MR Brain Image Segmentation Based on Unsupervised and Semi-Supervised Fuzzy Clustering Methods

Publisher:
IEEE
Publication Type:
Conference Proceeding
Citation:
2016 International Conference on Digital Image Computing: Techniques and Applications, DICTA 2016, 2016, pp. 1 - 7
Issue Date:
2016-12-22
Full metadata record
Files in This Item:
Filename Description Size
07797066.pdfPublished version342.69 kB
Adobe PDF
© 2016 IEEE.In medical imaging applications, the segmentation of Magnetic Resonance (MR) brain images plays a crucial role for measuring and visualizing the anatomical structures of interest. In general, the brain image segmentation aims to divide the image pixels into non-overlapping homogeneous regions for analyzing the changes in the brain for surgical planning. Several supervised and unsupervised clustering methods have been developed over the years to segment the magnetic resonance brain image. However, most of these methods have certain limitations such as requiring user interaction and high computational complexity. In this context, this paper proposes a methodology that combines semi-supervised and unsupervised classification techniques for achieving efficient and fully-automatic segmentation of brain images. Firstly, the algorithm applies a median filter to remove the noise inherent in MR images prior to the clustering step. Secondly, the background of the MR image is removed by using a global thresholding technique. Thirdly, we utilize the subtractive clustering method to overcome the deficiency of randomly initialized Fuzzy C-Means (FCM) parameters. This method is used for estimating the clustering number and to generate the initial centers, which is used as initialization parameter for FCM clustering. Finally, a semi-supervised algorithm with Standard Fuzzy Clustering is selected to divide the brain MR image into different classes based on the generated membership function from FCM. The efficiency of the proposed method is demonstrated on various MR brain images and compared with some of the well-known clustering techniques.
Please use this identifier to cite or link to this item: