StratoLAMP-2: A Microfluidics-Free, Deep-Learning Platform for Multiplex Digital Molecular Diagnostics

Accurate and accessible molecular diagnostics are critical for early disease detection, infection control, and personalized medicine. Digital nucleic acid testing offers absolute quantification by isolating individual amplification events into discrete compartments. However, most current approaches require microfluidics to generate monodisperse droplets and fluorescence for multiplex detection, limiting scalability and accessibility. Here, we present StratoLAMP-2, a microfluidics-free, label-free digital nucleic acid quantification platform for multiplex digital molecular diagnostics, using vortex-generated polydisperse droplets and magnesium pyrophosphate precipitate as the visual readout. Target identity is encoded through primer concentration-driven stratification of precipitate levels, distinguishing droplets containing Target 1, Target 2, or both. To address the analytical challenges posed by droplet size variability and precipitate heterogeneity, we first developed a rule-based classification method using an optical density-droplet diameter phase diagram, followed by a deep learning approach for droplet segmentation, tracking, and classification, which achieved superior accuracy and robustness. Combined with volume-aware Poisson modeling, StratoLAMP-2 enables multiplexed absolute quantification of nucleic acids over a broad dynamic range. This approach offers a clinically relevant, low-cost, and scalable alternative to conventional digital assays, paving the way for point-of-care molecular diagnostics in decentralized and underserved environments.