Droplet microfluidics with image texture quantification for detection of rare antibiotic-resistant subpopulations from bloodstream infections

Abstract Heteroresistance (HR) is an antibiotic resistance phenotype characterized by the presence of rare resistant subpopulations (frequency ≈ 10 −7 to 10 −4 ) within a main susceptible bacterial population. During antibiotic exposure, these subpopulations can be enriched and cause treatment failure. Standard antibiotic susceptibility testing (AST) often fails to detect HR, and the current gold-standard population analysis profile (PAP) test is labor-intensive and time-consuming. We present a digital phenotyping approach combining droplet microfluidics with image texture to detect HR from clinical isolates, including Gram-negative ( Klebsiella pneumoniae, Pseudomonas aeruginosa, Acinetobacter baumannii ) and Gram-positive ( Staphylococcus aureus ) bacteria isolated from bloodstream infections. Our method achieves detection at subpopulation frequencies as low as 10 −6 in 12 to 30 h, depending on bacterial species, which is faster than the PAP test, together with single-cell resolution and high-throughput. This computationally assisted microfluidic platform enables rapid and accurate identification of HR, representing a step toward targeted antibiotic therapy in critical infections.