Deep learning enabled rapid detection of live bacteria in the presence of food debris - PubMed (original) (raw)

Deep learning enabled rapid detection of live bacteria in the presence of food debris

Hyeon Woo Park et al. NPJ Sci Food. 2025.

Abstract

The contamination of food with pathogenic bacteria is a major public health concern, requiring rapid and accurate detection methods. Conventional approaches, such as culture-based or molecular assays, are time-consuming, labor-intensive, and often demand specialized expertise. Here, we developed a deep learning-based strategy for rapid detection and classification of live bacteria using simple white-light microscopic images of microcolonies, even in the presence of morphologically similar food debris. The model, based on ResNet50 with a Region Proposal Network, was trained on Escherichia coli, Listeria monocytogenes, Bacillus subtilis, and debris from chicken, spinach, and cheese. The model trained on bacteria misclassified debris as bacteria (24.2% false positives), whereas the model trained on both bacteria and food debris achieved 0% false positives with 100% precision and 94.4% recall. Validation with GFP-producing B. subtilis in food matrices further confirmed robust performance (mPrecision 94.6%, mRecall 92.5%). This cost-effective method enables reliable bacterial detection in complex foods within 3 h.

© 2025. The Author(s).

PubMed Disclaimer

Conflict of interest statement

Competing interests: The authors declare no competing interests.

Figures

Fig. 1

Fig. 1

Pipeline of the AI-based bacterial detection model, illustrating the process from sample preparation to model application for the detection of target bacteria.

Fig. 2

Fig. 2

Representative images of bacterial microcolonies and food debris on agar plate captured using a phase-contrast microscope with a 60× objective.

Fig. 3

Fig. 3. Bacterial classification using the deep convolutional neural network trained on Listeria monocytogenes, Escherichia coli, and Bacillus subtilis.

a Representative images of object detection using the AI model to detect target bacteria; b Confusion matrices at different confidence levels.

Fig. 4

Fig. 4. Object detection in food debris images using deep convolutional neural networks trained with and without debris.

a Predictions based on the model trained using bacterial microcolonies; b Predictions based on the model trained using both bacteriaand food debris.

Fig. 5

Fig. 5. Influence of food debris on bacterial detection and classification performance.

a True negative rates of food debris images at different confidence levels. Precision and recall curves of bacterial species in relation to the true negative rate of the corresponding food debris: b Listeria monocytogenes and spinach, c Escherichia coli and Cotija cheese, and d Bacillus subtilis and chicken breast. “Unannotated” and “annotated” refer to models trained on bacteria only and on bacteria with food debris, respectively.

Fig. 6

Fig. 6. Validation of deep learning-based detection of Bacillus subtilis microcolonies in the presence of various food debris.

a Phase-contrast image, b Fluorescence image, and c Model inference. Fluorescence images were used to locate B. subtilis microcolonies in the presence of food debris, serving as a reference for validating the deep learning-based bacterial detection.

Fig. 7

Fig. 7

Architecture of the bacterial microcolony detection model using white light microscopy images as input, followed by a convolutional backbone network, a regional proposal network, and object detection and classification.

References

    1. Bakhtiary, F. et al. Evaluation of bacterial contamination sources in meat production line. J. Food Qual.39, 750–756 (2016). -DOI
    1. Chatterjee, A. & Abraham, J. Microbial contamination, prevention, and early detection in food industry. in Microbial Contamination and Food Degradation 21–47 (Elsevier, 2018).
    1. Park, H. W. & Park, S. Inactivation kinetics of Escherichia coli O157:H7 on baby-cut carrots during gaseous chlorine dioxide treatment: the impact of relative humidity. Food Bioprocess Technol.17, 4195–4203 (2024). -DOI
    1. Park, H. W., Balasubramaniam, V. M., Heldman, D. R., Cai, S. & Snyder, A. B. Computational fluid dynamics analysis of superheated steam’s impact on temperature and humidity distribution within enclosed dry food processing spaces. J. Food Eng.360, 111718 (2024). -DOI
    1. Ma, L., Yi, J., Wisuthiphaet, N., Earles, M. & Nitin, N. Accelerating the detection of bacteria in food using artificial intelligence and optical imaging. Appl. Environ. Microbiol.89, e01828–22 (2023). -PMC -PubMed

Grants and funding

LinkOut - more resources