A framework for ovarian cancer diagnosis based on amino acids using fuzzy-rough sets with SVM

Abstract

Ovarian cancer is diagnosed in nearly a quarter of a million women globally each year. It has the highest mortality rate of all cancers for women. The prognosis for ovarian cancer patients is poor, particularly about 20% of ovarian cancers are found at an early stage. CA-125 test is used as a tumor marker. A high level of CA-125 could be a sign of ovarian cancer or other conditions. We use a machine learning technique to explore better knowledge and most important factors useful for detecting early-stage ovarian cancer by evaluating the significance of data between the amino acids and the ovarian cancer. Therefore, we propose a Fuzzy Rough with Support Vector Machine (SVM) classification model to mine suitable rules. In pre-processing stage, we use Fuzzy Rough set theory for feature selection. In post-processing stage, we use SVM to merit of dealing with real and complex data, performing quick learning and having good classification performance.