Automatic Diagnosis of Ovarian Cancer Based on Relative Entropy and Neural Network

Abstract

Ovarian Cancer is one of the most common causes of death for women in developing countries. Screening and early diagnoses of OC are urgently needed. Early diagnosis would help in consequence procedures and treatment. Mass spectrometry (MS) data is been used as an effective component of cancer diagnosis tools. However, these valuable data have a large number of dimensions that can affect the learning process in addition to time-consuming considerations. Feature selection plays an important role in reducing information redundancy, and deals with the invalidation that occurs in basic classification algorithms when there are too many features and huge datasets. To improve the automatic system diagnosis accuracy, entropy-based selection features are proposed. These features are combined with the novel learning capabilities of neural networks to achieve higher diagnostic accuracy. Experiments have been performed using different feature selection algorithms and machine learning classification approaches. Experimental results have proved that the proposed system performs better based on the measure of accuracy.