We systematically and comprehensively searched EMBASE, PubMed, and Web of Science. The search terms were as follows: (1) (“functional magnetic resonance imaging” [MeSH] OR (resting state)) AND (“mild cognitive impairment” [MeSH]) AND [(DAN) OR (attention network)] AND [(functional connectivity) OR (FC)]. (2) (“functional magnetic resonance imaging” [MeSH] OR (resting state)) AND (“mild cognitive impairment” [MeSH]) AND [(regional homogeneity) OR (ReHo)]. (3) (“functional magnetic resonance imaging” [MeSH] OR (resting state)) AND (“mild cognitive impairment” [MeSH]) AND [(amplitude of low frequency fluctuation) OR (ALFF)].

Our entry criteria were included (1) reported significant alterations of ALFF, ReHo, or FC in the DAN, (2) made comparisons between MCI and healthy control (HC), (3) reported information about the space in Talairach or Montreal Neurological Institute (MNI) coordinates, and (4) were published in English in peer-reviewed journals.

The patients with MCI met the following criteria: (a) attention to cognitive change, (b) impairment of one or more cognitive domains, (c) maintain functional independence in daily life, and (d) not demented (Albert et al., 2011).

Exclusion criteria were as follows: (1) patients were diagnosed with other disease such as Parkinson’s disease, (2) meta-analysis, review, and case report, (3) a lack of regular control group or comparison related coordinates.

A total of 691 publications were initially retrieved. After careful screening, a total of 38 publications were included in the final analysis (Figure 1). The included studies met the criteria of 38 MCI (17 ALFF, 14 ReHo, and 7 FC).

The two researchers in our group extracted data from the literature. First, we included patients with MCI with the specific criteria. Second, we read each study to determine whether it had healthy control group or comparison related coordinates. Then, whether it was a study of ALFF, ReHo, or FC in the DAN. Finally, we extracted coordinates of the DAN in the literature and worked with the method in the form of MNI coordinates. If two current researchers disagree on the adoption of the article, the third researcher will vote on the decision.

We divided MCI subjects into increased and decreased groups on three indexes (ALFF, ReHo, and FC): (1) increased ALFF (321 subjects, 22 foci, and 13 experiments); decreased ALFF (201 subjects, 26 foci, and 9 experiments); (2) increased ReHo (216 subjects, 13 foci, and 7 experiments); decreased ReHo (344 subjects, 29 foci, and 12 experiments); (3) increased FC (115 subjects, 12 foci, and 5 experiments); decreased FC (77 subjects, 7 foci, and 3 experiments).

This study used Java-based version of Ginger ALE 2.3.6¹ to assess the junction of the difference between MCI and HC group in terms of foci across the studies (Eickhoff et al., 2012). We format the collected foci which were performed in MNI coordinates into six text files. We imported them into the software by setting a threshold at p < 0.05. The ALE map was performed with a cluster-level family-wise error (FWE) correction at p < 0.05 and 1,000 threshold permutations. The FWE correction threshold is set to an ALE value that does not exceed this value for the specified portion of the distribution. FWE thresholds are conservative, so a 5% randomized study or p < 0.05 is recommended (Eickhoff et al., 2012). The maps were covered into MNI152 and visualized with the BrainNet Viewer² (Xia et al., 2013) in the Matlab R2013b. The results are shown in Figure 2. The meta-analysis was registered in advance on PROSPERO (registration number: CRD42021287958).³

The study characteristics and results are summarized in Table 1.

The brain areas of the DAN were summarized as follows: (1) use independent component analysis (ICA): inferior occipital gyrus, superior occipital gyrus, superior parietal lobule (SPL), inferior temporal gyrus (Zhang et al., 2020), right superior/middle frontal gyrus (S/MFG), right inferior parietal lobule (IPL), left precentral gyrus (preCG), left IPL/angular gyrus (Qian et al., 2015), dorsolateral prefrontal cortex (dlPFC) and SPL (Avelar-Pereira et al., 2017), IPS, middle temporal gyrus (MTG; Ding et al., 2019), inferior precentral sulcus (Baggio et al., 2015), and ventral IPS (Lei et al., 2014); (2) use seed ROI: MTG, SPL (Liang et al., 2014), dorsal anterior cingulate cortex (dACC; Esposito et al., 2018).

