Deep Learning-Enhanced Hand-Driven Spatial Encoding Microfluidics for Multiplexed Molecular Testing at Home

Abstract

The frequent global outbreaks of viral infectious diseases have significantly heightened the urgent demand for molecular testing at home. However, the labor-intensive sample preparation and nucleic acid amplification steps, along with the complexity and bulkiness of detection equipment, have limited the large-scale application of molecular testing at home. Here, we propose artificial intelligence-enhanced hand-driven microfluidic system (MACRO) based on RPA and CRISPR technologies for home diagnosis of multiple types of infectious diseases. Leveraging a multidimensional space hourglass structure design, precise spatiotemporal control of fluids can be achieved simply by flipping the chip. Through dual chemical reactions, the system eliminates the need for nucleic acid extraction and purification, simplifying sample preparation and obviating the reliance on heating equipment. The MACRO achieves attomolar sensitivity within 60 min from sample input to result, and 100% specificity for 27 HPV subtypes. Clinical validation using 140 cervical swab specimens demonstrated 98.57% accuracy with 100% specificity. Further, we validated MACRO through multiplex detection of three clinically critical respiratory pathogens (SARS-CoV-2, Influenza A, and Influenza B) in 70 samples, achieving 100% diagnostic concordance. To circumvent subjective errors and enable real-time data collection, we further developed a mobile health platform based on the YoLov8 image recognition algorithm to ensure rapid and precise result output. With the performance of cost-effectiveness ($1.34 per target), and independence from instrument support, MACRO provides a comprehensive solution for molecular testing at home, offering significant implications for enhancing early warning systems for major epidemics and improving public health emergency response capabilities.