Alzheimer's disease (AD) is a progressive neurodegenerative disorder, which is characterized by global cognitive decline, including progressive loss of memory, reasoning and language. There has been much anatomical and functional evidence pointing toward AD as a kind of disconnection syndrome 
. Some studies have shown disruption of a distributed network, especially the linkages between the hippocampus, the prefrontal cortex and other brain areas 
. Moreover, a number of studies have reported the compensatory mechanism in AD, as indicated by the increased functional connectivity within the prefrontal lobe or between prefrontal regions and other brain regions 
. Mild cognitive impairment (MCI) refers to a transitional stage between normal aging and early AD 
. MCI has a high probability of evolving toward AD at a rate of 10–15% per year 
. It is thus of theoretical and practical significance to examine if MCI show similar effects of functional disconnection and compensation with AD.
The study of resting state brain function is especially applicable to the study of patients because of the practical advantages of resting-state fMRI in terms of ease of clinical application. Functional connectivity during rest, measured by correlation in low frequency fluctuations (LFF) (<0.1 Hz) of the blood oxygen level-dependent (BOLD) signal, has been used to characterize intrinsic architecture of large-scale brain systems, including motor 
, visual 
, auditory 
, language 
, attention 
and default mode network (DMN) 
Several resting state fMRI studies have demonstrated the abnormalities of functional integrity in MCI patients 
. For example, based on posterior cingulate cortex (PCC) functional connectivity analysis, Bai et al. 
indicated regions of decreased functional connectivity were identified in aMCI patients, most notably in temporal cortex (BA 21, 37, 39), whilst significantly increased functional connectivity was mainly in frontal cortex (BA 9, 11, 47), as compared to healthy controls. Using independent component analysis (ICA) approach, Qi et al. 
compared the default mode network (DMN) activity between MCI patients and healthy elderly, and showed that MCI patients exhibited decreased functional activity in the DMN regions, including the bilateral precuneus/PCC, right inferior parietal lobule, and left fusiform gyrus, while increased activity in the left prefrontal cortex (BA 8) and other parietal, temporal regions, as compared to the healthy elderly. Most of these studies showed that functional disconnection and compensation coexisted in MCI patients, however, they mainly focused on the DMN, i.e., the task-negative network.
The dorsolateral prefrontal cortex (DLPFC), different from DMN regions (e.g., PCC/precuneus, medial prefrontal cortex, etc.), is an important region of task-positive. Numerious evidences have demonstrated that DLPFC plays crucial roles in various attention-demanded cognitive tasks such as working memory and executive function 
, episodic memory 
, attention 
, decision making 
, reasoning 
and problem solving 
, and so on. Functional and structural studies also indicated that DLPFC is connected to a variety of brain areas, including the thalamus, basal ganglia, the orbitofrontal cortex and primary and secondary association areas of neocortex, including posterior temporal, parietal, and occipital areas 
. Although abnormal activities of the DLPFC as well as functional disconnection involving DLPFC have been observed in MCI and AD 
, the precise and detailed connection patterns of the DLPFC in MCI are still unclear. Considering its local dysfunction and roles in various neural circuits relevant to the physiological mechanisms of cognitive impairment, the functional connectivity of the DLPFC in MCI should be especially emphasized. However, to our knowledge so far, no study has examined the DLPFC functional connectivity (DLPFC-FC) in MCI patients.
Using resting state functional connectivity and graph theory analysis, Dosenbach et al. 
proposed that there are two commanders in the brain: fronto-parietal and cingulo-opercular circuits. The fronto-parietal circuit mainly comprises of premotor cortex (BA 6), DLPFC (BA 46, 47), IPL (BA 40) and ACC (BA 24/32). Two previous resting state fMRI studies observed the similar fronto-parietal network by using probabilistic independent component analysis (PICA) 
or seed-based correlation analysis 
. Moreover, in an automatic meta-analysis of 825 neuroimaging articles representing 3402 experiments, Toro et al. 
obtained a similar co-activation map of fronto-parietal network. The fronto-parietal network in these studies was called as “task-positive” network, which was activated during performance of attention-demanding cognitive tasks. The cingulo-opercular circuit is composed of the medial frontal cortex (BA 6, 8), insula (BA 13), anterior prefrontal cortex (aPFC; BA 11/10), and thalamus. This pathway was recruited in many cognitive tasks, e.g., cognitive control 
, learning 
and planning 
The first goal of the present study is then to investigate the resting-state functional connectivity patterns of the DLPFC in MCI patients. It was hypothesized that the effects of functional disconnection of the two circuits and compensation within lobes coexisted in AD can be detected in its early stage, i.e., MCI, by using DLPFC-FC analysis.
The second goal of this study is to examine the asymmetry of DLPFC-FC in MCI patients. The different role of the bilateral DLPFC is well-known in working memory 
, episodic memory 
for healthy adults. We thus expect to observe the functional asymmetry of DLPFC-FC in healthy controls, and are interested to test whether the functional asymmetry of DLPFC-FC in MCI patients is altered as compared to healthy controls.