLS-DA

Nondestructive Classification of Yellow and Orange Colored Yolk Eggs using Hyperspectral Imaging Combined with PLS-DA | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Nondestructive Classification of Yellow and Orange Colored Yolk Eggs using Hyperspectral Imaging Combined with PLS-DA Alin Khaliduzzaman, Jason Lee Emmert, Mohammad Kamruzzaman This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6207045/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Nondestructive yolk color detection holds significant potential in the egg and poultry industries due to its critical role in shaping consumer preferences, nutritional perceptions, and marketability. A consistent yolk color may support product branding, quality assurance, and adherence to market-specific standards. Thus, this research study aimed to develop an nondestructive approach using hyperspectral imaging (HSI) combined with partial least squares discriminant analysis (PLS-DA) to separate orange yolk-colored eggs from yellow yolk-colored eggs. A hyperspectral camera in the visible range of 400 nm to 1000 nm was used for the spectral information of the eggs. A total of 146 white eggshell infertile eggs were collected from the poultry farm of the University of Illinois at Urbana-Champaign and were used for the investigation. The total dataset was divided into 70% for training and 30% for testing purposes. A classification model was developed using PLS-DA with various spectral preprocessing techniques. An accuracy of 100% on the testing set was achieved using spectral preprocessing with standard normal variate (SNV), first derivative (FD) and second derivative (SD) data. The key 9 variables (wavelengths) in the classification model were found in the range 520–680 nm, which indicated the variation in the types and amount of carotenoid pigments deposited in egg yolks, which are influenced by hen feed and metabolic efficiency. These findings suggested that HSI combined with multivariate analysis could be used to grade chicken eggs based on their internal yolk color for the future egg and poultry industry. Agricultural Engineering Animal Science Egg grading internal quality yolk color consumer preference EI 4.0 Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 1. Introduction In countries such as the United States, yellow yolk is maintained on poultry farms by controlling feed formulation and breed preferences to meet consumer preferences (Ortiz et al., 2021 ). However, in a preliminary experiment involving egg yolk, a few orange-colored yolks were found within the same lot and breed, possibly due to individual variations in feed intake, metabolic efficiency and thus yolk pigmentation. The lipids of the yolk are exclusively associated with lipoprotein aggregates composed of triglycerides, phospholipids, and cholesterol. Less than 1% of yolk lipids are carotenoids, which give the yolk its hue ranging from pale yellow to dark bright orange (Dansou et al., 2023 ). This yolk color is also influenced by the genotype and the rate of egg production in hens (Hanusova et al., 2015 ; Sokołowicz et al., 2018 ). There is some variation between breeds, strains, and individual hens in terms of their ability to absorb and deposit oxycarotenoids in egg yolk (H. Karunajeewa R. J. Hughes and Shenstone, 1984). Therefore, there is a need to develop nondestructive, fast, and precise methods for objectively measuring yolk color for the benefit of the egg and poultry industry as well as consumer markets. The functional properties (e.g., antioxidants) of carotenoids in egg yolk may also improve the quality and nutritional value of eggs. The egg industry is undergoing a significant transformation through the adoption of Industry 4.0 technologies, emphasizing automation and digitalization to enhance efficiency and product quality (Ahmed et al., 2023 ). Although automated egg grading systems utilizing advanced sensors and machine learning algorithms have been developed to assess external characteristics such as size, weight, and shell integrity, limited research has been conducted for evaluating internal egg qualities, specifically yolk attributes due to advanced technological challenges. Yolk attributes are important for both table and hatching eggs, as they influence the gender, hatchability and various quality parameters of day-old chicks (Valcu et al., 2020 ; Khaliduzzaman et al., 2021 ; Rahman et al., 2021 ). Traditional methods often require destructive sampling, limiting their applicability in large-scale operations. Although nondestructive techniques, such as hyperspectral imaging (HSI) and near-infrared (NIR) spectroscopy, have been recently reported to predict internal qualities, such as the yolk ratio and albumen freshness (Syduzzaman et al., 2019a ; Loffredi et al., 2021 ), there is no single study is conducted so far on yolk color based eggs classification using non-destructive means. Yolk color is a critical quality parameter that influences consumer preferences and perceptions of nutritional value. The risk is primarily determined by the hen's diet, particularly the intake of carotenoids, which impart hues ranging from pale yellow to deep orange. The major pigments responsible for yolk color are lutein, zeaxanthin, ß -cryptoxanthin and canthaxanthin (Ortiz et al., 2021 ; Mrowicka et al., 2022 ). The source of zeazanthins is a corn-based diet, and the type of corn influences the yolk color, where the diet from orange corn is responsible for more orange color pigmentation on the yolk (Ortiz et al., 2021 ). Developing protocol for nondestructive yolk color detection could revolutionize quality control system in both the table egg and hatching egg sectors. Such technologies might enable producers to meet consumer demands more effectively, ensuring consistency in product appearance without compromising the integrity of the eggs during inspection. Yolk coloration, ranging from pale yellow to deep orange, is closely linked to consumer preferences, nutritional quality, freshness, and premium value, thus may drive demand and enable higher pricing. In some other countries, such as Japan, the deep orange color of the yolk is often preferred for aesthetic reasons, especially when eating eggs raw in dishes such as "tamago-kake-gohan”. Furthermore, yolk pigmentation serves as a biomarker for hens' dietary carotenoid intake, offering insights into feed quality and management practices. Therefore, the objective of this study is to develop a noninvasive approach using hyperspectral imaging (HSI) and multivariate analysis techniques, such as PLS-DA, to distinguish orange yolk eggs from yellow yolk eggs. HSI integrates the advantages of spectroscopy and conventional imaging techniques in one system to provide both spectral and spatial information about an object. Hyperspectral imaging is a noninvasive and high-resolution method for capturing detailed spectral information from chicken eggs (Ahmed et al., 2025 ). Thus, an HSI in the visible range can be used to convey yolk and yolk color information when light passes through the egg reaches the camera sensor. PLS-DA is a supervised multivariate statistical method used for classification that involves maximizing the separation between predefined groups while explaining variance in predictor variables (Kamruzzaman et al., 2012 ). It works by projecting predictor variables onto a new space that correlates with the response variable, enabling clear group discrimination. Thus, HSI together with multivariate analysis could be a good combination to classify eggs based on yolk for the future egg industry. 2. Materials and methods 2.1 Materials The animal experiment protocol was followed and approved by the Office of the Vice-Chancellor for Research IACUC online protocols, University of Illinois (protocol #: 22224). A total of 146 White Leghorn eggs (117 yellow yolk eggs and 29 orange yolk eggs) were collected from the poultry farm of the University of Illinois at Urbana-Champaign; these eggs included both fertile and infertile white eggs and were used for internal yolk color-based egg classification. The total dataset was divided into 70% for training ( n = 102, 80 eggs for the yellow class and 22 for the orange yolk class) and 30% for testing (n = 44, 37 for the yellow class and 7 for the orange class) sets using random sate 42 (in python). The average size of the eggs was 59.47 g, and the major diameter and minor diameter were 57.04 and 43.10 mm, respectively. The variability of egg mass and size is an important criterion for the classification model and feasibility of the proposed technique when applied to actual cases and industrial settings. Two types of yolk color were found in all the egg samples, among which yellow yolk eggs were the major class (Fig. 1). The yolks were defined as yellow or orange based on their yolkfan scores (yellow: 5–6; orange: 11–13). In this study, approximately 20% of the orange yolk eggs were produced via mass production. 2.2 Hyperspectral imaging system A hyperspectral camera (Model: PIKA-L, line scanning) in the visible range of 400 nm to 100 nm was used for the spectral and spatial information of the eggs (Fig. 2). The PIKA-L hyperspectral camera works based on the line-scanning principle called the push-broom method, as it sweeps over the target line by line. A scanning speed of 0.06 cm/s with a frame rate of 9.9 fps and an exposure time of 100 ms were maintained during hyperspectral image acquisition. The output of an HSI system is a data-rich hyperspectral image, which is a three-dimensional data cube also known as voxel I (x, y, λ). The data cube can be interpreted as a collection of monochrome images I (x, y) for each wavelength λ or as a spectrum I (λ) for each pixel (x, y). The key components of a line scanning hyperspectral camera include a spectrograph with an input slit, a grayscale camera and an objective lens. A hyperspectral image of an object is obtained when a spectrograph is added to the camera system. In this case, the spectrograph consists of an input slit, collimating optics, a dispersive unit and a focusing lens. The input slit limits the incoming information from the object by limiting the input light, which allows only a single line. The collimating lens collates the light into dispersive light, and the dispersive unit converts the light into spectra. The focusing lens then focuses the dispersed light into the grayscale camera, which measures the intensity of light at different wavelengths. Thus, a hyperspectral camera collects spectral data for each pixel in a line across the target area simultaneously. The target moves (or the sensor moves) to capture the next line, gradually building up a full hyperspectral image as the lines are combined. 