earticle

영어

This paper develops a new diagnosis model using urine biomarker proteins for early ovarian cancer diagnosis. The optimum marker combination that best classifies the benign tumor and cancer was determined from 15 biomarkers and the performance was evaluated. Samples consist of 121 patients with benign tumor, and 55 patients with ovarian cancer. 15 urine biomarkers showing specific reaction to ovarian cancer was the concentration value obtained from xMAP™ bead-based technology (Luminex Corp.). The area under the curve (AUC) of ROC was evaluated to determine the optimum marker combination showing the best performance in difference between patients diagnosed with ovarian cancer and benign. The performance of the selected combination was confirmed with logistic regression. For each combination of two, three, and four biomarkers showed the highest AUC of 85.71%, 86.38%, and 86.7%, respectively. The highest accuracy of distinguishing benign to cancer was 82.58% for a single biomarker, and 84.27%, 84.83%, and 85.29% for each combination of two, three, and four biomarkers, respectively. It was confirmed that the ovarian cancer diagnosis model utilizing the optimum biomarker combination determined in this research showed better performance and higher accuracy than using a single biomarker.