【作者】 翁晓光；

【导师】 王惠南；

【作者基本信息】 南京航空航天大学 ， 精密仪器及机械， 2010， 博士

【摘要】 人类大脑具有最复杂的体系框架,是一个高度复杂的网络系统,应用脑成像技术分析脑的功能与结构特性是认知科学中的重要研究内容。功能磁共振成像技术(functional magnetic resonance imaging,fMRI)基于血氧水平依赖原理(blood-oxygen level dependent,BOLD),具有无辐射性、无创性、高时空分辨率等多重优势,已成为脑科学研究的重要工具,受到临床、心理、神经和认知等领域的极大关注。静息态功能磁共振技术不需要被试做出反应,也不需要复杂精细的实验设计,容易被患者接受,非常适合临床上的应用和研究。大量研究表明,在静息状态下人脑存在着一种低频振荡现象。尽管目前还不清楚它的具体生理意义,但是它与某些疾病的生理病理以及静息状态人脑的脑功能有着密切的关系。本文围绕fMRI在脑科学研究中的应用,以静息状态下的fMRI数据研究方法为突破口,对独立成分分析(independent component analysis,ICA)、局域一致性(regional homogeneity,ReHo)、低频振荡振幅算法(amplitude of low frequency fluctuations,ALFF)等静息状态fMRI脑图像分析方法,进行了系统性的深入研究和探讨。以此为基础,结合认知科学和临床研究中的前沿课题,开展了对弱视患者的视觉皮层功能异常研究,从而加深对弱视神经机制的理解和认识。全文的主要内容如下:1、提出了具有三阶收敛的快速ICA算法。ICA方法可以从观测混合信号中分离出相互之间统计独立的源信号。本文引入负熵作为独立性的估计准则,给出ICA的一个优化模型。在此基础上,采用一种改进的牛顿迭代型的独立成分分析算法来提取fMRI信号中的各个独立成分;为加快收敛速度,对标准的牛顿迭代进一步修正,使算法具有三阶收敛。将该算法应用于fMRI大型数据的盲分离,并与目前广泛使用的另外两种算法比较,本文算法具有更快的收敛速度。2、提出了基于ReHo选取种子区域并进行功能连接的新方法。传统的功能连接分析中,种子区域的选择常常是基于激活图或是基于解剖知识,而没有利用到静息状态的功能数据。本文利用局域一致性分析静息状态下的fMRI数据,将感兴趣区域中ReHo值最大的点所在的小区域作为种子进行功能连接。将脑的功能连接研究和局域一致性分析相结合,为研究人脑静息状态功能连接网络提供了新的思路。应用该方法,发现并验证了正常被试静息视觉网络的存在。3、提出了利用静息状态fMRI数据,从功能连接的角度对屈光参差性弱视患者静息视觉网络进行研究的新思路。目前fMRI对弱视的研究多是基于有视觉刺激的情况下,本文采用ICA算法分离静息状态fMRI数据,针对ICA算法本身无法自动识别成分顺序问题,引入Goodness-of-fit方法提取静息状态下弱视患者和正常被试的静息视觉网络,将结果进行组内和组间分析。结果表明在屈光参差性弱视患者的静息视觉网络中,纹状皮层和纹外皮层均发生了明显的功能损害,其功能连接程度显著低于正常组,并且纹外皮层比纹状皮层损害更加严重,为深入研究弱视初、高级视觉皮层功能损害的神经机制提供了新的思路。4、提出了采用ALFF方法研究弱视患者皮层功能损害的新方法。目前绝大部分静息fMRI研究关注的是不同脑区低频振荡的同步性即功能连接,功能连接的异常说明其功能整和出现变化,但是无法说明是哪一个脑区有不同的神经元活动,而低频振荡幅度方法可以提示局部神经元的自发活动,说明神经元活动的能量。本文首先采用ALFF方法检测出人脑默认模式网络,所得结果与已知研究结果具有很大的一致性,表明ALFF是一个有效的研究大脑静息状态自发性低频振荡的方法。然后使用该方法研究弱视患者闭眼、健眼刺激、患眼刺激三种静息状态下大脑的激活情况。研究结果表明在三种静息状态下,大脑出现了不同的激活情况,健眼刺激时受到的外界视觉刺激最多,因此在视觉皮层纹外区,其ALFF值最大;而在患眼刺激时,患者需要更多的选择性注意,因此在与注意有关的额上回区域,其ALFF值最大,显示该技术在大脑皮层功能定位上具有良好的应用价值,从另一个角度为研究弱视神经机制提供了新的方法。更多还原

