基于序列间先验约束和多视角信息融合的肝脏CT图像分割

摘要

The accurate segmentation of liver in medical Computed Tomography (CT) sequence images is important prerequisite for computer-assisted liver surgery. However, the presence of tissue lesions, the blurred or missing boundary and the adhesion between different organs/tissues poses great challenges to liver segmentation. To address these problems, this paper presents a semi-automatic segmentation method based on the sequential constraints of image sequences, and introduces further a multi-view information fusion method to achieve the accurate segmentation of the liver. One advantage of this approach is that it does not need extensive data collection and complicated prior training. The validation and comparison results on the Sliver07 public data show that the proposed method shows competitive performance, especially when there is liver tumor, blurred or missing liver boundary.%医学电子计算机断层扫描(CT)序列图像中肝脏的准确分割是实现计算机辅助肝手术的重要前提,然而图像中存在的组织病变、边界模糊或缺失、不同组织间的粘连给肝脏分割带来极大挑战.针对这些问题,该文提出一种基于图像序列间先验约束的半自动分割方法,并进一步采取了多视角信息融合的方式实现肝脏的准确分割.该方法的优势在于无需大量数据的收集和复杂的先验训练.在Sliver07公开数据集合的验证结果显示,和领域内主要方法相比,该方法具有较高的分割准确度,特别是当肝脏区域存在病灶、边界模糊或缺失的情况下具有明显提升.