This study is aimed at evaluating the feasibility of a screening method for the pulmonary adenocarcinoma nodules through surface-enhanced Raman spectroscopy (SERS). Objective. Using SERS to measure serum from pulmonary nodules and healthy subjects, intraoperative biopsy pathological diagnosis was regarded as the gold standard for labeling serum samples. To explore the application value of SERS in the differential diagnosis of pulmonary adenocarcinoma nodules, benign nodules, and healthy, we build a machine learning model. Method. We collected 116 serum samples from patients. Radiographically confirmed nodules less than 3 cm in maximum diameter in all patients, including 58 cancer (pathologic diagnosis: adenocarcinoma nodules, labeled as cancer) patients, 58 pathologic diagnoses as benign nodule (labeled as benign) patients, and 63 healthy (labeled as normal) people from the clinical laboratory of Sichuan Cancer Hospital. Gold nanorods were employed as SERS substrates. Support vector machine (SVM) was used to classify the normal, benign, and cancer sample groups, and SVM model evaluated using cross-validation. Results. The average SERS spectra of serum were significantly different between the normal group and the cancer/benign group. While the average SERS spectra of the cancer group and the benign group differed slightly, for the cancer, benign, and normal groups, SVM models can predict with 93.33% accuracy. Conclusion. This exploratory study demonstrates that the SERS technique based on nanoparticles in conjunction with SVM has great potential as a clinical auxiliary diagnosis and screening for pulmonary adenocarcinoma nodules.