A target-oriented segmentation method for specific tissues in MRI images of the brain
A target-oriented segmentation method for specific tissues in MRI images of the brain
复制标题
脑MRI图像中特定组织的面向目标的分割方法
DOI：
10.1007/s11042-017-5484-1
复制
发表时间：
--
影响因子：
3.6
通讯作者：
Chih-Cheng Hung
中科院分区：
文献类型：
--
作者：
Enmin Song;Yueing Qian;Hong Liu;Meng Yan;Huimin Song;Chih-Cheng Hung
基于多图谱的分割方法可以实现对磁共振成像（MRI）中人脑特定组织的准确分割。该方法中正确的图像配准和融合方案直接影响分割的准确性。与任何传统的刚性配准方法类似，我们在我们提出的面向目标的配准中使用相同的方法来进行目标图像和图谱图像之间的粗配准。然而，为了提高待分割区域的配准精度，我们提出了一种面向目标的图像配准方法进行细化。在基于稀疏块的标签融合过程中，我们采用组织（待分割）的分布概率。我们的目的是确定所提出的配准方法是否有助于分割精度，以及哪种标签融合方法适合这种面向目标的配准。为了评估我们所提出的方法的效率，我们比较了多数投票法（MV），非本地补丁为基础的方法（Nonlocal-PBM）和稀疏补丁为基础的方法（Sparse-PBM）的性能。实验结果表明，本文提出的配准方法可以获得更准确的分割结果。该结果可为临床提供更准确的诊断信息。
The multi-atlas based segmentation method can achieve the accurate segmentation of specific tissues of the human brain in the magnetic resonance imaging (MRI). The correct image registration and fusion scheme used in this method have an impact on the accuracy of segmentation. Similar to any traditional rigid registration method, we use the same method in our proposed target-oriented registration for the coarse registration between the target image and atlas image. However, to improve the registration accuracy in the area to be segmented, we propose a target-oriented image registration method for the refinement. We employ the distribution probability of the tissue (to be segmented) in the sparse patch-based label fusion process. Our aim is to determine if the proposed registration method can contribute the segmentation accuracy and which label fusion method is a good fit with this target-oriented registration. To evaluate the efficiency of our proposed method, we compare the performance of the majority voting method (MV), the nonlocal patch-based method (Nonlocal-PBM) and the sparse patch-based method (Sparse-PBM). Experimental results show that more accurate segmentation results can be obtained with the proposed registration method in this study. This result can provide more accurate clinical diagnosis information.