轻度认知障碍患者脑结构MRI分析与分类研究

发布时间：2017-12-26 18:09

本文关键词：轻度认知障碍患者脑结构MRI分析与分类研究　出处：《华南理工大学》2016年硕士论文　论文类型：学位论文

【摘要】：阿尔茨海默病(Alzheimer’s disease,AD)是一种多发于老年人的神经退行性病变,轻度认知障碍(mild cognitive impairment,MCI)具转化为AD的高风险性,它包括稳定型MCI(stable MCI,s MCI)和转化型MCI(converted MCI,c MCI)。分析s MCI、c MCI患者的脑结构异同,以及它们和正常老年人、AD患者的区别,对揭示MCI的转化规律与过程,控制和干预疾病进程,具有十分重要的意义。本研究选取225人的脑结构MR影像,其中包括72例正常老年人作为对照组(normal control,NC)、46例s MCI、40例c MCI和67例AD。每一组人群均在基线期(初次检查)、基线期后12个月和24个月,共采集三次MR影像。首先,对数据进行预处理并计算其脑皮层特征(皮层厚度、灰质体积、表面积和平均曲率),随后,应用双样本T检验法,比较NC、s MCI、c MCI和AD组间脑皮层特征差异。为了更有效地区分不同人群,根据不同脑皮层特征、脑皮层的基线和纵向变化特征,分别采用递归特征消除法(RFE)和稀疏约束降维法(SCDR)两种方法进行降维,再利用支持向量机(SVM)进行自动分类。将SVM分别与RFE、SCDR结合,得到SVM-RFE模型和SVM-SCDR模型。最后,对不同皮层特征、两个模型的分类性能进行比较和分析。结果发现,从NC发展为AD的过程中,患者的脑皮层、表面积和脑沟(平均曲率表征脑沟)是一个渐进性变化的过程。AD脑结构变化的关键脑叶是颞叶,颞叶脑区的皮层厚度和平均曲率在s MCI时期已发生显著变化,其灰质体积显著减小是在c MCI时期,表面积变化在AD阶段逐渐明显,而颞叶的脑结构变化随着病程进展而逐步加重。从s MCI开始,患者的脑结构变化逐渐扩散到额叶、顶叶、枕叶等脑叶的大部分脑区和扣带回,至AD阶段,脑结构萎缩扩展到各脑叶的重要脑区。在组间分类性能方面,对分类最有效的皮层特征是皮层厚度和灰质体积,表面积和平均曲率次之;尤其多皮层特征,以及基线与纵向变化特征融合,显著提高分类性能。其中,SVM-RFE模型的分类性能优于SVM-SCDR。研究结果说明,NC、s MCI、c MCI和AD的脑皮层特征及其变化,存在显著差异,其结构萎缩是一个渐进的过程,脑皮层特征可作为区分AD不同阶段、评估MCI转化风险的重要指标。
[Abstract]:Alzheimer's disease (Alzheimer 's disease, AD) is a multiple of senile neurodegenerative disease, mild cognitive impairment (mild cognitive, impairment, MCI) has transformed into the high risk of AD, which includes a stable type MCI (stable MCI, s MCI) and MCI (converted MCI, into C MCI). Analyzing the similarities and differences of brain structure in patients with s MCI and C MCI and their difference from normal elderly patients and AD patients is very important for revealing the transformation rule and process of MCI, controlling and interfering in the course of disease. The study selected 225 people's brain structure MR images, including 72 normal aged people as the control group (normal control, NC), 46 cases of s MCI, 40 cases of C MCI and 67 cases of AD. Three MR images were collected at baseline (first examination), 12 months after baseline and 24 months in each group. First, preprocessing the data and calculate the cerebral cortex characteristics (cortical thickness, gray matter volume, surface area and mean curvature), then test using two sample T, NC, MCI, s difference between C MCI and AD groups of cortex. In order to more effectively distinguish different groups according to different characteristics of cerebral cortex cerebral cortex, baseline and longitudinal variation, respectively using recursive feature elimination (RFE) and sparse constraint dimensionality reduction (SCDR) to reduce the dimensions of the two methods, and then use the support vector machine (SVM) classification. The SVM-RFE model and the SVM-SCDR model are obtained by combining SVM with RFE and SCDR respectively. Finally, the classification performance of different cortical features and two models was compared and analyzed. The results showed that the cerebral cortex, surface area and cerebral sulcus (mean curvature of the brain) were a gradual process during the development of NC to AD. The key of lobar AD structural brain changes is the temporal lobe cortex thickness and the average curvature of temporal lobe regions has changed significantly in s MCI period, the gray matter volume decreased significantly in C MCI period, changes in the surface area of AD phase gradually obvious, and structural brain changes in the temporal lobe with the course of development and gradually increase. From the beginning of s MCI, brain structure changes of patients gradually spread to most brain regions of the frontal lobe, parietal lobe and occipital lobe lobe and cingulate gyrus, to the stage of AD, atrophy of brain structure extended to brain area of each lobe. In terms of inter group classification performance, the most effective cortical characteristics of classification were cortical thickness and gray matter volume, surface area and mean curvature followed. In particular, multi cortical characteristics and baseline and vertical variation characteristics fusion significantly improved classification performance. Among them, the classification performance of SVM-RFE model is better than that of SVM-SCDR. The results showed that NC, s MCI, C MCI and AD had significant differences in their cerebral cortex characteristics and their changes, and their structural atrophy was a gradual process. Cerebral cortex characteristics can be used as an important index to distinguish AD from different stages and assess MCI transformation risk.
【学位授予单位】：华南理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：R749.1

【相似文献】