Food Control最新文献

{"title":"Detecting vegetable oils in black soldier fly larvae oil using infrared spectroscopy coupled with one-class and hierarchical modelling","authors":"J.P. Cruz-Tirado ,&nbsp;Ramon Sousa Barros Ferreira ,&nbsp;Eduardo Augusto Caldas Batista ,&nbsp;Lara Honório ,&nbsp;José Manuel Amigo ,&nbsp;Douglas Fernandes Barbin","doi":"10.1016/j.foodcont.2026.111961","DOIUrl":"10.1016/j.foodcont.2026.111961","url":null,"abstract":"<div><div>The Black Soldier Fly Larvae (BSFL) oil is a by-product from the animal feed industry. BSFL oil is rich in medium-chain fatty acids, which has led to an increase in demand for pharmaceutical or animal feed industry. This study evaluated the authenticity of BSFL oil and its adulteration with vegetable oils using two portable near infrared (NIR) spectrometers (<em>spectrometer 1</em> (<em>900 –</em> 1700 nm) and <em>spectrometer 2</em> (1350–2500 nm)) and Fourier-transform middle infrared (FTIR) spectroscopy (4000–650 cm⁻¹) combined with chemometric methods. Principal Component Analysis (PCA) showed partial separation between pure and adulterated samples, but some mixtures, like BSFL + sunflower and BSFL + corn, overlapped due to similar composition. Data Driven Soft Independent Modelling Class Analogy (DD-SIMCA) was used to authenticate pure BSFL oil, reaching up to 100 % sensitivity and specificity in test set, especially with <em>spectrometer 2</em> and <em>FTIR</em>. Multi-class PLS-DA models were able to classify BSFL oil and its mixtures, although they showed low classification rates (True Positive Rate (TPR) &lt; 30 %) for BSFL oil adulterated with canola or corn oil, especially using <em>spectrometer 1</em>. Hierarchical PLS-DA models were developed to improve discrimination, reaching better classification rates for all adulterated classes. Overall, DD-SIMCA, PLS-DA and hierarchical models combined with NIR or FTIR spectroscopy are promising tools to detect BSFL oil adulteration in a fast and non-destructive way, with potential applications in quality control and fraud prevention.</div></div>","PeriodicalId":319,"journal":{"name":"Food Control","volume":"183 ","pages":"Article 111961"},"PeriodicalIF":6.3,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145939325","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A new approach for identifying the origin of brown sugar: Multiple chemical profiles combined with machine learning forms an accurate and reliable product traceability strategy","authors":"Xinyao Li ,&nbsp;Yashu Zhang ,&nbsp;Haiqin Lu,&nbsp;Kai Li,&nbsp;Zhi Huang","doi":"10.1016/j.foodcont.2025.111889","DOIUrl":"10.1016/j.foodcont.2025.111889","url":null,"abstract":"<div><div>The traceability of brown sugar for origin remains a challenge for market regulation. To establish a reliable system for origin identification, this study integrated multi-dimensional chemical profiling with machine learning techniques to preliminarily classify brown sugar produced in five regions of China. Mineral and phenolic compound profiles were obtained by using inductively coupled plasma optical emission spectrometry and liquid chromatography–tandem mass spectrometry, respectively. Chemometric analysis identified potassium as the most effective element for distinguishing among brown sugar production regions. Principal component analysis determined six principal components for initial explanation of origin differences. Machine learning models were used to further enhance reliability of origin identification. Upon validation, the Random Forest and Support Vector Machine models achieved classification accuracies of 96 % and 98 %, respectively. These results provide a preliminary scientific basis and methodological foundation for the origin traceability and authentication of brown sugar.</div></div>","PeriodicalId":319,"journal":{"name":"Food Control","volume":"183 ","pages":"Article 111889"},"PeriodicalIF":6.3,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665485","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Unravelling the genetic threads: Comparative genomic analysis of Salmonella from food and clinical cases in England, 2015–2019","authors":"Beata Orzechowska ,&nbsp;Jim McLauchlin ,&nbsp;Caroline Willis ,&nbsp;Frieda Jorgensen ,&nbsp;Anaïs Painset ,&nbsp;Sandra Lai ,&nbsp;Heather Aird ,&nbsp;Amina Ismail Ahmed ,&nbsp;Marie Anne Chattaway","doi":"10.1016/j.foodcont.2025.111928","DOIUrl":"10.1016/j.foodcont.2025.111928","url":null,"abstract":"<div><div>Despite advances in <em>Salmonella</em> testing, surveillance, and improved food safety practices, salmonellosis remains prevalent in the UK. The primary challenge of preventing infections is determining how to ensure that food safety management plans are effective and to prioritise food sampling activities before the outbreaks occur.