Food Analytical Methods最新文献

The physical and chemical properties of coffee beans are drastically changed during the roasting process. The quality of coffee beans is directly determined by their color, texture, and brightness. The roasting level is essential for food safety, commercial value, and product quality. Traditional classification techniques are time-consuming, expensive, and prone to human error. These qualities have generated interest in deep learning. The features extracted from the Vision Transformer (ViT), which can understand advanced global contexts, enable detailed and hierarchical analysis of images. The aim of this study was to classify features extracted from images of the roasting levels of coffee beans using ViT and machine learning (ML) models. Two publicly available datasets were used for the study, including images of coffee beans in grayscale (Dataset1) and RGB (Dataset2). Random Forest (RF), Decision Tree, and K-Nearest Neighbors models were used to classify the features extracted by the BEiT, FlexiViT, MobileViT, and DeiT models. An Adaptive ViT feature fusion, combining features from ViT models, was also evaluated with these ML models. The experimental results revealed that the Adaptive ViT-RF achieved high accuracy rates of 0.825 on Dataset1 and 0.992 on Dataset2. The SHapley Additive exPlanations (SHAP) analysis was performed to evaluate the decision mechanism of the model. It was found that high feature values in the Green and Light classes and low feature values in the Dark and Medium classes were effective in classification. These results indicate that the Adaptive ViT-RF approach successfully and accurately classifies the roast level of coffee beans.

A novel approach for extracting curcumin was introduced, which integrates the advantages of microextraction with emulsive liquid–liquid and solidification of floating organic droplets processes. The extraction solvent, 1-undecanol, and deionized water, were mixed to generate an emulsion prior to their rapid injection into the sample solution. To improve the phase separation, centrifugation was used. The target analyte-containing organic drops that floated were then collected for analysis of curcumin after they solidified by cooling. The experimental variables that influence the extraction performance were explored and adjusted. Quantitative extraction was attained at pH 7.0 utilizing 100 μL of 1-undecanol as extraction solvent emulsified in 400 μL of deionized water, with the addition of 0.3 mol L⁻¹ NaCl. The procedure is selective in the presence of concomitant ions and dyes. The procedure exhibits linearity up to 2000 μg L⁻¹, with an enrichment factor of 19.8 for 10.0 mL sample. The detection and quantification limits were 1.6 μg L⁻¹ and 4.8 μg L⁻¹, respectively. The inter-day and intra-day precision varied between 1.7 and 2.6%. Ultimately, the developed approach succeeded in quantifying curcumin at a trace level in herbal tea and spice samples, with a recovery rate ranging from 95.0 to 101.0%.

Graphical Abstract

Volatile flavor compounds in rice wine were analyzed using gas chromatography-ion mobility chromatography (GC-IMS). The optimal analytical conditions were determined by systematically optimizing the incubation parameters of Huangjiu. The results revealed that the optimal incubation conditions were 60 °C, 15 min, and 0.6 g NaCl. Furthermore, 31 volatile compounds were identified, including esters, alcohols, aldehydes, ketones, and miscellaneous monomers. Subsequently, semiquantitative analysis showed esters were the predominant group, followed by alcohols. Comparative analysis of different Huangjiu varieties revealed distinct differences and underlying similarities in their characteristic volatile flavor compounds, providing valuable theoretical insights into the differentiation of Guangdong Huangjiu varieties.

