Ag–Au Nanorods for the Diagnosis of Cervical Cancer by Label-Free Serum Surface-Enhanced Raman Scattering

Developing cancer screening methods, particularly by Surface-enhanced Raman scattering (SERS) of serum samples, is crucial for early cancer diagnosis, given SERS's high biocompatibility and comprehensive spectra fingerprints. However, development of SERS as a cancer diagnostic tool is impeded by the lack of a high-quality serum database and robust, adaptable diagnostic models. In our study, we optimized SERS active Ag–Au nanorods and serum pretreatment to obtain high-quality spectra from 215 women's serum samples and then normalized the spectra to create a reliable serum database. Quadratic discriminant analysis (QDA) of this database could classify cancer patients from healthy individuals with an accuracy of 91.2%. By the kernel density estimation (KDE) algorithm on QDA scores, we calculated the probability distribution function (PDF) for two groups with which the diagnostic accuracy was significantly enhanced, exemplified by a 139/143 accuracy rate at 90% probability decision threshold. Notably, both PDF and QDA models can be refined iteratively with an increase in database increment. Our findings underscore the potential of label-free serum SERS in noninvasive cancer screening, particularly for distinguishing cervical cancer patients from healthy individuals.