Establishment and comparison of in situ detection models for foodborne pathogen contamination on mutton based on SWIR-HSI.

Introduction: Rapid and accurate detection of food-borne pathogens on mutton is of great significance to ensure the safety of mutton and its products and the health of consumers. Objectives: The feasibility of short-wave infrared hyperspectral imaging (SWIR-HSI) in detecting the contamination status and species of Escherichia coli (EC), Staphylococcus aureus (SA) and Salmonella typhimurium (ST) contaminated on mutton was explored. Materials and methods: The hyperspectral images of uncontaminated and contaminated mutton samples with different concentrations (108, 107, 106, 105, 104, 103 and 102 CFU/mL) of EC, SA and ST were acquired. The one dimensional convolutional neural network (1D-CNN) model was constructed and the influence of structure hyperparameters on the model was explored. The effects of different spectral preprocessing methods on partial least squares-discriminant analysis (PLS-DA), support vector machine (SVM) and 1D-CNN models were discussed. In addition, the feasibility of using the characteristic wavelength to establish simplified models was explored. Results and discussion: The best full band model was the 1D-CNN model with the convolution kernels number of (64, 16) and the activation function of tanh established by the original spectra, and its accuracy of training set, test set and external validation set were 100.00, 92.86 and 97.62%, respectively. The optimal simplified model was genetic algorithm optimization support vector machine (GA-SVM). For discriminating the pathogen species, the accuracies of SVM models established by full band spectra preprocessed by 2D and all 1D-CNN models with the convolution kernel number of (32, 16) and the activation function of tanh were 100.00%. In addition, the accuracies of all simplified models were 100.00% except for the 1D-CNN models. Considering the complexity of features and model calculation, the 1D-CNN models established by original spectra were the optimal models for pathogenic bacteria contamination status and species. The simplified models provide basis for developing multispectral detection instruments. Conclusion: The results proved that SWIR-HSI combined with machine learning and deep learning could accurately detect the foodborne pathogen contamination on mutton, and the performance of deep learning models were better than that of machine learning. This study can promote the application of HSI technology in the detection of foodborne pathogens on meat.

Adulteration Detection of Pork in Mutton Using Smart Phone with the CBAM-Invert-ResNet and Multiple Parts Feature Fusion.

To achieve accurate detection the content of multiple parts pork adulterated in mutton under the effect of mutton flavor essence and colorant by RGB images, the improved CBAM-Invert-ResNet50 network based on the attention mechanism and the inversion residual was used to detect the content of pork from the back, front leg, and hind leg in adulterated mutton. The deep features of different parts extracted by the CBAM-Invert-ResNet50 were fused by feature, stitched, and combined with transfer learning, and the content of pork from mixed parts in adulterated mutton was detected. The results showed that the R2 of the CBAM-Invert-ResNet50 for the back, front leg, and hind leg datasets were 0.9373, 0.8876, and 0.9055, respectively, and the RMSE values were 0.0268 g·g-1, 0.0378 g·g-1, and 0.0316 g·g-1, respectively. The R2 and RMSE of the mixed dataset were 0.9264 and 0.0290 g·g-1, respectively. When the features of different parts were fused, the R2 and RMSE of the CBAM-Invert-ResNet50 for the mixed dataset were 0.9589 and 0.0220 g·g-1, respectively. Compared with the model built before feature fusion, the R2 of the mixed dataset increased by 0.0325, and the RMSE decreased by 0.0070 g·g-1. The above results indicated that the CBAM-Invert-ResNet50 model could effectively detect the content of pork from different parts in adulterated mutton as additives. Feature fusion combined with transfer learning can effectively improve the detection accuracy for the content of mixed parts of pork in adulterated mutton. The results of this study can provide technical support and a basis for maintaining the mutton market order and protecting mutton food safety supervision.

Simulated and Verification of Mass and Heat Transfer Coupled Model of Jujube Slices Dried by Hot Air Combined with Radio Frequency Heat Treatment at Different Drying Stages.

