基于嗅皮层变性及脑结构MRI的AD分类及MCI转化预测研究

The early non-invasive detection of individuals at risk of developing Alzheimer's disease（AD）in clinical settings is of important significance for the clinicians and researchers. Olfactory impairment (OI) has been regarded as a potential biomarker of early stage and preclinical AD in recent studies. In contrast, there are limited studies on the functional activity decline of olfaction and the corresponding pathological changes in brain structures. In the current study we provide a validation of the decline in olfactory activity in the primary olfactory cortex（POC）was correlated with the structural atrophy using olfactory functional magnetic resonance imaging (fMRI). The AD classification model based on advanced automated segmentation techniques for structural magnetic resonance imaging (sMRI) analysis data is robust and efficient. However, the role of POC in AD classification is not clear. In the further study, multivariate analysis (MVA) method will be used to investigate the value of combining olfaction cortex degeneration and structural segmentation of MRI for the classification of AD and mild cognitive impairment (MCI). Furthermore, multiple future time points follow up of MCI will externally validate the AD vs. CTL classification model, and a new identification way based on multimode MRI for early AD classification will be provided theoretically.

探索阿尔茨海默病（AD）早期个性化无创性分类诊断模型，并对MCI患者是否会进展为AD进行预测具有重要的临床及科研意义。近年来研究表明嗅觉损害可作为AD早期及临床前期的潜在生物学标志物，但是基于嗅功能活动减低及反应其相应病理变化的脑结构改变的研究有限。课题组已通过嗅觉脑功能磁共振（fMRI）研究证实AD及轻度认知功能障碍（MCI）患者初级嗅皮层（POC）区激活减低，并与该区脑萎缩程度有关。基于先进的脑结构磁共振成像（sMRI）自动分割结果建立的AD分类模型有效且稳健，但嗅皮层变性在AD分类诊断中的作用尚不明确。本课题拟采用多元分析方法（MVA）联合嗅皮层变性及脑结构自动分割结果建立AD vs. CTL分类模型，对其在AD及MCI早期个性化诊断中的价值做进一步研究；并对MCI进行长期多时间点随访，对MCI病人是否转化为AD进行外部验证，为建立基于多模态MRI技术的AD分类诊断模型提供新思路。

近年来研究表明嗅觉损害可作为AD早期及临床前期的潜在生物学标志物，但是基于嗅功能活动减低及反应其相应病理变化的脑结构改变的研究有限。本课题组通过嗅觉脑功能磁共振（fMRI）研究证实AD及轻度认知功能障碍（MCI）患者初级嗅皮层（POC）区激活减低，并与该区脑萎缩程度有关。同时对脑结构及脑白质高信号（WMH）自动分割及定量测量进行可重复性进行研究，并对病灶填充是否影响脑结构自动分割进行研究，进而获得稳健的脑结构及WMH自动分割方法。本课题组进一步采用机器学习方法，联合嗅皮层变性特征（嗅觉脑fMRI POC区BOLD信号激活百分比、POC区脑体积）及基于FreeSurfer脑sMRI自动分割技术得到的皮层厚度及局部脑体积作为特征，分别并联合建立AD vs. CTRL、MCI vs. CTRL及AD vs. MCI分类诊断模型。发现联合嗅皮层变性及脑体积特征可以提高上述疾病的分类诊断效力，为建立基于多模态MRI技术的AD分类诊断模型提供新思路。