随着全世界经济发展和医疗卫生的改善，人口老龄化已经成为许多国家所面临的严峻问题。作为拥有庞大人口基数的发展中国家，中国在这方面的形势同样不容乐观。老年阶段是多种疾病的高发时期，庞大的老年人口数量也就意味着庞大的医疗和看护成本。正确认识老年疾病的发生发展机理，开发有效的疾病预防和治疗手段，是我们积极应对社会老龄化的重要举措。 在众多的老年疾病中，阿尔兹海默病（Alzheimer’s Disease, AD）因其给老人及其家庭带来的巨大身心危害而广受关注。作为一种退行性神经疾病，AD能够极大恶化个体认知能力衰退和大脑老化的发展轨迹，导致失智失能的严重后果。约15%的老年人（>65岁）深受其困扰，并且年龄越大发病率越高。随着其它疾病的有效干预和治疗手段不断发展，老年人的死亡原因中AD所占的比例正在逐年增加。这主要是因为当前AD的药物开发始终无法突破临床三期试验的瓶颈，难以实现AD的有效治疗。因此，研究人员将目光投向AD临床前的早期阶段，以期通过早发现早干预来实现对AD病理的有效改善。如何准确地筛选和评估AD早期风险人群便成了研究者所关注的焦点。 已有的研究表明，在AD临床确诊30年以前，大脑中便可以检测出相关的病理性变化，在此后的发展过程中，大脑的神经结构逐步发生病理性改变，成为了AD早期很好的标志物。已经有许多研究对AD临床前的轻度认知障碍（Mild Cognitive Impairment, MCI）阶段进行了大量的探索，证实该阶段患者会表现出典型的AD类大脑结构和功能改变。但该阶段患者已经有了明显的认知障碍，并在高速向AD发展，对于干预手段和药物的应用可能仍然不是一个十分理想的阶段。主诉记忆障碍（Subjective Memory Impairment, SMI）是一个比MCI更为早期的阶段，在这个阶段老人虽然可以主观地感受到记忆能力损伤，但却能够在实际测量中保持正常的记忆力水平。SMI老人以极高的MCI和AD转化率，被认为可能是目前AD病理可观测的最早期阶段。如果能够对这一目标人群进行准确的筛选评估，及时发现AD病理在超早期对大脑结构和功能的影响模式，将极大地方便干预手段和治疗药物的开发应用。 目前，已有部分研究描述了SMI个体在大脑结构和功能上表现出来的相异于健康老年人的神经变化，为SMI作为AD早期阶段提供了证据。也有一些研究通过对部分影像指标的分析，建立了认知能力和神经影像间的关系。相信随着更多研究结果得到报道，会有越来越多的研究者重视对这一阶段的探索，并最终为AD的干预和治疗找到突破口。 但是，当前的SMI研究也还存在着许多不足，有待进一步探索。首先，SMI研究中部分相同模态的结果报告不同，研究结论还缺乏一致性。其次，既往文献的报道大多只针对单一的问题进行局部的研究，缺乏对SMI在各个模态神经特征的系统性探索。最后，对SMI及AD早期病理发展神经机制的阐释较为片面，研究指标之间缺乏关联性，这不利于建立SMI的综合理论体系。以上不足使得目前我们在全面认识SMI神经特征方面还存在一定的局限性。本研究从北京社区纳入了健康老人（Health Control, HC），SMI老人和遗忘型轻度认知障碍（Amnestic Subjective Memory Impairment, aMCI）患者三组被试，对他们进行了成套的神经心理测验和多种模态的磁共振扫描。多维度地描述了SMI老人各认知领域的状态，以及大脑结构和功能网络的神经特征，并尝试建立神经影像与认知功能变化的联系，最后结合前人研究提出SMI的综合模型。本研究共分为4个部分。 研究1. 主诉记忆障碍的高级认知功能及灰质结构变化特点。本研究从北京老年脑健康（Beijing Aging Brain Rejuvenation Initiative, BABRI）数据库筛选了23名HC老人、22名SMI老人和24名aMCI患者，对他们的记忆、注意、视空间、语言和执行功能等认知能力进行了组间比较。结果发现在各项认知成绩上aMCI组都表现出与HC组间显著的差异，而SMI老人的行为表现介于两者之间，显示其较HC老人出现了一定程度的认知减退。SMI老人的灰质皮层厚度及灰质体积都和HC老人较为相似，二者和aMCI患者在内侧颞叶有显著差异，并且该处灰质体积与记忆显著相关，仅在右侧脑岛和左侧三角部额下回SMI老人较HC老人有一定灰质结构萎缩。 研究2. 主诉记忆障碍的白质结构变化特点。采集研究1样本中每个被试的白质结构像，对三组被试进行组间的白质属性和白质网络特征比较，观察这些指标与认知间的相关，并通过机器学习的方法探索这些指标在组别分类中的预测作用。在白质属性比较中，本研究发现多条白质纤维束出现了多种属性的显著组间差异。SMI老人在双侧丘脑前辐射区纤维束的表现和HC老人更为相似，但在海马回扣带束和HC老人有显著差异。在白质网络特征比较中，本研究发现SMI老人的网络全局特征保存相对完好，但中、后扣带等一些节点的局部效率已经呈现出类似aMCI的变化。机器学习发现白质属性的平均弥散率（Mean Diffusivity, MD）及轴向弥散率（Axial Diffusivity, DA）两者结合可以最有效地将SMI老人和aMCI患者从HC老人中区分出来，白质网络全局特征则只在aMCI患者和HC老人的区分中效果显著。 研究3. 主诉记忆障碍的功能网络变化特点。采集研究1样本中每个被试的静息态和工作记忆任务态功能核磁共振影像。通过独立成分分析（Independent Component Analysis, ICA）的方法提取默认网络（Default Mode Network, DMN）并进行组间比较，此后再以差异区为感兴趣区做全脑功能连接的组间比较。任务态下也进行了相似的功能连接组间比较。结果发现在静息态下SMI老人的DMN还保持着与HC老人较为相似的完整性。但SMI老人从该核心节点到额叶的部分功能连接表现出了与HC老人显著不同的状态，并且这些功能连接强度和认知能力具有显著的相关性。SMI老人在工作记忆任务上也能够保持像HC老人一样的表现。三组老人在多种任务条件下的激活没有显著差异，但在任务条件减去基线的情况下，大脑后部区域表现出显著的组间差异，且这一位置与静息条件下所出现区域较为相似。以静息态和任务态共同敏感的区域为种子点做全脑功能连接对比分析，SMI和HC老人在前额叶部分区域均显著强于aMCI患者。此外，这些功能连接的强度与1back的任务表现有显著的相关。 研究4. 疾病发展中结构、功能影像和认知能力的关联模型。在前述研究中我们发现了许多影像指标与记忆能力的显著相关，因此本研究将这些指标归纳到一起，通过结构方程模型SEM（Structural Equation Modeling）的方法，尝试建立一个随着疾病阶段的变化，大脑结构和功能影像对记忆能力的影响模型。结果发现在从HC到SMI再到aMCI的发展过程中，左侧海马回扣带束白质纤维损伤可引起楔前叶/中后扣带位置静息态下DMN内部功能活动变化，而该部位活动的变化既可以直接引起记忆能力的衰退，也可以通过改变其与大脑前部的功能连接对记忆功能带来影响。 综合以上研究，我们发现：（1）SMI的认知能力虽然还未达到MCI的诊断标准，但其多项认知能力可能较HC老人有显著的减退；（2）SMI老人灰质结构与HC老人较为相似，两者在内侧颞叶与aMCI患者具有显著差异；（3）SMI老人双侧丘脑前辐射区白质纤维束的完整性显著好于aMCI患者，但其左侧海马回扣带束白质纤维束的完整性较HC老人受到显著损伤，同时SMI老人部分节点的局部特征相对HC老人发生了显著损伤，但在白质网络全局属性上保存较为完好；（4）SMI静息态下的默认网络还能保持与HC老人相似的完整性，但其前-后脑间的功能连接受到显著的破坏；（5）SMI工作记忆任务态下的前-后脑功能连接还能够保持和HC老人较为相似的表现；（6）左侧海马回扣带束白质纤维束、静息态下默认网络内部功能活动和静息态下前-后脑间的功能连接形成的通路，可以显著地预测记忆功能的变化，这一路径模型可能特异于从HC到SMI再到aMCI的AD病理早期发展阶段。我们的研究多方位地详细描述了SMI所具有的大脑结构和功能网络特征，探索了AD早期阶段变化过程中多影像指标对认知能力产生影响的通路，对AD早期阶段敏感神经影像标记物的确立和疾病早期干预及评估具有重要的理论意义和临床价值。

