Label‐Free Discrimination of Liver Cancer Using Plasma Proteins Surface‐Enhanced Raman Spectroscopy and Multivariate Statistical Algorithm

ABSTRACT Plasma proteins are considered an important indicator for cancer diagnosis and prognosis. Currently, commonly used detection methods are unable to achieve rapid and highly sensitive detection of plasma proteins, making them insufficient to meet the demands of high‐sensitivity cancer screening. To address this issue, we propose a rapid detection method for plasma proteins using surface‐enhanced Raman spectroscopy (SERS) based on particulate triacetate cellulose (TCA), referred to as TCA‐SERS. To verify the feasibility of TCA‐SERS for liver cancer (LC) screening, we analyzed the SERS spectra of plasma samples from 60 LC patients and 60 healthy individuals using two multivariate statistical algorithms: PCA‐LDA and PLS‐SVM. The classification accuracy of PCA‐LDA in distinguishing LC patients from healthy individuals was 84.2%, with a sensitivity of 81.7%, specificity of 86.7%, and an AUC value of 0.911. In contrast, the classification accuracy of PLS‐SVM was 95.8%, with a sensitivity of 96.7%, specificity of 95%, and an AUC value of 0.999. The diagnostic performance of PLS‐SVM surpassed that of PCA‐LDA, indicating that TCA‐SERS demonstrates excellent potential for LC detection. Moreover, the results highlight the greater potential of the PLS‐SVM algorithm in analyzing SERS spectra of plasma samples in TCA‐SERS. These findings are of great significance for developing label‐free, noninvasive LC screening methods.