Graphical Abstract

Rhus coriaria L. (sumac) is an important plant for both its recognized use as a spice and as a rich source for bioactive compounds. However, its phytochemical profile is sensitive to climate factors of its geographical location, making quality standardization a major challenge. This research therefore aims to quantify the effect of climate variables (altitude, temperature, and precipitation) of different geographical locations on the chemical makeup of sumac. Additionally, it aims to provide climate-calibrated multivariate chemometric discrimination of the fruits from different locations based on selected physicochemical traits. For this purpose, 80 mature fruit samples of sumac collected from 16 geographically distinct locations across the Kurdistan Region of Iraq were analyzed for eight quality markers: refractive index (RI), pH, titratable acidity (TA), electrical conductivity (EC), total dissolved solids (TDS), oxidation–reduction potential (ORP), total phenolic content (TPC), and radical-scavenging activity (RSA). Multivariate analysis employing principal component analysis (PCA) reduced the data to two components that explain 89.7% of total variance: PC1 (70.3%) separates low-temperature, high-altitude sites (high pH, low TA, EC, TDS, and ORP) from lower altitude locations (low pH, high TA, EC, TDS, and ORP), while PC2 (19.4%) orders samples along an altitude-driven phenolic gradient. Hierarchical clustering analysis (HCA) yielded four main clusters that coincide with altitude and temperature gradient, confirming geographical discrimination. Temperature is the dominant quantitative driver for seven of the eight traits (p ≤ 0.007), while precipitation modulates bioactive traits, and altitude contributes through interactions. The study provides the first quantitative climate–chemistry map for Kurdish sumac and a chemometric tool for geographical discrimination and selection of optimum environmental conditions.

Honey is a natural sweetener produced by honeybees and is widely appreciated by consumers because of its multiple beneficial properties. Because of its high value, honey is placed as a targeted product for fraudulent practices. In this work, LC-LRMS fingerprinting was employed for classifying honey samples from 10 countries. Good classification and prediction performance were achieved based on a classification decision tree by consecutive paired PLS-DA models using a hierarchical model builder (HMB), obtaining sensitivity and specificity values higher than 83.3% and 92.6%, respectively, except for the case of China versus Japan. Tentative association of some phenolic compounds was accomplished, which provides useful chemical markers for country discrimination. For instance, methoxyphenylacetic acid, previously identified in New Zealander honeys, was tentatively annotated to m/z 165.0, detected in honey from New Zealand and Australia. The prediction of “unknown” samples was successful for most cases, obtaining sensitivity and specificity values of 100% for most countries. Good classification based on the continent of production was also accomplished, obtaining perfect discrimination among samples produced in Oceania and good classification performance was observed in Asian and European samples. Finally, the obtained fingerprints demonstrated to be useful chemical descriptors to quantify, as a proof of concept, adulterated Spanish honey with honey from Italy, China, and Serbia using partial least squares (PLS) regression, obtaining internal and external validation prediction errors lower than 23%.

The extensive use of antibiotics in dairy farming has raised significant concerns regarding the potential presence of their residues in milk. This poses serious health risks, including allergic reactions and the emergence and development of antimicrobial resistance (AMR). The study is aimed at the quantification of antibiotic residues, specifically oxytetracycline, penicillin G, sulfadiazine, and tetracycline, in raw cow milk collected from Adama, Ethiopia. A multiresidue LC-MS/MS analytical method was optimized and validated according to the standards set by the European Commission (EU 2021/808). Sample preparation was performed by solvent extraction that contains McIlvaine buffer (pH = 4) followed by solid phase extraction (SPE) that utilized Oasis® Hydrophile-Lipophile Balance (HLB) cartridges as well as extracting the antibiotic residues by methanol. Excellent performance characteristics of the method were demonstrated in terms of recovery rates and calibration linearity within the range from 0 to 250 μg/kg concentration. A total of 162 raw milk samples randomly collected from intensively and semi-intensively managed dairy farms in Adama, Ethiopia, were analyzed, with 8% of them testing positive above the decision limit for residues. Notably, penicillin G and oxytetracycline were detected in 5.6% and 2.5% of the samples, respectively. The concentrations found range from 13.15 to 142.38 μg/kg, which exceed the maximum residue levels (MRLs) established by EU Commission Regulation no. 37/2010 standards in several samples. The results raised public health concerns, especially with milk samples exceeding MRLs permitted. This is due to the potential health risks earlier highlighted. Therefore, the study findings necessitate the need for strict enforcement of regulatory frameworks and improved veterinary antibiotic use practices. Such measures will mitigate risks to public health and ensure food safety. Moreover, ongoing monitoring and consumer education are essential to effectively address the antibiotic residue issue.