With the growing of world economy and improving of health care, the aging of population has become a serious problem faced by many countries. As a large developing country with a huge number of population, China’s situation is not optimistic either. As the high incidence of many disease in the phase of old age, the large number of the elderly population also means a large number of medical and nursing costs. It is important for us to correctly understand the occurrence and development mechanism of geriatric diseases and develop effective disease prevention methods. Among the geriatric diseases, Alzheimer's disease (AD) is widely concerned for the great physical and mental harm it brings to the elderly people and their families. As a neurodegenerative disease, AD can greatly aggravate the progress of aging, leading to serious consequences of dementia. It affects about 15% of the population above 65 years old, and the number rises with age. With the development of effective intervention and treatment for other geriatric diseases, the proportion of AD in the cause of death of the elderly is increasing year by year. This is mainly because there has been no breakthrough of the current AD drug development. Therefore, researchers begin to focus on the early stage of pre-clinical AD, hoping to achieve effective improvement of AD condition through early detection and intervention. How to accurately screen and evaluate the population at risk of AD in the early stage becomes particularly important. Previous studies have shown that relevant pathological changes can be detected in the brain 30 years before the clinical diagnosis of AD. In the subsequent development process, the structure and function of the brain gradually changed pathologically, becoming a good neural marker in the early stage of AD. The Mild Cognitive Impairment (MCI) which is the stage before the clinical diagnosis of AD, has been extensively explored in many studies, and it has been confirmed that patients at this stage show typical brain structural and functional changes of AD. However, patients at this stage have already shown obvious cognitive impairment and are developing towards AD at a high speed, so it may still not be an ideal stage for the development and application of interventions and drugs. Subjective Memory Impairment (SMI) is a much earlier stage than MCI. In this stage, although the elderly can subjectively feel the impairment of memory ability, they can maintain normal memory level in actual measurement. SMI elderly with extremely high rates of MCI and AD conversion are considered to be the earliest stage of AD. If accurate screening and evaluation can be carried out for this target population and the influence pattern of AD pathology on brain structure and function in this early stage can be found in time, it will greatly facilitate the development of interventions and drugs. Some studies have described the neurological changes of SMI individuals in brain structure and function different from that of healthy elderly people, providing evidence for SMI as an early stage of AD. And there also have some studies established the relationship between cognitive ability and neuroimaging. It is believed that as more research results are reported, more and more researchers will pay attention to the exploration of this stage, and finally find a breakthrough for the intervention and treatment of AD. However, there are still many deficiencies in the current SMI research, which needs to be further explored. First of all, the results of SMI studies with