Forty-three epithelial ovarian cancerpatients and seventy-seven controls were included in the study from October 2003 to March 2005 in Sanggye Paik Hospital. Potential tumor biomarkers in sixty serum samples were screened, including twenty-one ovarian cancers and thirty-nine controls. Proteomic pattern was analyzed by SELDI-ToF-MS and optimal discriminating m/z value with proper cutoff of log-normalized intensity was determined by decision treeanalysis (Phase I). Another sixty serum samples were obtained from twenty-two ovarian cancers and thirty-eight controls. Through analysis using SELDI-ToF-MS, the performance of diagnosing ovarian cancer was determined by applying previously adopted cutoff log-normalized intensity of m/z value determined in Phase I experiment (Phase II).