改进局部自适应的快速FCM肺结节分割方法

Research on Fast FCM Pulmonary Nodule Segmentation Algorithm Using Improved Self-adaption

进行肺部肿瘤计算机辅助诊断的关键问题是实现对病变组织的正确、快速分割,为此,提出了一种能够有效提高局部邻域像素自适应程度的快速模糊C均值聚类肺结节分割方法.首先构造像素与邻域窗口空间关系的二维向量表示,获得不同向量值的统计分布规律;然后用改进的空间函数综合考虑中心像素与单个相邻像素间的灰度相似度、与邻域窗口的空间相似度对模糊隶属度的贡献,动态地调整邻域像素的隶属度对中心像素的影响;最后给出该方法在迭代计算效率和局部自适应方面的改进.实验结果表明,该方法对血管粘连型、胸膜粘连型和毛玻璃肺结节的分割效果优于其他典型算法.

The key problem of computer-aided diagnosis （CAD） of the lung cancer is to segment the pathological changed tissues fast and accurately. As pulmonary nodules are potential manifestation of the lung cancer, we propose a fast fuzzy C-means clustering pulmonary nodules segmentation method that can effectively improve the local neighborhood self-adaptability. First, the algorithm constructs two-dimensional vectors of spatial relations between pixels and neighborhood, for getting the statistical distribution pattern of different vector. Then, the enhanced spatial function considers both the gray- scale similarity and spatial similarity, updates the cluster centers iteratively and fuzzy membership degree, for adjusting the effect of membership degree from the neighborhood pixels. Finally, algorithm analysis shows improvement for iteration computing efficiency and local self-adaption. The experimental results show that the proposed algorithm can achieve more accurate segmentation of vascular adhesion, pleural adhesion and ground glass opacity （GGO） pulmonary nodules, and performs better in convergence efficiency and error rates.