Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
AJNR Am J Neuroradiol. Author manuscript; available in PMC 2017 December 30.
Published before final editing as:
PMCID: PMC5201447

Disrupted brain connectivity pattern in patients with type 2 diabetes


Background and Purpose

Type 2 diabetes is associated with an increased risk for dementia. This study investigated the global connectivity patterns in the brains of patients with type 2 diabetes using a functional MRI technique.

Materials and Methods

Forty patients and 43 age-, sex- and education-matched healthy controls underwent resting-state functional imaging in a 3T MR unit. Degree centrality, a commonly employed measurement of global connectivity, was computed for a full-brain exploration of the regions influenced by type 2 diabetes. We then examined the functional connectivity of each region using the seed-based approach. Finally, voxel-wise correlation analyses were performed to explore the relationship between the connectivity changes, cognitive performance and diabetes-related variables.


Patients exhibited decreased degree centrality in the left lingual gyrus, and increased centrality in the right insula and dorsal anterior cingulate cortex (corrected P <0.05). The occipital network anchored in the lingual gyrus showed extensively reduced connectivity, while the network connectivity of insula and cingulate cortex (mostly included in the salience network) was significantly elevated (corrected P <0.05). Correlational analyses revealed that in the diabetic group, impaired visual memory and executive-function performance were correlated with occipital hypoconnectivity, while higher fasting plasma glucose levels and better executive functioning was related to anterior cingulate cortex hyperconnectivity (all corrected P values <0.05). Similar effects were not detected in the controls.


This preliminary study shows that network connectivity is altered in patients with type 2 diabetes, which may provide critical insights into the neural substrate of diabetes-related cognitive decline.

1. Introduction

Type 2 diabetes mellitus (T2DM) has been suggested to be associated with cognitive dysfunction and a higher risk of dementia, especially in elderly subjects1, 2. However, the neural substrate of T2DM-related cognitive impairment remains elusive. Because brain pathologies may precede clinically measurable cognitive deficits3, identifying vulnerable brain regions might be advantageous to track the early effects of T2DM on cognitive functioning.

The human brain consists of spatially distributed but functionally interacting regions that form an efficient functional connectome to support normal cognitive functioning4. Recent studies on patients with Alzheimer’s disease (AD) suggest that altered functional connectivity (FC) between regions may represent early deleterious outcomes that occur prior to structural changes and clinical manifestations5. However, only a limited number of studies have explored FC changes in relation to T2DM so far: one study reported a decreased FC of the hippocampus, while another study observed alterations in thalamocortical connections6, 7. Other studies on this topic focused on either the changes within the default mode network810 or in the attentional network11. Notably, these studies relied on seed-based approaches or an independent component analysis (ICA), focusing their examinations within specific brain networks. Based on pathological evidence, neural degeneration in the brains of those with diabetes is diffusely distributed12. Therefore, a voxel-wise whole-brain FC analysis may yield a more comprehensive understanding of the functional alterations related to the disease.

Degree centrality (DC) is a commonly employed graph theory-based measurement of global connectivity. It computes each voxel’s correlation with all other voxels in the entire brain, resulting in a connectivity map that represents the sum (density) of connections at each voxel13. Recent studies from our group have demonstrated DC to be highly associated with regional cerebral blood flow and metabolism, which establishes the physiological basis of this topological measurement14. So far, DC analyses have been applied to various mental disorders, such as Alzheimer’s disease, schizophrenia and autism1517. In this study, we used DC measurements to perform a full-brain exploration of brain regions with altered connectivity density. The FC pattern of each region was subsequently examined using a seed-based approach. Finally, correlations between the connectivity in each brain circuit, cognitive performance and T2DM-relevant variables were explored to better understand the link between network dysfunction and cognitive deficits.

