MorphoAST: A Machine Learning Workflow for Sub-Doubling Time Single Cell Antimicrobial Susceptibility Testing

Abstract

Multidrug-resistant Enterobacteriaceae are among the most urgent global public health threats associated with various life-threatening infections. In the absence of a rapid method to identify antimicrobial susceptibility, empirical use of broad-spectrum antimicrobials such as carbapenem monotherapy has led to the spread of resistant organisms. Rapid determination of antimicrobial resistance is urgently needed to overcome this issue. By capturing dynamic single-cell morphological features of over thirty-nine thousand cells from nineteen strains of Klebsiella pneumoniae, we evaluated strategies based on time and concentration differentials for classifying its susceptibility to a commonly used carbapenem, meropenem, and predicting their minimum inhibitory concentrations (MIC). We report morphometric antimicrobial susceptibility testing (MorphoAST), an image-based machine learning workflow, for rapid determination of antimicrobial susceptibility by single-cell morphological analysis within sub-doubling time. We demonstrated that our algorithm has the ability to predict MIC/antimicrobial susceptibility in a fraction of the bacterial doubling time (<50 min.). The classifiers achieved as high as 97% accuracy in 20 minutes (two-fifths of the doubling time) and reached over 99% accuracy within 50 minutes (one doubling time) in predicting the antimicrobial response. A regression model based on the concentration differential of individual cells from nineteen strains predicted the MIC with 100% categorical agreement and essential agreement for seven unseen strains, including two clinical samples from patients with urinary tract infections with different responsiveness to meropenem. The expansion of this innovation to other drug-bug combinations could have significant implications for future development of rapid antimicrobial susceptibility testing.