2.3 Spectral extraction from the HSI cube The HSI cube contains a large amount of information about the egg sample in spectral and spatial dimensions. Thus, a series of operations are required to extract meaningful information from the HSI cube to enhance interpretability and prepare it for further analysis to develop prediction models. The average spectrum of an egg was obtained from an image cube based on the ROIs. ROIs were selected by applying a threshold to pixel intensity values. Pixels below the threshold are set as the background, while others form the object. This technique is straightforward and computationally efficient. Eggs typically have clear boundaries, making thresholding-based segmentation suitable for accurate separation from the background. Hyperspectral images stored in band interleaved-by-line (BIL) format were read using the MATLAB code multibandread . The hyperspectral cube dimensions were 790 × 900 × 300, where 790 represents the height (rows), 900 represents the width (columns), and 300 represents the spectral bands. The DN (digital number) value corresponding to band 150 was used as a threshold to filter relevant pixels. Pixels with DNs > 600 were selected for spectral extraction. The sum of the spectral values from all the selected pixels was computed, and the total number of selected pixels was subsequently used to compute the average spectral response. The extracted spectra were stored in a declared variable, R , which was a matrix containing spectral information from all processed images. Finally, R was reshaped to maintain the correct dimensions [ n , 300], where n represents the number of images (egg sample) processed. 2.4 Data analysis 2.4.1 Principal component analysis Although principal component analysis (PCA) is used for dimensionality reduction of data while retaining valuable information (Wold et al., 1987 ; Gewers et al., 2021 ). PCA is also very helpful for observing the class differences in spectral data and feasibility of further classification models since PCA is an unsupervised dimensionality reduction technique (Islam and Khaliduzzaman, 2022 ). PCA does not use class labels or target variables; it only analyzes the input features (independent variables). It finds new axes (principal components) that maximize variance in the data, helping reduce dimensionality while preserving as much information as possible. Therefore, PCA can potentially be used for exploring data structure and visualization. PCA transforms data into a new coordinate system where the greatest variance lies along the first principal component, the second greatest variance along the second component, and so on. 2.4.2 Spectral preprocessing Preprocessing is a common prestep process in spectral data processing since it considerably improves the performance of multivariate models (Islam et al., 2016 ). Despite extensive studies on various preprocessing approaches, trial and error methods, together with performance, are often used to determine the optimal technique. In this study, four widely used preprocessing techniques, standard normal variate (SNV) and Savitzky–Golay (SG) first-order and second-order derivatives with window sizes of 15 and polynomial order of 3, were used for PLS–DA calibration model development. The SNV effectively eliminates nonuniform scattering and particle size effects from the spectra. The first derivative mitigates baseline shifts and standardizes absorption rates, enhancing the comparability of spectral peaks and troughs. Second derivatives convert local maxima peaks to minima below the baseline, flattening it. The second derivative of a spectrum enhances subtle spectral features by highlighting changes in curvature, making it a valuable tool in spectral analysis. This approach sharpens overlapping peaks by reducing the effect of broad baseline variations. Therefore, small variations in absorbance or reflectance due to chemical composition become more prominent. The SNV is used to standardize features by removing the mean and scaling to unit variance. Standardization is an essential preprocessing step in many machine learning models, especially those sensitive to feature scales, such as PLS. Model performance can be improved by ensuring that each feature contributes equally, regardless of its original scale. In SNV, each feature is centered by subtracting its mean and scaling to unit variance, dividing each feature by its standard deviation. The transformation results in a mean of 0 and a standard deviation of 1 (Grisanti et al., 2018 ). The SNV removes baseline offsets and intensity variations, making the spectral differences between the two classes more chemically relevant than just due to sample properties. The effectiveness of preprocessing methods can vary depending on the specific spectral data and analytical goals, necessitating careful selection and optimization to ensure consistent and accurate results. However, in some cases, spectral preprocessing may lead to potential information loss, the introduction of artifacts, and increased complexity. 2.4.3 PLS-DA model Partial least squares regression (PLS-R) is an important methodology for solving both regression and classification problems. PLS-DA is a supervised classification technique derived from partial least squares regression (PLS-R). It is particularly useful when dealing with high-dimensional, collinear data, such as spectral datasets (Kamruzzaman et al., 2012 ). Unlike traditional classification techniques, PLS-DA finds the latent variables that best separate predefined classes while maximizing the covariance between the predictor matrix ( X ) and the response variable ( Y ). PLS-DA follows a similar approach to PLS-R but modifies the response variable ( Y ) to represent class labels instead of continuous outcomes (Islam and Khaliduzzaman, 2022 ). Class 0 was used for the yellow yolk class, and 1 was used for the orange yolk class. The threshold value for class separation using PLS-DA was set to 0.5. Thus, PLS-DA extracts latent variables that maximize the separation between classes while explaining most of the variance in X . The latent variables serve as projections that enhance class separability. The response variable Y of PLS-DA is a set of binary variables that relates to the categories or classes of the sample. PLS-R or PLS-DA is expressed by Eq. 1. Y = Xb + E. … … … (Eq. 1) where X is an ( n x p ) matrix that holds the spectral values of each class, b = [ b 1 b 2 ... b p ] T is an ( p x1) dimensional column vector that contains the regression coefficients, T denotes the transpose of a vector or matrix, and E is the error term. After PLS projection, classification was performed based on the predicted values of Y . 2.4.4 Important variables selection Important variables were selected based on the peak position (peak value > 0.1) in the PC loading plot to identify important wavelengths for classifying orange yolks from yellow yolk eggs. These peak positions in the PC loading plot should correspond to important spectral wavelengths (var) containing class information, as clear separation was observed in the PCA (PC1 vs PC3 plot). Variables with high absolute loadings on a specific PC are considered the most important contributors to that component, meaning that they are strongly correlated with the direction of variation represented by that PC (Wold et al., 1987 ). 3. Results and discussion 3.1 Spectral characteristics of egg classes The spectral properties of yellow egg yolk and orange yolk showed clear and typical differences in the range of 520–600 and 640–680 nm, as shown in Fig. 3. The 520–600 nm wavelength ranges from green (~ 520 nm) to yellow (~ 570 nm) to orange (~ 600 nm), whereas the 640–680 nm wavelength ranges from deep orange to red, indicating the presence of carotenoids in the egg yolk. Carotenoids might absorb in the green to yellow region and reflect or emit yellow to orange colors depending on the type and concentration since yolk has a much greater absorbance coefficient below 600 nm (Syduzzaman et al., 2019b ). This variation might be due to differences in the color of the various carotenoids deposited along with the amount of pigments in the yolk and the concentration of each class. The egg yolk color comes primarily from xanthophylls, which are oxygenated carotenoids. The main xanthophylls present in egg yolks are lutein (C₄₀H₅₆O₂), zeaxanthin (C₄₀H₅₆O₂), cryptoxanthin (C 40 H 56 O) and canthaxanthin (C 40 H 52 O 2 ). Lutein is a dihydroxycarotenoid with two hydroxyl (-OH) groups that is found mainly in leafy greens and corn, whereas Zeaxanthin is structurally similar to lutein but with slight differences in the position of double bonds (Mrowicka et al., 2022 ). Cryptoxanthin is a monohydroxycarotenoid with one hydroxyl (-OH) group that also serves as a precursor to vitamin A, and canthaxanthin is a diketocarotenoid with two ketone (C = O) groups that is often used as a feed additive to enhance yolk pigmentation (Burri, 2015 ). Although the absorbance wavelengths of the main yolk pigments and catenoids range from 400–600 nm, the emission spectra wavelengths of carotenoids in egg yolks generally fall within the 600–680 nm range, depending on the specific carotenoid and its environment. This difference might be due to the variation in key carotenoids (lutein, zeaxanthin and canthazanthin) found in egg yolks and their absorbances, reflectances and/or emission wavelengths. The excitation states of lutein and zeaxanthin (responsible for the yellow color of the yolk) are ~ 430–500 nm, and the emission wavelength is ~ 560 nm (Polívka and Sundström, 2004 ). On the other hand, it is widely established that the feed additive canthazanthin provides an orange hue to the yolk color. These wavelengths correspond to the yellow, orange, and red regions of the visible spectrum, which contribute to the observed yolk color. The exact emission spectrum can shift due to factors such as solvent polarity, molecular interactions, and the presence of proteins or lipids in the yolk. 3.2 Preprocessed spectra The yellow and orange yolk classes showed clearer differences in the 520–600 and 660–680 nm spectral regions after SNV and SD, suggesting that these regions were more important for distinguishing pigment composition than was being affected by scattering artifacts (Fig. 4). This approach helped in highlighting true absorption differences, making classification models more robust. Since yellow yolks and orange yolks have different pigment compositions, their spectral features differ. The peaks and valleys in the second derivative spectra correspond to changes in absorption at specific wavelengths. A shift in peak positions or intensity differences in the second derivative plot indicated variations in pigment concentration or type. The second derivative spectra of the two classes showed distinct trends at specific wavelengths and could be used as spectral markers to classify yolk color more effectively. 