论文部分内容阅读
肺区分割是计算机辅助诊断肺癌的前提。当肿块与胸壁粘连时,由于两者的计算机断层成像(CT)值接近,基于局部低级特征的传统分割方法不能得到正确结果;而且由于肿块体积大,造成了肺区内正常组织的大面积缺失,故以往含胸壁粘连型肺结节(直径小于3cm)的肺区分割方法不再适用,需要采用能结合先验形状和低级特征的主动形状模型(ASM)来分割含胸壁粘连型肺肿块的肺区。但传统ASM的搜索步骤是一种基于最小二乘的优化方法,该方法对异常标记点敏感,会使轮廓更新到正常肺组织和肿块的过渡区域而不是真正的肺边缘。针对这一问题,提出了改进的ASM算法:首先基于距离特征识别异常标记点,然后赋予异常标记点和正常标记点不同的搜索函数。搜索过程在设定的包围核(VOI)内进行。用所提出的ASM方法分割30个含胸壁粘连型肿块的肺区,与金标准的重叠度为93.6%。实验结果表明针对含胸壁粘连型肿块的肺区分割问题,改进的ASM算法能得到较好的分割结果,并且算法的运行时间是在临床可接受的范围内。
Pulmonary segmental segmentation is a prerequisite for computer-assisted diagnosis of lung cancer. When the mass is attached to the chest wall, the traditional segmentation method based on the local low-level features can not get correct results due to the close computed tomography (CT) values of both, and due to the bulky mass, a large area of normal tissue in the lung area is lost , The past classification of lung region with thoracic adhesions (less than 3 cm in diameter) is no longer applicable, and an active shape model (ASM) that incorporates a priori shape and low-level features is required to segment the thoracic wall-attached lung mass Lung area. However, the traditional ASM search step is a least-squares-based optimization method that is sensitive to anomalous marker points and updates the profile to the transitional area of normal lung tissue and mass instead of the true lung margin. To solve this problem, an improved ASM algorithm is proposed. Firstly, the abnormal marker points are identified based on the distance features, and then different search functions are given to the abnormal marker points and the normal marker points. The search process is performed within the set VOI. Using the proposed ASM method to segment the lung region of 30 thoracic adhesions, the overlap with the gold standard was 93.6%. The experimental results show that the improved ASM algorithm can get a better segmentation result and the running time of the algorithm is within the clinically acceptable range for the lung region segmentation problem of the thoracic adhesions.