摘要
目的比较区域生长和模糊C均值聚类两种经典分割算法在颅脑CT图像脑实质和脑脊液分割中的应用。方法搜集49例CT颅脑检查病例图像,分别采用区域生长法和模糊C均值聚类法,实现颅脑CT图像脑实质和脑脊液的分割,并对实验结果进行分析。结果区域生长法相对完整地分割了脑实质和脑脊液,而模糊C均值聚类算法提取的脑实质部分区域(主要是灰质)被错误地分割为脑脊液,造成脑实质提取不完整。结论在颅脑CT图像脑实质和脑脊液的分割中,区域生长法优于模糊C均值聚类法,但在运行时间和自动化程度上,还需进一步改进。
Objective To compare two methods （region growth and fuzzy C-means cluster） for segmentation of brain parenchyma and cerebrospinal fluid based on brain CT images. Methods Two methods were analyzed and applied to the segmentation of 49 series of CT brain images. Results Region growth algorithm gave the better results with a good labeling of brain parenchyma and cerebrospinal fluid. The algorithm of fuzzy C-means cluster resulted in imperfect attraction of brain parenchyma. Conclusions The experiment results shows that based on CT images, region growth algorithm works better in the segmentation of brain parenchyma and cerebrospinal fluid as far as the segmentation quality is concerned. However, there are limitations in the running time and automation that can be improved.
出处
《临床医学工程》
2013年第2期134-135,共2页
Clinical Medicine & Engineering
基金
江苏省高等学校大学生实践创新训练计划一般项目(项目编号:1015)
关键词
区域生长
模糊C均值聚类
脑实质
脑脊液
Region growth
Fuzzy C-means cluster
Brain parenchyma
Cerebrospinal fluid
