摘要
目的探讨不同滤波方法和分割算法对MRI脑肿瘤图像分割精度的影响,寻找最适合脑部胶质瘤MRI图像的滤波方法和分割算法。资料与方法通过MATLAB编程,分别使用非局部均值滤波、中值滤波、各向异性滤波和改进均值漂移分割、模糊C均值分割、分水岭分割算法对39幅脑肿瘤图像进行分割,其中39幅图像为39例不同患者的胶质瘤图像。以医师手动分割结果作为"金标准",评价不同方法分割精度。结果非局部均值滤波信噪比为7.9243,中值滤波信噪比为6.2160,各向异性滤波信噪比为6.5426;改进均值漂移分割算法精确度为92.31%,模糊C均值分割精确度为88.03%,分水岭分割精确度为84.93%。结论各种滤波方法和分割算法中非局部均值滤波效果优于中值滤波和各向异性滤波,改进均值漂移算法分割精度高于分水岭算法和模糊C均值算法,精确度高达92.31%。
Purpose To explore the segmentation accuracy of different filtering and segmentation methods in brain tumor MRI, and to identify the best algorithm for brain glioma. Materials and Methods Using the nonlocal average filtering, median filtering, the anisotropic filtering and improved mean shift algorithm segmentation, the watershed segmentation algorithm, fuzzy c-means segmentation algorithm to realize image segmentation in MATLAB program, 39 glioma images from different patients were analyzed. Pathology manual segmentation was used as gold standard to evaluate different segmentation precision. Results The signal-to-noise ratio was 7.9243, 6.2160 and 6.5426 for different filter methods, respectively. The segmentation methods accuracy was 92.31%, 88.03% and 84.93%, respectively. Conclusion The nonlocal average filtering effect is more accurate than median filtering and the anisotropic filtering. The improved mean shift algorithm segmentation is more accurate than watershed segmentation algorithm and fuzzy c-means segmentation algorithm with precision of 92.31%.
出处
《中国医学影像学杂志》
CSCD
北大核心
2015年第7期553-556,560,共5页
Chinese Journal of Medical Imaging
基金
四川省科技支撑项目(2015KJT0002-2014SZ0124)
关键词
脑肿瘤
神经胶质瘤
磁共振成像
图像处理
计算机辅助
算法
Brain neoplasms
Glioma
Magnetic resonance imaging
Image processing
computer-assisted
Algorithms