2. Methods and materials

2.1 Participants

The current study was approved by the ethics committee of the institutional review board. Written informed consent was obtained from all subjects prior to their participation. Patients with T2DM and age-, sex-, and education-matched healthy controls (HCs) were recruited via advertisement in the local community. All subjects were between 50 and 75 years old, with a minimum education of 6 years. The exclusion criteria were as follows: a history of alcohol or substance abuse, Mini-Mental State Examination (MMSE) score < 2418, Hamilton Depression Rating Scale (HAM-D) score ≥ 719, history of brain lesion such as tumor or stroke, unrelated psychiatric or neurological disorder, or MRI contraindications.

The diagnosis of T2DM was made based on medical histories, medication used, or fasting plasma glucose (FPG) levels (≥ 7 mmol/L)20. None of the patients reported any history of hypoglycemic episodes or had been diagnosed with clinically detectable complications such as retinopathy, nephropathy, and peripheral neuropathy. HCs underwent an oral glucose tolerance test (OGTT) (75 g dextrose monohydrate in 250 ml water). HCs with fasting blood glucose levels ≥ 7.0 mmol/L or post-OGTT glucose levels ≥ 7.8 mmol/L were excluded.

Information on medical histories and medication use were collected, and weight, height and waist circumferences were carefully recorded for all subjects. Hypertension was defined as previously described21. Blood samples were collected after an 8-h fast to assess FPG, glycosylated hemoglobin (HbA1c), fasting insulin, and cholesterol levels. The homeostasis model assessment of insulin resistance (HOMA-IR) was used to assess the degree of insulin resistance for all HCs and non-insulin treated patients22.

The final study sample included 40 T2DM patients (mean age, 60.5 years ± 6.9 [standard deviation], 21 females) and 43 HCs (57.6 years ±6.6, 30 females) (Table 1), who were 50 to 75 years of age. Among the patients, disease duration ranged from 2 to 17 years (mean duration, 8.9 years ±5.0). Eight of the T2DM patients received insulin treatment, while the others were treated with either oral hypoglycemic agents (n = 22) or dietary restriction (n = 10).

Table 1
Demographics and clinical characteristics for T2DM and control groups

2.2. Neuropsychological test

All participants were subjected to a detailed battery of neuropsychological tests that covered multiple cognitive domains. Episodic memory regarding verbal and visual information was assessed via the Auditory Verbal Learning Test (AVLT) and the Rey-Osterrieth Complex Figure Test (CFT)-delayed recall trial, respectively. Working memory was measured via the forward and backward trials of the digit span test (DST). Attention was evaluated via the Trail Making Test part A (TMT-A), and executive functioning was assessed via Trail Making Test part B (TMT-B). Spatial processing ability was assessed via the Clock Drawing Test (CDT) and the CFT-copy trial. The total time required to complete all tests was approximately 60 min.

2.3 MRI data acquisition

The resting-state fMRI data were acquired in a Siemens 3T Trio scanner (Erlangen, Germany). Foam padding and earplugs were used to reduce head motion and scanner noise. Subjects were instructed to keep their eyes closed, remain awake and avoid specific thoughts during the scanning, which was later confirmed using a questionnaire. A 6-min whole-brain fMRI dataset based on blood oxygen level-dependent (BOLD) signals was acquired for each subject using a gradient-echo EPI sequence (volume = 240, slice number = 36, TR/TE = 2000/25 ms, slice thickness = 4 mm, flip angle = 90°, field of view = 240 mm, acquisition matrix = 64 × 64). High-resolution (1 mm3) T1-weighted images were acquired using a whole-brain 3D MPRAGE sequence (slice number = 176, TR/TE = 1900/2.48 ms, slice thickness = 1.0 mm, flip angle = 9°, field of view = 250 mm, acquisition matrix = 256 × 256). FLAIR images were also obtained (TR/TE = 8500/94 ms, slice = 20, slice thickness = 5 mm, with each voxel size of 1.3 × 0.9 ×5 mm3).

