A Deep Learning Assisted Smartphone Platform for Screening of Alzheimer’s Disease Using a Microfluidic Paper-based Analytical Device

This paper reports a smartphone platform assisted by deep learning algorithms for paper-based colorimetric enzyme-linked immunosorbent assay (c-ELISA) detection of blood biomarker, β-amyloid peptide 1-42 (Aβ 1-42), for early-stage screening of Alzheimer’s disease (AD). Unlike the conventional smartphone platform used with paper-based c-ELISA, our platform is free from the influences of ambient lights without using any additional accessories and can produce quantitative c-ELISA results automatically from raw phone-taken photos. We also developed an Android application for ease-of-interfacing. These advantages are attributed to the customized deep learning algorithms. Our smartphone platform is particularly suitable for being integrated with paper-based microfluidic devices, for AD screening in rural areas with low-cost, robustness and ease-of-operation. We performed a proof-of-concept experiment using Aβ 1-42 (n=6) in buffer solution and achieved an accuracy of 93.3%.