|KANG, RUI - Nanjing Agricultural University|
|OUYANG, QIN - Jiangsu University|
|CHEN, KUN-JIE - Nanjing Agricultural University|
Submitted to: Sensors and Actuators B: Chemical
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 1/29/2020
Publication Date: 2/10/2020
Publication URL: https://handle.nal.usda.gov/10113/6949597
Citation: Kang, R., Park, B., Eady, M.B., Ouyang, Q., Chen, K. 2020. Single-cell classification of foodborne pathogens using hyperspectral microscope imaging coupled with deep learning frameworks. Sensors and Actuators B: Chemical. https://doi.org/10.1016/j.snb.2020.127789.
Interpretive Summary: A series of foodborne pathogen outbreaks increase threat to the public, and major outbreaks of foodborne illness are caused by several pathogens includes Campylobacter, Escherichia coli, Listeria, Salmonella, and Staphylococcus. In this study, a rapid, high-throughput hyperspectral microscope imaging (HMI) technology with hybrid deep learning (DL) frameworks defined as “Fusion-Net” is proposed for rapid identification of foodborne bacteria at a cellular level. HMI method offers new opportunity to directly acquire both image and spectral information of bacterial cells without any fluorescent labeling materials. Based on the HMI data, the proposed advanced DL frameworks have the capability of processing information of spatial and spectral features from hypercubes simultaneously with high classification accuracy. Compared to the traditional machine learning algorithms, the DL-assisted HMI technology achieves fast prediction without redundant feature analysis, demonstrating the potential for classification of major foodborne bacteria rapidly and accurately.
Technical Abstract: A high-throughput hyperspectral microscope imaging (HMI) technology with hybrid deep learning (DL) frameworks defined as “Fusion-Net” is proposed for rapid identification of foodborne bacteria at a single-cell level. HMI technology is useful for characterization of bacterial cells, providing spatial, spectral and combined spatial-spectral profiles with high resolution, yet direct analysis of these high-dimensional HMI data is challenging. In this study, HMI data were decomposed into three features including morphology, intensity distribution, and spectral profiles of Campylobacter, E. coli, Listeria, Staphylococcus, and Salmonella. Multiple advanced DL frameworks such as long-short term memory (LSTM) network, deep residual network (ResNet), and one-dimensional convolutional neural network (1D-CNN) were employed for model development, achieving classification accuracies of 92.2%, 93.8%, and 96.2%, respectively. In addition, taking advantage of fusion strategy, individual DL framework was stacked to form “Fusion-Net” that processed aforementioned three features simultaneously, resulted in an improved classification accuracy of 98.4%. Our study demonstrates the ability of DL frameworks to assist HMI technology for single-cell classification as a diagnostic tool for rapid detection of foodborne bacteria.