研究目的
Investigating the use of hyperspectral microscope imaging (HMI) technology coupled with deep learning frameworks for rapid classification of foodborne bacteria at the single-cell level.
研究成果
The proposed HMI technology coupled with the 'Fusion-Net' DL framework demonstrated high accuracy in classifying foodborne bacteria at the single-cell level, offering a rapid, non-invasive, and label-free diagnostic tool. This approach eliminates the need for long-time culturing and promotes early bacteria detection, which is necessary for practical applications.
研究不足
The study focused on laboratory-prepared samples, and further testing is needed using food matrix for practical applications in the food industry and clinical laboratory.
1：Experimental Design and Method Selection:
The study employed a high-throughput hyperspectral microscope imaging (HMI) technology with a hybrid deep learning (DL) framework called 'Fusion-Net' for the classification of foodborne bacteria. The HMI data were decomposed into morphological features, intensity images, and spectral profiles for analysis.
2：Sample Selection and Data Sources:
Five common foodborne bacterial cultures (Campylobacter jejuni, generic Escherichia coli, Listeria innocua, Staphylococcus aureus, and Salmonella Typhimurium) were prepared and imaged using the HMI system.
3：List of Experimental Equipment and Materials:
The HMI system included an electron-multiplying charge-coupled device (EMCCD) camera, acousto-optic tunable filter (AOTF) spectrometer, upright microscope, and metal-halide light source. ImageJ software was used for data preparation.
4：Experimental Procedures and Operational Workflow:
Bacterial cultures were prepared, imaged using the HMI system, and the resulting hypercubes were processed to extract single-cell hypercubes. These were then decomposed into three types of data features for analysis.
5：Data Analysis Methods:
Multiple advanced DL frameworks (LSTM, ResNet, 1D-CNN) were utilized for classification, and a fusion strategy was employed to combine these frameworks into 'Fusion-Net' for improved accuracy.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
