Automatic Segmentation of Brain CT Image Based on Multiplicate Features and Decision Tree

To overcome the invalidation of classical segmentation approaches and the problem that man-machine mutual division is time-consuming, this paper describes an automatic tissue segmentation method for brain CT image. It includes some theories such as pattern recognition, fuzzy theory, anatomy, fractal and technology of CT imaging. We define a epidermal coefficient for the first time, design a subjection function and import the fractal dimension as features in this algorithm. At same time, we choose decision tree which classifies tissues in multistep form and has different features mentioned previously and different decision making rules on every step as a classifier to segment the brain CT image. Decision tree can simplify the process, reduce the unnecessary calculation of eigenvector and enhance the speed. The experiment results show that the method based on multiplicate features and decision tree can realize the automatic accurate segmentation of brain CT image.