面向fMRI数据的人脑功能划分

The human brain functional parcellation based on f MRI data

人脑是目前人类发现的最复杂和最具智能的功能组织系统之一.数以万计的神经元相互连接形成了复杂的脑结构,并通过相互间的作用表现出多样的智能活动.脑功能研究是脑科学研究中最重要的内容之一,尤其是功能磁共振成像(f MRI)技术的出现,为人脑功能的研究带来了新的发展机遇.近年来,开始利用基于f MRI数据的计算方法对人脑进行功能的划分,已有研究表明,功能划分所产生的人脑功能子区域比结构图谱中的脑区具有更高的功能一致性,因此更适合用于人脑的功能分析.进一步讲,把人脑功能划分产生的功能子区域用于人脑功能网络的构建能够获得更具可解释性的人脑功能网络,能为人脑疾病机理的研究和探索提供新的手段.本文以f MRI数据为基础,首先,介绍了f MRI数据采集、人脑功能划分及其基本流程;其次,给出了一种人脑功能划分方法的分类体系,并详细阐述了其中主要的人脑功能划分算法;再次,梳理了人脑功能划分中常用的相似性度量和评价指标;最后,总结了人脑功能划分的应用,并深入地分析了人脑功能划分中存在的挑战及未来的研究方向,以期对相关研究提供有益的参考.

The human brain is one of the most sophisticated and intelligent functional organization systems found by the human at present, and its complexity lies in：（i） Tens of thousands of neurons are interconnected by synapses to form the complex human brain structure, which is the material and physiological basis of the brain normal work.（ii） These neurons connected to each other exhibit complex and diverse brain activities by their interactions, such as visual information processing, logical reasoning and language expression and so on. More importantly after a long period of evolution, the human brain is highly intelligent, which makes that the human not only can adapt to the environment, but also can transform the world. This diversity and intelligence of the human brain function arouse great interest from researchers and have also been the most important contents in the study of the brain science. Last more than 20 years, nuclear magnetic imaging technologies have got rapid development on nuclear magnetic imaging technologies, such as electroencephalography, magneto encephalography and functional magnetic resonance imaging（f MRI）, which greatly promote the development of the human brain function research. Especially, fM RI has the advantages of no invasiveness, high resolution and the simplicity and repeatability of the operation, and has brought a new development opportunity for the research of the human brain function. Recently, based on fM RI data, some computational methods have been used to divide the human brain functionally where the functional parcellation methods of the human brain divide full brain regions or local brain regions into disjoint subregions according to the functional consistency measurements, thus depicts the segmentation of the human brain function and plays a fundamental role in the study of the brain function. Studies have shown that the functional consistency of the subregions obtained by the human functional parcellation is higher than that of brain regions acquired in the human structural atlases. Namely, the human brain functional parcellation is more suitable for the functional analysis of the human brain. Further, the more real brain functional networks and the more interpretability results could be obtained by applying divided functional subregions, which can provide a new approach for the research and exploration of the human brain disease mechanisms. In this review, the principle of fM RI, the acquisition process and characteristics of fM RI data are first introduced, which can help researchers understand in depth fM RI data and associated calculation methods. On this basis, the general flow of the human brain functional parcellation is elaborated, where the human brain functional parcellation is a key step. And then, the paper gives a classification system of the human brain functional parcellation methods and discusses the major human brain functional parcellation algorithms and their characteristics in detail, which enables researchers to have a more systematic and in-depth understanding on the research field. To make an objective comparison, the similarity measures and evaluations commonly used in the human brain functional parcellation algorithms are listed and their characteristics are also outlined to guide the further study of the human brain functions. Next, the paper summarizes the application of the human brain functional parcellation at the aspects of the human brain functional network construction, a neurological disease research and the state forecast of subjects. Finally, the challenging problems and future research directions are deeply analyzed from the perspective of both scientific and technical issues. This work is hopefully beneficial to the researchers engaged in the human brain function research.