Introduction

Apple has a wide range of applications in the food industry and remains a strong consumer favorite. In 2020, global apple production reached 86,442,716 tons, harvested from 4,622,366 hectares of orchards1. According to TUIK, Türkiye’s total apple production was 4,817,500 tons in 2022, while this amount decreased to 4,602,517 tons in 20232. Apples can be enjoyed in various forms, including dried fruit, juice, jam, or marmalade3,4. Regular apple consumption helps prevent neurodegenerative diseases and lowers the risk of asthma5,6,7,8.

Apples can be consumed fresh or as dried chips. Reducing their moisture content by drying processes can significantly extend their shelf life while maintaining quality9,10,11. While drying offers substantial benefits for processing and storing food products, the use of inappropriate methods—such as excessively high temperatures, prolonged drying durations, or insufficient control of process parameters—can degrade the quality of the final product, lead to excessive energy consumption, and increase production costs12,13. Currently, the main technologies used for drying apple slices include hot air drying, vacuum freeze drying, microwave drying, and infrared drying14,15,16,17. Various processing techniques, including drying, frying, puffing, and cooking, are widely utilized to enhance food shelf stability and develop innovative products18,19. However, each technique has its own advantages and challenges. For instance, conventional hot-air drying is an efficient and cost-effective method but often leads to significant nutrient loss and textural changes due to prolonged exposure to high temperatures. Freeze-drying, on the other hand, is considered a gentle method that preserves heat-sensitive compounds and retains structural integrity, yet it is energy-intensive and costly. Frying enhances flavor and texture but introduces high oil absorption, reducing the overall healthiness of the product. Puffing drying, particularly when combined with pre-treatments like ultrasound, has the potential to improve texture and rehydration properties while minimizing excessive thermal degradation. Understanding these trade-offs is essential for optimizing processing conditions to achieve the best balance between quality retention and process efficiency20. A range of innovative technologies, such as ionizing irradiation21 high hydrostatic pressure22 pulsed electric fields23 and cold plasma24 have been explored primarily as pre-treatments to enhance drying efficiency and improve the quality of dried products. These methods can modify the structural, chemical, or microbial properties of food before drying, thereby influencing drying kinetics and the final product characteristics. However, they are not standalone drying technologies but rather complementary techniques used to optimize conventional drying methods. Apart from these, explosion puffing (EP), also known as instant controlled pressure drop texturing—has emerged as a promising approach for processing various foods25,26. EP has been widely adopted in food processing, with drying being one of its primary applications. Traditional drying methods typically result in slow drying rates and significant energy use, which can cause color deterioration and nutrient depletion due to prolonged heat exposure27. In contrast, EP drying creates a porous structure, providing a crispy texture and improved rehydration properties28,29. In recent years, there have been several attempts to develop a multispectral imaging system to simultaneously estimate several spoilage parameters and determine the quality parameters of some foods30,31. Moreover, investigations on the quality characteristics and microstructures of apples, mangoes, purple sweet potatoes, and potatoes using different tools such as hyperspectral imaging, laser light backscatter imaging during drying have been reported in the literature11,32,33,34. However, existing research lacks comprehensive integration of multispectral imaging with advanced statistical modeling for predicting key quality attributes in dried fruit products. Additionally, while FTIR spectroscopy has been widely used to assess chemical composition, its potential for capturing microstructural and functional changes in dried apples remains underexplored. There is also a need for optimized processing conditions that balance nutritional retention with desirable texture properties. Addressing these gaps, this study aims to combine FTIR spectroscopy and multispectral imaging with chemometric analysis to provide a more precise evaluation of puffing-dried apples, offering novel insights into their physicochemical and bioactive properties. This study provided valuable insights into how puffing drying affects the quality, composition, and structure of apple slices. Integrating multispectral imaging, FTIR spectroscopy, and chemometric analysis demonstrated a comprehensive and non-destructive approach to monitoring drying-induced changes. The findings highlighted significant alterations in color, texture, moisture, and chemical composition, particularly in carbohydrates, lipids, and phenolic compounds, due to different drying techniques. The potential use of PCA and PLS regression to predict quality attributes accurately by using spectral data.

