炎症マーカーの非侵襲的検出のための1D畳み込みニューラルネットワークと統合された表面増強ラマン散乱-デジタルマイクロ流体バイオセンシングプラットフォーム

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.