Deep Learning Enabled Universal Multiplexed Fluorescence Detection for Point-of-Care Applications.

Abstract

There is a significant demand for multiplexed fluorescence sensing and detection across a range of applications. Yet, the development of portable and compact multiplexable systems remains a substantial challenge. This difficulty largely stems from the inherent need for spectrum separation, which typically requires sophisticated and expensive optical components. Here, we demonstrate a compact, lens-free, and cost-effective fluorescence sensing setup that incorporates machine learning for scalable multiplexed fluorescence detection. This method utilizes low-cost optical components and a pretrained machine learning (ML) model to enable multiplexed fluorescence sensing without optical adjustments. Its multiplexing capability can be easily scaled up through updates to the machine learning model without altering the hardware. We demonstrate its real-world application in a probe-based multiplexed Loop-Mediated Isothermal Amplification (LAMP) assay designed to simultaneously detect three common respiratory viruses within a single reaction. The effectiveness of this approach highlights the system's potential for point-of-care applications that require cost-effective and scalable solutions. The machine learning-enabled multiplexed fluorescence sensing demonstrated in this work would pave the way for widespread adoption in diverse settings, from clinical laboratories to field diagnostics.