SVM Based Feature Screening Applied To Hierarchical Cervical Cancer Detection
Jiayong Zhang, Yanxi Liu, Tong Zhao · 2003
We present a novel feature screening method by deriving relevance measures from the decision boundary of Support Vector Machine, which has several advantages over traditional screening methods based on Information Gain and Augmented Variance Ratio. The new algorithm is applied to a bottom-up approach to cervical cancer detection in multispectral PAP smear images that has been recently proposed by the authors. Comparative experiments show significant improvements on pixel-level classification accuracy using the new feature screening method.