摘要
目的:本文基于灰度共生矩阵和游程长矩阵方法研究脑梗塞患者MR图像的纹理特征,目的在于揭示脑梗塞患者与健康对照组MR图像纹理是否存在显著性差异,从而借助这一微观改变实现脑梗塞患者的早期诊断。方法:提取患病组和健康对照组纹理特征参量,利用fisher系数进行有效纹理特征参量的筛选,构建分类器。结果:LDA分类器识别率为88.31%。说明脑梗塞患者和对照组MR图像纹理特征存在差异。结论:利用基于统计学的纹理分析方法可以有效的揭示脑梗塞患者脑组织纹理的细微改变,进而实现对脑梗塞疾病的早期诊断。
Objective： Based on Co-occurrence Matrix and Run-length Matrix, we studied cerebral infarction patients＇ MR im- age texture characters. The aim is to investigate the differences of lesion textures characters between patients＇ groups and nor- mal control groups, so that we can use this tiny change to realize early diagnosis of cerebral infarction. Methods： Texture fea- tures were extracted from MR images of patients and normal control groups respectively. Fisher test was applied to choose valid textures characters and made features classifier. Results： Linear discriminant analysis can achieve 88.31% classification accuracy. This demonstrated that cerebral infarction patients and normal control groups have the differences of textures Char- acters in MR image. Conclusion： We can discover cerebral infarction patients＇ MR image texture characters change by texture analysis, so that early diagnosis of cerebral infarction would be realized.
出处
《中国医学物理学杂志》
CSCD
2010年第1期1607-1609,共3页
Chinese Journal of Medical Physics
基金
北京市重点实验室开放研究课题
国家自然科学资助项目(No.30670656)