3.3 Principal component analysis Principal component analysis (PCA) was performed to observe the differences between two groups, namely, yellow and orange yolk egg groups, and to determine the feasibility of the application of a further classification model, PLS-DA. The PCA score plot revealed that the first three principal components (PCs) captured 96.1% of the total spectral variance in yolk color differences, with PC1 explaining 68.26%, PC2 accounting for 21.95% and PC3 accounting for 5.88%. Orange yolk eggs exhibited higher values along PC3, indicating that the primary source of variance might be associated with differences in deeper orange pigmentation, cryptozanthin and canthazanthin. The pattern suggests that darker orange yolks absorb more light at 600 nm range due to the higher xanthophyll concentration. Therefore, PC3 might indicate the relative amount of deposited xanthophylls in egg yolk. 3.4 PLS-DA model A hyperspectral camera in the visible range of 400 nm to 100 nm was used for the spectral and spatial information of the eggs, and partial least squares discriminant analysis (PLS-DA) with spectral preprocessing was used to develop a classification model. The full-range spectra (400–1000 nm) of the eggs were utilized to develop a PLS-DA model for detecting orange yolk eggs considering 70% of the data were used for training and 30% were used for testing. The classification model was developed using PLS-DA after various spectral preprocessing procedures, such as raw, SNV, first derivative (FD) and second derivative (SD) data. An accuracy of 100% on the testing set was achieved with PLS-DA in the case of spectral preprocessing with SNV FD and SD (Table 1). Table-1: Classification using PLS-DA (test set) with various preprocessing methods (threshold 0.5). Preprocessing Yellow yolk eggs Orange yolk eggs Overall Accuracy Ratio* Accuracy (%) Ratio Accuracy (%) Ratio Accuracy (%) Raw 37/37 100 5/7 71.43 42/44 95.45 SNV 37/37 100 7/7 100 44/44 100 SG-FD 37/37 100 7/7 100 44/44 100 SG-SD 37/37 100 7/7 100 44/44 100 SG-SD (9var) 37/37 100 7/7 100 44/44 100 *Accurately classified eggs/Total eggs for each class The important wavelengths in the classification model ranged from 520–680 nm, which indicated variations in the light absorbance, reflectance and/or emission of the deposited yolk pigments. Yolk color variation in hens of the same breed is primarily driven by dietary carotenoids, metabolic efficiency, and environmental factors, making it an important indicator of diet composition and egg quality. These pigments come from the hen's diet, especially from yellow corn, alfalfa, marigold petals, and other feed ingredients rich in xanthophylls. Individual hens may absorb and metabolize these pigments differently due to variations in the gut microbiota and liver function, which can affect color intensity (Dansou et al., 2023 ). 3.5 Selection of important variables Based on the peak position in the PC loading plot of the SD processed spectra in Fig. 1, 518, 566, 595, 608, 625, 636, 644, 663, and 674 nm (loading coefficient > 0.1) were found to be important wavelengths in the range of 520–680 nm. The most important variables corresponding to the three largest peak positions according to the PLS regression coefficients were 525, 565, and 680 nm (absolute) (Fig. 9 ). The important wavelengths identified in the model for distinguishing yellow and orange yolks (ranging from 525 nm to 680 nm) lie in the visible region, primarily in the green to deep orange region, which was similar to the wavelengths extracted based on the PC loading plot. These wavelengths were significant because they correspond to the absorption and reflectance characteristics of carotenoid and xanthophyll pigments, which are responsible for yolk color. The yellow yolk reflects more light in the 550–560 nm range, making it appear brighter. Carotenoids absorb light primarily in the blue region (400–500 nm) and reflect light in the yellow to red region (520–700 nm), giving them their characteristic colors. The important variables corresponding to the peak positions (PC1 and PC3 loading coefficients > 0.1) were found at 518, 566, 595, 608, 625, 636, 644, 663, and 674 nm. PCA based on those wavelengths showed clear separation between the two classes (Fig. 10 ). Therefore, based on the above findings, important wavelengths, such as 520, 565, 610, 640 and 675 nm, might be used to distinguish orange yolk eggs from yellow yolk eggs. This indicated that the wavelength corresponding to the deposited carotenoid concentration varied between yellow and orange yolks. In the US, most commercial egg production relies on corn and soybean-based diets, which naturally produce yellow yolks due to the presence of lutein and zeaxanthin from corn. However, to achieve deeper orange yolks, producers may add feed additives or natural ingredients rich in xanthophylls originiated from orange corn and marigold petals or supplemented canthaxanthin. In addition, pasture-raised or free-range hens consuming a diet rich in grass, insects, and flowers may also produce deeper-colored yolks due to the wider variety of xanthophyll-rich natural diets. 4. Conclusions The research findings demonstrated the potential of hyperspectral imaging (HSI) combined with partial least squares discriminant analysis (PLS-DA) for separating orange yolk eggs from desired yellow yolk eggs. The developed model achieved a classification accuracy of 100% with only 9 selected variables (518, 566, 595, 608, 625, 636, 644, 663, and 674 nm) from PC (PC1 and PC3) loadings (peaks > 0.1). The key spectral range of 520–680 nm was identified as crucial for classification, highlighting the role of carotenoids and xanthophylls in yolk pigmentation. The ability to noninvasively trace yolk color without breaking the eggshell might be applied in the egg and poultry industry, particularly in product branding, quality assurance, and compliance with market-specific standards. These findings, which involved nondestructive egg grading and quality assessment, align with Industry 4.0 advancements in automation. Moreover, this research could provide a basis for the prediction of yolk pigmentation intensity and antioxidant concentration in the future for the egg and pharmaceutical industries. Additionally, real-time implementation of these multispectral-based optical sensing systems in industrial settings could further optimize the efficiency and economic viability of the egg industry. Declarations Author Contributions A.K. conceptualized the idea, conducted the experiment as a lead and drafted the whole manuscript. Formal analysis was performed by A.K. and M.K. J. L. E. reviewed and edited the manuscript. M.K. and J.L.E. were project administrators. All the authors reviewed, edited and approved the manuscript. Funding This work was supported by the USDA National Institute of Food and Agriculture, Award # 2023−67015−39154. Data Availability Statement Available upon request from the corresponding authors. Competing of interest The authors declare no competing financial interest. References Ahmed MW, Hossainy SJ, Khaliduzzaman A, Emmert JL, Kamruzzaman M (2023) Non-destructive optical sensing technologies for advancing the egg industry toward Industry 4.0: A review. Compr. Rev. Food Sci. Food Saf. 22:4378–4403 Available at https://ift.onlinelibrary.wiley.com/doi/abs/ 10.1111/1541-4337.13227 Ahmed MW, Khaliduzzaman A, Emmert JL, Kamruzzaman M (2025) An overview of recent advancements in hyperspectral imaging in the egg and hatchery industry. Comput. Electron. Agric. 230:109847 Available at https://www.sciencedirect.com/science/article/pii/S0168169924012389 Burri BJ (2015) Beta-cryptoxanthin as a source of vitamin A. J. Sci. Food Agric. 95:1786–1794 Available at https://scijournals.onlinelibrary.wiley.com/doi/abs/10.1002/jsfa.6942 Dansou DM, Zhang H, Yu Y, Wang H, Tang C, Zhao Q, Qin Y, Zhang J (2023) Carotenoid enrichment in eggs: From biochemistry perspective. Anim. Nutr. 14:315–333 Available at https://www.sciencedirect.com/science/article/pii/S2405654523000793 Gewers FL, Ferreira GR, De Arruda HF, Silva FN, Comin CH, Amancio DR, Costa LDF (2021) Principal component analysis: A natural approach to data exploration. ACM Comput Surv. 54 Grisanti E, Totska M, Huber S, Calderon CK, Hohmann M, Lingenfelser D, Otto M (2018) Dynamic localized SNV, Peak SNV, and partial peak SNV: Novel standardization methods for preprocessing of spectroscopic data used in predictive modeling. J. Spectrosc. 2018 Karunajeewa H, Hughes RJ, M. W. M., and, Shenstone FS (1984) A Review of Factors Influencing Pigmentation of Egg Yolks. Worlds. Poult. Sci. J. 40:52–65 Available at https://doi.org/10.1079/WPS19840006 Hanusova E, Hrnčár C, Hanus A, Oravcova M (2015) Effect of breed on some parameters of egg quality in laying hens. Acta Fytotech Zootech 18:12–24 Islam MH, Khaliduzzaman A (2022) Chemometric for Spectroscopy-Based Egg and Poultry Research. In: Khaliduzzaman A (ed) Informatics in Poultry Production: A Technical Guidebook for Egg and Poultry Education, Research and Industry. Springer Nature Singapore, Singapore, pp 111–126 Islam MH, Kondo N, Ogawa Y, Fujiura T, Suzuki T, Fujitani S (2016) Detection of infertile eggs using visible transmission spectroscopy combined with multivariate analysis. Eng. Agric. Environ. Food:1–6 Available at http://linkinghub.elsevier.com/retrieve/pii/S1881836616301434 (verified 6 February 2017) Kamruzzaman M, Barbin D, ElMasry G, Sun D-W, Allen P (2012) Potential of hyperspectral imaging and pattern recognition for categorization and authentication of red meat. Innov. Food Sci. Emerg. Technol. 