Feature Extraction from Segmented Lung Images and Feature Selection Through Soft Set Based Approach

摘要

Among all types of cancer lung tumor is extreme critical type of cancer having very fewer endurance rate. It is very hard to identify the cancer at its primary phase with the extreme tests. Many CAD structures need been developed to detect the tumor at its initial stage on ComputedTomographic (CT) lung images. Researchers understand that in order to develop effective detection system, feature extraction and selection are indispensable components. Feature extraction procedures analyze objects and images to extract the most prominent features that are representative ofthe various classes of objects. Feature selection plays a vital role in lung cancer detection and classification. This paper explores the feature selection process in CT lung cancer images using soft set theory. Here, a new soft set based unsupervised feature selection algorithmis projected. Gray Level Co-occurrence Matrix (GLCM) in four possible directions (0°, 45°, 90° and 135°) and Gray Level Different Matrix (GLDM) are used to extract nineteen texture features from segmented CT lung images. Experimental results show that the proposed approacheffectively removes the redundant features and selects the prominent features.