Results and discussion

Physicochemical analysis

Table 1 presents the physicochemical results of the control and PD samples. The puffing-drying method significantly influenced the physicochemical properties of apple slices. The L* (lightness) value of PD (43.62 ± 6.98) was significantly lower than the control samples (75.00 ± 2.76). This indicates that PD apples experienced more browning during processing, likely due to Maillard reactions induced by rapid temperatures and pressure fluctuations, as well as enzymatic browning occurring before inactivation of polyphenol oxidase during drying. Similarly, the a* (redness) value increased significantly in PD apples (12.32 ± 1.75) compared to the control (5.41 ± 1.35), possibly due to the formation of brown pigments via Maillard reactions and caramelization under puffing drying conditions. Conversely, the b* value, associated with yellowness, decreased in PD apples (32.55 ± 3.81) compared to the control (37.73 ± 2.97). This reduction in b* value may be due to the thermal degradation of yellow pigments such as carotenoids, as well as structural changes in the tissue that affect light scattering and color perception.

Table 1 Physicochemical properties and total phenolic content of dried apple samples.
Full size table

The moisture content of the PD samples (7.41 ± 1.29%) was significantly lower than the control (10.34 ± 0.07%), suggesting effective moisture removal during puffing. However, the water activity (aw) of PD samples (0.31 ± 0.009) was slightly higher than the control (0.28 ± 0.003), which may indicate differences in water-binding capacity due to structural changes.

The titratable acidity (expressed as malic acid) and pH of PD apples were significantly higher (0.49 ± 0.08% and 4.83 ± 0.25, respectively) compared to the control (0.37 ± 0.01% and 4.52 ± 0.02) (p ≤ 0.05). This increase may be attributed to the concentration effect caused by moisture loss and potential chemical modifications, such as acid release or changes in buffering capacity, during the puffing process.

Similarly, the soluble solid content (°Brix) was slightly higher in the PD samples (12.45 ± 0.16 °Bx) compared to the control (12.20 ± 0.09 °Bx), likely due to the concentration of sugars as a result of water removal. These findings align with previous studies on PD fruits, where moisture reduction led to an increase in titratable acidity and °Brix values28.

The total phenolic content was lower in PD samples (3307.28 ± 180.75 mg GAE/kg) compared to the control (3650.03 ± 85.50 mg GAE/kg), suggesting that thermal and pressure treatments during puffing may degrade phenolic compounds.

Hardness was significantly (7 times) higher in PD samples (2967.06 ± 304.61) compared to the control (422.06 ± 26.98), indicating a more rigid texture likely due to structural modifications during puffing. However, the fracturability values of both treatments were not significantly different, suggesting comparable breakage behavior.