This study investigates the impact of radio frequency (RF) heat treatment on heat and mass transfer during the hot air drying of jujube slices. Experiments were conducted at different drying stages, comparing single-hot air drying with hot air combined with RF treatment. Numerical models using COMSOL Multiphysics® were developed to simulate the process, and the results were compared to validate the models. The maximum difference between the simulated value of the center temperature and the experimental value was 6.9 °C, while the minimum difference was 0.1 °C. The maximum difference in average surface temperature was 1.7 °C, with a minimum of 0.3 °C. The determination coefficient (R2) between the simulated experimental values of HA and the early (E-HA + RF), middle (M-HA + RF), and later (L-HA + RF) groups was 0.964, 0.987, 0.961, and 0.977, respectively. The study demonstrates that RF treatment reduces drying time, enhances internal temperature, promotes consistent heat and mass transfer, and accelerates moisture diffusion in jujube slices. Furthermore, the later the RF treatment is applied, the greater the increase in internal temperature and the faster the decrease in moisture content. This research elucidates the mechanism by which RF heat treatment influences heat transfer in hot air-dried jujube slices.

Synergetic application of thermal imaging and CCD imaging techniques to detect mutton adulteration based on data-level fusion and deep residual network.

To improve the performance of single thermal imaging and single CCD imaging in detecting unknown adulterated meat samples, these two imaging techniques combined with a deep residual network were synergistically applied to detect mutton adulteration. Considering the importance of spatial and detailed information in improving stability and accuracy, three data-level fusion methods, namely, colour image stitching, grey image stitching and grey channel stacking, were proposed for the fusion of thermal images and CCD images. Classification and prediction models were further developed based on fusion images. The results showed that the models with colour image stitching achieved the best performance. For the external validation set, the accuracy of the best classification model in discriminating five categories was 99.30%. In predicting pork proportions, the R2, RMSE, RPD and RER of the best prediction model were 0.9717, 0.0238, 7.8696 and 21.28, respectively. The best prediction model for duck proportions had a R2 of 0.9616, RMSE of 0.0277, RPD of 5.1015, and RER of 14.44. Therefore, the synergetic application of thermal imaging and CCD imaging can provide a novel and promising tool to detect mutton adulteration and the quality of other food items.

Discrimination of Minced Mutton Adulteration Based on Sized-Adaptive Online NIRS Information and 2D Conventional Neural Network.

Single-probe near-infrared spectroscopy (NIRS) usually uses different spectral information for modelling, but there are few reports about its influence on model performance. Based on sized-adaptive online NIRS information and the 2D conventional neural network (CNN), minced samples of pure mutton, pork, duck, and adulterated mutton with pork/duck were classified in this study. The influence of spectral information, convolution kernel sizes, and classifiers on model performance was separately explored. The results showed that spectral information had a great influence on model accuracy, of which the maximum difference could reach up to 12.06% for the same validation set. The convolution kernel sizes and classifiers had little effect on model accuracy but had significant influence on classification speed. For all datasets, the accuracy of the CNN model with mean spectral information per direction, extreme learning machine (ELM) classifier, and 7 × 7 convolution kernel was higher than 99.56%. Considering the rapidity and practicality, this study provides a fast and accurate method for online classification of adulterated mutton.

Microstructure Analysis and Quality Evaluation of Jujube Slices Dried by Hot Air Combined with Radio Frequency Heat Treatment at Different Drying Stages.