</div><div>In this genomic study, we assessed the relative contribution of various foods over a five-year period to cases of human salmonellosis in England, aiming to rank the associated food risks based on genetic similarity. To achieve this, we conducted a comparative genomic analysis of <em>Salmonella</em> recovered from food (802) and environmental samples (131) against routinely tested <em>Salmonella</em> isolates from clinical cases (37,673) between 2015 and 2019.</div><div>The analysis revealed that of all <em>Salmonella</em>-positive food and environmental isolates, 19.7 % were genetically related to <em>Salmonella</em> from clinical cases. A significant portion of these isolates was collected due to reactive sampling, which occurs in response to recognised public health concerns. In terms of <em>Salmonella</em> from clinical cases, only 2.5 % were genetically similar to isolates from food and 2.6 % from the environment. This relatively low rate of similarity may be due to other infection pathways, such as travel-related exposure or contact with animals, as well as infection from food items that are not subject to regular microbiological testing by public health bodies in the UK.</div><div>We found that isolates from food products of animal origin, specifically eggs and poultry, were genetically linked to a higher number of isolates from clinical cases compared to non-animal origin products or composite foods, potentially indicating a greater public health risk. Meanwhile, non-animal origin foods were mostly associated with sporadic cases or clusters of fewer than five cases that had not previously been identified.</div><div>The findings from this study underscore the need to shift from primarily reactive to additional proactive, risk-based food testing to better mitigate foodborne hazards and reduce the potential for outbreaks. While our genomic approach offers a valuable method for ranking food-related risks through genetic linkages, it does not replace the need for integrated epidemiological analysis. Such data are essential for understanding exposure pathways and quantifying foodborne risks, but were beyond the scope of this study. Nevertheless, the proposed genomic framework has the potential to inform future microbiological risk assessments, guide prioritisation of food sampling activities, support outbreak investigations, and contribute to the development of genomics-based public health interventions.</div></div>","PeriodicalId":319,"journal":{"name":"Food Control","volume":"183 ","pages":"Article 111928"},"PeriodicalIF":6.3,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145880907","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Control of Yersinia enterocolitica and Escherichia coli in Geotjeori using essential oils extracted from kimchi seasoning ingredients","authors":"KyeHwan Byun,&nbsp;Ji-Hyoung Ha","doi":"10.1016/j.foodcont.2025.111918","DOIUrl":"10.1016/j.foodcont.2025.111918","url":null,"abstract":"<div><div><em>Yersinia enterocolitica</em> and <em>Escherichia coli</em> are significant microbial contaminants of non-fermented fresh kimchi, named <em>Geotjeori</em>, the control of which remains largely understudied. In this study, we evaluated the antibacterial effects of essential oils (EOs) extracted from seasoning ingredients of kimchi against these pathogens. Initially, the viability and biofilm-forming ability were assessed under typical storage conditions (temperature and pH) of <em>Geotjeori</em>. At 5 °C, viability of both bacteria gradually decreased and biofilm formation was not observed. However, at 10 °C, bacterial population increased and biofilms were developed. Subsequently, the antibacterial activity of the EOs extracted from the kimchi seasoning ingredients were evaluated. Among the five tested EOs, only those from leaf mustard (Mu) and ginger (Gi) exhibited significant antibacterial activity, with measurable minimum inhibitory/biofilm inhibitory/bactericidal concentrations. When <em>Geotjeori</em> juice mixed with MU and Gi EOs at their minimum bactericidal concentrations was stored at 5 °C, planktonic cell counts decreased significantly, whereas biofilm formation still occurred, which indicated that biofilm formation is a major survival strategy for these pathogens. Additionally, Mu and Gi EOs significantly reduced the populations of <em>Y. enterocolitica</em>, <em>E. coli</em>, and total aerobic bacteria without adversely affecting lactic acid bacteria. In conclusion, Mu and Gi EOs are promising natural preservatives for enhancing the microbial safety of fresh kimchi; however, the existence of pathogens in the form of biofilms, even in the presence of EOs, highlights the need for further investigation.