An Edge-AI Fluorescence Detection System for Digital Nucleic Acid Analysis

Digital nucleic acid analysis enables absolute quantification of nucleic acids compared to other detection methods. However, its reliance on expensive fluorescence microscopes and computing equipment for analyzing thousands of microscale reaction chambers limits its applications in point-of-care testing (POCT). To overcome this limitation, we developed a digital nucleic acid fluorescence detection system capable of fluorescence imaging on an array-based microfluidic chip containing 20,096 chambers with 100 μm in diameter. Our improved YOLOv11n model enables precise detection of positive chambers with 98.9% precision and 97.5% recall rates. By quantizing and deploying this model on an edge computing device, we accomplished the detection of positive reaction chambers across the entire microfluidic chip within 1 second.