Introduction: There is an urgent need for rapid, high-throughput phenotypic antimicrobial susceptibility testing (AST) capable of assessing a microbial sample's susceptibility to multiple antibiotics.

Objectives: In this study, we have established a multiplexed rapid AST platform that employs droplet microfluidics for high-throughput single-cell based analysis, 2D angle-resolved light scattering for growth detection, and fluorescence detection via optical fibers to identify the antibiotic condition within each droplet.

Methods: For this, multiple antibiotic conditions are coded with fluorescence dyes and encapsulated with single cells to enable the testing of multiple antibiotics in a single experiment. We utilize convolutional neural networks (CNNs) and statistical models to assess the growth of various Staphylococcus aureus strains and determine the probability of susceptibility to different antibiotics.

Results: Our platform achieved a 95% categorical agreement with the disc diffusion reference method after just three hours of incubation, demonstrating the same level of accuracy as the established VITEK 2 system for the tested strains and antibiotics. Notably, our platform reduced the incubation time by 5-11 h compared to VITEK 2 and by 13-17 h compared to the gold standard disc diffusion method.

Conclusions: With the presented innovations, our technology takes a big step towards realizing true phenotypic determination of antibiotic resistance profiles for timely antimicrobial treatment decisions.

Keywords: Angle-resolved light scattering imaging; Convolutional neural networks; Droplet microfluidics; Rapid antibiotic susceptibility testing; Staphylococcus aureus.

Copyright © 2025 The Author(s). Published by Elsevier B.V. All rights reserved.