the same modes are reported differently, and the conclusions are inconsistent. At the same time, the interpretation of the neuromechanism of SMI and the early pathological development of AD is one-sided, and the lack of correlation between research indicators is not conducive to establish the understanding of SMI. In this study, three groups of subjects including Health Control (HC), SMI elderly and aMCI patients were recruited from the community in Beijing, and they were given a complete set of neuropsychological tests and multimodal MRI scans. Multidimensionally describe the cognitive state of the elderly with SMI in various fields, as well as the neural characteristics of brain structure and function, and try to establish the correlation between multimodal neuroimaging and cognitive function changes. Finally, combined with previous studies, the comprehensive model of SMI is proposed. This study is divided into four parts. Study 1. Change characteristics of higher cognitive function and gray matter of SMI. In this study, 23 HC elderly, 22 SMI elderly and 24 aMCI patients were recruited from the Beijing Aging Brain Rejuvenation Initiative (BABRI) database, and their memory, attention, spatial processing, language and executive function were compared among three groups. We found that the cognitive result of aMCI had significant difference with HC. The thickness and volume of gray matter of subjects with SMI were both similar to HC, and they all in a better condition in the medial temporal region than subjects with aMCI. Only in the right insula and the triangle part of left inferior frontal lobe that SMI had gray matter atrophy compare to HC. Study 2. Change characteristics of white matter of SMI. We collected images of brain white matter of each subject in study 1, and compared the white matter integrity among three groups. We observed the correlation between these neuroimaging indicators and cognition, and machine learning was used to explore the predictive role of these indicators in group classification. In the comparison of white matter integrity, we found multiple white matter fiber bundles showed significant integrity differences among groups. The integrity of white matter in bilateral anterior thalamic radiation of SMI was similar to that of HC, but there was a significant difference of the white matter in the cingulum of hippocampus between SMI and HC. In the comparison of white matter network characteristics, we found that the global characteristics of SMI elderly people were relatively well preserved, but the local efficiency of some nodes, such as the middle and post cingulum, has shown similar changes to aMCI. The machine learning has found that the combination of mean diffusivity and axial diffusivity values of white matter can best distinguish the elderly with SMI and aMCI from HC, while the global characteristics of white matter network were only significant in the differentiation between HC and patients with aMCI. Study 3. Change characteristics of brain functional network of SMI. We collected images of the brain in resting state and working memory task state of each subject in study 1. The Default Mode Network (DMN) was estimated by Independent Component Analysis (ICA) and compared among groups. After that, the functional connectivity between the region with significant group difference and whole brain was compared among three groups. A similar comparison was also performed in the task state. The results showed that in the resting state, the DMN of SMI