Abstract: Pulmonary fissure detection in CT image plays a key role for early disease examination and later treatment. However, pulmonary fissures in 2D image are soft tissues which appear as thin, weak and varying structures, often undulating planar surfaces with blurred boundaries. This is further complicated by presence of surrounding vessels and other structures. To solve these problems, a new regional iteration and back-tracking algorithm based on geodesic voting is proposed. The main procedure is as follows：A geodesic distance fieldis constructed with the fast marching method and a histogram peak finding method is used to locate the fissure branch. Then fissure branches are detected iteratively with a regionback-tracking method. The topological structure with fine branches is preserved while the noise is simultaneously suppressed. The experiment results show that the proposed method is less sensitive to noise with a good fault tolerance performance. It also improves the accuracy and computation efficiency of pulmonary fissure detection compared with the conventional global back-tracking methods.

Key words: pulmonary fissure, fast marching, geodesic voting, histogram, back-tracing

摘要： 在医学图像中，某些重要的CT图像中肺裂的检测在早期疾病检查和后期治疗中起着关键作用。然而，在2D断面方向上肺裂表现为细薄的曲线，对比度较低，密度值分布不均匀；肺部血管、支气管的存在也会干扰肺裂的检测。针对这些问题，提出了一种新的基于测地线投票的区域迭代回溯方法。该方法的主要处理流程为：首先采用快速行进方法构造测地线距离场，用直方图峰值查找方法定位肺裂分支，然后用区域迭代回溯方法搜索肺裂分支；结合测地线投票，从而在抑制噪声干扰的同时保持分支拓扑结构。实验结果表明该方法对噪声不敏感，具有良好的抗噪容错性能，较传统的全局回溯方法，区域回溯提高了检测的准确率和计算效率。

关键词: 肺裂, 快速行进, 测地线投票, 直方图, 回溯