Online Discriminative Multi-atlas Learning for Isointense Infant Brain Segmentation
Multi-atlas labeling in a non-local patch manner has emerged as an important approach to alleviate both the possible misalignment and mis-match among patches for guiding accurate image segmentation. However, the relationship among candidate patches and their intra/inter-class variability are less investigated, which limits the discriminative power of these patches. To address these issues, we present a new online discriminative multi-atlas learning method for labeling the target patch by the best representative candidates in a sparse sense. Specifically, the online multi-kernel learning is firstly adopted to map the patches into a cascade of discriminative kernel spaces for producing corresponding probability maps to model a label of each sample in these spaces. Then the online discriminative dictionary learning is proposed to build the atlas that handles the intra-class compactness and inter-class separability simultaneously. Finally, sparse coding is used to select patches in the dictionary for label propagation. In this way, the multi-atlas information dynamically learned with the context probability maps is iteratively incorporated to build the atlas dictionary, for gradually excluding the misleading candidate patches. The proposed method is validated by experiments on isointense infant brain tissue segmentation, and achieves promising results in comparison with several different labeling strategies.
KeywordsSparse Representation Sparse Code Dictionary Learning Target Patch Label Fusion
Unable to display preview. Download preview PDF.
- 1.Wang, L., Shi, F., Li, G., Lin, W., Gilmore, J.H., Shen, D.: Integration of sparse multi-modality representation and geometrical constraint for isointense infant brain segmentation. In: Mori, K., Sakuma, I., Sato, Y., Barillot, C., Navab, N. (eds.) MICCAI 2013, Part I. LNCS, vol. 8149, pp. 703–710. Springer, Heidelberg (2013)CrossRefGoogle Scholar
- 8.Wu, G., Wang, Q., Liao, S., Zhang, D., Nie, F., Shen, D.: Minimizing joint risk of mislabeling for iterative patch-based label fusion. In: Mori, K., Sakuma, I., Sato, Y., Barillot, C., Navab, N. (eds.) MICCAI 2013, Part III. LNCS, vol. 8151, pp. 551–558. Springer, Heidelberg (2013)CrossRefGoogle Scholar
- 10.Bai, W., Shi, W., O’Regan, D.P., Tong, T., Wang, H., Jamil-Copley, S., Peters, N.S., Rueckert, D.: A probabilistic patch-based label fusion model for multi-atlas segmentation with registration refinement: Application to cardiac MR images. IEEE Trans. Med. Imaging 32(7), 1302–1315 (2013)CrossRefGoogle Scholar
- 13.Shalev-Shwartz, S., Srebro, N.: SVM optimization: inverse dependence on training set size. In: ICML, pp. 928–935 (2008)Google Scholar
- 14.Jie, L., Orabona, F., Fornoni, M., Caputo, B., Cesa-Bianchi, N.: OM-2: An online multi-class multi-kernel learning algorithm. In: CVPRW, pp. 43–50 (2010)Google Scholar
- 15.Yang, M., Zhang, L., Feng, X., Zhang, D.: Fisher discrimination dictionary learning for sparse representation. In: ICCV, pp. 543–550 (2011)Google Scholar
- 16.Mairal, J., Bach, F., Ponce, J., Sapiro, G.: Online dictionary learning for sparse coding. In: ICML, p. 87 (2009)Google Scholar
- 17.Liu, J., Ji, S., Ye, J.: SLEP: Sparse Learning with Efficient Projections. Arizona State University (2009)Google Scholar