AstraDx: Sub-Hour AST Via Computational Image Processing

PROJECT SUMMARY/ABSTRACT Antimicrobial susceptibility testing (AST) for bacterial bloodstream infections (BSIs) is dangerously slow. Starting from a positive blood culture, AST currently takes a median of 15 hours, and often takes days. Until results are available, there is no choice but to treat patients with empiric antibiotics and hope the pathogen is susceptible. This is unacceptable. BSIs are serious, rapidly fatal infections that kill a quarter-million Americans each year, including a third of all hospital deaths. Empiric antibiotics destroy patients’ normal flora, predisposing them to additional, untreatable infections. Every hour a patient is on the wrong antibiotic increases risk of death by 7.6%. This key problem has made rapid, affordable AST a global priority. To address this problem, AstraDx has developed a low-cost new AST device that we hypothesize can perform AST in under an hour. The purpose of this proposal is to test this hypothesis. Preliminary studies involving multiple experiments on 19 strains vs. 8 standard antibiotics have demonstrated growth detection in 31±18 minutes (95% range, 20-63 minutes). Strains tested included methicillin-resistant S. aureus (MRSA), vancomycin-resistant Enterococci (VRE), carbapenem-resistant A. baumannii (CRAB), carbapenem-resistant Enterobacterales (CRE), multidrug-resistant P. aeruginosa—all high priority per the CDC—as well as corresponding susceptible strains. The innovation that makes these results possible is a new computational image-processing pipeline, which enables extremely sensitive and robust detection of an antibiotic’s effect on growth, often in less than the strain’s doubling time. A second innovation is the use of inexpensive standard components, leading to an affordable device and consumables. These innovations support our long-term goal of making rapid AST universally available. We will test our hypothesis through three tightly focused specific aims. Aim 1 will generate training data on diverse bug-drug combinations, with bacteria from simulated- positive blood cultures grown in the presence of doubling dilutions of antibiotic on 384-well plates, similar to the current phenotypic gold standard, with multiple antibiotics on each plate. Aim 2 will use the images from Aim 1 to create a unified picture of each strain’s behavior in the presence of each antibiotic, which we will use to refine how we determine minimum inhibitory concentrations (MICs) in order to maximize categorical agreement with the CLSI gold standard. Finally, Aim 3 will validate the method from Aim 2 by testing additional bacterial strains and measuring categorical agreement with the gold standard, per FDA standard guidelines. The expected outcome is a device proven to achieve sub-hour AST with FDA-level performance on a wide variety of pathogens, establishing technical merit, feasibility, and commercial potential for an SBIR Phase II. Our advanced technology, a clearly identified product, will help meet the critical need for rapid, highly sensitive infectious-disease diagnostics to address antimicrobial resistance for life-threatening BSIs and sepsis.