FTIR analysis

FTIR spectra of the samples were shown in Fig. 1. Functional band regions and assigned bands were summarized for dried apple samples in the Table 2. The broadband (3291 cm− 1) in this region represented hydroxyl groups in both the control and puffed dried samples. Puffing drying resulted in a lower peak intensity compared to the control group, indicating a rearrangement of structural bonds due to the effect of the drying method. The bands at 2920 and 2850 cm− 1, typically associated with methyl and methylene groups, suggest structural modifications in lipids and carbohydrates due to the thermal processing35,36. The slight decrease in peak intensities compared to the control group reflects the disruption of carbon chains. The peak at 1736 cm− 1, representing ester or carbonyl groups linked to lipids and pectin-like compounds37decreased in PD samples. The bands at 1598 and 1411 cm− 1, associated with phenolic compounds and acids, showed altered peak intensities, suggesting partial degradation of polyphenols or organic acids due to the rapid temperature and pressure changes of puffing drying. Furthermore, the peaks at 1239 and 1145 cm− 1, related to carbohydrate structures38exhibited changes, supporting the idea that puffing drying alters carbohydrate composition and molecular arrangements. The peak at 1025 cm− 1 serves as further evidence of modifications in the chemical bonds of carbohydrates during puffing drying. The peak at 1025 cm− 1 can be considered as an indication of changes in the chemical bonds associated with carbohydrate structures during puffing drying. These regions (918, 865, 816, and 776 cm− 1) are associated with amorphous structures or crystalline forms. The bands observed between 1200 and 800 cm− 1 come from the combination of stretching vibrations of C–O, C–C, C–O–H, and bending of C–O–H of carbohydrate absorption39. The pressure and temperature during puffing drying could have led to microstructural alterations in the apple slices, as reflected by changes in these bands. Similarly, the bands in the fingerprint region were successfully used to identify carbohydrates and phenolic compounds to determine the effect of the process in fresh and freeze-dried apple samples40.

Fig. 1
figure 1

Average FTIR spectra of control and puffing-dried (PD) apple slices.

Full size image
Table 2 Functional groups assigned in the FTIR spectra of dried apple samples.
Full size table

The PCA score plot shows a clear distinction between the control and puffed-dried apple samples based on their FTIR spectra. The Eigen analysis of the correlation matrix highlights that PC1 accounts for 92.40% of the total variation, and PC2 adds an additional 6.10%, cumulatively explaining 98.50% of the variability in the dataset. This indicates that the primary variance in the spectral data is dominated by differences induced by the drying process. The drying process substantially alters the chemical and structural attributes of the apple slices, with puffing drying resulting in significant microstructural and compositional changes, as observed in the FTIR spectra (Fig. 1). The PCA indicates that FTIR spectra provide robust differentiation and that PC1 primarily represents moisture reduction and associated chemical rearrangements.

The PLS analysis (Table 3) establishes predictive models for key physicochemical properties using FTIR spectral data. The results reveal strong predictive relationships for most properties, particularly lightness (L*), redness (a*), and hardness, with varying degrees of correlation for other parameters. The PCA and PLS analyses corroborate the FTIR-based characterization of apple slices subjected to puffing drying. The PCA results confirm distinct compositional and structural shifts between the control and PD apple slices, emphasizing differences in moisture content, texture, and color attributes (Fig. 2). Additionally, the PLS models effectively predicted key quality parameters such as color (L, a*, b*), texture (firmness), and moisture content, demonstrating the potential of spectral data in monitoring and optimizing drying conditions. Similar leave-one-out PLS models were successfully developed for the prediction of soluble matters, such as sugars, acids, and phenolic compounds, in the different apple samples40. Moreover, PLSR models have been successfully used to quantify titrable acidity, pH, total phenolic content, and antioxidant levels in sun or oven-dried fruit leathers from apples41.

Table 3 PLS regression analysis for the prediction of dried apple samples using FTIR spectral data.
Full size table
Fig. 2
figure 2

(a) PCA score plot of the FTIR spectra of control and puffed-dried (PD) apple samples. (b) PLS regression plot of actual versus predicted for a* obtained by FTIR.

Full size image

Multispectral imaging

The average spectral data (Fig. 3) indicated noticeable differences in mean intensity between the control and puffing dried apple samples in the UV-NIR region. These differences reflected changes in chemical and physical properties induced by the puffing drying process. In the UV–Vis region (360–700 nm), the control samples exhibited higher mean intensity compared to the PD samples. This could be that puffing drying resulted in the degradation or transformation of light-absorbing compounds such as polyphenols, and other pigments which were known to absorb in the UV-Vis region and sensitive to heat or mechanical stress during drying. In the transition region (650–750 nm), the difference in intensity between the two treatments were less, so structural or chemical changes were minimal. In the NIR region (750–960 nm), PD samples showed slightly higher mean intensity than the control. The NIR region was often associated with combination bands of O–H, C–H, and N–H bonds. The increased reflectance in this region was related to moisture loss, microstructural changes, or alterations in carbohydrate and protein content due to puffing drying. The higher intensity in PD samples could also reflect increased light scattering caused by porous structures formed during puffing drying42. Multispectral imaging (Fig. 4) revealed noticeable differences in color and surface texture of control and PD apple slices, confirming the visual impact of structural and compositional changes induced by the drying process.

