基于三维区域增长和灰度积分投影法的肺实质分割

为了提高肺结节计算机辅助检测、辅助诊断的准确性,研究精确分割肺实质方法。首先,采用Otsu法实现阈值法粗分割;其次,设计三维区域增长方法有效去除粘连在肺实质上的主气管和支气管树;再次,采用灰度积分投影法以及行扫描方法对左右肺粘连的情况进行判别和处理,实现左右肺分离;最后,用滚球法修复肺区边界,实现肺实质分割。该方法有效解决了肺实质分割中粘连肺实质的气管、支气管树去除问题,高密度的肺部组织易丢失、粘连肺壁的肺结节容易被遗漏等问题。通过对多组LIDC-IDRI肺部图像数据集的实验表明,该方法能够精确高效地实现对于肺实质分割。

Segmentation of Lung Parenchyma Based on Three-Dimensional Region Growing and Gray-Level Integral Projection

Purpose In order to improve the computer-aided detection and diagnosis of pulmonary nodules, precise lung parenchyma segmentation method was studied. Materials and Methods To begin with, the Otsu threshold method was used to achieve coarse segmentation. Next, the use of three-dimensional region growing method was designed to remove lung trachea and bronchial tree which was attached to lung parenchyma. Furthermore, gray-level integral projection and scan line searching was used to judge adhesion of left and right lungs, and separate left and right lungs if adhesion occurs. Finally, rolling ball was used to repair lung boundaries, and lung parenchyma segmentation was achieved. Results The method is effective to solve those problems: lung trachea and bronchial tree was attached to lung parenchyma, high density lung tissue would be lost easily when segmentation, and juxta-pleural nodules was likely to be missed. Conclusion The experiment with multiple LIDC-IDRI sets shows that the method can implement the lung parenchyma segmentation accurately and efficiently.