Multimatrix Quant^AI-LFIA: An On-Device Deep-Learning Biosensing Platform with Charge-Transfer-Active Labels for...

Area of Science:

• Biomedical Engineering
• Analytical Chemistry
• Artificial Intelligence in Diagnostics

Background:

• Commercial gold nanoparticle-based lateral flow immunoassays (LFIAs) are common for point-of-care diagnostics but suffer from low sensitivity and poor quantification.
• Inefficient signal transduction and lack of intelligent interpretation limit the performance of current LFIAs.

Purpose of the Study:

• To develop an on-device platform integrating deep learning for intelligent quantitative detection of staphylococcal enterotoxin B (SEB).
• To improve sensitivity, accuracy, and adaptability of LFIAs across multiple biological matrices.

Main Methods:

• Utilized Fe-polydopamine (Fe-PDA) as a charge-transfer-active label for enhanced colorimetric signal transduction.
• Integrated a deep learning model for quantitative analysis, robust to varying illumination conditions.
• Employed transfer learning for model adaptation to different biological samples like serum and urine.

Main Results:

• Achieved a visual limit of detection (LOD) of 5 ng/mL within 15 minutes using Fe-PDA labels.
• Deep learning model extended the dynamic range to 0.1-800 ng/mL and reduced LOD to 0.1 ng/mL (50-fold improvement).
• Demonstrated strong correlations (R² > 0.90) in serum and urine samples; achieved an AUC of 0.924 in blind tests via a smartphone app.

Conclusions:

• The Multimatrix QuantAI-LFIA platform offers rapid, accurate, and matrix-adaptive quantification of SEB.
• The integration of deep learning and advanced labeling significantly enhances LFIA performance for next-generation POCTs.