16:316–325 Available at https://www.sciencedirect.com/science/article/pii/S1466856412001026 Khaliduzzaman A, Kashimori A, Suzuki T, Ogawa Y, Kondo N (2021) Research Note: Nondestructive detection of super grade chick embryos or hatchlings using near-infrared spectroscopy. Poult. Sci. 100:101189 Available at https://doi.org/10.1016/j.psj.2021.101189 Loffredi E, Grassi S, Alamprese C (2021) Spectroscopic approaches for non-destructive shell egg quality and freshness evaluation: Opportunities and challenges. Food Control 129:108255 Available at https://www.sciencedirect.com/science/article/pii/S0956713521003935 Mrowicka M, Mrowicki J, Kucharska E, Majsterek I (2022) Lutein and Zeaxanthin and Their Roles in Age-Related Macular Degeneration—Neurodegenerative Disease. Nutrients 14 Ortiz D, Lawson T, Jarrett R, Ring A, Scoles KL, Hoverman L, Rocheford E, Karcher DM, Rocheford T (2021) Biofortified orange corn increases xanthophyll density and yolk pigmentation in egg yolks from laying hens. Poult. Sci. 100:101117 Available at https://doi.org/10.1016/j.psj.2021.101117 Polívka T, Sundström V (2004) Ultrafast Dynamics of Carotenoid Excited States – From Solution to Natural and Artificial Systems. Chem. Rev. 104:2021–2072 Available at https://doi.org/10.1021/cr020674n Rahman A, Syduzzaman M, Khaliduzzaman A, Fujitani S, Kashimori A, Suzuki T, Ogawa Y, Kondo N (2021) Nondestructive Preincubation Sex Determination of Broiler Eggs Using Visible Transmission Spectroscopy Combined with Multivariate Analysis. Eng Agric Environ Food 14:61–72 Sokołowicz Z, Krawczyk J, Dykiel M (2018) The Effect of the Type of Alternative Housing System, Genotype and Age of Laying Hens on Egg Quality. Ann Anim Sci 18:541–555 Syduzzaman M, Rahman A, Alin K, Fujitani S, Kashimori A, Suzuki T, Ogawa Y, Kondo N (2019a) Noninvasive quantification of yolk content using Vis-NIR spectroscopy and its effect on hatching time and gender of broiler chicken. Eng Agric Environ Food 12:289–296 Syduzzaman M, Rahman A, Khaliduzzaman 藤谷伸一, 樫森 (2019b) 亜由子 鈴木哲仁 小川, and 雄一 近藤直. Noninvasive Quantification of Egg Yolk Ratio Using Vis-NIR Spectroscopy and Multivariate Regression. J. Japanese Soc. Agric. Mach. Food Eng. 81:168–176 Available at https://agriknowledge.affrc.go.jp/RN/2030927856 Valcu CM, Valcu M, Teltscher K, Kempenaers B (2020) Non-analytical methods for the estimation of total yolk carotenoids in passerine eggs. Ibis (Lond 1859:162:1075–1081 Wold S, Esbensen K, Geladi P (1987) Principal component analysis. Chemom. Intell. Lab. Syst. 2:37–52 Available at https://www.sciencedirect.com/science/article/pii/0169743987800849 (verified 7 February 2019) Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6207045","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":427489046,"identity":"5c9a10ea-6eb2-4693-b05d-3c8f5c29cafc","order_by":0,"name":"Alin Khaliduzzaman","email":"data:image/png;base64,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","orcid":"https://orcid.org/0000-0001-6621-8430","institution":"University of Illinois Urbana−Champaign","correspondingAuthor":true,"prefix":"","firstName":"Alin","middleName":"","lastName":"Khaliduzzaman","suffix":""},{"id":427489047,"identity":"1e7de1e7-4df3-490a-9316-78f6db6cf9ee","order_by":1,"name":"Jason Lee Emmert","email":"","orcid":"","institution":"University of Illinois Urbana−Champaign","correspondingAuthor":false,"prefix":"","firstName":"Jason","middleName":"Lee","lastName":"Emmert","suffix":""},{"id":427489048,"identity":"7347c821-a9fd-4a41-bf07-3f8b81c6bc38","order_by":2,"name":"Mohammad Kamruzzaman","email":"","orcid":"","institution":"University of Illinois Urbana−Champaign","correspondingAuthor":false,"prefix":"","firstName":"Mohammad","middleName":"","lastName":"Kamruzzaman","suffix":""}],"badges":[],"createdAt":"2025-03-11 21:52:45","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-6207045/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6207045/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":78663976,"identity":"e4e71eae-f77c-4e60-b79a-8782606520ab","added_by":"auto","created_at":"2025-03-17 10:47:50","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":388175,"visible":true,"origin":"","legend":"\u003cp\u003eTypical yolk color classes: yellow and orange yolks found in the experimental eggs.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6207045/v1/f2eee4f3c2582f34816d3d43.png"},{"id":78663978,"identity":"f1a09961-b159-4771-8363-0298e0fb0878","added_by":"auto","created_at":"2025-03-17 10:47:51","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":228866,"visible":true,"origin":"","legend":"\u003cp\u003eHyperspectral imaging system setup for scanning egg samples.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6207045/v1/a35b8e06aaf3e052a1f3be38.png"},{"id":78664777,"identity":"5902cc5a-958d-4110-9b85-a7edae2e3d6d","added_by":"auto","created_at":"2025-03-17 10:55:51","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":87178,"visible":true,"origin":"","legend":"\u003cp\u003eThe average spectra of yellow and orange yolk egg classes. Adifference in the spectral pattern was clearly visible at520-600 nm and 660-680 nm.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6207045/v1/65a1f4745e368f65e2819558.png"},{"id":78663981,"identity":"3205ab55-5b1a-4a86-9bee-24a701be8205","added_by":"auto","created_at":"2025-03-17 10:47:51","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":73176,"visible":true,"origin":"","legend":"\u003cp\u003ePreprocessed spectral (SNV and SG-2\u003csup\u003end\u003c/sup\u003e derivatives) differences between the two classes.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6207045/v1/ecc5ff901569c3644a2675a9.png"},{"id":78664780,"identity":"b1562aae-ab1f-4a2c-bb09-7e5f94f5c4d0","added_by":"auto","created_at":"2025-03-17 10:55:51","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":100176,"visible":true,"origin":"","legend":"\u003cp\u003ePCA of all the egg samples with standard deviation (SD) spectral preprocessing (red bullet indicates orange yolk). Clear separation was observed in the PC1 vs PC3 score plot.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6207045/v1/0798eeca820172c38f0ec6c7.png"},{"id":78665208,"identity":"84ab21a6-8054-4fe4-8525-4409ead71b54","added_by":"auto","created_at":"2025-03-17 11:03:51","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":48407,"visible":true,"origin":"","legend":"\u003cp\u003ePC loading plots for PC1 and PC3 after SD.\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-6207045/v1/85edcc40e65af8ecf2ead0de.png"},{"id":78663982,"identity":"0ac04077-85ef-4551-9002-a5cbdf2ef372","added_by":"auto","created_at":"2025-03-17 10:47:51","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":38671,"visible":true,"origin":"","legend":"\u003cp\u003eConfusion matrix of the best PLS-DA model with SNV, FD, and SD for classifying yellow and orange yolk eggs.\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-6207045/v1/8c598eece394ef7f65be0822.png"},{"id":78664779,"identity":"b982b836-ebe1-4186-82cd-74e04b6fab0e","added_by":"auto","created_at":"2025-03-17 10:55:51","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":31430,"visible":true,"origin":"","legend":"\u003cp\u003eFigure-9: Important variables in the PLS-DA model with SD-preprocessed spectra.\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-6207045/v1/d1a4c31ef48492799e514c3a.png"},{"id":78664781,"identity":"421c2f99-3e8d-446a-81e6-707761481947","added_by":"auto","created_at":"2025-03-17 10:55:51","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":48909,"visible":true,"origin":"","legend":"\u003cp\u003eFigure-10: PCA of the PC loading at the peaks (peak absolute value\u0026gt;0.1).\u003c/p\u003e","description":"","filename":"10.png","url":"https://assets-eu.researchsquare.com/files/rs-6207045/v1/58fdd54ff2e75b1b5f416277.png"},{"id":78666422,"identity":"d220ab17-9dbb-4f42-a1e4-99d26d7fb3a7","added_by":"auto","created_at":"2025-03-17 11:19:51","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1814371,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6207045/v1/a3570159-ca37-41fc-ada8-db9992fbdaec.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eNondestructive Classification of Yellow and Orange Colored Yolk Eggs using Hyperspectral Imaging Combined with PLS-DA\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eIn countries such as the United States, yellow yolk is maintained on poultry farms by controlling feed formulation and breed preferences to meet consumer preferences (Ortiz et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). However, in a preliminary experiment involving egg yolk, a few orange-colored yolks were found within the same lot and breed, possibly due to individual variations in feed intake, metabolic efficiency and thus yolk pigmentation. The lipids of the yolk are exclusively associated with lipoprotein aggregates composed of triglycerides, phospholipids, and cholesterol. Less than 1% of yolk lipids are carotenoids, which give the yolk its hue ranging from pale yellow to dark bright orange (Dansou et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). This yolk color is also influenced by the genotype and the rate of egg production in hens (Hanusova et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Sokołowicz et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). There is some variation between breeds, strains, and individual hens in terms of their ability to absorb and deposit oxycarotenoids in egg yolk (H. Karunajeewa R. J. Hughes and Shenstone, 1984). Therefore, there is a need to develop nondestructive, fast, and precise methods for objectively measuring yolk color for the benefit of the egg and poultry industry as well as consumer markets. The functional properties (e.g., antioxidants) of carotenoids in egg yolk may also improve the quality and nutritional value of eggs.