Jujubes have been favored by consumers because of their rich nutrition and wide use. Hot air drying has been commercially and typically used to prolong shelf life and acquire the dried produce. Jujube slices were dried with hot air combined with radio frequency (RF) at different drying stages, namely, early (0-2 h, E-HA + RF), middle (2-4 h, M-HA + RF), later (4-6 h, L-HA + RF), and whole (0-6 h, W-HA + RF) stages. This study aimed to investigate the effects of different RF application stages on the microstructure, moisture absorption rate, color, and ascorbic acid of jujube slices. Compared with the hot air drying (HA) group, the E-HA + RF group obtained the best results among the experimental groups because it reduced the cells with a roundness of less than 0.4 by 5%. Moreover, the M-HA + RF group showed better results than those of other groups, with an 18.6% and 48.8% reduction in cells for a cross-sectional area less than 200 µm2 and a perimeter less than 25 µm, respectively. The minimum total color difference (ΔE = 9.21 ± 0.31) and maximum retention of ascorbic acid (285.06 mg/100 g) were also observed in this group. Therefore, the method of hot air drying assisted by phased RF is viable in the drying industry to improve the quality of dried agricultural products and reduce energy consumption.

Evaluation of Mutton Adulteration under the Effect of Mutton Flavour Essence Using Hyperspectral Imaging Combined with Machine Learning and Sparrow Search Algorithm.

The evaluation of mutton adulteration faces new challenges because of mutton flavour essence, which achieves a similar flavour between the adulterant and mutton. Hence, methods for classifying and quantifying the adulterated mutton under the effect of mutton flavour essence, based on near-infrared hyperspectral imaging (NIR-HSI, 1000-2500 nm) combined with machine learning (ML) and sparrow search algorithm (SSA), were proposed in this study. After spectral preprocessing via first derivative combined with multiple scattering correction (1D + MSC), classification and quantification models were established using back propagation neural network (BP), extreme learning machine (ELM) and support vector machine/regression (SVM/SVR). SSA was further used to explore the global optimal parameters of these models. Results showed that the performance of models improves after optimisation via the SSA. SSA-SVM achieved the optimal discrimination result, with an accuracy of 99.79% in the prediction set; SSA-SVR achieved the optimal prediction result, with an RP2 of 0.9304 and an RMSEP of 0.0458 g·g-1. Hence, NIR-HSI combined with ML and SSA is feasible for classification and quantification of mutton adulteration under the effect of mutton flavour essence. This study can provide a theoretical and practical reference for the evaluation and supervision of food quality under complex conditions.

Adulteration discrimination and analysis of fresh and frozen-thawed minced adulterated mutton using hyperspectral images combined with recurrence plot and convolutional neural network.

This paper presented a method to detect adulterated mutton using recurrence plot transformed by spectrum combined with convolutional neural network (RP-CNN). For this, 100 adulterated samples of mutton mixed with different proportions (0.5-1-2-5-10% (w/w)) of pork and 20 pure mutton samples were prepared. The results of the classification model of adulterated mutton and the quantitative prediction model of pork content established by this method were comparable for fresh, frozen-thawed and mixed datasets. It shows that the classification accuracies of adulteration mutton on three datasets were 100.00%, 100.00% and 99.95% respectively. Moreover, for the pork content prediction of adulterated mutton, the R2 on three datasets of fresh, frozen-thawed and mixed samples were 0.9762, 0.9807 and 0.9479, respectively. Therefore, the hyperspectral combined with RP-CNN proposed in this paper shows great potential in the classification of adulterated mutton and the pork content prediction of adulterated mutton.

Detection of adulteration in mutton using digital images in time domain combined with deep learning algorithm.

A novel method based on digital images in time domain combined with convolutional neural network (CNN) is proposed for discrimination and analysis of the adulterated mutton. For this, 195 sample images during the constant temperature heating process (about 10 min) were combined with CNN for qualitative discrimination and quantitative prediction of adulterated mutton. Furthermore, the hypothesis that temperature disturbance can improve the detection ability of adulterated mutton was confirmed by comparing the model performance of the initial heating stage and the entire heating process. The experimental results show that the performance of the latter was superior to that of the former. The accuracy of the qualitative discriminant model was increased by 7.33%, the R2 and RPD of the quantitative prediction model of the duck/pork in adulterated mutton were increased by 0.08/0.07 and 0.85/0.87 respectively, while the RMSE decreased by 0.01/0.01. Consequently, the proposed method can be used for detecting adulterated mutton effectively and accurately.