Compared to HCs, patients with MCI showed increased ALFF in the left MFG (BA 9), right SFG (BA 6), left superior temporal gyrus (STG) (BA 39), and right STG (BA 41). Patients with MCI showed decreased ALFF in the left IPL (BA 40) and right MTG (BA 22, 21). Patients with MCI showed increased ReHo in the left preCG (BA 4), left SPL (BA 7), and right preCG (BA 44). Patients with MCI showed decreased ReHo in the left IPL (BA 39, 40). Patients with MCI showed increased FC in the left, right preCG (BA 6) and left, right MFG (BA 9). Patients with MCI showed decreased FC in the left inferior frontal gyrus (IFG) (BA 9) and right IPL (BA 40) (Figure 2).

More details about clusters from the ALE analysis are summarized in Table 2.

The increased ALFF changes showed in the left MFG, right SFG, and STG. A study found that patients with MCI showed increased functional connectivity between the seed regions including bilateral IFG, bilateral MFG, and SFG (Shi et al., 2020). At the same time, the increased changes in the MFG were associated with reduced episodic memory in MCI (Zhao et al., 2019). This can indicate that brain regions that include MFG and SFG with increased ALFF might be closely related to memory loss in the progression of MCI. STG is the key to extract the meaningful language features from the speech input. A recent study also showed increased activity in the STG in MCI, which was consistent with our results (Agosta et al., 2012). To sum up, the increased ALFF in the MFG, SFG, and STG has a certain correlation with language impairment in MCI.

Both the increase and decrease of ALFF indicate the changes in the spontaneous activity of neurons. We described the relationship between the decreased changes of ALFF in specific brain regions and symptoms of patients with MCI below. Right MTG and left IPL demonstrated decreased ALFF changes. The subregions of MTG are related to human episodic memory (Berron et al., 2020). A study indicated that compared with the healthy control group, the stimulation-related activation of MTG in patients with MCI was lower, which is consistent with the indicator’s decline. It is suggested that the attention and cognitive control mechanism of patients with CI may be seriously damaged and became the basis of cognitive defects in this clinical group (Staffen et al., 2012). IPL has a close connection with episodic memory. So, reduced IPL activity indicated impaired memory functional system in patients with MCI which can be the critical early marker for prodromal stages (Zhao et al., 2014). Above all, early reduced ALFF in these brain regions in MCI could be associated with early clinical symptoms, such as impaired memory and attention.

Increased ReHo especially showed in the left, right PreCG, and left SPL. The preCG is located at the primary motor cortex. Its mechanism is to initiate the purposeful movement by integrating the information sent by the sensory motor cortex (Bahmani et al., 2019). An article proves that the impairment in cognitive domains such as working memory and behavioral flexibility can be associated with prefrontal cortex (Parnaudeau et al., 2018). Some studies have shown that blueberry-treated participants exhibited increased blood oxygen level-dependent (BOLD) activation in the left preCG during working memory load condition (Boespflug et al., 2018). SPL has been also related to properties that deal with visuospatial and spatial motion (Banaszkiewicz et al., 2021). An rs-fMRI study also showed increased ReHo in part parietal lobes in MCI, consistent with our findings (Min et al., 2019). Thus, we may hypothesize that the elevation of increased ReHo value in the SPL also plays an essential role in the transition process of AD.

Decreased ReHo especially showed in the left IPL. ReHo mainly explored the consistency of neuronal activity in the local brain area. At the same time, there is both decreasing change in the IPL in MCI. A study demonstrated that M50 sensory gating (SG) deficits in the IPL were related to the poorer performance in the immediate recall of logic memory (LM). Obviously, patients with MCI showed lower auditory short-term memory function with the deficit in the IPL in clinical manifestations (Cheng et al., 2020). It can be concluded that decreased ReHo can be related to the clinical manifestations of patients with MCI.