</div></div>","PeriodicalId":319,"journal":{"name":"Food Control","volume":"183 ","pages":"Article 111918"},"PeriodicalIF":6.3,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145798123","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Deep transfer learning for codling moth damage detection in ‘Conference’ pear using X-ray radiography","authors":"Jiaqi He ,&nbsp;Tim Belien ,&nbsp;Ammar Alhmedi ,&nbsp;Ann Schenk ,&nbsp;Dany Bylemans ,&nbsp;Pieter Verboven ,&nbsp;Bart Nicolai","doi":"10.1016/j.foodcont.2025.111923","DOIUrl":"10.1016/j.foodcont.2025.111923","url":null,"abstract":"<div><div>Early detection of pest infestation is essential for ensuring postharvest fruit quality and compliance with export standards. This study investigates deep transfer learning for detecting codling moth (<em>Cydia pomonella</em>) larval damage in ‘Conference’ pears using non-destructive X-ray radiography. Data were collected from both controlled laboratory and field infestations across multiple growing seasons, and augmented via a radiographic simulation framework. We evaluated three convolutional neural networks (ResNet50, DenseNet121, and InceptionV3) under three transfer learning strategies (feature extractor, partial, and full fine-tunning) using pretrained weights from ImageNet, RadImageNet, or random initialization. A classical machine learning baseline combining adaptive thresholding segmentation and handcrafted features with Support Vector Machine (SVM), k-Nearest Neighbors (kNN), and Logistic Regression (LR), was also implemented for comparison. Transfer learning strategy had the greatest influence on model performance, with full fine-tuning yielding the highest accuracy. Pretrained weights, especially from ImageNet, consistently improved performance over random initialization. The differences between model architectures were smaller under full fine-tunning and pretrained weights from ImageNet. The top-performing configuration of ResNet50 with ImageNet weights and full fine-tuning, reached 91–97 % accuracy across internal test sets and showed 20 % improvement in accuracy, precision and recall compared to the benchmark method. Performance remained robust on an independent test set, with an approximately 10 % improvement over the benchmark method. Guided Grad-CAM heatmaps confirmed that the infestation regions were correctly localized by deep learning models, supporting modal interpretability and reliability. These findings underline the potential of deep transfer learning for accurate and interpretable detection of internal pest damage in pear fruit and support the deployment of automated X-ray-based inspection systems in postharvest quality control workflows.</div></div>","PeriodicalId":319,"journal":{"name":"Food Control","volume":"183 ","pages":"Article 111923"},"PeriodicalIF":6.3,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145798093","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"An approach to regulate Maillard reaction and caramelization reaction in brown sugar: the addition of epicatechin naturally occurring in sugarcane","authors":"Jie Liu ,&nbsp;Haoran Zhang ,&nbsp;Runyu Jiao ,&nbsp;Xinwen Liang ,&nbsp;De-Wei Chen","doi":"10.1016/j.foodcont.2025.111901","DOIUrl":"10.1016/j.foodcont.2025.111901","url":null,"abstract":"<div><div>Maillard reaction and caramelization reaction occurred in brown sugar are unavoidable during the process, contributing to its characteristic odor. Products with strong or mild odor are preferred by different consumers. Epicatechin (EC), a flavonoid, a polyphenolic compound naturally occurring in sugarcane, has impacts on both Maillard reaction and caramelization reaction. The aim of this work was to propose a strategy for inhibiting these reactions in brown sugar through EC addition. Twenty-seven odorants, including oxygenated heterocyclic compounds, N-containing compounds, phenolic, benzene-containing compounds, ketones, other compounds, and sulfur compounds, derived from Maillard reaction and/or caramelization reaction were detected in brown sugar using GC-MS, and EC addition significantly reduced the abundance of most odorants. Model systems were constructed using fructose or glucose, either alone or in combination with asparagine, with or without EC. The results of model systems also revealed that EC addition markedly suppressed the formation of key odorants. The inhibitory mechanism was further investigated using ultra-performance liquid chromatography coupled with Q-Exactive mass spectrometry (UPLC-Q-Exactive-MS) and nuclear magnetic resonance (NMR) spectroscopy. Adducts formed between EC and reactive intermediates of Maillard reaction and caramelization reaction, such as 3-deoxyglucosone, 3,4-dideoxyglucosone-3-ene, and 5-hydroxymethylfurfural were detected, confirming the inhibitory effect of EC was attributed to its binding with intermediate compounds. This study provides a feasible approach to regulate Maillard reaction and caramelization reaction in brown sugar, thereby modulating the formation of key odorants, which offers a scientific basis for controlling its sensory profile to meet diverse consumer preferences.