\u003c/p\u003e \u003cp\u003eThe egg industry is undergoing a significant transformation through the adoption of Industry 4.0 technologies, emphasizing automation and digitalization to enhance efficiency and product quality (Ahmed et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Although automated egg grading systems utilizing advanced sensors and machine learning algorithms have been developed to assess external characteristics such as size, weight, and shell integrity, limited research has been conducted for evaluating internal egg qualities, specifically yolk attributes due to advanced technological challenges. Yolk attributes are important for both table and hatching eggs, as they influence the gender, hatchability and various quality parameters of day-old chicks (Valcu et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Khaliduzzaman et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Rahman et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Traditional methods often require destructive sampling, limiting their applicability in large-scale operations. Although nondestructive techniques, such as hyperspectral imaging (HSI) and near-infrared (NIR) spectroscopy, have been recently reported to predict internal qualities, such as the yolk ratio and albumen freshness (Syduzzaman et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2019a\u003c/span\u003e; Loffredi et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), there is no single study is conducted so far on yolk color based eggs classification using non-destructive means. Yolk color is a critical quality parameter that influences consumer preferences and perceptions of nutritional value. The risk is primarily determined by the hen's diet, particularly the intake of carotenoids, which impart hues ranging from pale yellow to deep orange. The major pigments responsible for yolk color are lutein, zeaxanthin, \u003cem\u003e\u0026szlig;\u003c/em\u003e-cryptoxanthin and canthaxanthin (Ortiz et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Mrowicka et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). The source of zeazanthins is a corn-based diet, and the type of corn influences the yolk color, where the diet from orange corn is responsible for more orange color pigmentation on the yolk (Ortiz et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eDeveloping protocol for nondestructive yolk color detection could revolutionize quality control system in both the table egg and hatching egg sectors. Such technologies might enable producers to meet consumer demands more effectively, ensuring consistency in product appearance without compromising the integrity of the eggs during inspection. Yolk coloration, ranging from pale yellow to deep orange, is closely linked to consumer preferences, nutritional quality, freshness, and premium value, thus may drive demand and enable higher pricing. In some other countries, such as Japan, the deep orange color of the yolk is often preferred for aesthetic reasons, especially when eating eggs raw in dishes such as \"tamago-kake-gohan\u0026rdquo;. Furthermore, yolk pigmentation serves as a biomarker for hens' dietary carotenoid intake, offering insights into feed quality and management practices. Therefore, the objective of this study is to develop a noninvasive approach using hyperspectral imaging (HSI) and multivariate analysis techniques, such as PLS-DA, to distinguish orange yolk eggs from yellow yolk eggs.\u003c/p\u003e \u003cp\u003eHSI integrates the advantages of spectroscopy and conventional imaging techniques in one system to provide both spectral and spatial information about an object. Hyperspectral imaging is a noninvasive and high-resolution method for capturing detailed spectral information from chicken eggs (Ahmed et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Thus, an HSI in the visible range can be used to convey yolk and yolk color information when light passes through the egg reaches the camera sensor. PLS-DA is a supervised multivariate statistical method used for classification that involves maximizing the separation between predefined groups while explaining variance in predictor variables (Kamruzzaman et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2012\u003c/span\u003e). It works by projecting predictor variables onto a new space that correlates with the response variable, enabling clear group discrimination. Thus, HSI together with multivariate analysis could be a good combination to classify eggs based on yolk for the future egg industry.\u003c/p\u003e"},{"header":"2. Materials and methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003e2.1 Materials\u003c/h2\u003e\n \u003cp\u003eThe animal experiment protocol was followed and approved by the Office of the Vice-Chancellor for Research IACUC online protocols, University of Illinois (protocol #: 22224). A total of 146 White Leghorn eggs (117 yellow yolk eggs and 29 orange yolk eggs) were collected from the poultry farm of the University of Illinois at Urbana-Champaign; these eggs included both fertile and infertile white eggs and were used for internal yolk color-based egg classification. The total dataset was divided into 70% for training (\u003cem\u003en\u003c/em\u003e\u0026thinsp;=\u0026thinsp;102, 80 eggs for the yellow class and 22 for the orange yolk class) and 30% for testing (n\u0026thinsp;=\u0026thinsp;44, 37 for the yellow class and 7 for the orange class) sets using random sate 42 (in python). The average size of the eggs was 59.47 g, and the major diameter and minor diameter were 57.04 and 43.10 mm, respectively. The variability of egg mass and size is an important criterion for the classification model and feasibility of the proposed technique when applied to actual cases and industrial settings. Two types of yolk color were found in all the egg samples, among which yellow yolk eggs were the major class (Fig. 1). The yolks were defined as yellow or orange based on their yolkfan scores (yellow: 5\u0026ndash;6; orange: 11\u0026ndash;13). In this study, approximately 20% of the orange yolk eggs were produced via mass production.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\n \u003ch2\u003e2.2 Hyperspectral imaging system\u003c/h2\u003e\n \u003cp\u003eA hyperspectral camera (Model: PIKA-L, line scanning) in the visible range of 400 nm to 100 nm was used for the spectral and spatial information of the eggs (Fig. 2). The PIKA-L hyperspectral camera works based on the line-scanning principle called the push-broom method, as it sweeps over the target line by line.\u003c/p\u003e\n \u003cp\u003eA scanning speed of 0.06 cm/s with a frame rate of 9.9 fps and an exposure time of 100 ms were maintained during hyperspectral image acquisition. The output of an HSI system is a data-rich hyperspectral image, which is a three-dimensional data cube also known as voxel \u003cem\u003eI\u003c/em\u003e (x, y, \u0026lambda;). The data cube can be interpreted as a collection of monochrome images \u003cem\u003eI\u003c/em\u003e (x, y) for each wavelength \u0026lambda; or as a spectrum \u003cem\u003eI\u003c/em\u003e (\u0026lambda;) for each pixel (x, y).\u003c/p\u003e\n \u003cp\u003eThe key components of a line scanning hyperspectral camera include a spectrograph with an input slit, a grayscale camera and an objective lens. A hyperspectral image of an object is obtained when a spectrograph is added to the camera system. In this case, the spectrograph consists of an input slit, collimating optics, a dispersive unit and a focusing lens. The input slit limits the incoming information from the object by limiting the input light, which allows only a single line. The collimating lens collates the light into dispersive light, and the dispersive unit converts the light into spectra. The focusing lens then focuses the dispersed light into the grayscale camera, which measures the intensity of light at different wavelengths. Thus, a hyperspectral camera collects spectral data for each pixel in a line across the target area simultaneously. The target moves (or the sensor moves) to capture the next line, gradually building up a full hyperspectral image as the lines are combined.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n \u003ch2\u003e2.3 Spectral extraction from the HSI cube\u003c/h2\u003e\n \u003cp\u003eThe HSI cube contains a large amount of information about the egg sample in spectral and spatial dimensions. Thus, a series of operations are required to extract meaningful information from the HSI cube to enhance interpretability and prepare it for further analysis to develop prediction models. The average spectrum of an egg was obtained from an image cube based on the ROIs. ROIs were selected by applying a threshold to pixel intensity values. Pixels below the threshold are set as the background, while others form the object. This technique is straightforward and computationally efficient. Eggs typically have clear boundaries, making thresholding-based segmentation suitable for accurate separation from the background.\u003c/p\u003e\n \u003cp\u003eHyperspectral images stored in band interleaved-by-line (BIL) format were read using the MATLAB code \u003cem\u003emultibandread\u003c/em\u003e. The hyperspectral cube dimensions were 790 \u0026times; 900 \u0026times; 300, where 790 represents the height (rows), 900 represents the width (columns), and 300 represents the spectral bands. The DN (digital number) value corresponding to band 150 was used as a threshold to filter relevant pixels. Pixels with DNs\u0026thinsp;\u0026gt;\u0026thinsp;600 were selected for spectral extraction. The sum of the spectral values from all the selected pixels was computed, and the total number of selected pixels was subsequently used to compute the average spectral response. The extracted spectra were stored in a declared variable, \u003cem\u003eR\u003c/em\u003e, which was a matrix containing spectral information from all processed images. Finally, \u003cem\u003eR\u003c/em\u003e was reshaped to maintain the correct dimensions [\u003cem\u003en\u003c/em\u003e, 300], where \u003cem\u003en\u003c/em\u003e represents the number of images (egg sample) processed.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n \u003ch2\u003e2.4 Data analysis\u003c/h2\u003e\n \u003cdiv id=\"Sec7\" class=\"Section3\"\u003e\n \u003ch2\u003e2.4.1 Principal component analysis\u003c/h2\u003e\n \u003cp\u003eAlthough principal component analysis (PCA) is used for dimensionality reduction of data while retaining valuable information (Wold et al., \u003cspan class=\"CitationRef\"\u003e1987\u003c/span\u003e; Gewers et al., \u003cspan class=\"CitationRef\"\u003e2021\u003c/span\u003e). PCA is also very helpful for observing the class differences in spectral data and feasibility of further classification models since PCA is an unsupervised dimensionality reduction technique (Islam and Khaliduzzaman, \u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e). PCA does not use class labels or target variables; it only analyzes the input features (independent variables). It finds new axes (principal components) that maximize variance in the data, helping reduce dimensionality while preserving as much information as possible. Therefore, PCA can potentially be used for exploring data structure and visualization. PCA transforms data into a new coordinate system where the greatest variance lies along the first principal component, the second greatest variance along the second component, and so on.