MRI brain segmentation using cellular automaton approach

Abstract

In this article new approach to the MRI brain segmentation is presented. It is based on the Cellular Automaton (CA). The method is truly interactive, and produces very good results comparable to those achieved by Random Walker or Graph Cut algorithms. It can be used for CT and MRI images, and is accurate for various types of tissues. Discussed here version of the algorithm is developed especially for the purpose of the MRI brain segmentation. This method is still in the phase of development and therefore can be improved, thus final version of the algorithm can differ from the one presented here. The method is extensible, allowing simple modification of the algorithm for a specific task. As we will also see it is very accurate for two-dimensional medical images. Three-dimensional cases require some unsophisticated data post processing [1], or making some modifications in the manner in which the automaton grows into the third dimension from the two-dimensional layer. To prove quality of this method, some tests results are presented at the end of this paper. © 2010 Springer-Verlag Berlin Heidelberg.