Deep Learning Assisted Ultra-Accurate Smartphone Testing of Paper-Based ELISA Assays

This paper reports a deep learning assisted smartphone platform for ultra-accurate testing of paper-based microfluidic colorimetric enzyme-linked immunosorbent assay (c-ELISA). Unlike existing platforms, our platform is capable of fully automated extracting c-ELISA features from raw photos and achieved ultra-high accuracy (>97%) for quantitatively classification/prediction of c-ELISA results. The superior performance of our platform can be attributed to the customized deep learning algorithms. The unique feature of our platform further strengthens the advantages of paper-based microfluidics for being used by laypersons in poor resources settings and holds great promise for real-world use.