Label Fusion for Multi-atlas Segmentation Based on Majority Voting
Multi-atlas based segmentation is a popular approach in medical image analysis. Majority voting, as the simplest label fusion method in multi-atlas based segmentation, is a powerful tool for segmentation. In this paper, a novel majority voting-based label fusion algorithm is proposed by introducing a patch-based analysis for automatic segmentation of brain MR images. The proposed approach, by comparing the similarity between patches, avoids the over-segmentation problem of the majority fusion. The approach is successfully applied to the segmentation of hippocampus, and the experimental results demonstrate significant improvement over three state-of-the-art approaches in the literature.
KeywordsMulti-atlas segmentation Majority voting Label fusion
The work is partly supported by the NSERC, Kansas NASA EPSCoR Program, and the NSFC (61273282).
- 1.Wang, H., Sun, J., Pluta, J., Craige, J., Yushkevich, P.: Multiatlas segmentation with joint label fusion. IEEE Trans. Pattern Anal. Mach. Intell. 35(3), 311–623 (2013)Google Scholar
- 4.Coupé, P., Manjón, J.V., Fonov, V., Pruessner, J., Robles, M., Collins, D.L.: Nonlocal patch-based label fusion for hippocampus segmentation. In: Jiang, T., Navab, N., Pluim, J.P.W., Viergever, M.A. (eds.) MICCAI 2010, Part III. LNCS, vol. 6363, pp. 129–136. Springer, Heidelberg (2010) CrossRefGoogle Scholar