Vis/NIR reflectance spectroscopy for non-destructive diagnosis of Fusarium spp. infection in postharvest potato tubers (Solanum tuberosum)

Conventional methods of detecting Fusarium spp. infection, which causes significant economic losses in potato production, are time-consuming and expensive. This study explored rapid and non-destructive detection techniques using visible/near-infrared (Vis/NIR) spectroscopy. Potato seeds of the Granola L variety were intentionally infected with Fusarium spp. by fungal inoculation, then stored at 12°C, 25°C, and a combination of both. Healthy potatoes were stored under the same conditions in containers for 30 days and monitored every five days. Principal component analysis-linear discriminate analysis (PCA-LDA) was used to classify potato tubers based on their infection status. PCA-LDA analysis revealed significant spectral differences between healthy and infected potato seeds across all storage temperatures. Calibration reliability values were 95.87% (for samples stored at 12°C), 97.52% (stored at 25°C), and 98.35% (for the combination of 12°C and 25°C). Similar trends were observed for accuracy: 91.96% (12°C), 98.29% (25°C), and the highest accuracy of 98.65% for the combined temperature. These techniques facilitate rapid identification of infections, aiding farmers and producers in implementing more efficient preventive actions, resulting in decreased crop losses and waste products and enhanced productivity in the agricultural sector.


Introduction
Potatoes are globally recognized as the fourth most cultivated staple crop, contributing to food security through their high yield and nutritional value.They constitute a crucial source of carbohydrates, proteins, and vitamins for human nutrition.The worldwide production of potatoes reached a significant 376 million tonnes in the year 2021 based on data from FAO (Food and Agriculture Organization).Notably, China emerged as the leading potato-producing nation with a staggering output of 94 million tonnes, closely followed by India with 54 million tonnes.The FAO further reveals that the total area utilized for potato cultivation globally in 2021 amounted to 18,132,694 hectares, which yielded a substantial production volume reaching 378 million tons in 2021.Potatoes stand as one of the world's largest crops.Approximately 5% of this production is dedicated to seedlings [1].Various approaches have been proposed to assess potato quality during commercial production, harvesting, and storage.
In the last 10 years, Fusarium spp.dry rot has emerged as a significant potato disease, impacting tubers in storage and seed pieces post-planting, causing crop establishment reduction by terminating developing potato sprouts, leading to losses of up to 25 percent, with over 60 percent of tubers getting infected during storage [2].Conventional approaches to evaluating potatoes' internal quality are frequently damaging and inefficient.An ideal system for assessing quality in practical potato cultivation should be precise, rapid, and cost-effective.Visible/near-infrared spectroscopy (Vis/NIR) techniques are popular due to their non-destructive procedure, efficiency, speed, accuracy, absence of pollution, and affordability.Vis/NIR radiation is the electromagnetic spectrum between 350 and 2500 nm.These Vis/NIR spectroscopy techniques make it possible to analyze the intricate composition of food samples by acquiring spectra of reflection, absorption, and transmittance at specific wavelengths [3].The spectral characteristics change due to scattering and absorption processes.The change depends on the product's composition and light-scattering properties.Statistical techniques are used to extract information from the spectra.Vis/NIR spectroscopy is a non-destructive technique used to measure quality attributes of fruits and vegetables [4].
One potential use for visible/near-infrared (Vis/NIR) spectroscopy lies in its ability to ascertain the physical attributes of potatoes [5][6][7].Employing this analytical technique makes it possible to evaluate and assess the overall quality of potatoes by analyzing their spectral data.Near-infrared spectroscopy (NIR) has effectively demonstrated the ability to forecast the proportion of overall fungal presence and the occurrence of Fusarium verticillioides infection in maize kernels [8].NIR technology additionally permits the assessment of the overall concentration of soluble solids and the measure of titrable acidity in tomatoes [9].Estimating dry matter and reducing sugars in intact potato tuber has been successfully demonstrated using the MicroNIR portable device, showcasing NIR spectroscopy's effectiveness and efficiency in this particular application [2].
Therefore, the primary objective of this study is to evaluate the potential of cost-effective and adaptable modular Vis/NIR spectrometers for detecting internal damage in potatoes.Our investigation focuses on the accuracy and effectiveness of Vis/NIR spectroscopy in combination with principal component analysis (PCA)-linear discriminant analysis (LDA) for identifying Fusarium spp.infection in potatoes.We anticipate that our proposed methodology will significantly improve postharvest quality management and reduce losses in potato production.

