AI-Enhanced Lateral Flow Assay Enables 3-Minute Quantitative Detection with Laboratory-Grade Accuracy.

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Diagnostics
• Point-of-Care Testing

Background:

• Lateral flow immunoassays (LFAs) are common for point-of-care testing (POCT) but often lack speed and quantitative accuracy.
• Traditional LFAs can take up to 30 minutes and provide only qualitative results, limiting their use in time-sensitive diagnostic scenarios.

Purpose of the Study:

• To develop a novel quantitative lateral flow assay (RAD-LFA) that overcomes the speed and accuracy limitations of conventional LFAs.
• To integrate deep learning modules for rapid and precise quantification of target analytes.

Main Methods:

• Developed RAD-LFA incorporating a Residual Network (ResNet) for spatial feature extraction and a DyFormer module for dynamic temporal modeling.
• Validated RAD-LFA using datasets for Coronavirus disease 2019 (COVID-19) and hepatitis B virus (HBV).
• Evaluated performance through qualitative accuracy, quantitative correlation (R^2), and clinical blind tests.

Main Results:

• RAD-LFA achieved precise quantification within 3 minutes, significantly reducing assay time.
• Improved qualitative detection accuracy by 15% compared to expert visual interpretation.
• Demonstrated robust quantitative performance with R^2 values up to 0.9985 in clinical tests, alongside 94% overall accuracy, 95% sensitivity, and 92% specificity.

Conclusions:

• RAD-LFA offers a significant advancement in POCT, providing laboratory-grade quantification rapidly and reliably.
• The deep learning-based approach enhances both speed and accuracy, making it a promising solution for decentralized diagnostics.
• RAD-LFA effectively bridges the gap between rapid point-of-care testing and precise laboratory quantification.