still maintained a relatively similar integrity to that of HC. However, the functional connectivity from this core node to the frontal lobe of SMI showed a significantly different state from HC, and the strength of these functional connectivity had a significant correlation with cognitive ability. There was no significant difference in activation among the three groups under various task conditions, but when the task conditions were subtracted from the baseline, there was a significant difference among groups in the posterior part of the brain, which was similar to the area in the resting condition. A analysis of whole-brain functional connections was conducted from the region sensitive to both resting state and task state. The functional connectivity of SMI and HC were significantly stronger in partial frontal lobe regions than aMCI. In addition, the strength of these functional connections were significantly correlated with the task performance of 1back. Study 4. Correlational models of structural and functional imaging with cognitive ability in disease development. In the studies above, we found that many neuroimaging indicators are significantly correlated with memory ability, so we tried to summarize these indicators together and build a model of the influence of brain structure and functional network on memory ability with the change of disease stage through Structural Equation Modeling (SEM). We found that as the disease stage development, the integrity damage of the white matter in the left hippocampus of cingulate can cause the functional activity impairment in precuneus/ middle posterior cingulum in resting state, and the change of the activities can directly cause the decline of memory ability, and also can impair the memory ability by changing the functional connectivity with the front of the brain. In summary, our research found several interesting results as follows: (1) Multiple cognitive abilities of SMI may have significantly decline compare with HC; (2) The gray matter of SMI elderly is similar to that of HC elderly, and there is significant difference between them and aMCI patients in the medial temporal lobe; (3) The white matter integrity in bilateral anterior thalamic radiation of SMI were significantly better than aMCI, but the white matter integrity in left hippocampus of cingulate was significantly worse than HC, the global characteristics of SMI elderly people were relatively well preserved, but the local efficiency of some nodes has shown similar changes to aMCI; (4) The DMN of SMI in resting state can still maintain the integrity similar to HC, but the functional connectivity between the anterior and posterior brain is significantly damaged; (5) The functional connectivity between the anterior and posterior brains of SMI in working memory task state can maintain similar performance as HC; (6) The changes in memory ability were significantly predicted by the white matter integrity in the left hippocampus of cingulate, the functional activities inside the DMN in resting state, and the functional connectivity between the anterior and posterior brains in resting state. Our research described the characteristics of brain structure and functional network of SMI in detail, and explored the pathway of the influence of multiple neuroimaging makers on cognitive ability in the change process of pre-clinical AD. It may have important theoretical significance and clinical value for the establishment of sensitive neuroimaging markers in the early stage of AD.