Collaborative Internal Cavity Effect and Interfacial Modulation Mechanism for Boosting Deep Learning-Powered Immunochromatographic Pathogen Detection

Nanoenabled immunochromatographic assay (ICA) emerges as a powerful tool for pathogen diagnosis, yet current nanotechnologies are still constrained by inadequate light-matter interaction efficiency, sluggish nanomaterial flow dynamics, and inefficient immunorecognition. Herein, we present a deep learning-enhanced immunoassay synergistically leveraging the internal cavity effect of hollow carbon nanospheres (h-CNSs) and interfacial antibody orientation modulation for the ultrasensitive detection of S. typhimurium. The h-CNSs exhibit significantly enhanced light absorption (molar extinction coefficients 5.4 × 1011 vs. 3.7 × 1011 L mol–1 cm–1 for counterpart) and photothermal conversion efficiency (66.78% vs. 43.37%) due to internal light reflection within the hollow cavity, while the reduced density (0.05 g mL–1) optimizes lateral flow kinetics. Further interfacial modification with 3,5-dicarboxybenzeneboronic acid enables directional antibody immobilization through boronate affinity, improving antibody binding affinity by 83-fold (Ka = 2.95 × 109 vs. 3.55 × 107 M–1). Integrated into an ICA, D-h-CNSs achieve visual detection limits of 500 CFU mL–1 (colorimetric) and 100 CFU mL–1 (photothermal), surpassing conventional ICA (104 CFU mL–1) and demonstrating high specificity, robust stability, and reliable performance in spiked milk and lettuce. By integration with a convolutional neural network (CNN), the developed nanoplatform achieves 100% accuracy for S. typhimurium detection with augmented training, providing a paradigm for amplifying biosensing signals through nanomaterial design and intelligent data analysis.