Department of Electrical and Computer Engineering, University of Calgary, Calgary, AB, Canada.
Int J Comput Assist Radiol Surg. 2012 Jan;7(1):111-23. doi: 10.1007/s11548-011-0632-y. Epub 2011 Jun 22.
The major hurdle for three-dimensional display of lung lobes is the automatic recognition of lobar fissures, boundaries of lung lobes. Lobar fissures are difficult to recognize due to their variable shape and appearance, along with the low contrast and high noise inherent in computed tomographic (CT) images. An algorithm for recognizing the major fissures in human lungs was developed and tested.
The algorithm employs texture analysis and fissure appearance to mimic the way that surgeons/radiologists read CT images in clinical settings. The algorithm uses 3 stages to automatically find the major fissures in human lungs: (a) texture analysis, (b) fissure region analysis, and (c) fissure identification.
The algorithm's feasibility was evaluated using isotropic CT images from 16 anonymous patients with varying pathologies. Compared with manual segmentation, the algorithm yielded mean distances of 1.92 ± 2.07 and 2.07 ± 2.37 mm, for recognizing the left and right major fissures, respectively.
An automatic recognition algorithm for major fissures in human lungs is feasible, providing a foundation for the future development of a complete segmentation algorithm for lung lobes.
肺部三维显示的主要障碍是叶裂的自动识别，即肺叶的边界。由于叶裂的形状和外观多变，且 CT 图像固有对比度和噪声低，因此很难识别。本文开发并测试了一种用于识别人肺主要叶裂的算法。
该算法采用纹理分析和叶裂外观模拟外科医生/放射科医生在临床环境中阅读 CT 图像的方式。该算法使用 3 个阶段自动在人肺中找到主要叶裂：（a）纹理分析，（b）叶裂区域分析，和（c）叶裂识别。
该算法的可行性使用来自 16 名具有不同病理的匿名患者的各向同性 CT 图像进行了评估。与手动分割相比，该算法识别左右主叶裂的平均距离分别为 1.92 ± 2.07mm 和 2.07 ± 2.37mm。
一种用于识别人肺主要叶裂的自动识别算法是可行的，为未来开发完整的肺叶分割算法奠定了基础。
