Kidney stones are a common urological disease with an increasing incidence worldwide. Traditional diagnostic methods for kidney stones are relatively complex and time-consuming, thus necessitating the development of a quicker and simpler diagnostic approach. This study investigates the clinical screening of kidney stones using Surface-Enhanced Raman Scattering (SERS) technology combined with multivariate statistical algorithms, comparing the classification performance of three algorithms (PCA-LDA, PCA-LR, PCA-SVM). Urine samples from 32 kidney stone patients, 30 patients with other urinary stones, and 36 healthy individuals were analyzed. SERS spectra data were collected in the range of 450–1800 cm−1 and analyzed. The results showed that the PCA-SVM algorithm had the highest classification accuracy, with 92.9% for distinguishing kidney stone patients from healthy individuals and 92 % for distinguishing kidney stone patients from those with other urinary stones. In comparison, the classification accuracy of PCA-LR and PCA-LDA was slightly lower. The findings indicate that SERS combined with PCA-SVM demonstrates excellent performance in the clinical screening of kidney stones and has potential for practical clinical application. Future research can further optimize SERS technology and algorithms to enhance their stability and accuracy, and expand the sample size to verify their applicability across different populations. Overall, this study provides a new method for the rapid diagnosis of kidney stones, which is expected to play an important role in clinical diagnostics.