Machine learning‐based SERS serum detection platform for high‐sensitive and high‐throughput diagnosis of colorectal precancerous lesions

Colorectal precancerous lesions (CRP) are early signs of cancer development, and early detection helps prevent progression to colorectal cancer (CRC), reducing incidence and mortality rates. This study developed a serum detection platform integrating surface‐enhanced Raman scattering (SERS) with machine learning (ML) for early detection of CRP. Specifically, a microarray chip with Au/SnO2 nanorope arrays (Au/SnO2 NRAs) substrate was designed for SERS spectral measurement of serum. The Principal Component Analysis (PCA)‐Optimal Class Discrimination and Compactness Optimization (OCDCO) model was proposed to identify CRP spectra. The results demonstrated that the microarray chip exhibited superior portability, SERS activity, stability, and uniformity. Through PCA‐OCDCO, the serum samples from healthy controls, CRP patients, and CRC patients were effectively classified, and several key spectral features for distinguishing different groups were identified. The established PCA‐OCDCO model achieved outstanding performance, with an accuracy of 97%, a sensitivity of 95%, a specificity of 97%, and an AUC of 0.96. This study suggests that the platform, integrating SERS with the PCA‐OCDCO model, holds potential for the early detection of CRP, providing an approach for CRP prevention and clinical diagnostics.