意识障碍患者的脑成像研究不但为临床诊断和预后提供了重要指导，也为更好地理解人类意识提供了有效途径。在这个快速发展的科学领域中并存着两个主要研究思路，分别为考察意识“内部觉知”的神经基础和意识的“交流特性”。本论文兼顾这两个研究思路，一方面以意识障碍患者为研究对象，探究脑损伤引起的内部觉知神经活动异常；另一方面以正常人为研究对象，试图建立一个意识“交流特性”的fMRI检测范式。 从意识的“内部觉知”出发，研究一选取了11例意识障碍患者与12位正常被试,考察自我参照加工任务引起的皮层反应及静息态脑功能。研究发现自我参照任务可以引起意识障碍患者大脑中线结构如前扣带回膝部（perigenual cingulate cortex，PACC）和后扣带回（posterior cingulate cortex，PCC）等脑区的活动,但激活强度显著低于正常对照组。另外,患者在PACC的活动强度与意识水平呈正相关。重要的是，对于静息态的网络特性-功能连接（functional connectivity）与局部特性-低频振幅（amplitude of low frequency fluctuations），意识障碍患者在PACC和PCC两个脑区均表现出异常。 从意识的“交流特性”出发，研究二试图从大脑的BOLD-fMRI信号中“读取”人类对问题的真实回答。结合任务-线索范式和多变量模式分析技术（MVPA），本研究发现大脑皮层的左侧背外侧前额叶（额中回的BA9区）编码真实回答（是/否）。通过进一步的跨条件（cross-condition）以及条件内（within-condition）解码分析，揭示了此区对真实回答的编码并不受主观意图(诚实与否)的影响。另外，此区的功能定位具有较高的被试间一致性，而且对刺激模态（视觉/听觉）与问题类型（常识问题/自传体问题）没有依赖性。 总之，研究一首次发现意识障碍患者自我参照加工异常伴随静息态脑功能的异常，两者聚焦于大脑中线的默认网络结构中。这不但具有潜在的临床价值，也为更好地理解自我、静息态以及意识之间的关系提供了新视角。研究二首次表明大脑左侧背外侧前额叶采用不同的空间活动模式对真实回答进行编码，此研究范式不但为测谎提供了新思路，也为临床实践提供了可能的途径：无需借助患者外在的行为反应，直接考察脑活动所体现的交流能力进而检测患者的意识状态。
Results of neuroimaging studies on disorders of consciousness (DOC) not only have practical significance for clinical diagnosis and prognosis, but also have theoretical implications for our understanding of the neural basis of consciousness. There are two major lines in this rapidly progressing field, i.e. to examine the neural aspect of internal awareness and to characterize the communication features of consciousness. Along the above two lines, this dissertation presents two fMRI experiments, which respectively examined the neural abnormalities of internal awareness in patients with DOC and developed a new paradigm for detecting the communication features of consciousness in healthy subjects. Focusing on the neural aspect of internal awareness, Study 1 investigated both task-related neural activities during self-referential processing and various measures of resting state activity in 11 patients with DOC and 12 healthy subjects. We found that the self-referential task elicited task-specific activation in anterior and posterior midline regions, including the perigenual and posterior cingulate cortex (PACC and PCC), although the intensity signal change was significantly lower than that in healthy subjects. Moreover, the activation in the PACC predicted the level of consciousness in DOC patients. Importantly, the same midline regions, PACC and PCC, showed severe abnormalities in network features measured by functional connectivity and local features measured by amplitude of low frequency fluctuations of resting state activity in DOC patients. Aiming to establish a reliable fMRI paradigm for detecting the communication feature of consciousness, Study 2 examined whether truthful answers can be decoded from the spatial patterns of BOLD-fMRI signals. By combined use of a task-cueing paradigm and multivariate pattern searchlight analysis, we identified a cortical region in dorsolateral prefrontal cortex (DLPFC, the BA9 portion of the left middle frontal gyrus) that carried accurate information for predicting the truthful “Yes” and “No” answers. Cross-condition and within-condition classification analyses further revealed that the encoding of the truthful answers in this region was not modulated by the intention to answer honestly or dishonestly. In addition, the location of this region was consistent across participants, stimuli modalities (visual/auditory) and question types (simple-fact/autobiographical questions). Taken together, Study 1 shows for the first time neural abnormalities both during self-referential processing and resting state in the midline regions in DOC patients, which may shed important light on future clinical utility and our understanding of the relationships among self, resting state, and consciousness. Study 2 for the first time demonstrates that truthful answers are encoded in spatial patterns of neural activity in the prefrontal cortex. The paradigm developed in this study not only offers an alternative approach to fMRI lie-detection, but also permits the identification of communication related brain activity (thus of consciousness) in behaviorally non-communicative, brain-injured patients, without the need of measuring overt motor responses.