Deep-learning-enhanced automated coherent-light diffraction system for high-speed, highly accurate strain-specific foodborne bacterial recognition

Authors

• Yang TianFollow
• Ziyu Liu, University of Arkansas, Fayetteville
• Yinuo Huang, University of Arkansas, Fayetteville
• Ramesh Bahadur Bist, University of Arkansas, Fayetteville
• Yiting Xiao, University of Arkansas, Fayetteville
• Samantha Marrianne Howe, University of Arkansas, Fayetteville
• Tsung Cheng Tsai, University of Arkansas, Fayetteville
• Terry Howell, University of Arkansas, Fayetteville
• Jeyam Subbiah, University of Arkansas, Fayetteville
• Michael L. Looper, University of Arkansas, Fayetteville
• Dongyi Wang, University of Arkansas, FayettevilleFollow

Document Type

Article

Publication Date

12-2025

Keywords

Laser diffraction, Deep learning, Bacterial strain differentiation, Automated colony targeting

Abstract

Rapid, accurate differentiation of bacterial pathogens at both species and strain levels is critical for clinical diagnostics, food safety, and epidemiological surveillance. We report a fully automated laser-diffraction platform that achieves rapid, label-free identification of bacterial pathogens at both species and strain levels. The system integrates dual cameras, an automated X-Y stage, and deep-learning classifiers to deliver sub-second, nondestructive diagnostics without reagents or culture. Using diffraction patterns from eight strains, including Escherichia coli O157:H7, K12, ATCC 25922, F18, Salmonella enterica, Listeria monocytogenes, Listeria innocua, and Staphylococcus aureus, a custom YOLO11x model attained 100 % Top-1 accuracy under five-fold stratified cross-validation with zero misclassifications. PCR testing further verified results in poultry samples. Whole-genome sequencing and comparative annotation (Prokka, PhyloPhlAn, DIAMOND against CARD/VFDB/mobileOG-db) confirmed phylogenetic relationships and functional gene profiles. The system's modular design allows extension to mixed cultures and potential interventions such as photoablation, representing a cost-effective, high-throughput solution for clinical diagnostics, food safety monitoring, and epidemiological surveillance.

Comments

Web of Science

Elsevier

Citation

Yang Tian, Ziyu Liu, Yinuo Huang, Ramesh Bahadur Bist, Yiting Xiao, Samantha Marrianne Howe, Tsung Cheng Tsai, Terry Howell, Jeyam Subbiah, Michael L. Looper, Dongyi Wang, Deep-learning-enhanced automated coherent-light diffraction system for high-speed, highly accurate strain-specific foodborne bacterial recognition, Journal of Agriculture and Food Research, Volume 24, 2025, 102357, ISSN 2666-1543, https://doi.org/10.1016/j.jafr.2025.102357.

Creative Commons License

This work is licensed under a Creative Commons Attribution-NonCommercial-No Derivative Works 4.0 International License.