Fig. 3
figure 3

Average UV-NIR spectra of control and puffing-dried (PD) apple slices.

Full size image
Fig. 4
figure 4

Multispectral images of control (a) and puffing-dried apple (b) slices.

Full size image

The PCA score plot (Fig. 5) of the UV-NIR spectral data further emphasized the clear separation between the two sample groups, suggesting substantial structural and compositional modifications. The strong loadings associated with color-related spectral regions indicate significant pigment alterations, likely due to Maillard reactions and caramelization.

Fig. 5
figure 5

(a) PCA score plot of the UV-NIR spectra of control and puffed-dried apple samples. (b) PLS regression plot of actual versus predicted for a* obtained by videometer data.

Full size image

The PLS regression results (Table 4) demonstrated high predictive accuracy for L* (RCV = 0.90448) and a* (RCV = 0.86380), supporting the effectiveness of multispectral imaging in assessing color changes. Hardness exhibited one of the highest predictive correlations (RCV = 0.95382), confirming the structural transformations induced by puffing drying. Moisture content and water activity were also well predicted (RCV = 0.83057 and RCV = 0.87611, respectively), reflecting the drying efficiency. However, parameters such as pH and titratable acidity showed relatively lower predictive power (RCV < 0.52), likely due to their weaker spectral signatures in the measured range. These attributes may not exhibit strong absorbance in the spectral regions analyzed, leading to limited model accuracy. Despite these limitations, the integration of spectral analysis with multivariate modeling remains a valuable approach for non-destructive quality assessment in food processing. While it performs well for predicting parameters strongly associated with structural and compositional changes (e.g., color, texture, and moisture content), its accuracy for pH and acidity measurements may require complementary techniques such as direct chemical analysis or extended spectral range coverage to enhance predictive performance.

Table 4 PLS regression analysis for the prediction of dried apple samples using videometer data.
Full size table

Conclusion

The study demonstrated that the puffing drying method significantly influences the physicochemical, structural, and compositional properties of apple slices. Puffing drying resulted in increased textural attributes such as hardness and color properties such as redness, along with reduced moisture content, compared to the control samples. However, it also led to the browning and degradation of phenolic compounds, likely due to thermal and pressure effects during processing. FTIR spectroscopy revealed clear structural modifications, particularly in carbohydrates and lipids, and a decline in total phenolic content, indicating thermal degradation. PCA analysis further confirmed the marked differences between control and puffed-dried samples, with PC1 capturing most of the variability. PLS models effectively predicted key quality parameters, emphasizing the utility of FTIR data for monitoring drying-induced changes. Collectively, these findings underscore the potential of puffing drying to enhance certain textural and color properties of apple slices while also highlighting the importance of optimizing process conditions to mitigate nutrient losses and maintain desirable product quality. Multispectral imaging and UV-NIR spectroscopy, combined with chemometric techniques, effectively captured the physicochemical changes, confirming reduced lightness and increased heterogeneity in apple slices subjected to puffing drying. PCA successfully differentiated control and puffed samples, while PLS regression provided robust predictive models for key quality attributes such as color, moisture content, and hardness. Chemometric analysis, including PCA and PLS regression, successfully distinguished between control and treated samples and accurately predicted the quality parameters from spectral data. To the best of our knowledge, this was the first report to integrate FTIR spectroscopy and multispectral imaging with chemometric analysis to comprehensively assess microstructural and functional changes in puffed-dried apple slices. However, this study showed that spectral imaging, such as FTIR without any imaging techniques, did not always provide sufficient information to fully determine various quality parameters in puffed-dried apples. These results highlighted the use of multispectral imaging, supported by chemometric tools, as a powerful, non-invasive method for evaluating quality in dried fruit processing. This methodology could be further developed for real-time, in-line quality control during food processing. With proper optimization, such spectral and imaging technologies may enable faster, more efficient monitoring across various dried fruit products, contributing to improved consistency, safety, and consumer acceptance in industrial applications. Future studies should focus on combined use of imaging and spectral techniques to further improve the predictive accuracy of chemical parameters and textural properties.