Potato tuber samples
Two hundred and forty (240) potato tubers cultivar Granola L of uniform size were cleaned and dried before analysis.Then, they were sterilized with a 1% NaClO solution for 10 minutes, rinsed with distilled water, and dried again [10].Standard cuts of 30 mm × 30 mm were made on all tubers using sterile electric drills.The fungal pathogen Fusarium spp. was prepared using the cork borer wounding (CBW) method [11].One hundred and forty samples (140) were infected with the Fusarium spp.inoculum, while one hundred samples (100) were kept as the control group.Three storage variations are used: 12°C, 25°C, and combined temperature (combination of 12°C and 25°C).The temperature combination is a treatment of potatoes that are stored at 12°C for 10 days and then transferred to 25°C room temperature for up to 30 days.The relative humidity is maintained at 85% for all variations.The potato tubers are stored in containers for thirty (30) days in a dark storage facility and observed every five (5) days.
Figure 1 shows the process of acquiring spectra, which consists of performing spectra capture settings.To gather reflectance spectra in the Vis-NIR and SW-NIR regions, we employed Ocean View 2.0.12 software.The integration time for the spectra acquisition process was configured to be 470 ms (Vis-NIR) and 800 ms (NIR); the average scans were set at 20 (Vis-NIR) and 12 (NIR), and A boxcar width of 1, respectively.Each sample was subjected to ten spectra acquisitions.

Chemometrics
PCA is a chemometric method that employs an unsupervised approach to reduce the dimensionality of a dataset with many features.This dimension reduction makes the dataset more manageable for analysis and enhances spectral visualization.LDA is commonly used in supervised classification tasks.In the case of PCA-LDA, LDA was applied to the collected PCA data, enabling the classification of noninfected and infected potato tubers.Using PCA-LDA, an evaluation was conducted to assess the effectiveness of the classification model through prediction.The results obtained from these predictions highlighted the benefits of the PCA-LDA model.

Vis/NIR spectral data of potato tuber
This study uses a modular spectrometer with two probes, one for visible and short-wave near-infrared (Vis-NIR) light with wavelengths of 400-1000 nm and the other for near-infrared (NIR) light with wavelengths of 900-1700 nm.In Vis/NIR spectroscopy, after plotting a spectrum from 350 to 1100 nm, the wavelength ranges below 400 nm and above 1000 nm are often noisy or show interference, so it is expected to cut them manually.Decreasing wavelength points can speed up spectral detection but also reduce resolution and slightly affect the model's accuracy [12].
The original spectra for Noninfected (N) and infected (I) potato tubers during various storage (12°C, 25°C) and combined temperatures (combined storage at 12°C and 25°C) are shown in Figure 2 (a), 2 (c), and 2 (e).All spectra have a similar pattern, with an absorbance at 480 nm associated with carotenoids, the most important pigments in potatoes [13,14].A lower reflectance at 670 nm suggests chlorophyll in noninfected (N) potatoes.This means they absorb more energy in this spectrum than infected (I) ones.This aligns with prior research showing higher chlorophyll content in healthy tomatoes than those infected with Fusarium spp.[15].The water content in the samples caused significant absorbance peaks at 995 nm, 1210 nm, and 1435 nm [16,17].The absorbance of noninfected (N) potatoes was higher than that of infected (I) potatoes at 990 nm and 1210 nm, which could be attributed to their higher water content.show the spectra after preprocessing them with Savitzky-Golay 2 nd Derivative/SGD2, which enhanced spectral resolution and revealed critical information about potato parameters, such as sugar content and pH.Infected (I) and Noninfected (N) potatoes show significant variation in absorbance between 400 -700 nm and 700 -950 nm.High reflectance at 690 nm in potato spectra correlates with chlorophyll, while reflectance at 770 nm correlates with fructose and glucose [7].A significant increase in peak absorption resolution between 1065 and 1335 nm represents the sugar content of glucose and fructose ) [18].The factors of harvest age and postharvest handling affect the sugar content of glucose in potatoes.This study found that potatoes stored at 12°C lasted 30 days, while those stored at 25°C spoiled by day 15 [5].Other absorption peaks or reflectance are seen at wavelengths of 1370 nm.SW-NIR bands in the 1355-1400 nm region, seen in all three potato spectra with different storage temperature treatments, represent the second overtone of the C-H combination (carbohydrates, fats) [19].

