A New Fuzzy Clustering Algorithm for Brain MR Image Segmentation Using Gaussian Probabilistic and Entropy-Based Likelihood Measures

Medical image segmentation plays a crucial role in medical image analyses, computer-guided surgical planning, abnormality detection, etc. The magnetic resonance (MR) image segmentation process is much more challenging as the contour of the soft tissue regions are vague or uncertain. This paper presents a new fuzzy clustering algorithm to address the class uncertainty associated with each pixel in the image region. In particular, the class uncertainty is handled by integrating the Shannon's entropy within the objective function. In addition, the objective function also includes Gaussian probabilistic measure to estimate the membership function. The proposed algorithm is validated on several synthetic brain MR images with varying noise and inhomogeneity. Additionally, we have also validated the method on in-vivo (real-patient) human brain MR images. The empirical results of the proposed algorithm are compared with some competent image segmentation methods and found superior to them.