A Surface-Enhanced Raman Scattering-Digital Microfluidics Biosensing Platform Integrated with a 1D-Convolutional Neural Network for Noninvasive Detection of Inflammation Markers

The noninvasive detection of multiple inflammatory markers (IMs) for early sepsis diagnosis is challenging. Therefore, herein, we innovatively developed an integrated sensing platform based on surface-enhanced Raman scattering (SERS) and digital microfluidics (DMF). The platform allows highly sensitive and automated analysis of the inflammatory marker C-reactive protein (CRP) in sweat. It uses functionalized Au@Fe₃O₄ as a magnetic capture core and Raman-reported molecule-modified SERS tags to form a "AuMNPs/IMs/SERS tag" sandwich immunocomplex. Combined with the inherent precision of the DMF chip in droplet manipulation, the synergy between these two technologies achieves automated, high-efficiency enrichment and in situ detection of trace analytes in sweat samples. We observed that the SERS-DMF platform can efficiently detect CRP within 20 min, achieving a detection limit of 0.77 pg/mL. These values are substantially below clinical thresholds. Consistent with the gold-standard ELISA, our platform demonstrates high reliability. Moreover, the establishment of a one-dimensional convolutional neural network model considerably improved the accuracy and robustness for quantitatively analyzing the CRP. In the test set, the coefficient of determination for the CRP was 0.994, with a root mean square error of 0.135. Based on linear assumptions, our technology significantly outperforms traditional analytical methods. This study integrates the application of SERS-DMF technology for the noninvasive, rapid screening of early-stage sepsis. Using advantages such as ultralow sample consumption and complete process automation provides an innovative technical foundation for developing diagnostic devices for IMs.