The Detection And Classification Of Foodborne Pathogens Using Hyperspectral Microscope Imaging Technology Coupled With Deep Learning Frameworks

Foodborne pathogens are currently the most important issue in the food industry and threaten the health and life of consumers around the world seriously.They lurk in many kinds of food and enter the body silently through eating,causing a series of bacterial infections or diseases.According to the report from World Health Organization,nearly 600 million people get sick from eating contaminated food every year,about 420,000 people die.Most of the victims are caused by various microbial contamination.The presence of foodborne pathogens now becomes a heavy burden on medical system,causing huge economic losses to various food manufacturers.Although traditional isolation and manual observation are still the golden standard for the detection of foodborne pathogens,the experiments are extremely complicated and time-consuming,and often balked the best period to control the outbreak.Nucleic acid-based detection methods such as polymerase chain reaction(PCR)have been popular in food-borne rapid detection nowadays.However,the expensive cost,and strict experimental environment limited its usage as a routine detection method.In addition,although the PCR method is highly specific and sensitive,due to capability of destructive biochemical detection,it failed in effectively distinguish live and dead bacteria,while the prevention is the principle for the detection of food-borne pathogens.Therefore,governments,inspection agencies,and food safety departments of various food companies are seeking simple,efficient,accurate,and highly automated food-borne pathogen detection solutions.This study proposes a rapid detection and classification method for foodborne pathogens based on hyperspectral microscope imaging(HMI)and deep learning(DL)technology,which can efficiently capture high-quality hyperspectral images of living cells of various foodborne pathogens in the early stage.Combined with various customized DL algorithms,non-destructive and non-invasive digital diagnosis of food-borne pathogens is realized.The main contents of this research are as follows:(1)Design and optimization of HMI system.We designed and optimized a special hyperspectral system for the detection of foodborne pathogens.The study configured the specific software and hardware for the HMI system.All the parameters were optimized in order to obtain high-quality cellular hyperspectral data.To obtain a clear image of the system,a suitable gain value was selected for the HMI system,and data analysis and verification were performed.In addition,in order to determine the optimal light source for the HMI system,we conduct a classification task of different serogroups of Salmonella.The influences of two different light sources(metal halide and tungsten halogen lamps)on the classification task were summarized and discussed.The research results indicate that,when the camera gain was set to 9 and the exposure time was set to 250 ms,the signal-to-noise ratio of the image reached the highest value.The highest signal-to-noise ratios value of the metal halide lamp and tungsten halogen lamp were 14.14 and 4.49,respectively.In addition,the research results showed that both metal halide and tungsten halogen lamps were suitable for HMI systems.Although the spectral data generated by these two light sources were different,the classification accuracy of the PCALDA classifier were both 100%,indicating that both of them were capable of providing highquality spectra.(2)Identifying Non-O157 Shiga toxin-producing Escherichia coli(STEC)using deep learning methods with hyperspectral microscope imaging.Non-O157 Shiga toxin-producing Escherichia coli(STEC)serogroups such as O26,O45,O103,O111,O121 and O145 often cause illness to people,whereas the conventional identification of these "Big-Six" are complex.The label-free hyperspectral microscope imaging(HMI)method,which provides spectral "fingerprints" information of bacterial cells,was employed to classify serogroups at the cellular level.In spectra analysis,principal component analysis(PCA)method and stacked auto-encoder(SAE)method were conducted to extract principal spectral features for classification task.Based on these features,multiple classifiers including linear discriminant analysis(LDA),support vector machine(SVM)and soft-max regression(SR)methods were evaluated.Different sizes of datasets were also tested in search for the suitable classification models.Among the results,SAE-based classification models performed better than PCA-based models,achieving classification accuracy of SAELDA(93.5%),SAE-SVM(94.9%)and SAE-SR(94.6%),respectively.In contrast,classification results of PCA-based methods such as PCA-LDA,PCA-SVM and PCA-SR were only 75.5%.85.7%and 77.1%,respectively.The results also suggested the increasing number of training samples have positive effects on classification models.Taking advantage of increasing dataset,the SAE-SR classification model finally performed better than others with average accuracy of 94.9%in classifying STEC serogroups.Specifically,O103 serogroup was classified with the highest accuracy of 97.4%,followed by O111(96.5%),O26(95.3%),O121(95%).O145(92.9%)and O45(92.4%),respectively.Thus,the HMI technology coupled with SAE-SR classification model has the potential for "Big-Six"identification.