PCA Result
Figure 3 (a) shows a score plot representing the differentiation between groups of potato samples that were noninfected (N) and infected (I) with Fusarium spp.fungus, achieved using PCA with PC-2 (13%) and PC-5 (2%).PC-5 can group potatoes at storage temperatures of 12°C without fungal infection (N-12) in PC less than 0 (PC<0) with potatoes infected with fungi (I-12) in PC greater than 0 (PC>0).Noninfected potatoes tend to be PC-5 negative in quadrants 3 and 4. PC-2 and PC-5 spectra show different absorbance and reflectance, with the pattern in PC-5 being more significant and more straightforward to identify.PC2 and PC5 capture important data features that are easier to interpret than others.Based on the loadings of PCA shown in Figure 3 (b), the discrimination and diagnosis of infected potato tubers with Fusarium spp.may be attributed to the wavelengths 405 nm, 407 nm, 436 nm, 545 nm, 121 nm, and 1163 nm.The absorptions at wavelengths of 405 nm, 407 nm, and 436 nm exhibit a connection to carotenoids [13].At a wavelength of 545 nm, other peaks, influenced by carotenoids, anthocyanins, and chlorophyll, can be observed.To differentiate pathogenic reactions from healthy potatoes, such as the Fusarium spp.fungus effect, Omori et al. [20] employed absorbances at a wavelength of 405 nm in their analysis.Additional absorbance and reflectance peaks at wavelengths 1121 nm and 1163 nm demonstrate a correlation with second overtone O-H combinations [21].Figure 4 (a) depicts a score plot using PC-1 and PC-5, suggesting a possible distinction between potato samples free from fungal infection and those affected by fungal infection.Potatoes without fungal infection tend to be PC-5 negative in quadrants 3 and 4, while a PC-5 positive reading indicates that potatoes are infected with the Fusarium spp.fungus in quadrants 1 and 2. Figure 4 (b) shows the wavelength 1140 nm is likely related to the stretching of C-O and C-C bonds, harmonic spectra of both C-H and C-O-H bonds, and carbohydrate deformation in fungal cell walls [22].The phenomena are consistent with the effect of Fusarium spp.The sugar content parameter analysis shows fungus inoculation on artificially infected potato samples.The wavelength 1380 nm is likely correlated with a combination of stretching and deformation of the C-H cluster [23].PCA methods for grouping potato samples in combined temperature for classification using PC-1 (42%) and PC-2 (15%)) are shown in Figure 5 (a).However, the plots show overlapping patterns, which does not allow for clear grouping.From Figure 5 (b), according to loadings, can be identified wavelengths likely related to the analysis of sugar content (glucose, fructose, sucrose, and inverted sugar) and detection of Fusarium spp.fungi in potatoes at 960-1100 nm and 1420-1600 nm.These wavelengths indicate stretching of both N-H and first overtones O-H observed at 960-1100 nm (reflectance) and 1420-1600 nm (absorbances), which might related to the protein and water content of the sample [24].Table 1 and Figure 7 (b) present the confusion matrix for model development using calibration sample sets (n = 121) and prediction sample sets (n = 44) for potato samples stored at 25°C.Savitsky-Golay's 1st Derivative preprocessing spectra yielded the best classification model for potato samples stored at 25°C.The PCA-LDA analysis results indicate that samples with and without fungal infection were correctly classified in potatoes (n = 55 and n = 63, respectively) with an accuracy of 97.52% and a reliability of 98.29%.When tested using prediction sample sets, the PCA-LDA model correctly classified samples with a prediction set accuracy of 95.45% and a prediction set reliability of 95.83%.This indicates that the PCA-LDA model is reliable.This is illustrated in Figure 7 (b), which shows the PCA-LDA modelling results for potato samples without and with fungal infection stored at 25°C.Table 1 and Figure 8 (b) show the confusion matrix development of a PCA-LDA model of potato using combined temperature, using the best spectral analysis of the Savitzky-Golay 2 nd Derivatives preprocess.The results of PCA-LDA analysis for the calibration set (n = 122) showed that the samples had been correctly classified in potatoes without fungal infection (n = 70) and potatoes infected with