As to the FC, the increased changes in the MFG were associated with reduced episodic memory in MCI which were the same as increased ALFF (Zhao et al., 2019). The increased FC changes in the preCG were the same as the increased ReHo. A study showed that ReHo focused on the consistency of neuronal activity in local brain areas while FC focused on the connective relationship of two brain regions (Zuo et al., 2018). Thus, the increased FC meant that the connectivity between MFG, preCG, and other regions was higher, making up for deficits in MCI. The decreased FC of the IPL and IFG showed the dysfunction of DAN connectivity, which could be a biomarker to suggest the occurrence of MCI. The activation of IFG was essential for residual language function. At the same time, some task-related functional neuroimaging studies indicated MCI-related low activation in the left IFG (Winhuisen et al., 2005). In a word, with the development of the disease, the brain’s normal function was affected, and partial compensation of functional connectivity was necessary.

According to the specific imaging abnormal changes in DAN, transcranial magnetic stimulation (TMS) and other timely interventions can be carried out. During working memory load conditions, blueberry-treated participants exhibited increased BOLD activation in the left preCG and left IPL (Boespflug et al., 2018). At the same time, the effects of exercise and fitness seem to mainly affect brain structures sensitive to neurodegeneration, especially including frontal and parietal regions (Haeger et al., 2019). To carry out early intervention treatment for patients with MCI, we can carry out a practical course of blueberry taking, exercise, and fitness, which can draw from the above schemes.

Reviewing the results of ALE analysis of the FC in the DAN, interactions of the DAN with DMN, ECN, and frontoparietal network (FPN) had been observed in MCI group. The main functions of the four networks are different, but they all have overlapping regions and co-activation. The key area that shows anti-interaction with DMN included the IPL in the DAN. At the same time, the one that shows interactions with ECN was mainly in the MFG. Finally, the FPN showed a close connection with DAN, especially in the IPL and IFG.

The DMN is a task-negative network associated with deactivating arduous tasks during attention execution (Weiler et al., 2014). The DAN is also called the task-positive network because its central regions are commonly activated when attention and mind control are required. A study suggested that a reduced anticorrelated activity between DMN and DAN was a part of the normal aging process, and that MCI status was associated with more evident inter-network functional connectivity changes (Esposito et al., 2018). So, the decreased FC of the IPL in the DAN in MCI might be interpreted as a compensation mechanism by impairments of IPL. The interaction among networks may be associated with the structural location.

The ECN has an executive function, which includes problem-solving and working memory, and plays a key role in cognitive regulation and sensory information integration (Xu et al., 2020). The ECN, especially the dlPFC, is structurally connected with the frontal cortices. Therefore, the network is well located and can support a wide range of cognitive processes (Chand et al., 2018). A study highlighted the increased resting-state functional connectivity (rsFC) in the ECN and DAN as neuroimaging indications of disease progression in AD (Joshi et al., 2019). At the same time, the DAN in the MFG showing increased FC also supported this conclusion. Apparently, the increased FC was predictive of impaired episodic memory in MCI and may reflect a dysfunctional change within the episodic memory-related neural network (Zhang et al., 2016).

The FPN is involved in top-down attentional control and allocation of available neural resources to important cognitive processes and motor planning and motor execution (Hsu et al., 2019). The FPN comprises of multiple regions spanning the frontal and parietal cortices, which includes IPL, dlPFC, and preCG (Agosta et al., 2012; Liang et al., 2014). In this study, FPN showed that reduced connectivity in IPL and IFG may lead to the dysfunction of logic, regulating behavior, complex planning, and learning, which patients with MCI can exist (Zhao et al., 2019). This further supported the abnormality of FC as a biomarker for monitoring disease progression.

The interactions related to FC between DAN and other networks, such as DMN, ECN, and FPN, were mainly present in the IPL, IFG, MFG, and PreCG. A study showed for the first time that theta and alpha frequency repetitive transcranial magnetic stimulations (rTMSs) were able to modulate FC in DAN. With theta frequency band in left dlPFC, the memory performs better in a sustained attention task (Kazemi et al., 2020). Thus, changes in the interactions with these networks provided targets for early intervention treatment, which delayed the occurrence of MCI.

Although we have acquired valuable outcomes, some details still need improvement. First of all, the subjects’ age, sex, years of education, and other factors are heterogeneous. However, these factors have no practical impact on the results. What is more, given the limited number of studies included in the analysis, the findings from our meta-analysis should be confirmed in the future research. Last but not least, the selection of seed points of DAN is affected by the subjective idea of the operator, and the selection of different seed points will affect the results to a certain extent. The literature on the coordinates of such seed points can enrich our results in some ways.