Feature Extraction and Selection for Decision Making

Abstract

This chapter presents and discusses useful algorithms and techniques of feature extraction and selection as well as the relationship between the image features, their discretization and distance functions to maximize the image representativeness when executing similarity queries to improve medical image processing, mining, indexing and retrieval. In particular, we discuss the Omega algorithm combining both, feature selection and discretization, as well as the technique of association rule mining. In addition, we present the Image Diagnosis Enhancement through Associations (IDEA) framework as an example of a system developed to be part of a computer-aided diagnosis environment, which validates the approaches discussed here.