Meta-Heuristic Algorithms in Medical Image Segmentation: A Review

Meta-Heuristic Algorithms in Medical Image Segmentation: A Review

Nilanjan Dey (Techno India College of Technology, India) and Amira S. Ashour (Tanta University, Egypt)
Copyright: © 2018 |Pages: 19
DOI: 10.4018/978-1-5225-4151-6.ch008
OnDemand PDF Download:
List Price: $37.50


Artificial intelligence is the outlet of computer science apprehensive with creating computers that perform as humans. It compromises expert systems, playing games, natural language, and robotics. However, soft computing (SC) varies from the hard (conventional) computing in its tolerant of partial truth, uncertainty, imprecision, and approximation, thus, it models the human mind. The most common SC techniques include neural networks, fuzzy systems, machine learning, and the meta-heuristic stochastic algorithms (e.g., Cellular automata, ant colony optimization, Memetic algorithms, particle swarms, Tabu search, evolutionary computation and simulated annealing. Due to the required accurate diseases analysis, magnetic resonance imaging, computed tomography images and images of other modalities segmentation remains a challenging problem. Over the past years, soft computing approaches attract attention of several researchers for problems solving in medical data applications. Image segmentation is the process that partitioned an image into some groups based on similarity measures. This process is employed for abnormalities volumetric analysis in medical images to identify the disease nature. Recently, meta-heuristic algorithms are conducted to support the segmentation techniques. In the current chapter, different segmentation procedures are addressed. Several meta-heuristic approaches are reported with highlights on their procedures. Finally, several medical applications using meta-heuristic based-approaches for segmentation are discussed.
Chapter Preview


Image analysis, pattern recognition, image disciplines are the foremost domains of computer engineering and computer science in several domains, such as medical, military, astronomy and real-world applications. In the medical applications, image-guided therapy is one of the vital methods for accurate diagnosis. Medical image computing has a rising prominence for medical diagnosis. Image analysis systems incorporating with innovative image computing procedures carried out to extract quantifiable parameters from the medical image in order to support the diagnosis and treatment. In order to achieve accurate clinical routine, automated and robust medical image computing techniques become an active research area. Model-based image analysis as well as image-based modelling methods becomes significant tools for accurate assessable analysis of the objects in the medical image. These methods require earlier information about the medical images’ structures, including bones, tumours, tissue, vessels and organs. Afterward, the image-based modelling approaches applied to extract the significant features automatically. For complex visualization and quantitative measures, processing of digital indicative imaging data carried out to support disease progression monitoring, diagnosis and pre-operative planning. Nevertheless, successful image analysis requires optimized and complex processing systems, which is a challenging aspect. Currently, research in medical image analysis pursued by an ongoing stream of successful new clinical applications to achieve robust solutions based on computing techniques (Dey,& Ashour, 2016; Kotyk et al., 2016; Saba et al. (2016); Ahmed et al., 2017; Ashour et al., 2016; Dey et al., 2017).

Medical image segmentation and classification are considered the main image processing approaches. High soft tissue contrast of magnetic resonance (MR) images segmentation has a significant role for evaluating the brain tumors’ therapy. Manual segmentation by physicians still the segmentation gold standard of atypical brain images, however, it is disposed to human bias/error as well as its tedious process. An endeavor for reliable computerization of medical image segmentation is consequently extremely desired. This leads to the necessity to used clustering algorithms to label the pixels of the medical image into a prearranged number of clusters, where the number of clusters is known a priori in most of the anatomic structures. Furthermore, medical image classification is extensively used to discriminate the abnormal and normal images. Automated classification process is highly desired to support the physicians in the analysis monotonous task.

Complete Chapter List

Search this Book: