Deep Learning-Enabled Multiplexed Point-of-Care Sensor using a Paper-Based Fluorescence Vertical Flow Assay

Area of Science:

• Biomedical Engineering
• Point-of-Care Diagnostics
• Biosensing Technologies

Background:

• Acute cardiac injury diagnosis relies on timely biomarker quantification.
• Current diagnostic methods can be time-consuming and require specialized equipment.
• Multiplexed biomarker detection offers a more comprehensive assessment of cardiac status.

Purpose of the Study:

• To develop and validate a multiplexed computational sensing platform for simultaneous quantification of three cardiac injury biomarkers.
• To assess the performance of a paper-based fluorescence vertical flow assay (fxVFA) coupled with neural network-based inference.
• To evaluate the potential of this point-of-care (POC) assay for resource-limited settings.

Main Methods:

• Development of a paper-based fluorescence vertical flow assay (fxVFA) for multiplexed biomarker detection.
• Utilized a low-cost mobile reader for data acquisition and analysis.
• Employed trained neural networks for biomarker quantification from serum samples.
• Validated the assay using human serum samples for myoglobin, creatine kinase-MB, and heart-type fatty acid binding protein.

Main Results:

• Achieved a limit-of-detection below 0.52 ng mL⁻¹ for all three cardiac biomarkers with minimal cross-reactivity.
• Demonstrated high correlation (>0.9 linearity) and low coefficient of variation (<15%) compared to ground truth concentrations.
• Quantification was performed in under 15 minutes using only 50 µL of serum.
• Blind testing on 46 cartridges confirmed reliable biomarker concentration measurements.

Conclusions:

• The multiplexed computational fxVFA is a promising point-of-care diagnostic tool for acute cardiac injury.
• Its paper-based design, low cost, and rapid results enhance accessibility to diagnostics.
• This technology has significant potential for deployment in resource-limited healthcare settings.