To assess small vessel disease (SVD), we evaluated the white matter lesions (WMLs) and lacunar infarcts on FLAIR images23. The brain was divided into five regions in each hemisphere, and the WML score was determined separately for each region on a 4-point scale (from 0 to 3), resulting in final scores ranging from 0–3023. Participants with a score of 3 in any region were considered to have severe SVD and were thus excluded. The final results indicated that none of the participants met the WML exclusion criteria.

2.4 Data preprocessing

Functional MRI data were preprocessed using DPARSF software ( and SPM8 ( After discarding the first 10 functional volumes, slice-timing correction and realignment were subsequently performed (subjects with head motions of > 2.0 mm of translation or >2.0° of rotation were excluded). Structural and functional images were then coregistered and normalized to the standard Montreal Neurological Institute (MNI) space. Afterwards, nuisance covariates, including 6 motion parameters, white matter and cerebrospinal fluid signals were regressed out, followed by detrending and band-pass filtering (0.01–0.1 Hz) procedures. To minimize motion-induced artificial correlations, functional images were cleaned by applying a scrubbing procedure that removed images (frames) with > 0.5 mm frame-wise displacement (FD) along with the two frames immediately before and after it24. The number of scrubbed imaging volumes for each individual is summarized in Supplementary Figure 1. The mean FD of each subject was included as a motion covariate in the statistical analyses.

2.5 DC analyses

DC maps can be represented as either weighted or binarized graphs, with the former focusing on the sum of weights from edges connecting to a node and the latter on the number of the connected edges13. Here, both graphs were computed to obtain more comprehensive information. The DC maps were generated using Pearson correlations in MATLAB (The MathWorks Inc., Natick, Massachusetts). Specifically, the time course of each voxel within a grey matter (GM) mask (obtained from the overlap of all subjects’ segmented GM) was extracted and correlated with that of every other voxel within the mask to generate a correlation matrix. After thresholding each correlation at R > 0.25, DC was computed as either the sum of connections (binarized) or the sum of the weights of connections (weighted) for each voxel. The resulting voxel-wise DC map was subsequently converted into a z-score map by subtracting the global mean DC and dividing by the standard deviation of the whole-brain DC13. Finally, the DC maps were smoothed with a Gaussian kernel of 4 mm to obtain the individual DC maps for each subject.

Further analyses for DC maps were performed using the AFNI software package ( The smoothed individual DC maps of all subjects in each group were analyzed using one-sample t-test to identify the spatial distribution of DC values in each group. The statistical threshold was set at P < 0.05, with a family-wise error (FWE) correction. Independent two-sample t-tests were also performed based on the individual DC maps to examine the differences in DC between groups. Age, sex, years of education, and head motion (mean FD) were included as covariates. To further exclude the confounding effects of SVD, we also controlled for the WML score and the existence of hypertension during the comparison. The Puncorrected value was set at 0.01, with a cluster size of 30 voxels (determined via Monte Carlo simulation) corresponding to a Pcorrected of 0.05.

2.6 Seed-based FC analyses

The identified brain regions with altered DC were chosen as the seeds to examine their FC changes. The time series of each region was averaged and correlated with that of every other voxel within the GM mask. Correlation coefficients between the seed and every other voxel were then converted using Fisher’s r to z transform, yielding variates that were approximately normally distributed. One-sample t-tests were performed on the individual z-transformed FC maps in each group to identify the spatial distribution of each brain circuit (P < 0.05 with FWE correction). Voxel-wise group comparisons on each z-transformed FC map were then carried out, which were confined to a mask obtained by combining the results of the within-group analyses in each group. Statistical significance was determined based on Puncorrected < 0.01 and Pcorrected < 0.05. The same covariates used during the DC analyses were included.

2.7 Statistical analyses

2.7.1 Clinical data

Statistical analyses were performed using SPSS software (ver. 18.0; SPSS, Inc., Chicago, IL, USA). Normal distributions were tested for using the Kolmogorov-Smirnov test. Group comparisons of clinical parameters were explored using independent two-sample t-tests for normally distributed variables, nonparametric Mann-Whitney U test for asymmetrically distributed variables, and χ2-tests for categorical variables. A P value of 0.05 or less was considered statistically significant.

