Diagnosis and staging of cervical cancer using label-free surface-enhanced Raman spectroscopy and BWRPCA-TLNN model

Abstract

This paper presents a label-free and highly accurate classification serum analytical platform, which will be used to identify cervical cancer at different stages. In detail, the microarray chip fabricated based on the ordered Ag-AuNRs substrate was prepared to measure the surface-enhanced Raman scattering (SERS) spectra of serum of healthy subjects and cervical cancer patients (Ⅰ, Ⅱ, Ⅲ, and Ⅳ), and then a binary weight robust principal component analysis (BWRPCA)-two-layer nearest neighbor (TLNN) model was designed as the diagnosis and recognition model of SERS spectra. The results revealed that the microarray chip can realize rapid, sensitive, high-throughput detection of SERS spectra of serum. The BWRPCA-TLNN successfully differentiated the SERS spectra and accurately captured the key characteristics for classification. The established BWRPCA-TLNN model achieved an excellent classification accuracy of 91.2 %, a diagnostic sensitivity of over 84.0 % and a specificity of over 97.0 %. This exploratory work demonstrated that SERS combined with BWRPCA-TLNN as a diagnostic screening method has potential for improving cervical cancer detection and screening.