Indian Journal of Science and Technology
DOI: 10.17485/ijst/2015/v8i34/85361
Year: 2015, Volume: 8, Issue: 34, Pages: 1-4
Original Article
S. U. Aswathy1* , G. Glan Deva Dhas2 and S. S. Kumar2
1 Department of Computer Science, Noorul Islam University, Kanyakumari - 629180, Tamil Nadu, India; [email protected]
2 Electronics and Instrumentation, Noorul Islam University, Kanyakumari - 629180, Tamil Nadu, India; [email protected], [email protected]
In medical image processing, Brain tumor segmentation from MRI scan slice without human intervention has become one of the most challenging area in research.MR image slices usually contain a significant amount of noise caused by operator interactions, environmental or external factors, machines used etc, which in turn may cause serious segmentation inaccuracy.Our main aim is to recognize a tumour and its quantification from a specific MRI scan of a brain image to obtain the best segmentation in minimum time. In this paper, we are trying to develop a robust segmentation technique, by combining two segmentation algorithms, expectation maximization and level set. This algorithm framework composed of three stages. 1. Pre-processing is used for background separation of brain; it can be also called as pre-segmentation method 2. Abnormal regions in brain is detected by using Expectation Maximization (EM) algorithm and in 3. A level set method to sharpen the segmented EM output to get sharp and accurate boundaries. At the end of the process, the tumour region is extracted from the MR images and its exact position and shape is determined with minimum time. The Experimental results clearly define the effectiveness of our approach in its accuracy and computation time when compared with other EM based method.
Keywords: Expectation-Maximization, Feature Extraction, Image Segmentation, Level Set, Pre-processing
Subscribe now for latest articles and news.