Quantitative Detection of Chromium Pollution in Biochar Based on Matrix Effect Classification Regression Model.

Returning biochar to farmland has become one of the nationally promoted technologies for soil remediation and improvement in China. Rapid detection of heavy metals in biochar derived from varied materials can provide a guarantee for contaminated soil, avoiding secondary pollution. This work aims first to apply laser-induced breakdown spectroscopy (LIBS) for the quantitative detection of Cr in biochar. Learning from the principles of traditional matrix effect correction methods, calibration samples were divided into 1-3 classifications by an unsupervised hierarchical clustering method based on the main elemental LIBS data in biochar. The prediction samples were then divided into diverse classifications of calibration samples by a supervised K-nearest neighbor (KNN) algorithm. By comparing the effects of multiple partial least squares regression (PLSR) models, the results show that larger numbered classifications have a lower averaged relative standard deviations of cross-validation (ARSDCV) value, signifying a better calibration performance. Therefore, the 3 classification regression model was employed in this study, which had a better prediction performance with a lower averaged relative standard deviations of prediction (ARSDP) value of 8.13%, in comparison with our previous research and related literature results. The LIBS technology combined with matrix effect classification regression model can weaken the influence of the complex matrix effect of biochar and achieve accurate quantification of contaminated metal Cr in biochar.

Hyperspectral Imaging Detection of Total Viable Count from Vacuum Packing Cooling Mutton Based on GA and CARS Algorithms.

In the process of spectral modeling, spectral extraction of characteristic bands with different variable screening algorithms is an important step for improving the model effects. Total viable count of cooling mutton under vacuum packing condition was chosen as the research index in this paper, while the influence of 2 variable screening algorithms on its hyperspectral PLS model effects was compared. Mutton muscle spectra of Regions of interest (ROIs) were extracted and preprocessed. Subsequently, Genetic Algorithm (GA) and Competitive Adaptive Reweighted Sampling (CARS) were applied to extract characteristic bands from preprocessed spectra at full band range of 473～1 000 nm. Model effects of GA-PLS, CARS-PLS and W-PLS with corresponding bands selection were contrasted and analyzed. The results indicated that both model effects of GA-PLS, CARS-PLS were better than that of W-PLS, and CARS-PLS model effect was optimal. As for the CARS-PLS model, the determination coefficient (R2c) and root mean square error (RMSEC) of calibration set was 0.96 and 0.29, and the determination coefficient (R2cv) and root mean square error (RMSECV) of leave-one-out cross validation was 0.92 and 0.46, respectively. Meanwhile, the determination coefficient (R2p), root mean square error of prediction (RMSEP) and the ratio of standard deviation to standard error of prediction (RPD) of prediction set was 0.92 and 0.47 and 3.58, respectively. Therefore, hyperspectral imaging (HSI) technology combined with CARS-PLS can achieve quick, non-destructive and accurate detection of mutton total viable count.

Clinical value of mammography in diagnosis and identification of breast mass.

OBJECTIVE: To study the effect and clinical value of mammography in the diagnosis of breast lump so as to improve the diagnosis level of breast cancer. METHODS: A retrospective analysis was carried out on clinical data of 110 patients with mammary lump confirmed by pathology to study the compliance of mammography diagnosis and Pathology diagnosis in breast lump, and the detection of microcalcifications, phyllode, and observe the image performance of mammography. Taking infitrating ductal carcinoma (IDC) as an example, the correlation of image performance and clinical pathological features of different types was studied so as to predict if mammography performance was effective in the treatment and prognosis in breast cancer. RESULTS: Taking Breast Imaging Reporting and Data System (BI-RADS) grade 4A as the critical point, the sensitivity, specificity and accuracy of mammography was 90.80% (109/120), 84.60% (126/149) and 87.40% (235/269); taking BI-RADS grade 4B as the critical point, the sensitivity, specificity and accuracy of mammography was 85.00% (102/120), 93.30% (139/149) and 89.60% (241/269); the correlation analysis suggested that, there was some kind of correlation between the mammography performance and clinical features of breast cancer. CONCLUSION: Mammography is worth being promoted in clinic for its significant clinical value in diagnosing and identifying breast lump.

