Automatic Medical Image Segmentation Based On Finite Skew Gaussian Mixture Model

A novel methodology for segmenting the brain Magnetic Resonance Imaging (MRI) images using the finite skew Gaussian mixture model has been proposed for improving the effectiveness of the segmentation process. This model includes Gaussian mixture model as a limiting case and we believe does more effective segmentation of both symmetric and asymmetric nature of brain tissues as compared to the existing models. The segmentation is carried out by identifying the initial parameters and utilizing the Expectation-Maximization (EM) algorithm for fine tuning the parameters. For effective segmentation, hierarchical clustering technique is utilized. The proposed model has been evaluated on the brain images extracted from the brain web image database using 8sub-images of 2 brain images. The segmentation evaluation is carried out using objective evaluation criterion viz. Jacquard Coefficient (JC) and Volumetric Similarity (VS). The performance evaluation of reconstructed images is carried out using image quality metrics. The experimentation is carried out using T-1 weighted images and the results are presented. We infer from the results that the proposed model achieves good segmentation results when used in brain image processing.