</div></div>","PeriodicalId":319,"journal":{"name":"Food Control","volume":"183 ","pages":"Article 111901"},"PeriodicalIF":6.3,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145749548","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Probiotics as a biocontrol strategy against Salmonella in poultry: A global meta-analysis","authors":"Mingjuan Qu ,&nbsp;Lisheng Zhang ,&nbsp;Jianlong Zhang ,&nbsp;Xin Yu ,&nbsp;Linlin Jiang ,&nbsp;Hongwei Zhu ,&nbsp;Xingxiao Zhang","doi":"10.1016/j.foodcont.2025.111952","DOIUrl":"10.1016/j.foodcont.2025.111952","url":null,"abstract":"<div><div>This study systematically evaluated the efficacy of probiotics as a biocontrol strategy against <em>Salmonella</em> in poultry production, with the aim of reducing <em>Salmonella</em> contamination at the source within the food chain. A comprehensive literature search was conducted in databases including PubMed and Web of Science for studies published up to March 2025. Fifty animal experimental studies were included in the meta-analysis. The effects of probiotics on inflammatory cytokines, <em>Salmonella</em> colonization levels, and poultry health indicators were assessed. Subgroup analyses were performed to examine the influence of probiotic strain type, administration method, intervention duration, and the probiotic-to-<em>Salmonella</em> dose ratio <strong>(PSDR)</strong>. The results indicated that probiotic interventions significantly decreased levels of inflammatory cytokines such as interleukin-6 (IL-6) and IL-1β, reduced <em>Salmonella</em> colonization in the spleen, liver, and cecum, and improved body weight (BW). Subgroup analyses revealed that <em>Lactobacillus</em> and <em>Bacillus</em> strains exhibited superior anti-inflammatory effects, while <em>Bifidobacterium</em> was most effective in promoting weight gain. Oral gavage emerged as the most effective administration method for mitigating inflammation and reducing colonization. Optimal outcomes were achieved with intervention duration of 21–30 days and a PSDR maintained between 1 and 1000. Furthermore, multivariate meta-regression refined these insights by revealing that while lower PSDR and therapeutic administration specifically maximize anti-inflammatory benefits, feed-based delivery was identified as superior for promoting growth. The findings demonstrate that probiotics are effective in controlling <em>Salmonella</em> infections in poultry, thereby contributing to the reduction of <em>Salmonella</em> entry into the food supply. Efficacy is influenced by strain type, administration method, intervention duration, and dose ratio, underscoring the potential of probiotics as a sustainable alternative to antibiotics for enhancing food safety from farm to fork.</div></div>","PeriodicalId":319,"journal":{"name":"Food Control","volume":"183 ","pages":"Article 111952"},"PeriodicalIF":6.3,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145880831","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Portable photonic immunosensor for the simultaneous detection of Escherichia coli and Salmonella typhimurium in milk samples","authors":"Dimitra Kourti ,&nbsp;Michailia Angelopoulou ,&nbsp;Eleni Makarona ,&nbsp;Anastasios Economou ,&nbsp;Panagiota Petrou ,&nbsp;Konstantinos Misiakos ,&nbsp;Sotirios Kakabakos","doi":"10.1016/j.foodcont.2025.111944","DOIUrl":"10.1016/j.foodcont.2025.111944","url":null,"abstract":"<div><div>Milk can be contaminated with pathogenic bacteria such as Salmonella <em>typhimurium</em> and <em>Escherichia coli</em>, which can cause acute foodborne illnesses. In this study, we present an optical immunosensor designed for the simultaneous detection of these two bacteria in milk samples. The sensor is based on a silicon chip that incorporates two U-shaped silicon nitride waveguides configured as Mach-Zehnder Interferometers (MZIs). The sensing windows of the MZIs, located at one end of the chip, differ in length and allow for sequential immersion into reagent