Brain MR Image Segmentation and Bias Field Correction through Class-K HMRF Model and EM Algorithm
Accurate brain tissue segmentation from magnetic resonance (MR) images plays an important role in both clinical practice and neuroscience research. In this paper, we extend the hidden Markov random field (HMRF) model, and propose a novel model, called Class-K HMRF model, to further improve the segmentation accuracy by incorporating more contextual information during classification. This model simultaneously takes account of spatial dependencies between image pixels and bias field, and hence can overcome the difficulties caused by noise and intensity inhomogeneity. By comparing our algorithm with state-of-the-art approaches, the experimental results demonstrate that the proposed algorithm can produce more accurate and reliable segmentations.
KeywordsMRI segmentation bias field correction hidden Markov random field expectation-maximization
Unable to display preview. Download preview PDF.
- 1.Li, S.Z.: Markov Random Field Modeling in Computer Vision. Springer, London (2001)Google Scholar
- 2.Deng, H.W., Clausi, D.A.: Unsupervised image segmentation using a simple MRF model with a new implementation scheme. Patt. Recogn. 37, 2323–2335 (2004)Google Scholar
- 13.Li, C.M., Gatenby, C., Wang, L., Gore, J.: A robust parametric method for bias field estimation and segmentation of MR images. In: IEEE Conference on CVPR, pp. 218–223 (2009)Google Scholar
- 14.BrainWeb - Simulated Brain Database, http://www.bic.mni.mcgill.ca/brainweb/