Figure 1 .
Figure 1.Spectra acquisition schema for potato tuber sampling

Figure 2 .
Figure 2. Reflectance spectra of potato tuber (a) original spectra and (b) SGD2 spectra of 12°C, (c) original spectra and (d) SGD2 spectra of 25°C, (e) original spectra and (f) SGD2 spectra stored at 12°C and 25°C Figures 2 (b), 2 (d), and 2 (f) show the spectra after preprocessing them with Savitzky-Golay 2 ndDerivative/SGD2, which enhanced spectral resolution and revealed critical information about potato parameters, such as sugar content and pH.Infected (I) and Noninfected (N) potatoes show significant variation in absorbance between 400 -700 nm and 700 -950 nm.High reflectance at 690 nm in potato spectra correlates with chlorophyll, while reflectance at 770 nm correlates with fructose and glucose[7].A significant increase in peak absorption resolution between 1065 and 1335 nm represents the sugar content of glucose and fructose )[18].The factors of harvest age and postharvest handling affect the sugar content of glucose in potatoes.This study found that potatoes stored at 12°C lasted 30 days, while those stored at 25°C spoiled by day 15[5].Other absorption peaks or reflectance are seen at wavelengths of 1370 nm.SW-NIR bands in the 1355-1400 nm region, seen in all three potato spectra with different storage temperature treatments, represent the second overtone of the C-H combination (carbohydrates, fats)[19].

Figure 5 .Figure 6 .Figure 7 .
PCA result of potatoes using combined temperature: (a) score plot and (b) loadings of PC-1 (42%) and PC-2 (15%)3.3.PCA -LDA AssessmentDeveloping a model that combines Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) plays a significant role in the process of determining the ideal count of principal components (PCs) to be employed within the declared model.The optimal number of PCs for PCA-LDA modelling of potatoes with varying storage temperatures differs.In the 12°C storage potatoes shown in Figure6(a), seven PCs explain 86% of the cumulative variance.These seven PCs can account for over 86% of the data variability.In the 25°C storage potatoes shown in Figure7(a), five PCs explain 89% of the cumulative variance.In the combined temperature potatoes shown in Figure 8 (a), six PCs explain 78% of the cumulative variance.Therefore, in developing the PCA-LDA model for potato storage temperatures of 12°C, 25°C, and combined temperature, 7 principal components (PCs), 5 PCs, and 6 PCs are input variables.PCA-LDA model assessment of potatoes at storage temperature 12 o C IOP Publishing doi:10.1088/1755-1315/1317/1/0120127 PCA-LDA model assessment of potatoes at storage temperature 25 o C (a) (b) Figure 8. PCA-LDA model assessment of potatoes using combined temperature Table 1 and Figure 6 (b) illustrate the confusion matrix of the model development outcomes when employing calibration sample sets (n = 121) and prediction sample sets (n = 44) for the storage of potato samples at 12°C.The most effective model for categorizing potato samples stored at 12°C was the Savitzky-Golay 2 nd Derivative preprocessing spectra.The findings indicate that all samples were accurately classified as either potatoes without fungal infection (n = 57) or potatoes infected with fungi (n = 59), resulting in an accuracy of 95.87%.Regarding the PCA-LDA model's reliability, the classification of potato samples stored at 12°C using a calibration set yielded a value of 91.96%.This measure of reliability in PCA-LDA assesses the consistency with which the model can classify data, thereby indicating its reliability.Furthermore, this PCA-LDA model, which was trained using a calibration sample set (n = 121), was also tested using a prediction sample set (n = 44).The results demonstrated that the model could accurately classify samples in the prediction set with an accuracy of 93.18% and a reliability of 92.68%.The result confirms the reliability of the PCA-LDA model.