Materials and methods

Materials

In this study, apples, commonly produced in the Niğde region and cultivated in various areas, have been used as the drying material. The Red Chief (Malus domestica Borkh.) variety, which is widely grown in Niğde, has been sourced from the Faculty of Agricultural Sciences and Technologies at Niğde Ömer Halisdemir University and stored at + 4 °C in the Food Engineering Laboratories of the university.

Methods

Drying

Apples were washed with tap water to remove surface contaminants, then sliced into 3 mm-thick pieces. The sliced apples were then placed in an ultrasonic bath (Kudos, China) operating at 37 kHz frequency and 100% ultrasonic power in pure water at room temperature for 30 min to preserve color, increase dehydration rate, increase hardness, and crispness. The samples were filtered using a food strainer, gently dried with a paper towel, and then subjected to a preliminary drying step. In this step, apple slices were dried in a hot air oven (Isotex, Türkiye) at 70 °C until their moisture content was reduced to 40%, which took approximately 6–7 h. Afterward, the pre-dried apple slices were puffed using the puffing drying method at 50 °C for 10 min under a pressure of 5 bar. The parameters used in puffing drying were optimized before (data not shown). Finally, the puffed apples underwent a second drying step in a hot air oven at 70 °C until the moisture content reached 10%, which took approximately 1–2 h. Apples that were dried solely in a hot air oven at 70 °C up to 12 h, without any prior treatment, served as the control group.

Physicochemical analysis

The dried apple samples were finely ground (Lavion, Türkiye), and 0.5 g of the resulting powder was mixed with 20 mL of distilled water. The mixtures were allowed to equilibrate at 20 °C for one hour to ensure uniform temperature distribution and stabilization before analysis. For titratable acidity determination, the solutions were titrated with 0.1 M NaOH until the pH reached 8.1, using a pH meter (Hanna Instruments, Italy). The acidity was expressed as grams of malic acid per 100 g of dry sample. After sample preparation, the pH values of the samples were measured with a pH meter (Hanna Instruments, Italy). The water-soluble dry basis content of the samples was determined by ABBE refractometer (A.Krüssoptronic Gmbh AR4, Germany) at room temperature. The results were expressed in degrees brix (°Bx). The moisture content of the samples was determined using the standard gravimetric method by drying in an oven at 105 °C for 24 h. The results were expressed on a dry basis43. Water activities (aw) of control and dried apple samples were determined at room temperature using a water activity analyzer (AXIS, Türkiye). The color values of the control and puffing-dried (PD) samples were determined with a colorimeter (Konica Minolta CR400, Japan) according to the CIELab color scale and 3 measurements were taken from different regions of each sample. Total phenolic content (TPC) was determined according to the Folin-Ciocaltaeu method44. Phenolic compounds were extracted from the samples using an 80% methanol solution containing 1% HCl. The mixture was shaken at 40 rpm for 4 h and then centrifuged (Nüve NR 800R model, Türkiye) at 9000 g at 4 °C for 15 min to separate the clear fraction. For analysis, 0.75 mL of Folin-Ciocalteu reagent (10% in water) was added to 100 µL of the sample and left at room temperature for 5 min. Subsequently, 0.75 mL of Na₂CO₃ solution (75 g/L in water) was added and mixed thoroughly. The mixture was kept at room temperature in the dark for 1 h, followed by measurement of the absorbance at 725 nm. The same protocol was followed for gallic acid solutions of varying concentrations to construct a calibration curve. The absorbance values of the extracts were compared against the calibration curve, and the TPC was expressed as gallic acid equivalent (mg GAE/kg)45. Texture analysis was conducted using a TA-XT Texture Analyzer (Stable Micro Systems, UK) equipped with a P/0.25 spherical tip designed for dry fruits. The pre-test speed, post-test speed, test speed, and strain were set at 3 mm/s, 10 mm/s, 3 mm/s, and 40%, respectively. Hardness and fracturability values were measured during the penetration test46.

