Title: Semi-automatic 3D segmentation of brain structures from MRI

Authors: Qing He, Kevin Karsch, Ye Duan

Addresses: Department of Computer Science, University of Missouri-Columbia, Columbia, MO 65211, USA. ' Department of Computer Science, University of Missouri-Columbia, Columbia, MO 65211, USA. ' Department of Computer Science, University of Missouri-Columbia, Columbia, MO 65211, USA

Abstract: We present a semi-automatic 3D segmentation method for brain structures from Magnetic Resonance Imaging (MRI). There are three main contributions. First, our method combines boundary-based and region-based approaches but differs from previous hybrid methods in that we perform them in two separate phases. This allows for more efficient segmentation. Second, a probability map is generated and used throughout the segmentation to account for the brain structures with low-intensity contrast to the background. Third, we develop a set of tools for manual adjustment after the segmentation. This is particularly important in clinical research because the reliability of the results can be ensured. The experimental results and validations on different data sets are shown.

Keywords: semi-automatic segmentation; MRI; magnetic resonance imaging; brain structures; deformable models; modelling; brain scanning; probability maps; 3D segmentation.

DOI: 10.1504/IJDMB.2011.039175

International Journal of Data Mining and Bioinformatics, 2011 Vol.5 No.2, pp.158 - 173

Published online: 24 Jan 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article