论文部分内容阅读
利用磁共振(MR)图像对阿尔茨海默病(AD)和健康对照(NC)进行分类识别,比较双侧海马在分类识别中的意义。选取AD患者和NC各25人,采用灰度共生矩阵和游程长矩阵提取每位受试者的海马部位的三维纹理特征。通过筛选得到组间存在显著差异的纹理特征参量,对主成分分析、线性判别分析和非线性判别分析3种方法得到的识别结果进行比较。利用反向传播(BP)神经网络建立识别模型,对AD和NC进行分类识别,采用相关性分析比较双侧海马纹理参数与简明智力状态检查(MMSE)评分的相关性。结果显示使用神经网络模型的非线性判别分析的分类识别正确率最高,右侧海马分类识别的正确率均高于左侧。两侧海马的纹理特征与MMSE评分均具有相关性且右侧海马的相关性系数均大于左侧。利用三维纹理特征的神经网络模型可分类识别AD组和NC组,并且采用右侧海马进行分类识别可能更有利于AD的诊断。
Magnetic resonance (MR) images were used to classify and identify Alzheimer’s disease (AD) and healthy controls (NC). The significance of bilateral hippocampus in classification and recognition was compared. 25 patients with AD and 25 NC patients were selected. The three-dimensional texture features of each subject’s hippocampus were extracted by using gray level co-occurrence matrix and run length matrix. The texture feature parameters with significant differences among the groups were obtained by screening, and the recognition results obtained by the three methods of principal component analysis, linear discriminant analysis and non-linear discriminant analysis were compared. The recognition model was established by back propagation (BP) neural network, and the AD and NC were classified and identified. Correlation analysis was used to compare the correlation between bilateral hippocampal texture parameters and Concise Mental Status Examination (MMSE). The results show that the neural network model based on the non-linear discriminant analysis of the highest classification accuracy, right hippocampus classification accuracy are higher than the left. The texture features of both hippocampus were correlated with MMSE scores and the correlation coefficient of the right hippocampus was greater than that of the left. Neural network model using three-dimensional texture features can classify AD group and NC group, and using the right hippocampus for classification and classification may be more conducive to the diagnosis of AD.