FTIR spectroscopy

Both control and PD samples were freeze-dried in a freeze dryer (Scanvac Coolsafe 95–15 Pro, Denmark) to prevent the functional bands from being closed by the water band remaining in the samples or absorbed by the samples after drying, and then powdered. The absorption spectra of the samples were obtained in the 400–4000 cm− 1 region using FTIR spectroscopy (Bruker, Germany) with an ATR cell and 128 scans at 2 cm− 1 resolution. Raw spectra were vector normalized for further analysis using the appropriate program (OPUS, v.8.1, Bruker Optics, Germany).

Multispectral imaging

Multispectral imaging (MSI) of the samples was conducted using a VideometerLab system (Videometer A/S, Hørsholm, Denmark). This system operates across 18 discrete wavelengths (375, 405, 435, 450, 470, 505, 525, 570, 590, 630, 645, 660, 700, 850, 870, 890, 940, and 970 nm), covering the visible to near-infrared (NIR) spectrum. The samples were placed inside a custom-designed integrating sphere coated with matte titanium paint to ensure uniform light distribution. A digital camera equipped with a monochrome charge-coupled device (CCD) sensor was positioned above the sphere to capture surface reflectance data. Light-emitting diodes (LEDs) arranged along the sphere’s rim provided narrowband spectral illumination, optimizing dynamic range while minimizing gloss artifacts, shadows, and specular reflections. Each LED strobe generated a monochromatic image with 32-bit floating-point precision and a spatial resolution of 2056 × 2056 pixels.

Regions of interest (ROI) segmentation was performed by exploiting the spectral differences between the various components of the sample. Following the segmentation process, spectral data were extracted through computer processing, which involved identifying and isolating spectral features from the segmented images. During segmentation, the sample was transformed into a segmented image, with a single, flat section of the apple selected as the primary ROI for spectral data collection. This approach ensured accurate and focused analysis of the sample’s key characteristics.

Chemometric analysis

Principal component analysis (PCA) and partial least squares (PLS) regression analyses were performed using Minitab 18 (Minitab Inc.). PCA was used to differentiate control and PD samples, and the results were presented as score plots. Total acidity, pH, water-soluble dry basis content, moisture content, color, and TPC of PD samples were estimated by PLS regression using FTIR spectral data and multispectral data. The spectral data were used as predictor variables (X), while the measured (color, moisture, water activity, titratable acidity, pH, °Brix, total phenolic content, hardness) served as the response variables (Y) in the development of Partial Least Squares (PLS) models. The PLS models were constructed to establish predictive relationships between spectral data and the analytical results, following the methodology outlined in previous studies47. Model performance was evaluated using cross-validation and statistical parameters such as the coefficient of determination (R²) and root mean square error (RMSE). Cross-validation (leave-one-out) was applied in all analyses to evaluate the calibration models, ensuring their robustness and ability to generalize effectively.

Statistical analysis

The statistical evaluation of the data obtained from the PD samples was obtained by comparing the differences between the means of the analysis results according to the treatments. The analysis was conducted using the Minitab Statistical Program (version 17, Minitab Inc., State College, PA, USA), and the Tukey multiple comparison test was performed to assess statistical differences between the means (p ≤ 0.05).