Detecting Escherichia coli contamination on plant leaf surfaces using UV-C fluorescence imaging and deep learning

Authors: VADDI, SNEHIT - University Of Florida; BURKS, THOMAS - University Of Florida; IQBAL, ZAFAR - University Of Florida; YADAV, PAPPU - South Dakota State University; FREDERICK, QUENTIN - University Of Florida; OBELLANENI, SATYA - University Of Florida; Qin, Jianwei; Kim, Moon; RITENOUR, MARK - University Of Florida; ZHANG, JIUXU - University Of Florida; VASEFI, FARTASH - Safetyspect Inc

Submitted to: Plants
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/28/2025
Publication Date: 10/31/2025
Citation: Vaddi, S., Burks, T., Iqbal, Z., Yadav, P.K., Frederick, Q., Obellaneni, S.A., Qin, J., Kim, M.S., Ritenour, M., Zhang, J., Vasefi, F. 2025. Detecting Escherichia coli contamination on plant leaf surfaces using UV-C fluorescence imaging and deep learning. Plants. 14(21). Article 3352. https://doi.org/10.3390/plants14213352.
DOI: https://doi.org/10.3390/plants14213352

Interpretive Summary: Bacterial pathogens, such as E. coli, can be transmitted through contaminated fruits and vegetables that can cause severe health issues for consumers. Rapid and reliable detection methods are important to reduce food safety risks and foodborne diseases. This study developed a method for detecting E. coli on plant leaves using a contamination, sanitization inspection and disinfection (CSI-D) handheld fluorescence imaging device that was commercialized based on an ARS patented technology. Fluorescence images were acquired from eight concentrations of E. coli contamination inoculated on citrus and spinach leaf samples. Image classifications were conducted using state-of-the-art deep learning algorithms, including EfficientNet, ConvNeXt, and YOLO. The best accuracies to classify E. coli concentration levels were achieved at 85.93% (citrus leaves) and 92.00% (spinach leaves) using the YOLO models. The combination of the CSI-D handheld imaging and deep learning techniques would benefit the food industry and the regulatory agencies by enabling timely interventions to prevent contaminated produce from reaching the consumers.

Technical Abstract: The transmission of Escherichia coli through contaminated fruits and vegetables poses serious public health risks and has led to several national outbreaks in the USA. To enhance food safety, rapid and reliable detection of E. coli on produce is essential. This study evaluated the performance of the CSI-D+ system combined with deep learning to detect varying concentrations of E. coli on plant leaves such as citrus and spinach. Eight levels of E. coli contamination (ranging from 0 to 108 colony forming units (CFU)/mL) were inoculated onto citrus and spinach leaf specimens, and fluorescence images were captured for each sample. After several post-processing operations (segmentation, quadrant division, wavelet denoising, and augmentation) on the captured images, multiple deep learning (DL) models, including EfficientNet, ConvNeXt, and five YOLO11 variants (n, s, m, l, x), were trained to classify E. coli concentration levels. Additionally, Eigen-CAM heatmaps were used to visualize the spatial responses of the models to bacterial presence. All YOLO11 models outperformed EfficientNet and ConvNeXt. In particular, YOLO11s-cls was identified as the best-performing model, achieving average validation accuracies of 88.43% (citrus) and 92.03% (spinach), and average test accuracies of 85.93% (citrus) and 92.00% (spinach). This model demonstrated an inference speed of 0.011 seconds per image with a size of 13 MB. These findings support the potential of fluorescence-based imaging combined with deep learning for rapid E. coli detection, enabling timely interventions to prevent contaminated produce from reaching consumers.