Segmentation of Medical Image Using a Statistical Technique and Its 3D Visualization

Abstract

We present an automatic segmentation and its visualization method for medical image. First, the statistical segmentation consists of two steps: number detection of clusters composing an image and parameter estimation of a statistical model. Here we use the morphological operations to determine automatically the number of clusters or objects composing a given image without any prior knowledge and adopt the Gaussian mixture model (GMM) to characterize an image statistically. Next, the Deterministic Annealing Expectation Maximization algorithm is employed to estimate the parameters of the GMM. Finally, we use a modified marching cubes algorithm to visualize the extracted images. The experimental results show that our proposed method can extract and visualize exactly the human organs from the CT image.