Hardware-Free Testing for Antimicrobial Resistance Using Artificial Intelligence

ABSTRACT

Antimicrobial resistance (AMR) is one of the most challenging public health problems, and implementation of effective and accessible testing solutions is an ever-increasing unmet need. Artificial intelligence (AI) offers a promising avenue for enhanced testing performance and accuracy. We introduce an AI system specifically designed for rapid AMR testing, eliminating the requirement for bulky hardware and extensive automation. Our system incorporates a novel approach for nanotechnology-empowered intelligent diagnostics (NEIDx), leveraging nanoparticles to enable novel AI-based advanced systems for detection. We employ catalytic nanoparticle-based NEIDx coupled with magnetic separation to facilitate the direct detection of AMR-associated enzymes from blood samples. This is achieved through the formation of easily visible and detectable large bubbles, a process streamlined by AI running on a cellphone. We evaluated the performance of our AI system using two clinically relevant AMR enzymes: Klebsiella pneumoniae carbapenemase-2 (KPC-2) and Sulfhydryl variable-1 (SHV-1) β-lactamases. The system demonstrated qualitative detection with a sensitivity of 82.61% (CI of 79.7 - 85.5%) and a specificity of 92.31% (CI of 90.3 - 94.3%) in blood samples, respectively. This innovative approach holds significant promise for advancing point-of-care diagnostics and addressing the urgent need for rapid and accessible AMR testing in diverse healthcare settings.