A portable deep-learning-assisted digital single-particle counting biosensing platform for amplification-free nucleic acid detection using a lens-free holography microscope

Abstract

The digital single-particle assay has emerged as a highly promising technology in various detection applications, such as food safety inspection, environmental monitoring, and in vitro diagnosis. Conventional digital assays rely on pre-amplification and expensive equipment, which limits their practical applications to point-of-care testing. Herein, we report a deep-learning-assisted digital single-particle counting biosensing platform for nucleic acid detection without pre-amplification using a portable and low-cost lens-free holography microscope. This device can perceive the number change of signal probes and capture microsphere probe holograms, which is ultra-lightweight (∼ 318 g), and has a low cost (∼ $70) and an ultrawide field of view of 24.396 mm2. The improved YOLOv7-based deep learning algorithm is trained to detect small objects (∼ 10 μm) in high-resolution images with high throughput. As a proof of concept, our strategy has successfully distinguished viable and nonviable Salmonella typhimurium quantitatively with high sensitivity (72 CFU/mL) without pre-amplification using phage-mediated DNA extraction and has been verified in various real samples. It has also been successfully applied to detecting non-nucleic acid targets in real samples, including procalcitonin and chloramphenicol. As a versatile and multi-functional platform, this platform exhibits excellent potential for point-of-care multi-type target detection in resource-limited settings.