An SVM Based Automatic Segmentation Method for Brain Magnetic Resonance Image Series

To segment magnetic resonance image series is an interdisciplinary topic that involves both medical and computer science. It is one of the most important steps for medical diagnosis and quantitative analysis. This paper proposes an automatic segmentation method based on support vector machine (SVM). Feature vectors are generated according to both grayscale value and texture pattern of MR brain images. To speed up, some results are acquired directly from the segmentation model trained in adjacent layers. Further more, morphological image processing is introduced to refine the image contour. The experiment shows that our method can achieve good segmentation results in a fast way.