基于T-prim模型的肺气管分割算法

针对现有高阶支气管分割算法计算成本过大或分割精度不足等问题,提出一种基于T-prim模型的肺气管树分割算法。通过形态学灰度重建对CT图像进行初步处理,使用区域生长算法得到主支气管;从马尔可夫随机场的角度对分水岭算法分割框架进行优化,得到优化的分割框架;利用主气管骨架提取自动获得种子节点,算法迭代构造出T-prim模型,利用优化的分割框架得到完整的肺气管树。通过与两种EXACT09竞赛算法的对比实验证明了该算法在不依赖于种子点的人工选择,不需要训练集的条件下,能以极低的泄漏量获得更完整的分割结果。

LUNG AIRWAY SEGMENTATION ALGORITHM BASED ON T-PRIM MODEL

Aiming at the problems such as excessive calculation cost or insufficient segmentation accuracy of the existing high-generations bronchial segmentation algorithm,this paper proposes a lung tracheal tree segmentation algorithm based on T-prim model.The CT image was preliminarily processed by morphological gray level reconstruction,and the main bronchi was obtained by region growing algorithm.Then,the segmentation framework of watershed algorithm was optimized from the perspective of random Markov random field,and the optimized segmentation framework was obtained.Finally,we used the skeleton of the main trachea to extract the seed nodes automatically,and constructed T-prim model iteratively.The complete lung-trachea tree was obtained from the optimized segmentation framework.Compared with two EXACT09 contest algorithms,the experimental results show that the proposed algorithm can achieve more complete segmentation results with extremely low leakage without relying on the artificial selection of seed points under the condition of no training set.