\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec8\" class=\"Section3\"\u003e\n \u003ch2\u003e2.4.2 Spectral preprocessing\u003c/h2\u003e\n \u003cp\u003ePreprocessing is a common prestep process in spectral data processing since it considerably improves the performance of multivariate models (Islam et al., \u003cspan class=\"CitationRef\"\u003e2016\u003c/span\u003e). Despite extensive studies on various preprocessing approaches, trial and error methods, together with performance, are often used to determine the optimal technique. In this study, four widely used preprocessing techniques, standard normal variate (SNV) and Savitzky\u0026ndash;Golay (SG) first-order and second-order derivatives with window sizes of 15 and polynomial order of 3, were used for PLS\u0026ndash;DA calibration model development. The SNV effectively eliminates nonuniform scattering and particle size effects from the spectra. The first derivative mitigates baseline shifts and standardizes absorption rates, enhancing the comparability of spectral peaks and troughs. Second derivatives convert local maxima peaks to minima below the baseline, flattening it. The second derivative of a spectrum enhances subtle spectral features by highlighting changes in curvature, making it a valuable tool in spectral analysis. This approach sharpens overlapping peaks by reducing the effect of broad baseline variations. Therefore, small variations in absorbance or reflectance due to chemical composition become more prominent.\u003c/p\u003e\n \u003cp\u003eThe SNV is used to standardize features by removing the mean and scaling to unit variance. Standardization is an essential preprocessing step in many machine learning models, especially those sensitive to feature scales, such as PLS. Model performance can be improved by ensuring that each feature contributes equally, regardless of its original scale. In SNV, each feature is centered by subtracting its mean and scaling to unit variance, dividing each feature by its standard deviation. The transformation results in a mean of 0 and a standard deviation of 1 (Grisanti et al., \u003cspan class=\"CitationRef\"\u003e2018\u003c/span\u003e). The SNV removes baseline offsets and intensity variations, making the spectral differences between the two classes more chemically relevant than just due to sample properties. The effectiveness of preprocessing methods can vary depending on the specific spectral data and analytical goals, necessitating careful selection and optimization to ensure consistent and accurate results. However, in some cases, spectral preprocessing may lead to potential information loss, the introduction of artifacts, and increased complexity.\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec9\" class=\"Section3\"\u003e\n \u003ch2\u003e2.4.3 PLS-DA model\u003c/h2\u003e\n \u003cp\u003ePartial least squares regression (PLS-R) is an important methodology for solving both regression and classification problems. PLS-DA is a supervised classification technique derived from partial least squares regression (PLS-R). It is particularly useful when dealing with high-dimensional, collinear data, such as spectral datasets (Kamruzzaman et al., \u003cspan class=\"CitationRef\"\u003e2012\u003c/span\u003e). Unlike traditional classification techniques, PLS-DA finds the latent variables that best separate predefined classes while maximizing the covariance between the predictor matrix (\u003cem\u003eX\u003c/em\u003e) and the response variable (\u003cem\u003eY\u003c/em\u003e). PLS-DA follows a similar approach to PLS-R but modifies the response variable (\u003cem\u003eY\u003c/em\u003e) to represent class labels instead of continuous outcomes (Islam and Khaliduzzaman, \u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e). Class 0 was used for the yellow yolk class, and 1 was used for the orange yolk class. The threshold value for class separation using PLS-DA was set to 0.5.\u003c/p\u003e\n \u003cp\u003eThus, PLS-DA extracts latent variables that maximize the separation between classes while explaining most of the variance in \u003cem\u003eX\u003c/em\u003e. The latent variables serve as projections that enhance class separability. The response variable \u003cem\u003eY\u003c/em\u003e of PLS-DA is a set of binary variables that relates to the categories or classes of the sample. PLS-R or PLS-DA is expressed by Eq. 1.\u003c/p\u003e\n \u003cp\u003e\u003cem\u003eY\u003c/em\u003e\u0026thinsp;=\u0026thinsp;\u003cem\u003eXb\u003c/em\u003e\u0026thinsp;+\u0026thinsp;\u003cem\u003eE.\u003c/em\u003e \u0026hellip; \u0026hellip; \u0026hellip; (Eq. 1)\u003c/p\u003e\n \u003cp\u003ewhere \u003cem\u003eX\u003c/em\u003e is an (\u003cem\u003en\u003c/em\u003e x \u003cem\u003ep\u003c/em\u003e) matrix that holds the spectral values of each class, \u003cem\u003eb\u003c/em\u003e = [\u003cem\u003eb\u003c/em\u003e\u003csub\u003e1\u003c/sub\u003e \u003cem\u003eb\u003c/em\u003e\u003csub\u003e2\u003c/sub\u003e ... \u003cem\u003eb\u003c/em\u003e\u003csub\u003ep\u003c/sub\u003e] \u003csup\u003e\u003cem\u003eT\u003c/em\u003e\u003c/sup\u003e is an (\u003cem\u003ep\u003c/em\u003e x1) dimensional column vector that contains the regression coefficients, \u003cem\u003eT\u003c/em\u003e denotes the transpose of a vector or matrix, and \u003cem\u003eE\u003c/em\u003e is the error term. After PLS projection, classification was performed based on the predicted values of \u003cem\u003eY\u003c/em\u003e.\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec10\" class=\"Section3\"\u003e\n \u003ch2\u003e2.4.4 Important variables selection\u003c/h2\u003e\n \u003cp\u003eImportant variables were selected based on the peak position (peak value\u0026thinsp;\u0026gt;\u0026thinsp;0.1) in the PC loading plot to identify important wavelengths for classifying orange yolks from yellow yolk eggs. These peak positions in the PC loading plot should correspond to important spectral wavelengths (var) containing class information, as clear separation was observed in the PCA (PC1 vs PC3 plot). Variables with high absolute loadings on a specific PC are considered the most important contributors to that component, meaning that they are strongly correlated with the direction of variation represented by that PC (Wold et al., \u003cspan class=\"CitationRef\"\u003e1987\u003c/span\u003e).\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e"},{"header":"3. Results and discussion","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\n \u003ch2\u003e3.1 Spectral characteristics of egg classes\u003c/h2\u003e\n \u003cp\u003eThe spectral properties of yellow egg yolk and orange yolk showed clear and typical differences in the range of 520\u0026ndash;600 and 640\u0026ndash;680 nm, as shown in Fig. 3. The 520\u0026ndash;600 nm wavelength ranges from green (~\u0026thinsp;520 nm) to yellow (~\u0026thinsp;570 nm) to orange (~\u0026thinsp;600 nm), whereas the 640\u0026ndash;680 nm wavelength ranges from deep orange to red, indicating the presence of carotenoids in the egg yolk. Carotenoids might absorb in the green to yellow region and reflect or emit yellow to orange colors depending on the type and concentration since yolk has a much greater absorbance coefficient below 600 nm (Syduzzaman et al., \u003cspan class=\"CitationRef\"\u003e2019b\u003c/span\u003e). This variation might be due to differences in the color of the various carotenoids deposited along with the amount of pigments in the yolk and the concentration of each class.\u003c/p\u003e\n \u003cp\u003eThe egg yolk color comes primarily from xanthophylls, which are oxygenated carotenoids. The main xanthophylls present in egg yolks are lutein (C₄₀H₅₆O₂), zeaxanthin (C₄₀H₅₆O₂), cryptoxanthin (C\u003csub\u003e40\u003c/sub\u003eH\u003csub\u003e56\u003c/sub\u003eO) and canthaxanthin (C\u003csub\u003e40\u003c/sub\u003eH\u003csub\u003e52\u003c/sub\u003eO\u003csub\u003e2\u003c/sub\u003e). Lutein is a dihydroxycarotenoid with two hydroxyl (-OH) groups that is found mainly in leafy greens and corn, whereas Zeaxanthin is structurally similar to lutein but with slight differences in the position of double bonds (Mrowicka et al., \u003cspan class=\"CitationRef\"\u003e2022\u003c/span\u003e). Cryptoxanthin is a monohydroxycarotenoid with one hydroxyl (-OH) group that also serves as a precursor to vitamin A, and canthaxanthin is a diketocarotenoid with two ketone (C\u0026thinsp;=\u0026thinsp;O) groups that is often used as a feed additive to enhance yolk pigmentation (Burri, \u003cspan class=\"CitationRef\"\u003e2015\u003c/span\u003e).\u003c/p\u003e\n \u003cp\u003eAlthough the absorbance wavelengths of the main yolk pigments and catenoids range from 400\u0026ndash;600 nm, the emission spectra wavelengths of carotenoids in egg yolks generally fall within the 600\u0026ndash;680 nm range, depending on the specific carotenoid and its environment. This difference might be due to the variation in key carotenoids (lutein, zeaxanthin and canthazanthin) found in egg yolks and their absorbances, reflectances and/or emission wavelengths. The excitation states of lutein and zeaxanthin (responsible for the yellow color of the yolk) are ~\u0026thinsp;430\u0026ndash;500 nm, and the emission wavelength is ~\u0026thinsp;560 nm (Pol\u0026iacute;vka and Sundstr\u0026ouml;m, \u003cspan class=\"CitationRef\"\u003e2004\u003c/span\u003e). On the other hand, it is widely established that the feed additive canthazanthin provides an orange hue to the yolk color.\u003c/p\u003e\n \u003cp\u003eThese wavelengths correspond to the yellow, orange, and red regions of the visible spectrum, which contribute to the observed yolk color. The exact emission spectrum can shift due to factors such as solvent polarity, molecular interactions, and the presence of proteins or lipids in the yolk.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n \u003ch2\u003e3.2 Preprocessed spectra\u003c/h2\u003e\n \u003cp\u003eThe yellow and orange yolk classes showed clearer differences in the 520\u0026ndash;600 and 660\u0026ndash;680 nm spectral regions after SNV and SD, suggesting that these regions were more important for distinguishing pigment composition than was being affected by scattering artifacts (Fig. 4). This approach helped in highlighting true absorption differences, making classification models more robust.\u003c/p\u003e\n \u003cp\u003eSince yellow yolks and orange yolks have different pigment compositions, their spectral features differ. The peaks and valleys in the second derivative spectra correspond to changes in absorption at specific wavelengths. A shift in peak positions or intensity differences in the second derivative plot indicated variations in pigment concentration or type. The second derivative spectra of the two classes showed distinct trends at specific wavelengths and could be used as spectral markers to classify yolk color more effectively.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n \u003ch2\u003e3.3 Principal component analysis\u003c/h2\u003e\n \u003cp\u003ePrincipal component analysis (PCA) was performed to observe the differences between two groups, namely, yellow and orange yolk egg groups, and to determine the feasibility of the application of a further classification model, PLS-DA. The PCA score plot revealed that the first three principal components (PCs) captured 96.1% of the total spectral variance in yolk color differences, with PC1 explaining 68.26%, PC2 accounting for 21.95% and PC3 accounting for 5.88%.