목적: Head and neck cancer (HNC) is often diagnosed at an advanced stage due to the absence of effective early screening methods. Saliva, a biofluid rich in metabolites, has emerged as a promising non-invasive diagnostic medium. Surface-enhanced Raman spectroscopy (SERS) amplifies Raman signals of biomolecules, allowing for highly sensitive metabolomic profiling. This study aims to identify HNC-specific metabolic signatures in saliva using SERS and evaluate its feasibility as a diagnostic tool. 방법:Saliva samples were collected from head and neck cancer (HNC) patients undergoing surgery at a single institution. Prior to treatment, patients provided approximately 1cc of saliva in a fasting state, and the samples were immediately frozen for storage. Healthy control samples were also collected under similar conditions. Surface-enhanced Raman spectroscopy (SERS) was used to acquire the Raman spectra of these samples. Multivariate statistical methods, including principal component analysis (PCA), and machine learning algorithms such as logistic regression and support vector machine (SVM), were employed to analyze the spectral data. Key Raman peaks indicative of metabolic alterations were identified and correlated with known cancer-related biomarkers. 결과:Preliminary SERS analysis revealed distinct spectral differences between HNC and control groups. Specific Raman peaks corresponding to lipids, nucleotides, polyamines, and lactate showed significant alterations in HNC patients. Machine learning-based classification models demonstrated high accuracy in differentiating HNC from healthy individuals, highlighting the potential of saliva-based SERS analysis as a screening tool. 결론:This preliminary study suggests that SERS-based saliva metabolomics can serve as a promising approach for non-invasive HNC screening. Further large-scale validation and clinical trials are required to establish its diagnostic reliability and integration into routine cancer screening programs.