Urinary Cell-Free DNA detection platform for bacterial cystitis diagnosis via spot-based machine learning

This is the authors' abstract. We don't add key points for this paper.

• Urinary cfDNA is a sensitive biomarker for cystitis, with elevated levels strongly correlating with disease severity. • A PEI-functionalized fluorescent biosensor enables rapid, high-throughput cfDNA analysis, revealing distinct spot size distributions that differentiate bacterial cystitis. • Integration with machine learning-assisted fluorescent biosensor achieves 94.43% diagnostic accuracy, offering a rapid, pretreatment-free point-of-care solution for precise bacterial cystitis diagnosis. Bacterial cystitis is a prevalent infectious condition that is primarily caused by uropathogenic Escherichia coli ( E. coli ). The rapid and precise identification of this pathogen is crucial for guiding timely and effective clinical intervention. Urinary cell-free DNA (cfDNA) has emerged as a promising biomarker for cystitis, yet its complex origins from both inflammatory and infectious processes pose challenges in distinguishing bacterial from non-bacterial etiologies. Here, we developed an innovative image-based, machine learning-assisted fluorescent detection system for simultaneous quantitative and qualitative analysis of urinary cfDNA to diagnose bacterial cystitis. The detection system utilizes polyethyleneimine (PEI)-functionalized glass slides for efficient cfDNA enrichment, followed by fluorescence imaging with propidium iodide labeling. The performance of the detection system was rigorously validated using a bacterial cystitis mouse model and human urine specimens. This platform enables rapid quantification of cfDNA within 20 min, with a calibration range of 31.25–4000 ng mL −1 and a detection limit of 20.29 ng mL −1 . Integration with the YOLOv5 object detection algorithm facilitates automated analysis of cfDNA fluorescence spot patterns, revealing distinct size distributions that differentiate bacterial (77.12 % smaller spots) from non-bacterial cystitis (13.87 % smaller spots). Applied to 148 clinical urine samples, this approach achieved a diagnostic accuracy of 94.43 %, with 94.45 % sensitivity and 94.35 % specificity for bacterial cystitis, markedly improving upon conventional methods (65.96 % specificity, 78.22 % sensitivity). Requiring no urine pretreatment, this rapid, on-site platform represents a transformative advance in point-of-care testing (POCT) for bacterial cystitis, with potential applications for digital biomarker detection in diverse inflammatory and infectious diseases.