Deep Learning-Enabled Magnetic Bead-Capture Fluorescence In Situ Hybridization Enables Multiplex Identification and Digital Counting of Pathogenic Bacteria

Abstract Urinary tract infections (UTIs) are among the most prevalent infectious diseases worldwide. For complicated and antimicrobial-resistant cases, timely diagnosis is critical for guiding targeted therapy and improving patient outcomes. While conventional urine culture suffers from prolonged turnaround times and low positive detection rates, fluorescence in situ hybridization (FISH) has emerged as a rapid, highly specific alternative. However, its clinical application remains hindered by complicated sample pretreatment, insufficient sensitivity that often requires pre-enrichment by culture, and subjective bias in manual interpretation. To address these bottlenecks, we developed deep learning-enabled magnetic bead-capture fluorescence in situ hybridization (DeepMag-FISH), a culture-free platform that employs artificial intelligence (AI)-assisted digital counting for rapid identification and single-bacterium enumeration of pathogens. We adopted soluble β2-glycoprotein I (sβ2GPI)-coated magnetic beads for efficient pathogen enrichment, and utilized a modular dual-probe labeling strategy for multiplexed hybridization, thus enabling highly specific molecular labeling. Additionally, we integrated this sample processing pipeline with a YOLOv11-based deep learning framework for automatic multiscale feature extraction and precise single-bacterium counting. Validated using spiked urine specimens, DeepMag-FISH clearly differentiated Escherichia coli, Vibrio parahaemolyticus, and Klebsiella pneumoniae with negligible spectral crosstalk. The platform reaches a detection capability of 30 CFU/mL, completing the entire workflow, including all magnetic separation and washing steps, within approximately 3.5 h. By combining magnetic bead-based pathogen enrichment with AI-driven image analysis, this approach mitigates matrix interference and subjective bias in conventional methods, offering a rapid and quantitative approach for pathogen detection in spiked urine samples. Beyond UTI management, DeepMag-FISH may also be explored for potential applications in other fields, such as bloodstream infection diagnosis and food safety monitoring.