【Abstract】 Human brain is a huge network system with the most complex structure. In cognitive neuroscience, the application of brain imaging techniques plays a very important role to understand the characteristics of brain. The functional magnetic resonance imaging (fMRI) based on the theory of blood-oxygen level dependent (BOLD) is a non-invasive and non-radiative technology, and it has high temporal and spatial resolution. The technology has become an important tool to do scientific researches on brain and has been concerned by many science branches, such as neuroscience, cognition, psychology and clinic et al. Resting-state functional MRI could be an advantageous choice for applications and clinical researches because it can be more easily accepted by patients without subjects’response and intricate experiment design.A lot of fMRI studies have demonstrated that there exist spontaneous low frequency fluctuations (LFFs) in the resting brain. Although the neurophysiological mechanisms of LFFs remain unclear, it has been suggested that LFFs are physiologically meaningful for keeping normal brain function and are associated with pathophysiology of the diseases. In this work, focused on the application of fMRI in brain network, aimed at clinical applications in diseases and understandings of brain complexity, systematic researches and explorations are conduced to analysis methods of resting-state fMRI, including independent component analysis (ICA), regional homogeneity (ReHo), amplitude of low frequency fluctuations(ALFF). On the basis, dysfunction study of the visual cortex in patients with amblyopia is studied so as to enhance the understanding of the pathological mechanism of amblyopia. The main contents in this dissertation can be stated as follows:1、A fast ICA algorithm with cubic convergence is prorosed. ICA is a new statistical signal processing technique. The goal of ICA is to recover independent sources given only sensor observations that are unknown linear mixtures of the unobserved independent source signals. In this paper, negentropy is introduced as measure of independence, and an optimization model for ICA is presented. A new Newton iterative algorithm based on the model is proposed. In order to accelerate the convergence, an improvement on Newton method is made which makes the convergence of the new algorithm cubic. Applying the algorithm and two other algorithms to invivo fMRI data, the results show that the new algorithm separates independent components stably. It has faster convergence speed and less computation than the other two algorithms.2、A novel approach of selecting“seed”regions based on ReHo in functional connectivity is proposed. In the previous methods,“seed”is selected utilizing prior anatomical information (i.e. knowledge-based) or activation maps (i.e. activation-based) without taking the natures of resting-state data into account. In this dissertation, ReHo is used to select the region with maximum ReHo value as“seed”in functional connectivity studies using“seed”correlation analysis. This provides a novel way to study the resting-state functional networks. By using the new method combing ReHo and functional connectivity, the resting-state visual network of human brains is identified in normal subjects.3、A new method to study the resting-state visual network of the anisometropic amblyopia patients is advanced from the perspective of functional connectivity only using the resting-state fMRI data. ICA is used to separate the fMRI data. Considering that ICA algorithm can hardly choose the optimal one of the separated components, goodness-of-fit method is introduced to extract the resting-state visual networks of the anisometropic amblyopia patients and normal controls. Intra-group analysis and inter-group analysis are performed. The results show that there are remarkable deficits on different levels of visual cortex in anisometropic amblyopia predominantly in the extrastriate cortex rather than the striate cortex. The resting-state fMRI provides a new way to investigate visual cortex deficit in amblyopia.4、A new method to study the functional impairment in amblyopia using ALFF is proposed. Most studies of resting-state fMRI have investigated LFFs from the aspect of temporal synchronization, i.e. functional connectivity between brain regions. Although a result of abnormal functional connectivity between two remote areas is comprehensive and integrative, one could not draw any conclusion about which area is abnormal from such an examination. The ALFF signal can reflect cerebral spontaneous neuronal activity, indicating the energy of neuronal activity. The dissertation explores default mode network in the resting brain using ALFF. Our result shows a large overlap with those known suggesting ALFF is an effective method to study spontaneous LFFs of the resting-state. In addition, the brain activation of amblyopia patients is studied for the three resting states: eye-closed, healthy eye-stimulated, suffering eye-stimulated. The results show different brain activations for the three resting states: healthy eye-stimulated has the most increased ALFF in the extrastriate areas of the visual cortex due to the most visual input; suffering eye-stimulated has the most increased ALFF in the superior frontal gyrus which is related to selective attention, due to poor sight demanding more efforts to attention. ALFF has potential in the clinical applications of localization of the brain cortex function and provides a different way to investigate the pathophysiological mechanisms in amblyopia.更多还原