\u003c/p\u003e\n \u003cp\u003eOrange yolk eggs exhibited higher values along PC3, indicating that the primary source of variance might be associated with differences in deeper orange pigmentation, cryptozanthin and canthazanthin. The pattern suggests that darker orange yolks absorb more light at \u0026lt;\u0026thinsp;600 nm and reflect or emit more light in the \u0026gt;\u0026thinsp;600 nm range due to the higher xanthophyll concentration. Therefore, PC3 might indicate the relative amount of deposited xanthophylls in egg yolk.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\n \u003ch2\u003e3.4 PLS-DA model\u003c/h2\u003e\n \u003cp\u003eA hyperspectral camera in the visible range of 400 nm to 100 nm was used for the spectral and spatial information of the eggs, and partial least squares discriminant analysis (PLS-DA) with spectral preprocessing was used to develop a classification model. The full-range spectra (400\u0026ndash;1000 nm) of the eggs were utilized to develop a PLS-DA model for detecting orange yolk eggs considering 70% of the data were used for training and 30% were used for testing.\u003c/p\u003e\n \u003cp\u003eThe classification model was developed using PLS-DA after various spectral preprocessing procedures, such as raw, SNV, first derivative (FD) and second derivative (SD) data. An accuracy of 100% on the testing set was achieved with PLS-DA in the case of spectral preprocessing with SNV FD and SD (Table\u0026nbsp;1).\u003c/p\u003e\n \u003cp\u003eTable-1: Classification using PLS-DA (test set) with various preprocessing methods (threshold 0.5).\u003c/p\u003e\n \u003cdiv class=\"gridtable\"\u003e\n \u003ctable id=\"Tabf\" border=\"1\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\" rowspan=\"2\"\u003e\n \u003cp\u003ePreprocessing\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colspan=\"2\"\u003e\n \u003cp\u003eYellow yolk eggs\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colspan=\"2\"\u003e\n \u003cp\u003eOrange yolk eggs\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\" colspan=\"2\"\u003e\n \u003cp\u003eOverall Accuracy\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eRatio*\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAccuracy (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eRatio\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAccuracy (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eRatio\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAccuracy (%)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eRaw\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e37/37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5/7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e71.43\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e42/44\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95.45\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSNV\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e37/37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7/7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e44/44\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSG-FD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e37/37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7/7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e44/44\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSG-SD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e37/37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7/7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e44/44\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSG-SD (9var)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e37/37\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"char\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e7/7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e100\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e44/44\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003e100\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003ctfoot\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"7\"\u003e*Accurately classified eggs/Total eggs for each class\u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tfoot\u003e\n \u003c/table\u003e\n \u003c/div\u003e\n \u003cbr\u003e\n \u003cp\u003eThe important wavelengths in the classification model ranged from 520\u0026ndash;680 nm, which indicated variations in the light absorbance, reflectance and/or emission of the deposited yolk pigments. Yolk color variation in hens of the same breed is primarily driven by dietary carotenoids, metabolic efficiency, and environmental factors, making it an important indicator of diet composition and egg quality.\u003c/p\u003e\n \u003cp\u003eThese pigments come from the hen\u0026apos;s diet, especially from yellow corn, alfalfa, marigold petals, and other feed ingredients rich in xanthophylls. Individual hens may absorb and metabolize these pigments differently due to variations in the gut microbiota and liver function, which can affect color intensity (Dansou et al., \u003cspan class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n \u003ch2\u003e3.5 Selection of important variables\u003c/h2\u003e\n \u003cp\u003eBased on the peak position in the PC loading plot of the SD processed spectra in Fig. 1, 518, 566, 595, 608, 625, 636, 644, 663, and 674 nm (loading coefficient\u0026thinsp;\u0026gt;\u0026thinsp;0.1) were found to be important wavelengths in the range of 520\u0026ndash;680 nm. The most important variables corresponding to the three largest peak positions according to the PLS regression coefficients were 525, 565, and 680 nm (absolute) (Fig. \u003cspan class=\"InternalRef\"\u003e9\u003c/span\u003e). The important wavelengths identified in the model for distinguishing yellow and orange yolks (ranging from 525 nm to 680 nm) lie in the visible region, primarily in the green to deep orange region, which was similar to the wavelengths extracted based on the PC loading plot.\u003c/p\u003e\n \u003cp\u003eThese wavelengths were significant because they correspond to the absorption and reflectance characteristics of carotenoid and xanthophyll pigments, which are responsible for yolk color. The yellow yolk reflects more light in the 550\u0026ndash;560 nm range, making it appear brighter. Carotenoids absorb light primarily in the blue region (400\u0026ndash;500 nm) and reflect light in the yellow to red region (520\u0026ndash;700 nm), giving them their characteristic colors.\u003c/p\u003e\n \u003cp\u003eThe important variables corresponding to the peak positions (PC1 and PC3 loading coefficients\u0026thinsp;\u0026gt;\u0026thinsp;0.1) were found at 518, 566, 595, 608, 625, 636, 644, 663, and 674 nm. PCA based on those wavelengths showed clear separation between the two classes (Fig. \u003cspan class=\"InternalRef\"\u003e10\u003c/span\u003e). Therefore, based on the above findings, important wavelengths, such as 520, 565, 610, 640 and 675 nm, might be used to distinguish orange yolk eggs from yellow yolk eggs. This indicated that the wavelength corresponding to the deposited carotenoid concentration varied between yellow and orange yolks. In the US, most commercial egg production relies on corn and soybean-based diets, which naturally produce yellow yolks due to the presence of lutein and zeaxanthin from corn. However, to achieve deeper orange yolks, producers may add feed additives or natural ingredients rich in xanthophylls originiated from orange corn and marigold petals or supplemented canthaxanthin. In addition, pasture-raised or free-range hens consuming a diet rich in grass, insects, and flowers may also produce deeper-colored yolks due to the wider variety of xanthophyll-rich natural diets.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"4. Conclusions","content":"\u003cp\u003eThe research findings demonstrated the potential of hyperspectral imaging (HSI) combined with partial least squares discriminant analysis (PLS-DA) for separating orange yolk eggs from desired yellow yolk eggs. The developed model achieved a classification accuracy of 100% with only 9 selected variables (518, 566, 595, 608, 625, 636, 644, 663, and 674 nm) from PC (PC1 and PC3) loadings (peaks\u0026thinsp;\u0026gt;\u0026thinsp;0.1). The key spectral range of 520\u0026ndash;680 nm was identified as crucial for classification, highlighting the role of carotenoids and xanthophylls in yolk pigmentation. The ability to noninvasively trace yolk color without breaking the eggshell might be applied in the egg and poultry industry, particularly in product branding, quality assurance, and compliance with market-specific standards. These findings, which involved nondestructive egg grading and quality assessment, align with Industry 4.0 advancements in automation. Moreover, this research could provide a basis for the prediction of yolk pigmentation intensity and antioxidant concentration in the future for the egg and pharmaceutical industries. Additionally, real-time implementation of these multispectral-based optical sensing systems in industrial settings could further optimize the efficiency and economic viability of the egg industry.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAuthor Contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA.K. conceptualized the idea, conducted the experiment as a lead and drafted the whole manuscript. Formal analysis was performed by A.K. and M.K. J. L. E. reviewed and edited the manuscript. M.K. and J.L.E. were project administrators. All the authors reviewed, edited and approved the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by the USDA National Institute of Food and Agriculture, Award # 2023\u0026minus;67015\u0026minus;39154.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Availability Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAvailable upon request from the corresponding authors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting of interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing financial interest.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAhmed MW, Hossainy SJ, Khaliduzzaman A, Emmert JL, Kamruzzaman M (2023) Non-destructive optical sensing technologies for advancing the egg industry toward Industry 4.0: A review. Compr. Rev. Food Sci. Food Saf. 22:4378\u0026ndash;4403 Available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://ift.onlinelibrary.wiley.com/doi/abs/\u003c/span\u003e\u003cspan address=\"https://ift.onlinelibrary.wiley.com/doi/abs/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1111/1541-4337.13227\u003c/span\u003e\u003cspan address=\"10.1111/1541-4337.13227\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAhmed MW, Khaliduzzaman A, Emmert JL, Kamruzzaman M (2025) An overview of recent advancements in hyperspectral imaging in the egg and hatchery industry. Comput. Electron. Agric. 