Feature Selection for Classification of Abnormalities in Medical Images – A Review

Abstract

In medical imaging, Computer Aided Diagnosis (CAD) has become one of the major research topics but is still in the infancy stage due to the lack of its full potential for applications to analyze the lesions obtained from various modalities. Pattern recognition and computer vision plays a significant role in clinical procedures for detecting and diagnosing different human diseases through the processing and analyzing of images acquired through various medical imaging modalities. In many cases of medical applications having high dimensional data characterized by huge number of features require large amount of memory and computation power. In order to tackle this problem, the aim is to construct a combination of feature that builds a unique model to provide better classification performance and accuracy. In this paper, we have conducted a survey on widely used approaches for feature selection and analyzed the purpose to investigate the strength and weakness of existing methods used in different types of modalities of images. Most of the work discussed in this literature review faces many limitations such as accuracy, cost, time and storage when dealing with huge amount of data. Our prime intention is to tackle these problems by building a uniform modal for feature selection to rank the features which are extracted from different medical image modalities to detect and diagnose the abnormalities present in those images in a most efficient way.