2.7.2 Relationship between FC and clinical variables

The correlations between the FC of each seed-relevant network and the clinical variables, including neurocognitive performance (those tasks performed significantly worse by patients than by HCs) and T2DM-related parameters (HbA1c, FPG and HOMA-IR), were explored in a whole-brain linear regression model using the AFNI’s 3dLME command. The FC maps and clinical variables in both groups were entered into the models along with age, sex, education, head motion and SVD parameters as covariates of no interest. To identify the regions where the FC was differentially correlated with clinical variables in patients vs. HCs, we focused on the group×variable statistical interaction maps (Pcorrected < 0.05, Pvoxel-wise < 0.01). Brain regions with significant interaction effects were considered important in contributing to T2DM-related cognitive impairment.

3. Results

3.1 Demographic and clinical parameters

All clinical parameters are summarized in Table 1. The two groups were matched in terms of age, sex, education and their micro-head motion index. As expected, patients exhibited significantly higher FPG, HbA1c and HOMA-IR values than the HCs. No differences were observed in blood pressure, lipid levels or degree of SVD. Patients performed significantly worse on CFT-delay and TMT-B than HCs (Table 2), which mainly involve the visual memory and executive functioning domains. Among the diabetes-related variables, disease duration was significantly correlated with the scores in two tasks after controlling for age (CFT-copy, R = −0.45, P < 0.01; CFT-delay, R = −0.40, P = 0.02), while the HbA1c level was associated with the CFT-copy score (R = −0.35, P = 0.02).

Table 2
Cognitive test results for T2DM and control groups

3.2 DC results

Due to the highly consistent results of the weighted and binarized measurements, the findings present are primarily based on the weighted maps.

The spatial distribution of the weighted DC maps is shown in Fig. 1. In both groups, the DC in the posterior cingulate cortex, cuneus, visual cortex, medial prefrontal cortex (MPFC) and insula were significantly higher than the global mean value. The binarized map shows a similar pattern (Supplementary Fig. 2).

Figure 1
Spatial distribution of weighted DC maps in T2DM patients and HCs (P < 0.05, FWE corrected)

T2DM patients exhibited decreased DC in the left lingual gyrus and increased DC in the right anterior insula (rAI) and dorsal anterior cingulate cortex (dACC) (Fig. 2). The results shown inthe binarized DC map are highly consistent with the results in the weighted map (Supplementary Fig. 3). Detailed information for the identified brain regions are summarized in Table 3 and Supplementary Table 1.

Figure 2
Group differences of weighted DC maps between T2DM patients and HCs (P < 0.05, AlphaSim corrected)
Table 3
Brain regions with significant differences in weighted DC maps between T2DM patients and HCs

3.3. Seed-based FC analyses

The FC pattern of each region (i.e., dACC, rAI and left lingual gyrus) is shown in Fig. 3. Specifically, the dACC was connected with the cingulate cortex, anterior insula and sensorimotor cortex, while the rAI was connected with the entire insula, dACC and the adjacent frontal, temporal, and sensorimotor cortices. Meanwhile, the lingual gyrus was mainly connected with the visual cortex, and the superior/middle temporal and sensorimotor cortices.

Figure 3
Spatial pattern of the network anchored in the regions with altered DC (P < 0.05, FWE corrected)

In the diabetic subjects, the dACC showed stronger connectivity with the bilateral ventral ACC/MPFC (Fig. 4, first row), while the rAI had increased interactions with the right posterior insula and left superior temporal gyrus (Fig. 4, second row). In contrast, diffusely decreased connectivity was observed in the lingual gyrus-related visual network, especially in the higher-order visual cortex and the sensory areas (Fig. 4, third row).

