Performance Evaluation of Region-Based Segmentation Algorithms for Brain MR Images
Segmentation of an image is crucial for dividing it into different classes so that it can be made useful for information extraction for image classification and dissemination in medical imaging and diagnosis. There are various region-based segmentation al
- PDF / 661,077 Bytes
- 13 Pages / 439.37 x 666.142 pts Page_size
- 80 Downloads / 219 Views
Abstract Segmentation of an image is crucial for dividing it into different classes so that it can be made useful for information extraction for image classification and dissemination in medical imaging and diagnosis. There are various region-based segmentation algorithms that are used to detect the required region of interest in MR image. The hard and soft clustering algorithm along with an unsupervised classification method based on expectation maximization for image segmentation is illustrated in this paper. These methods are used on the brain MR image to confer the result for medical diagnosis purposes. The best method is chosen by computing certain performance evaluating parameters such as rand index, variation of information and global consistency error. As per the performance measure, expectation maximization result proved to be the best when it was compared with the ground truth image. Thus, it propels to be used in medical diagnosis. Keywords Hard and soft clustering · Expectation maximization · Region of interest · Unsupervised classification
1 Introduction In digital image processing, a digital image is transformed into its integer containing elements, known as pixels, having a particular location and value, for example, scene luminance, stored in an advanced memory, and primed by any advanced machinery T. Sahoo (B) · R. K. Pradhan · K. K. Das · S. Sahu ITER, Siksha O Anusandhan (Deemed to be University), Bhubaneswar 751030, Odisha, India e-mail: [email protected] R. K. Pradhan e-mail: [email protected] K. K. Das e-mail: [email protected] S. Sahu e-mail: [email protected] © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 R. Sharma et al. (eds.), Green Technology for Smart City and Society, Lecture Notes in Networks and Systems 151, https://doi.org/10.1007/978-981-15-8218-9_44
539
540
T. Sahoo et al.
like digital computer. The representation of an image is done by f (x, y), i.e., a twodimensional function [1], where the amplitude of f is known as the intensity of the image at any pair of coordinates (x, y). An image is said to be digital when x, y and intensity values of f (positive scalar quantity) are all finite and discontinuous in nature [1]. Differentiating amid image processing and its various domains and equivalent area computer vision is found to be challenging. The study related with the science of artificial systems that has high level of understanding and unique feature to extract information from digital images is known as computer vision [2]. The intensity of MRI is similar to that of gray images in the range between 0 and 255. The GM, WM and CSF are the three main tissue types of brain MRI’s elements [3]. The GM encompasses of cortex, which shapes brain’s external surface, the deep internal gray nuclei, basal ganglia and thalami. The WM encompasses of neural axons, which interlinks various sections of the brain, serving like a connection to the body. The CSF is a watery fluid that flows withi
Data Loading...