阿尔茨海默症的脑灰质萎缩研究

发布时间：2018-09-10 17:35

【摘要】：阿尔茨海默症(Alzheimer’s disease,AD)是一种多发于老年人的神经退行性病变,临床表现为记忆及其他认知功能的衰退。阿尔茨海默症患者大多是在出现明显痴呆症状后才进行治疗,难以起到遏制认知下降的作用,因此需要将治疗阶段前移。目前临床上对阿尔茨海默症的诊断主要是依靠认知评分量表,而结构磁共振成像揭示的灰质萎缩有助于我们对阿尔茨海默症进行早期诊断。本文基于结构磁共振图像,分析了阿尔茨海默症不同阶段的脑灰质体积差异及灰质萎缩模式,并根据脑灰质体积建立了阿尔茨海默症的分类及认知评分预测模型。应用基于体素的形态学分析,我们比较了正常对照(normal control,NC)、非转化型轻度认知障碍(stable mild cognitive impairment,s MCI)、转化型轻度认知障碍(converted MCI,c MCI)、阿尔茨海默症之间的灰质体积,同时通过不同时间点数据揭示灰质萎缩模式。然后将组间的群体差异体素作为原始特征,利用结合特征选择的支持向量机对体素进一步筛选,建立NC/MCI、NC/AD、MCI/AD、s MCI/c MCI分类模型。除此之外,我们还根据纵向变化中涉及的体素点建立简易精神状态量表(MMSE)评分的预测模型。通过组间双样本T检验发现阿尔茨海默症的灰质萎缩是一个渐进的过程,萎缩范围从颞叶、岛叶、基底核开始,然后增加顶叶、枕叶、扣带回,最后再增加额叶。通过组内单因素方差分析发现转化型轻度认知障碍与阿尔茨海默症具有类似的灰质萎缩模式,萎缩速率快萎缩范围大。非转化型轻度认知障碍与正常对照具有类似的灰质萎缩模式,萎缩速率慢萎缩范围小。结合基线灰质体积和萎缩速率发现非转化型轻度认知障碍与转化型轻度认知障碍的差异最开始并不明显,随着时间推移差异逐渐加大。分类模型对NC/MCI、NC/AD、MCI/AD、s MCI/c MCI的分类准确率分别达到84.9%、100%、81.4%、73.9%。内外侧颞叶在四组分类中都是特征体素的重要部位。感兴趣区分析和特征体素分析说明扣带回后部、缘上回、楔前叶可能是早期区分非转化型轻度认知障碍与转化型轻度认知障碍的关键脑区。MMSE评分预测模型在交叉验证下的均方误差为2.5460,预测值与实际值的相关系数为0.8191。
[Abstract]:Alzheimer's disease (Alzheimer's disease,AD) is a neurodegenerative disorder that occurs in the elderly and is characterized by a decline in memory and other cognitive functions. Most patients with Alzheimer's disease are treated only after obvious dementia symptoms, which is difficult to curb cognitive decline, so the treatment phase needs to be moved forward. At present, the clinical diagnosis of Alzheimer's disease mainly depends on cognitive scale, and the gray matter atrophy revealed by structural magnetic resonance imaging is helpful for the early diagnosis of Alzheimer's disease. Based on the structural magnetic resonance images, the difference of gray matter volume and the pattern of gray matter atrophy in different stages of Alzheimer's disease were analyzed, and the prediction model of classification and cognitive score of Alzheimer's disease was established according to the volume of gray matter. Using morphologic analysis based on voxel, we compared the gray matter volume between normal control (normal control,NC) and non-transforming mild cognitive impairment (stable mild cognitive impairment,s MCI),) with mild cognitive impairment (converted MCI,c MCI),). At the same time, the pattern of gray matter atrophy was revealed by data at different time points. Then the group differential voxels among groups are taken as the original features and the support vector machine (SVM) combined with feature selection is used to further screen voxels and establish a NC/MCI,NC/AD,MCI/AD,s MCI/c MCI classification model. In addition, we also established a predictive model of the (MMSE) score of the mini-mental state scale according to the voxel points involved in the longitudinal changes. A double-sample T test showed that the gray matter atrophy of Alzheimer's disease was a gradual process, beginning with temporal lobe, insular lobe, basal nucleus, then increasing parietal lobe, occipital lobe, cingulate gyrus, and finally increasing frontal lobe. Univariate analysis of variance within the group showed that the transforming mild cognitive impairment had a similar gray matter atrophy pattern with Alzheimer's disease, and the rapid atrophy rate had a wide range of atrophy. Non-transforming mild cognitive impairment had a similar pattern of gray matter atrophy, and the atrophy rate was small. Combined with baseline gray matter volume and atrophy rate, it was found that the difference between non-transforming mild cognitive impairment and transforming mild cognitive impairment was not obvious at the beginning, and the difference gradually increased with the passage of time. The classification accuracy of classification model for NC/MCI,NC/AD,MCI/AD,s MCI/c MCI was 84.9 and 81.4, respectively. The internal and external temporal lobes are important parts of the characteristic voxel in the four groups. The analysis of region of interest and characteristic voxel analysis showed that the posterior cingulate gyrus, superior marginal gyrus, The mean square error of MMSE prediction model is 2.5460 and the correlation coefficient between the predicted value and the actual value is 0.8191, which may be the key area to distinguish the non-transforming mild cognitive impairment from the transforming mild cognitive impairment in the early stage. The mean square error of the prediction model under cross-validation is 2.5460, and the correlation coefficient between the predicted value and the actual value is 0.8191.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：R749.16

【相似文献】