Figure 4
Group differences of network connectivity based on the seed regions identified in the DC comparison (P < 0.05, AlphaSim corrected)

Voxel-wise correlation analyses identified significant effects of the clinical variables on several brain regions in the T2DM group (Fig. 4). The occipital connectivity was positively correlated with the CFT-delay score (Fig. 4A, R = 0.48, P = 0.002, group×performance interaction, P = 0.001) and negatively correlated with the time spent on the TMT-B (Fig. 4C, R = −0.46, P = 0.003, group× performance interaction, P = 0.87). The hyperconnectivity of the dACC was correlated with higher FPG levels (Fig. 4B, R = 0.65, P < 0.001, group×FPG interaction, P = 0.001) and better TMT-B performance (Fig. 4D, R = −0.52, P = 0.001, group×performance interaction, P = 0.04). Due to the significant correlation between disease duration and neurocognitive performance, we further controlled for disease duration and reanalyzed the correlations. This reanalysis did not significantly affect our findings (data not shown). Similar effects were not detected in the HCs.

4. Discussion

Using graph theory-based analyses, the present study provides the initial evidence of altered global connectivity in the brains of T2DM patients. Patients showed decreased DC in the occipital region and increased DC in the higher-order cognitive control regions. Seed-based analytical approaches revealed that the brain circuits anchored in these regions were also affected, which was correlated with altered neurocognitive performance, suggesting that brain connectivity might be a potential imaging marker for T2DM-associated cognitive impairment.

Decreased DC and within-network connectivity were observed in the lingual gyrus. Consistently with these results, previous imaging studies on T2DM patients have also reported occipital alterations. For example, the occipital lobe has been shown to not only have impaired cerebrovascular reactivity but also decreased overall volume26,27. Studies on a similar diabetic population also reported a decrease in neural intensity and coherence around the lingual gyrus28, 29. The lingual gyrus and its associated occipital regions are regions linked to processing vision-related information and encoding visual memories30. Given the positive correlations of occipital connectivity with visual memory and executive performance demonstrated in the current results, we suggest that decreased occipital connectivity might play an important role in the reduced performance in vision-dependent tasks in T2DM patients. However, due to the small sample size and the lack of visual measurements, it is difficult to determine whether the hypoconnectivity is a neural alteration prior to or is a reflection of reduced visual input induced by potential diabetic retinopathy. Future studies are warranted to clarify the underlying neuropathology of these findings.

In the T2DM group, increased DC was observed in two critical brain hubs, the rAI and dACC. Moreover, the connectivity of the networks anchored in these two regions were also elevated. The rAI and dACC have strong reciprocal connectivity, forming the core of the salience network (SN), which facilitates higher-order cognitive control and behavioral adaption via the “bottom-up” signal detection and the “top-down” transmission of control signals31. Although cognitive functions are often affected in T2DM patients, the effect sizes are smaller in middle-aged adults compared with the effects in those older than 65 years2. This is probably due to a greater brain reserve capacity in younger subjects, reflected by an increased neuronal interconnection32. Moreover, longer diabetes duration in elder patients is often associated with a higher prevalence and severity of diabetic complications and comorbidities, which may further contribute to the worsening of cognitive function. Therefore, the increased connectivity in hub regions probably represents a compensatory mechanism in the younger diabetic population, in cases where more neural resources were required to successfully accomplish relevant tasks. Our results showing a significant correlation between dACC hyperconnectivity and executive performance in the TMT-B task may also support such an assumption. However, whether this compensatory role ceases to be effective in elderly populations needs to be examined in future studies.

Previous fMRI studies primarily focused on the changes within the DMN, attentional network and thalamocortical connections have also identified differences in FC between T2DM and HCs68, 10, 11. Unlike the seed-based or ICA analyses, the graph theory-based approach adopted in the current study takes into account the entire functional connectome instead of relying on apriori seeds or blind source separation13. The discrepancies in the methods used and the brain regions observed might be the cause of the differences in the findings between the current study and those of previous studies. In addition, our results suggest more prominent alterations in the occipital and SN-related regions in relation to whole-brain functional interactions. Nevertheless, such results remain to be confirmed by future studies with a larger sample size.