[Study on the Rapid Evaluation of Total Volatile Basic Nitrogen (TVB-N) of Mutton by Hyperspectral Imaging Technique].

Total Volatile Basic Nitrogen (TVB-N) was usually taken as the physicochemical reference value to evaluate the mutton freshness. In order to explore the feasibility of hyperspectral (HSI) imaging technique to detect mutton freshness, 71 representative mutton samples were collected and scanned using a diffuse reflectance hyperspectral imaging (HSI) system in the Visible-Near infrared (NIR) spectral region (400-1 000 nm), and the chemical values of TVB-N content were determined using the semimicro Kjeldahl method according to the modified Chinese national standard. The representative spectra of mutton samples were extracted and obtained after selection of the region of interests (ROIs). The samples of calibration set and prediction set were divided at the ratio of 3 : 1 according to the content gradient method. Optimum HSI calibration models of the mutton (TVB-N) were established and evaluated by comparing different spectral preprocessing methods and modeling methods, which included Stepwise Multiple Linear Regression (SMLR), Partial Least Squares Regression (PLSR) and Principal Component Regression (PCR) methods. The results are that through the utilization of Multiplicative Scatter Correction (MSC), first derivative, Savitzky-Golay (S-G) smoothing and mean-centering together, both PLSR and PCR were able to achieve quantitative detection of mutton TVB-N. As for the PLSR model of mutton TVB-N established, the spectral pretreatment methods chosen included MSC, first derivative, S-G (15,2) smoothing and mean-centering, and the latent variables (LVs) number used was 11. As for the calibration set of PLSR model of mutton TVB-N, the correlation coefficient (r) and root mean square error of calibration (RMSEC) were 0.92 and 3.00 mg x (100 g)(-1), respectively. As for the prediction set of PLSR model of mutton TVB-N, the correlation coefficient (r), Root Mean Square Error of Prediction (RMSEP), and ratio of standard deviation to standard error of prediction (RPD) were 0.92, 3.46 mg x (100 g)(-1) and 2.35, respectively. The study demonstrated that the rapid and accurate analysis of TVB-N, the key freshness attribute, could be implemented by using the hyperspectral imaging (HSI) technique. The study provides the basis for further rapid and non-destructive detection of other mutton freshness attributes by using the hyperspectral imaging (HSI) technique, the improvement of current modeling effect of TVB-N content and the application involved of the technique in the practical production.

Application value of magnetic resonance imaging in diagnosing central nervous system lymphoma.

OBJECTIVE: To describe the magnetic resonance imaging (MRI) appearance of central nervous system lymphoma. METHODS: We retrospectively reviewed MRI images of 40 patients who had pathologically proven primary central nervous system lymphoma (PCNSL) and received treatment in Binzhou People's Hospital, Shandong, China from January to December in 2014. Location, size and form of tumor was observed and relevant data were recorded for analysis. RESULTS: Foci of 40 cases of PCNSL all located in brain, among which. 18 cases were single (45.0%) and 22 cases were multiple (55.5%). Of 96 Foci, 84 were supratentorial, 12 were subtentorial. Enhanced MRI scanning showed that, most Foci had significant homogenous enhancement, shaping as multiple nodular or lumpy, and few had ring-enhancement. MRI suggested that, T1 signal of most Foci concentrated on low signal segment and T2 signal gathered on high signal segment, suggesting a significant homogeneous enhancement; moreover, mild and medium edema surrounded the tumor. They were pathologically confirmed as B cell derived non-hodgkin lymphoma. Except one case of Burkitt lymphoma, the others were all diffuse large B cell lymphoma which was observed with diffuse distribution of cancer cells (little cytoplasm, large nucleus, rough perichromatin granule) in same size. Fifteen cases were observed with sleeve-like infiltration of cancer cells around blood vessels. No case was found with hemorrhage, necrosis or calcification. CONCLUSION: Pathological foundation of PCNSL determines its characteristic MRI performance. Typical case of PCNSL can be diagnosed accurately by MRI.

