Chapter 5 - Feature extraction and feature selection in medical images

Abstract

Several life-killing diseases threaten the most valuable human life. Physicians commonly prefer taking an X-ray, CT scan, and MRI scan for disease identification. They need help to identify diseases at an early stage with the existing practices. Identification of disease at an early stage helps to save many human lives. Image processing is a technique to transform images into digital forms for performing various operations. It is possible with emerging technologies like image processing, machine learning (ML), and deep learning (DL). Image acquisition, enhancement, restoration, compression, and segmentation are a few steps involved in image processing. ML or DL process images using classification, prediction, and clustering techniques. The significant role of features is required to implement these models. Feature extraction and feature selection are essential to improve the model’s accuracy. Feature extraction creates a new set of minimum features from existing ones through functional mapping. Dimensionality reduction and visualization are the advantages of feature extraction. On the other way, feature selection is to decrease the feature space by selecting minimal features. It has the benefit of lowering dimensions and increasing the execution speed and accuracy of the model. These dimension reduction techniques help process the physician’s test image to analyze the disease with better accuracy. This chapter discusses the processing methods of medical images using various feature extraction and selection algorithms implemented in ML and DL models.