模糊聚类技术分割颅脑肿瘤MR图像的应用研究  

Application study of cerebral neoplasms MR images segmentation using fuzzy clustering technology

作　　者：段玉婷[1] 张梦龙[2] 姜玉芝[1] 谢国华[1] 

机构地区：[1]山东省聊城市人民医院影像科,山东聊城252000 [2]山东省医学影像学研究所,山东济南250021

出　　处：《医学影像学杂志》2009年第9期1091-1093,共3页Journal of Medical Imaging

摘　　要：目的:分析自编的模糊C均值(fuzzy cmeans,FCM)算法图像处理程序对颅脑肿瘤MRI图像的分割效果。方法:采集50幅不同脑肿瘤MRI平扫系列图像,用FCM程序对病灶自动分割,同时请影像学专家对相应病灶面积进行手工绘制,用重合率和误分割率两个指标进行比较分析两者的分割效果。结果:FCM程序分割病灶的重合率为91.2%,误分割率为10.8%。二者面积的差值无显著性差异(P>0.05)。结论:FCM程序可以无监督地自动分割脑肿瘤MRI图像,有较强的临床应用价值。Objective：To study the effects of fuzzy C-means（FCM） algorithm for image processing prepared by authors in the segmentation of cerebral neoplasms MR images.Methods：FCM was performed in 50 cases with different cerebral neoplasms MR images and compared to the hand-drawn images lesion areas by experts.The following indicators were extracted：coincidence rate,segmentation error rate,and analysed data.Results：The coincidence rate is 91.178% and the segmentation error rate is 10.8% by using FCM process.And no significant statistic differences （ P 〉 0.05） were shown between two areas. Conclusion： FCM procedure can be used as automatic and non-supervisory method in segrnentatior, of cerebral neoplasms MR images and it have a strong clinical value.

关 键 词：模糊C均值 图像分割 颅脑肿瘤 磁共振成像 

分 类 号：R739.4[医药卫生—肿瘤] R445.2[医药卫生—临床医学]