As in many cross-sectional clinical studies, the current study has several limitations in addition to the relatively small sample size. First, the medication and the duration of diabetes of the included patients were quite variable, which could exert confounding effects on the DC measurement. Treatment-naive patients from a narrower disease duration range should be recruited in future studies. Second, visual acuity was not adjusted for during the vision-dependent cognitive tasks. The inclusion of such measurements is crucial for understanding potential impairments of occipital connectivity, and these should be considered in future studies. Third, a questionnaire was used to confirm the state of the subjects during MR scanning. More objective and rigorous methods, such as visual fixation on a screen, should be performed to avoid such confounding effects. Finally, other FC measurements, such as dynamic and Granger causality connectivity, should be included to obtain more comprehensive information about the network changes in T2DM patients.

5. Conclusion

In summary, this preliminary study suggests that in patients with T2DM, the connectivity density is altered in several brain regions. Decreased DC was primarily found in the occipital lobe, which was correlated with impaired visual memory and executive performance. Hyperconnectivity was found in key nodes (dACC and AI) of the salience network, which is responsible for higher-order cognitive control, and this was correlated with better executive performance. The current results suggest the importance of network connectivity as a potential imaging marker of cognitive decline in T2DM and may provide valuable insights into the neuropathological process of T2DM-related brain alterations.

Figure 5
Voxel-wise correlation between network connectivity (z score) and clinical variables

Supplementary Material


We thank Prof. Shaohua Wang and Dr. Wenqing Xia, Department of Endocrinology, Affiliated Zhongda Hospital of Southeast University, for their assistance with the data collection.

This work was supported by grants from the Major State Basic Research Development Program of China (973 Program) (Nos. 2013CB733800, 2013CB733803), Jiangsu Provincial Special Program of Medical Science (BL2013029) and National Natural Science Foundation of China General Projects (Nos. 81230034, 81271739). SL, HG, YH, XL and YY were supported by the Intramural Research Program of the National Institute on Drug Abuse, the National Institutes of Health.

Grant support: Key Project of Jiangsu Province Natural Science Foundation of China (BK20130057) and National Natural Science Foundation of China General Projects (Nos. 81230034, 81271739).


Type 2 diabetes
healthy control
degree centrality
default mode network
small vessel disease
fasting plasma glucose
oral glucose tolerance test
glycosylated hemoglobin
complex figure test
trail making test
dorsal anterior cingulate cortex
right anterior insula


DISCLOSURES: Yihong YangRELATED: Other: Intramural Research Program of the National Institute on Drug Abuse, Comments: Yihong Yang is an employee of the Intramural Research Program of the National Institute on Drug Abuse, the National Institutes of Health.


