轻度认知障碍及健康对照的分类研究

摘要

Objective The aim of this study is to research the MR images 3D texture features of corpus callosum from the patients with mild cognitive impairment (MCI), and using the features to identify MCI and normal control (NC). Methods 3D texture analysis was performed on 20 MCI patients and 20 NC. The three-dimensional texture features were extracted from corpus callosum by gray-level co-occurrence matrix and run length matrix. The texture features that existed significant differences between MCI and NC were used as the features in a classification procedure. Back propagation (BP) neural network model were built to classify MCI patients from NC. Results The classification accuracy of the training set and test set was separately 95.83％ and 93.75％. Conclusion The back propagation neural network model with three-dimensional texture features can recognize MCI patients and NC.%目的 研究轻度认知障碍(Mild Cognitive Impairment,MCI)患者脑部MR图像的三维纹理特征,利用此特征对MCI和健康对照(Normal Control,NC)进行分类识别,以探索MCI诊断的新途径.方法 分别选取MCI患者和NC的脑部MR图像各20例进行三维纹理分析,采用灰度共生矩阵和游程长矩阵提取每位受试者胼胝体的三维纹理特征.通过筛选得到组间存在显著性差异的纹理特征参量,利用BP神经网络建立识别模型,对MCI和NC进行分类识别.结果 训练集和测试集的分类识别正确率分别为95.83％和93.75％.结论 利用三维纹理特征的BP神经网络模型可以分类识别MCI和NC.