Entropy Guided Clustering Improvements and Statistical Rule-Based refinements for Bone Segmentation of X-Ray Images
Jamshid Tamouk, Adnan Acan

During the recent decade, segmentation algorithms have been applied extensively to medical images in order to assist physician treating and diagnosing variety of diseases. Among all types of medical images, X-ray imaging is one of the frequently used imaging methods, especially for diagnosing bone diseases. Great numbers of segmentation methods have been applied on X-ray images aiming to segment bones with higher accuracy; however, it is hard to find a single method that is capable of segmenting all body parts with equal level of quality. Therefore, researchers put significant efforts on the combination different methods or improvement of existing methods using some techniques to achieve a desired segmentation outcome. In this research, improvement and refinement methods are proposed for three well-known clustering algorithms, namely the K-means, fuzzy c-means and spatial fuzzy c-means algorithms. The proposed methods include a novel entropy guided improvement strategy for clustering and a statistics based cellular rule engine for further refinement of pixel clusters. The improved algorithms are applied to X-ray images and experimental evaluations in comparison to well-known methods exhibit that significant improvements are achieved for all test cases.

Full Text: PDF     DOI: 10.15640/jcsit.v4n1a3