Deep Learning-Enhanced Hand-Driven Microfluidic Chip for Multiplexed Nucleic Acid Detection Based on RPA/CRISPR.

Abstract

The early detection of high-risk human papillomavirus (HR-HPV) is crucial for the assessment and improvement of prognosis in cervical cancer. However, existing PCR-based screening methods suffer from inadequate accessibility, which dampens the enthusiasm for screening among grassroots populations, especially in resource-limited areas, and contributes to the persistently high mortality rate of cervical cancer. Here, a portable system is proposed for multiplexed nucleic acid detection, termed R-CHIP, that integrates Recombinase polymerase amplification (RPA), CRISPR detection, Hand-driven microfluidics, and an artificial Intelligence Platform. The system can go from sample pre-processing to results readout in less than an hour with simple manual operation. Optimized for sensitivity of 10^-17 M for HPV-16 and 10^-18 M for HPV-18, R-CHIP has an accuracy of over 95% in 300 tests on clinical samples. In addition, a smartphone microimaging system combined with the ResNet-18 deep learning model is used to improve the readout efficiency and convenience of the detection system, with initial prediction accuracies of 96.0% and 98.0% for HPV-16 and HPV-18, respectively. R-CHIP, as a user-friendly and intelligent detection platform, has great potential for community-level HR-HPV screening in resource-constrained settings, and contributes to the prevention and early diagnosis of other diseases.