Foodborne diseases caused by bacterial pathogens are of major public health and zoonotic concern. The objective of the present study was to develop real-time TaqMan PCR assays for the detection and quantification of important bacterial foodborne pathogens, including Clostridium perfringens, Staphylococcus aureus, and Salmonella spp. For this purpose, primers and probes were designed from conserved regions of the target genes (cpa for C. perfringens, nuc for S. aureus, and invA for Salmonella spp.) and TaqMan assays were standardized. The analytical sensitivity of the developed real-time TaqMan assays using gel-purified PCR amplicons was determined to be 2.8 copies/μL, 3.5 copies/μL, and 7.0 copies/μL of DNA for C. perfringens, Salmonella spp., and S. aureus, respectively. The analytical sensitivity of the TaqMan assays was 10- to 1000-fold higher than that of conventional endpoint PCR. The standard curves showed good linearity with R² = 0.99 for all the pathogen-specific TaqMan assays developed and the assays were found to be reliable and reproducible. In spiking studies, the limit of detection (LoD) of the developed TaqMan assays under un-enriched conditions was 1.2 × 10⁵ CFU/g, 3.2 × 10⁸ CFU/g and 3.3 × 10⁴ CFU/g of meat for C. perfringens, Salmonella spp., and S. aureus respectively. After 6 h of enrichment, the LoD considerably improved to 1.2 CFU/g, 320 CFU/g, and 3.3 CFU/g of meat, respectively. The study highlights the need and importance of the enrichment step in the detection of FBPs. The developed real-time TaqMan assays may serve as rapid laboratory tools for the detection and quantification of C. perfringens, S. aureus, and Salmonella spp. in meat.

The formation of crack kernels, which compromises rice quality, is influenced by the maximum moisture content gradient (MMCG). Consequently, accurately modeling the MMCG as it relates to drying conditions is essential for optimizing drying processes. However, this gradient demonstrates complex, non-linear variations involving multiple variables, making it challenging and time-consuming to model using traditional methods. Artificial neural networks (ANNs) offer a powerful alternative due to their inherent ability to handle such complexities. This work presents an ANN model for predicting the MMCG within rice kernels during combined hot air and far-infrared drying. The model utilized a three-layer, fully connected feedforward network. The inputs were drying time, inlet air temperature, and far-infrared (FIR) intensity. The outputs predicted the average of moisture content (MC), MC at the short axis of the kernel (MCS), and MC at the kernel center, enabling the prediction of MMCG. The two hidden layers, containing 20 neurons, employed a tan-sigmoid transfer function. The Levenberg-Marquardt algorithm was used to train the network. Training data was generated from a finite element method (FEM) simulation based on Fick’s law of diffusion. The trained ANN was validated and tested using randomly generated data. To prevent overfitting, the training process incorporated an early stopping method. The results demonstrate the network’s ability to accurately predict MC and MMCG behavior, as indicated by root mean square error (RMSE) and R-squared (R²). The ANN model demonstrates high predictive accuracy, confirming its effectiveness in modeling moisture content and MMCG during rice drying.

This study presents the development of an electrochemical sensor based on nanochannels for the sensitive detection of the fungal toxin zearalenone (ZEN). The sensor incorporates gold nanoparticles (AuNPs) and multi-walled carbon nanotubes (cMWCNTs) onto anodic aluminum oxide (AAO), with nickel oxide (NiO) modification on the reverse side of the AAO, thereby forming a NiO-AAO@AuNPs-cMWCNTs/SPCE electrode system. The NiO modification facilitates the initial oxidation of ZEN, resulting in the generation of distinctive electrochemical oxidation peaks. To investigate the reaction mechanism of ZEN oxidation, electrochemical impedance spectroscopy (EIS) and cyclic voltammetry (CV) were employed. In the concentration range of 1 ~ 40 μg/mL, the sensor showed a clear linear correlation between ZEN concentration and the impedance response, expressed by the regression equation: R_{EIS (Ω)} = 40.8 + 2.7 C_{ZEN (μg/mL)} (R² = 0.997, n = 3). The limit of detection (LOD) was determined to be 0.1 μg/mL. This nanochannel-based electrochemical platform provides a reliable and efficient strategy for ZEN detection, demonstrating the potential of nanostructured materials in food safety monitoring.