solutions during assay procedures. At the opposite end, the chip is optically coupled to a broadband white LED and a spectrophotometer via a bifurcated optical fiber and a dedicated coupler. To enable selective detection, the sensing windows of the MZIs are functionalized with lipopolysaccharides from the outer membranes of S. <em>typhimurium</em> and <em>E. coli</em>, respectively. A competitive immunoassay, completed in just 15 min, enabled detection limits of 45 cfu/mL for S. <em>typhimurium</em> and 125 cfu/mL for <em>E. coli</em> in milk. With a pre-enrichment step of approximately 8 h, single-cell detection became possible for both bacteria. The sensor excellent performance, quantitative determinations capability, and compact design, achieved by eliminating the need for microfluidics or pumps, make it a powerful and practical tool for bacterial detection in milk.</div></div>","PeriodicalId":319,"journal":{"name":"Food Control","volume":"183 ","pages":"Article 111944"},"PeriodicalIF":6.3,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145880909","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Antibacterial and antioxidant tetracycline-functionalized carbon dots nanocomposite films for enhanced fruit preservation","authors":"Xueyu Fu ,&nbsp;Yao Chen ,&nbsp;Kairui Wang,&nbsp;Min Zheng","doi":"10.1016/j.foodcont.2025.111908","DOIUrl":"10.1016/j.foodcont.2025.111908","url":null,"abstract":"<div><div>This study developed a novel antibacterial and antioxidant nanocomposite by synthesizing carbon dots (CDs) and tetracycline-loaded CDs (TCDs), which were integrated into polyvinyl alcohol (PVA) to fabricate active preservation films. Comprehensive characterization confirmed that TCDs exhibited a spherical morphology (94.4 nm diameter), high drug-loading efficiency (73 %), and potent radical scavenging capacity (81.13 % DPPH inhibition). TCDs demonstrated a 1000-fold enhancement in antibacterial activity against <em>Staphylococcus aureus</em> (<em>S. aureus</em>) (MIC = 1.0 μg/mL) compared to pristine CDs. The optimized 3 %PVA composite film exhibited exceptional UV-blocking performance (62 % transmittance reduction), improved mechanical strength (2.5 MPa tensile strength), and tunable hydrophobicity (52.6° water contact angle). Preservation trials demonstrated that the film extended the shelf life of longans to 11 days and strawberries to 6 days, maintaining vitamin C content, and preserving total phenolic compounds. The composite films also displayed excellent hemocompatibility and plant safety. This work presents a multifunctional material platform for active food packaging applications.</div></div>","PeriodicalId":319,"journal":{"name":"Food Control","volume":"183 ","pages":"Article 111908"},"PeriodicalIF":6.3,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145749547","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Ensemble learning: a study on the border inspection prediction model for high-risk products adulterated with Sudan Red in Taiwan","authors":"Li-Ya Wu,&nbsp;Wei-Chih Cheng,&nbsp;King-Fu Lin,&nbsp;Jia-Jhen Lee,&nbsp;Wen-Chou Lin,&nbsp;Hsu-Yang Lin,&nbsp;Chia-Yen Lu,&nbsp;Chih-Kang Chiang","doi":"10.1016/j.foodcont.2025.111931","DOIUrl":"10.1016/j.foodcont.2025.111931","url":null,"abstract":"<div><div>Sudan Red (SR), an industrial azo dye classified as a Group 3 carcinogen, remains a food safety concern despite bans and Taiwan's Border Prediction Intelligent System (BPIS). To address the lack of validated models for rare hazard detection, we developed an ensemble learning model combining Random Forest, Naïve Bayes, and a Multilayer Perceptron with synthetic oversampling. Trained on 3699 inspection records and validated with 700 real cases, the model detected two of three SR-contaminated batches by sampling only 31 shipments (hit rate 6.45 %, coverage 66.67 %), outperforming BPIS (1/220; 0.45 %, 33.33 %). Statistical tests (risk difference +5.997 %, relative risk 14.19, <em>p</em> = 0.041; Bayesian posterior 99.3 %) confirmed superiority. Deep learning improved predictive robustness, and trade partnerships with importers were significantly linked to SR violations (χ² test, <em>p</em> &lt; 0.001). This hazard-specific, data-driven approach improves detection efficiency, reduces inspection burden, and provides an adaptive framework for risk-based border food safety control.</div></div>","PeriodicalId":319,"journal":{"name":"Food Control","volume":"183 ","pages":"Article 111931"},"PeriodicalIF":6.3,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145837627","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"农林科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}