Rapid culture-free diagnosis of clinical pathogens via integrated microfluidic-Raman micro-spectroscopy.

Abstract

Antimicrobial resistance (AMR) is a critical global health challenge, demanding rapid and accurate diagnostics to guide timely antimicrobial therapy. Current diagnosis is hindered by prolonged culturing and difficulties detecting low pathogen loads. Here, we present a culture-free diagnostic platform that integrates microfluidics, Raman micro-spectroscopy, and deep learning to deliver "sample-to-report" testing within 20 min. The microfluidic enrichment system employs dialysis-dielectrophoresis (DEP) technology to rapidly isolate pathogens directly from clinical samples with a detection limit as low as <2 colony forming unit (CFU)/ml. Combining a single-cell Raman fingerprint database of 342 clinical isolates from 29 bacterial and 7 fungal species with a 1D ResNet deep learning model, our approach achieved 95.1% accuracy in lab settings. Validated in a 305-patient clinical study involving primary urine and other clinical samples, it demonstrated 95.4% agreement with traditional culture methods and 98.5% sensitivity in diagnosing infections. While broader validation is needed for clinical implementation, the integrated, rapid diagnosis pipeline, as well as broad-spectrum detection, offer a promising solution for next-generation diagnostics for combating AMR.