Alzheimer病MR图像纹理和形态学特征的比较研究

Comparison of Texture Analysis with Morphological Measure Based on the MR Images of Alzheimer's Disease

目的比较MR图像纹理分析和形态学测量方法在鉴别阿尔茨海默病(Alzheimer’sdisease,AD)患者的实际效果,探讨纹理分析技术对AD早期诊断的价值。方法提取29例AD患者和19例健康对照者(NC)的海马纹理参数,并测量内侧颞叶最窄宽度和海马结构体积等形态学指标,在t检验基础上对纹理参数与形态学指标进行判别分析,然后采用Kappa参数值评价上述两种方法的一致性。结果 AD组与NC组间的和均值、能量等纹理参数均显著不同,右侧海马和均值与形态学测量指标内侧颞叶最窄宽度的判别准确度均为97.9%。经多元逐步判别分析,纹理参数联合判别的准确度为97.9%。纹理分析与形态学测量的Kappa值为0.957(p<0.001)。结论 MR图像的纹理分析与形态学测量具有较好的一致性,纹理分析可能有助于AD的早期诊断。

Objective To evaluate the positive value of texture analysis in early diagnosis for Alzheimer＇ s disease （AD） by means of comparing the discriminative ability of texture analysis and morphologic measure. Methods The texture parameters of hippocampus were firstly extracted from 29 patients with AD and 19 normal controls （NC） , and then the morphological indexes, such as the thinnest medial temporal lobe （tMTL）, the volumes of hippocampal formation （HF） were measured. Based on t test, discriminant analyses of both texture parameters in hippocampus and morphological indexes were respectively conducted. Finally, the value of Kappa was used to evaluate the consistency between two methods. Results The texture features such as sum average （SAV） and energy （ENG） in hippocampus were significantly different between AD and NC groups （p 〈 0.01 ）. The discriminations of SAV in hippocampus and the morphological indicator tMTL were both 97.9%. By a stepwise discriminant analysis, the classification accuracy of discriminant by combining texture parameters was 97.9%. With Kappa test, Kappa value between texture analysis and morphological measure was 0. 957 （p 〈 0. 001）. Conclusion Texture analysis of MR images is consistent well with the morphological measure. This study may be helpful in the early diagnosis of AD.