[Effects of Regions of Interest (ROIs) on Detection Models of Mutton pH Based on Hyperspectral Imaging].

Selection of Regions of interest (ROIs) and subsequent spectral extraction was a key step of non-destructive detection and analysis based on hyperspectral imaging (HSI). For the rapid and accurate detection of mutton pH, the study on the effects of 2 different ROIs on mutton pH models was carried out in the visible-near infrared region of 473～1 000 nm. 2 ROIs methods of Rectangle Regions (RR) and Image Segmentation (IS) were adopted to extract 122 corresponding representative spectra respectively. The influence of different preprocessing methods and ROIs methods on 3 pH models, including stepwise multiple linear regression (SMLR), principal component regression (PCR) and partial least squares regression (PLSR), was compared and analyzed. The results indicated that SMLR and PLSR model performance was optimal in 3 models established with spectral data extracted from Rectangle Regions (RR) and Image Segmentation (IS) respectively. As for the SMLR model, corresponding to the RR ROIs method, the correlation coefficient (Rcal) and root mean square error (RMSEC) of calibration set was 0.85 and 0.085 respectively, and the correlation coefficient (Rp) and root mean square error (RMSEP) of prediction set was 0.82 and 0.097 respectively. As for the PLSR model, corresponding to the IS ROIs method, the correlation coefficient(Rcal) and root mean square error (RMSEC) of calibration set was 0.95 and 0.050 respectively, and the correlation coefficient (Rp) and root mean square error (RMSEP) of prediction set was 0.91 and 0.071 respectively. By comparing the modeling results of spectral data extracted from 2 ROIs methods, the modeling performances of Image Segmentation (IS) were always better than Rectangle Regions (RR) in all the 3 modeling methods. The study shows that it is feasible to apply hyperspectral imaging technology combined with the ROIs method of Image Segmentation (IS) to accurate, fast and non-destructive detection of mutton pH.

[Study on Characteristic Bands Selection of Lamb pH Value Based on Hyperspectral Imaging and Partial Least Squares(PLS)].

Characteristic bands method selection and subsequent spectral extraction has a great influence on the hyperspectral model performance. For rapid and accurate detection of mutton pH value, the effects of 2 band-selection methods on PLS models of mutton pH based on HSI technique were carried out and discussed. Initially, the preprocessing method of second derivative (2D), multiplicative scatter correction (MSC) and mean-centering together was implemented on the representative spectra of mutton muscle portion. Then, 2 methods of synergy interval partial least square (siPLS) and the combination of synergy interval partial least squares with genetic algorithm (siPLS-GA) were used to extract the characteristic bands in the spectral range of 473～1 000 nm. Finally, 2 PLS models of lamb pH value were established with the corresponding characteristic bands, and were also compared with the effect of full-band PLS model. The results indicated that the effect of siPLS-GA-PLS model was the best. As for the siPLS-GA-PLS model, 56 characteristic wavelength points were chosen, the correlation coefficient(Rcal) and root mean square error(RMSEC) of calibration set was 0.96 and 0.043 respectively, and the correlation coefficient(Rp) and root mean square error(RMSEP) of prediction set was 0.96 and 0.048 respectively. Spectral variables were reduced and model accuracy was improved. It can be concluded that characteristic bands selection and rapid and accurate detection of lamb pH can be achieved using hyperspectral imaging technique combined with siPLS-GA method.