Automatic Segmentation of the Full Heart in Cardiac Computed Tomography Images Using a Haar Classifier and a Statistical Model
Automatic Segmentation of the Full Heart in Cardiac Computed Tomography Images Using a Haar Classifier and a Statistical Model
复制标题
DOI：
10.1166/jmihi.2016.1916
复制
发表时间：
2016-09
影响因子：
--
通讯作者：
Fei Ma;Jiquan Liu;Bin Wang-;H. Duan
中科院分区：
文献类型：
--
作者：
Fei Ma;Jiquan Liu;Bin Wang-;H. Duan
从CT图像中分割出心脏的主要解剖结构对心血管疾病的诊断和治疗具有重要意义。然而，由于心脏CT图像的高分辨率和多切片特性，心脏的手动分割提出了许多挑战。因此，有必要开发具有自动心脏分割方法的强大软件工具。本文提出了一种基于切片的CT图像全心脏自动分割方法。该方法首先训练一个Haar分类器和一个主动形状模型（ASM），然后，与Haar分类器，在每个输入图像中的心脏区域被检测。接下来，在每个心脏区域中应用并初始化预构建的ASM，并且对每个初始模型执行自适应形状配准以匹配心脏的边界。在该方法中，心脏CT图像中相邻帧之间的相似性被充分估计，并被用作一个重要的约束，以优化结果。将分割结果与人工分割结果进行了比较，结果之间的距离为3 ~ 5 mm，表明该方法是有效的。
The segmentation of the main anatomical structures of the heart from CT images is of great significance for the diagnosis and treatment of cardiovascular disease. However, due to the high-resolution and multi-slice properties of cardiac CT images, manual segmentation of the heart presents many challenges. Therefore, it is necessary to develop powerful software tools with automatic methods for heart segmentation. In this paper, a slice-based automatic method for full heart segmentation in CT images is proposed. The method starts with the training of a Haar classifier and an Active Shape Model (ASM); then, with the Haar classifier, the heart area in each input image is detected. Next, the pre-built ASM is applied and initialized in each heart area and an adaptive shape registration is performed on each initial model to match the boundaries of the heart. In the proposed method, the similarity between adjacent frames in the Cardiac CT images is fully estimated and is used as an important constraint to optimize the results. The segmentation results are compared with manual segmentation and the distance between the results is 3∼5 mm, which shows that the proposed method is effective.