(3)Classification of foodborne bacteria using hyperspectral microscope imaging technology coupled with convolutional neural networks.Foodborne pathogens have become ongoing threats in the food industry,whereas their rapid detection and classification at an early stage are still challenging.To address early and rapid detection,hyperspectral microscope imaging(HMI)technology combined with convolutional neural networks(CNN)was proposed to classify foodborne bacterial species at the cellular level.HMI technology can simultaneously obtain both spatial and spectral information of different live bacterial cells,while two CNN frameworks,U-Net and onedimensional CNN(1D-CNN)were employed to accelerate the data analysis process.U-Net was used for automating cellular regions of interest(ROI)segmentation,which generated accurate cell-ROI masks in a shorter timeframe(0.4s)than the conventional Otsu or Watershed methods.The segmented cellular ROIs of U-net were similar to ground truth ROIs,resulting in 98%in mean pixel accuracy(MPA)and 0.96 in mean intersection-over-union(MIOU).The highest MPA and MIOU of Otsu and Watershed method were 94%and 0.83,84%and 0.77,respectively.In addition,the 1D-CNN was employed for classifying the spectral profiles extracted from cell-ROI and resulted in a higher accuracy(90%)than kNearest Neighbor(81%)and Support Vector Machine(81%).Overall,the CNN-assisted HMI technology showed potential for foodborne bacteria detection.(4)Rapid identification of foodborne bacteria based on hyperspectral microscope imaging and long short-term memory algorithms.Traditional analysis of spectra was usually composed of two steps,first conducting principle component analysis(PCA)to realize feature reduction,then adopting different classifiers.Here,a tiny artificial intelligent algorithm named long-short term memory(LSTM)network was successfully developed to assist hyperspectral microscopy imaging(HMI)method to differentiate five common foodborne pathogens without feature reduction.HMI is extremely powerful for characterizing living cells,with every pixel of the cell region contained abundant spectral information.Three regions of interest(ROI),including wholecell ROI,boundary ROI,and center ROI,were investigated to explore their performances on the classification task.Compared to current classic based classifiers such as latent discriminant analysis(PCA-LDA,66.0%),k-nearest network(PCA-KNN,74.0%),and support vector machine(PCA-SVM,85.0%),the LSTM classifier achieved the highest accuracy of 92.9%on center ROI dataset.Furthermore,the classification results indicated that LSTM-assisted HMI is capable of predicting spectra instantly once trained,eliminating the two-steps PCA feature analysis,becoming an efficient tool for foodborne pathogen detection.(5)Single-cell classification of foodborne pathogens using hyperspectral microscope imaging with deep learning frameworks.A high-throughput hyperspectral microscope imaging(HMI)technology with hybrid deep learning(DL)framework defined as "Fusion-Net" was proposed for rapid classification of foodborne bacteria at single-cell level.HMI technology is useful in single-cell characterization,providing spatial,spectral and combined spatial-spectral profiles with high resolution.However,direct analysis of these high-dimensional HMI data is challenging.In this work,HMI data were decomposed into three parts as morphological features,intensity images,and spectral profiles.Multiple advanced DL frameworks including long-short term memory(LSTM)network,deep residual network(ResNet),and one-dimensional convolutional neural network(1D-CNN)were utilized,achieving classification accuracies of 92.2%,93.8%,and 96.2%,respectively.Taking advantage of fusion strategy,individual DL framework was stacked to form "Fusion-Net" that processed these features simultaneously with improved classification accuracy of up to 98.4%.Our study demonstrated the ability of DL frameworks to assist HMI technology in single-cell classification as a diagnostic tool for rapid detection of foodborne pathogens.