Sparse Representation Label Fusion Method Combining Pixel Grayscale Weight for Brain MR Segmentation

摘要

Multi-atlas based segmentation (MAS) methods have demonstrated superior performance in the field of automatic image segmentation, and label fusion is an important part of MAS methods. In this paper, we propose a sparse representation label fusion (SRLF) method combining pixel grayscale weight. We adopt a strategy for solving sparse coefficients multiple times and introduce pixel grayscale weight information in the label fusion process. In order to verify the segmentation performance, we apply the proposed method to segment subcuta-neous tissues in 3D brain MR images of the challenging publicly available IBSR datasets. The results show that our method effectively improves the defects of SRLF method and achieves higher segmentation accuracy. We also compared our methods with commonly used automatic segmentation tools and state-of-the-art methods, and the average Dice similarity coefficient (Dsc) of the subcutaneous tissues obtained by our method was significantly higher than that of the automatic segmentation tools and state-of-the-art methods.