230:109847 Available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.sciencedirect.com/science/article/pii/S0168169924012389\u003c/span\u003e\u003cspan address=\"https://www.sciencedirect.com/science/article/pii/S0168169924012389\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBurri BJ (2015) Beta-cryptoxanthin as a source of vitamin A. J. Sci. Food Agric. 95:1786\u0026ndash;1794 Available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://scijournals.onlinelibrary.wiley.com/doi/abs/10.1002/jsfa.6942\u003c/span\u003e\u003cspan address=\"https://scijournals.onlinelibrary.wiley.com/doi/abs/10.1002/jsfa.6942\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDansou DM, Zhang H, Yu Y, Wang H, Tang C, Zhao Q, Qin Y, Zhang J (2023) Carotenoid enrichment in eggs: From biochemistry perspective. Anim. Nutr. 14:315\u0026ndash;333 Available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.sciencedirect.com/science/article/pii/S2405654523000793\u003c/span\u003e\u003cspan address=\"https://www.sciencedirect.com/science/article/pii/S2405654523000793\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGewers FL, Ferreira GR, De Arruda HF, Silva FN, Comin CH, Amancio DR, Costa LDF (2021) Principal component analysis: A natural approach to data exploration. ACM Comput Surv. 54\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGrisanti E, Totska M, Huber S, Calderon CK, Hohmann M, Lingenfelser D, Otto M (2018) Dynamic localized SNV, Peak SNV, and partial peak SNV: Novel standardization methods for preprocessing of spectroscopic data used in predictive modeling. J. Spectrosc. 2018\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKarunajeewa H, Hughes RJ, M. W. M., and, Shenstone FS (1984) A Review of Factors Influencing Pigmentation of Egg Yolks. Worlds. Poult. Sci. J. 40:52\u0026ndash;65 Available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1079/WPS19840006\u003c/span\u003e\u003cspan address=\"10.1079/WPS19840006\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHanusova E, Hrnč\u0026aacute;r C, Hanus A, Oravcova M (2015) Effect of breed on some parameters of egg quality in laying hens. Acta Fytotech Zootech 18:12\u0026ndash;24\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIslam MH, Khaliduzzaman A (2022) Chemometric for Spectroscopy-Based Egg and Poultry Research. In: Khaliduzzaman A (ed) Informatics in Poultry Production: A Technical Guidebook for Egg and Poultry Education, Research and Industry. Springer Nature Singapore, Singapore, pp 111\u0026ndash;126\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIslam MH, Kondo N, Ogawa Y, Fujiura T, Suzuki T, Fujitani S (2016) Detection of infertile eggs using visible transmission spectroscopy combined with multivariate analysis. Eng. Agric. Environ. Food:1\u0026ndash;6 Available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://linkinghub.elsevier.com/retrieve/pii/S1881836616301434\u003c/span\u003e\u003cspan address=\"http://linkinghub.elsevier.com/retrieve/pii/S1881836616301434\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (verified 6 February 2017)\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKamruzzaman M, Barbin D, ElMasry G, Sun D-W, Allen P (2012) Potential of hyperspectral imaging and pattern recognition for categorization and authentication of red meat. Innov. Food Sci. Emerg. Technol. 16:316\u0026ndash;325 Available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.sciencedirect.com/science/article/pii/S1466856412001026\u003c/span\u003e\u003cspan address=\"https://www.sciencedirect.com/science/article/pii/S1466856412001026\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKhaliduzzaman A, Kashimori A, Suzuki T, Ogawa Y, Kondo N (2021) Research Note: Nondestructive detection of super grade chick embryos or hatchlings using near-infrared spectroscopy. Poult. Sci. 100:101189 Available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.psj.2021.101189\u003c/span\u003e\u003cspan address=\"10.1016/j.psj.2021.101189\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLoffredi E, Grassi S, Alamprese C (2021) Spectroscopic approaches for non-destructive shell egg quality and freshness evaluation: Opportunities and challenges. Food Control 129:108255 Available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.sciencedirect.com/science/article/pii/S0956713521003935\u003c/span\u003e\u003cspan address=\"https://www.sciencedirect.com/science/article/pii/S0956713521003935\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMrowicka M, Mrowicki J, Kucharska E, Majsterek I (2022) Lutein and Zeaxanthin and Their Roles in Age-Related Macular Degeneration\u0026mdash;Neurodegenerative Disease. Nutrients 14\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOrtiz D, Lawson T, Jarrett R, Ring A, Scoles KL, Hoverman L, Rocheford E, Karcher DM, Rocheford T (2021) Biofortified orange corn increases xanthophyll density and yolk pigmentation in egg yolks from laying hens. Poult. Sci. 100:101117 Available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.psj.2021.101117\u003c/span\u003e\u003cspan address=\"10.1016/j.psj.2021.101117\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePol\u0026iacute;vka T, Sundstr\u0026ouml;m V (2004) Ultrafast Dynamics of Carotenoid Excited States\u0026thinsp;\u0026ndash;\u0026thinsp;From Solution to Natural and Artificial Systems. Chem. Rev. 104:2021\u0026ndash;2072 Available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1021/cr020674n\u003c/span\u003e\u003cspan address=\"10.1021/cr020674n\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRahman A, Syduzzaman M, Khaliduzzaman A, Fujitani S, Kashimori A, Suzuki T, Ogawa Y, Kondo N (2021) Nondestructive Preincubation Sex Determination of Broiler Eggs Using Visible Transmission Spectroscopy Combined with Multivariate Analysis. Eng Agric Environ Food 14:61\u0026ndash;72\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSokołowicz Z, Krawczyk J, Dykiel M (2018) The Effect of the Type of Alternative Housing System, Genotype and Age of Laying Hens on Egg Quality. Ann Anim Sci 18:541\u0026ndash;555\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSyduzzaman M, Rahman A, Alin K, Fujitani S, Kashimori A, Suzuki T, Ogawa Y, Kondo N (2019a) Noninvasive quantification of yolk content using Vis-NIR spectroscopy and its effect on hatching time and gender of broiler chicken. Eng Agric Environ Food 12:289\u0026ndash;296\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSyduzzaman M, Rahman A, Khaliduzzaman 藤谷伸一, 樫森 (2019b) 亜由子 鈴木哲仁 小川, and 雄一 近藤直. Noninvasive Quantification of Egg Yolk Ratio Using Vis-NIR Spectroscopy and Multivariate Regression. J. Japanese Soc. Agric. Mach. Food Eng. 81:168\u0026ndash;176 Available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://agriknowledge.affrc.go.jp/RN/2030927856\u003c/span\u003e\u003cspan address=\"https://agriknowledge.affrc.go.jp/RN/2030927856\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eValcu CM, Valcu M, Teltscher K, Kempenaers B (2020) Non-analytical methods for the estimation of total yolk carotenoids in passerine eggs. Ibis (Lond 1859:162:1075\u0026ndash;1081\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWold S, Esbensen K, Geladi P (1987) Principal component analysis. Chemom. Intell. Lab. Syst. 2:37\u0026ndash;52 Available at \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.sciencedirect.com/science/article/pii/0169743987800849\u003c/span\u003e\u003cspan address=\"https://www.sciencedirect.com/science/article/pii/0169743987800849\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e (verified 7 February 2019)\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"University of Illinois at Urbana-Champaign","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":true,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Egg grading, internal quality, yolk color, consumer preference, EI 4.0","lastPublishedDoi":"10.21203/rs.3.rs-6207045/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6207045/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eNondestructive yolk color detection holds significant potential in the egg and poultry industries due to its critical role in shaping consumer preferences, nutritional perceptions, and marketability. A consistent yolk color may support product branding, quality assurance, and adherence to market-specific standards. Thus, this research study aimed to develop an nondestructive approach using hyperspectral imaging (HSI) combined with partial least squares discriminant analysis (PLS-DA) to separate orange yolk-colored eggs from yellow yolk-colored eggs. A hyperspectral camera in the visible range of 400 nm to 1000 nm was used for the spectral information of the eggs. A total of 146 white eggshell infertile eggs were collected from the poultry farm of the University of Illinois at Urbana-Champaign and were used for the investigation. The total dataset was divided into 70% for training and 30% for testing purposes. A classification model was developed using PLS-DA with various spectral preprocessing techniques. An accuracy of 100% on the testing set was achieved using spectral preprocessing with standard normal variate (SNV), first derivative (FD) and second derivative (SD) data. The key 9 variables (wavelengths) in the classification model were found in the range 520\u0026ndash;680 nm, which indicated the variation in the types and amount of carotenoid pigments deposited in egg yolks, which are influenced by hen feed and metabolic efficiency. These findings suggested that HSI combined with multivariate analysis could be used to grade chicken eggs based on their internal yolk color for the future egg and poultry industry.\u003c/p\u003e","manuscriptTitle":"Nondestructive Classification of Yellow and Orange Colored Yolk Eggs using Hyperspectral Imaging Combined with PLS-DA","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-03-17 10:47:46","doi":"10.21203/rs.3.rs-6207045/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"a4918750-b2e0-4cfe-914d-c3e044891ca5","owner":[],"postedDate":"March 17th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":45548322,"name":"Agricultural Engineering"},{"id":45548323,"name":"Animal Science"}],"tags":[],"updatedAt":"2025-03-17T10:47:46+00:00","versionOfRecord":[],"versionCreatedAt":"2025-03-17 10:47:46","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6207045","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6207045","identity":"rs-6207045","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}