Electro-Optical Multiclassification Platform for Minimizing Occasional Inaccuracy in Point-of-Care Biomarker Detection

Changhao Dai,Huiwen Xiong,Rui He,Chenxin Zhu,Pintao Li,Mingquan Guo,Jian Gou,Miaomiao Mei,Derong Kong,Qiang Li,Andrew Thye Shen Wee,Xueen Fang,Jilie Kong,Yunqi Liu,Dacheng Wei
DOI: https://doi.org/10.1002/adma.202312540
Abstract:On-site diagnostic tests that accurately identify disease biomarkers lay the foundation for self-healthcare applications. However, these tests routinely rely on single-mode signals and suffer from insufficient accuracy, especially for multiplexed point-of-care tests (POCTs) within a few minutes. Here, this work develops a dual-mode multiclassification diagnostic platform that integrates an electrochemiluminescence sensor and a field-effect transistor sensor in a microfluidic chip. The microfluidic channel guides the testing samples to flow across electro-optical sensor units, which produce dual-mode readouts by detecting infectious biomarkers of tuberculosis (TB), human rhinovirus (HRV), and group B streptococcus (GBS). Then, machine-learning classifiers generate three-dimensional (3D) hyperplanes to diagnose different diseases. Dual-mode readouts derived from distinct mechanisms enhance the anti-interference ability physically, and machine-learning-aided diagnosis in high-dimensional space reduces the occasional inaccuracy mathematically. Clinical validation studies with 501 unprocessed samples indicate that the platform has an accuracy approaching 99%, higher than the 77%-93% accuracy of rapid point-of-care testing technologies at 100% statistical power (>150 clinical tests). Moreover, the diagnosis time is 5 min without a trade-off of accuracy. This work solves the occasional inaccuracy issue of rapid on-site diagnosis, endowing POCT systems with the same accuracy as laboratory tests and holding unique prospects for complicated scenes of personalized healthcare.
What problem does this paper attempt to address?