1. Geijselaers SL, Sep SJ, Stehouwer CD, et al. Glucose regulation, cognition, and brain MRI in type 2 diabetes: a systematic review. Lancet Diabetes Endocrinol. 2015;3:75–89. [PubMed]
2. Biessels GJ, Deary IJ, Ryan CM. Cognition and diabetes: a lifespan perspective. Lancet Neurol. 2008;7:184–190. [PubMed]
3. Roberts RO, Knopman DS, Przybelski SA, et al. Association of type 2 diabetes with brain atrophy and cognitive impairment. Neurology. 2014;82:1132–1141. [PMC free article] [PubMed]
4. van den Heuvel MP, Hulshoff Pol HE. Exploring the brain network: a review on resting-state fMRI functional connectivity. Eur Neuropsychopharmacol. 2010;20:519–534. [PubMed]
5. Sheline YI, Raichle ME. Resting state functional connectivity in preclinical Alzheimer’s disease. Biol Psychiatry. 2013;74:340–347. [PMC free article] [PubMed]
6. Zhou H, Lu W, Shi Y, et al. Impairments in cognition and resting-state connectivity of the hippocampus in elderly subjects with type 2 diabetes. Neuroscience Letters. 2010;473:5–10. [PubMed]
7. Chen YC, Xia W, Qian C, et al. Thalamic resting-state functional connectivity: disruption in patients with type 2 diabetes. Metab Brain Dis. 2015;30:1227–1236. [PubMed]
8. Musen G, Jacobson AM, Bolo NR, et al. Resting-State Brain Functional Connectivity Is Altered in Type 2 Diabetes. Diabetes. 2012;61:2375–2379. [PMC free article] [PubMed]
9. Hoogenboom WS, Marder TJ, Flores VL, et al. Cerebral white matter integrity and resting-state functional connectivity in middle-aged patients with type 2 diabetes. Diabetes. 2014;63:728–738. [PMC free article] [PubMed]
10. Cui Y, Jiao Y, Chen HJ, et al. Aberrant functional connectivity of default-mode network in type 2 diabetes patients. Eur Radiol. 2015;25:3238–3246. [PMC free article] [PubMed]
11. Xia W, Wang S, Rao H, et al. Disrupted resting-state attentional networks in T2DM patients. Scientific reports. 2015;5:11148. [PMC free article] [PubMed]
12. Nelson PT, Smith CD, Abner EA, et al. Human cerebral neuropathology of Type 2 diabetes mellitus. Biochim Biophys Acta. 2009;1792:454–469. [PMC free article] [PubMed]
13. Zuo XN, Ehmke R, Mennes M, et al. Network centrality in the human functional connectome. Cereb Cortex. 2012;22:1862–1875. [PubMed]
14. Liang X, Zou Q, He Y, et al. Coupling of functional connectivity and regional cerebral blood flow reveals a physiological basis for network hubs of the human brain. Proc Natl Acad Sci U S A. 2013;110:1929–1934. [PubMed]
15. Buckner RL, Sepulcre J, Talukdar T, et al. Cortical hubs revealed by intrinsic functional connectivity: mapping, assessment of stability, and relation to Alzheimer’s disease. J Neurosci. 2009;29:1860–1873. [PMC free article] [PubMed]
16. Di Martino A, Zuo XN, Kelly C, et al. Shared and distinct intrinsic functional network centrality in autism and attention-deficit/hyperactivity disorder. Biol Psychiatry. 2013;74:623–632. [PMC free article] [PubMed]
17. Rubinov M, Knock SA, Stam CJ, et al. Small-world properties of nonlinear brain activity in schizophrenia. Hum Brain Mapp. 2009;30:403–416. [PubMed]
18. Galea M, Woodward M. Mini-Mental State Examination (MMSE) Aust J Physiother. 2005;51:198. [PubMed]
19. Hamilton M. A rating scale for depression. J Neurol Neurosurg Psychiatry. 1960;23:56–62. [PMC free article] [PubMed]
20. American Diabetes Association. Diagnosis and classification of diabetes mellitus. Diabetes Care. 2014;37(Suppl 1):S81–90. [PubMed]
21. Berg E, Reijmer YD, Bresser J, et al. A 4 year follow-up study of cognitive functioning in patients with type 2 diabetes mellitus. Diabetologia. 2009;53:58–65. [PMC free article] [PubMed]
22. Matthews DR, Hosker JP, Rudenski AS, et al. Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia. 1985;28:412–419. [PubMed]
23. Wahlund LO, Barkhof F, Fazekas F, et al. A new rating scale for age-related white matter changes applicable to MRI and CT. Stroke; a journal of cerebral circulation. 2001;32:1318–1322.
24. Jenkinson M, Bannister P, Brady M, et al. Improved optimization for the robust and accurate linear registration and motion correction of brain images. NeuroImage. 2002;17:825–841. [PubMed]
25. Cox RW. AFNI: software for analysis and visualization of functional magnetic resonance neuroimages. Comput Biomed Res. 1996;29:162–173. [PubMed]
26. Tchistiakova E, Anderson ND, Greenwood CE, et al. Combined effects of type 2 diabetes and hypertension associated with cortical thinning and impaired cerebrovascular reactivity relative to hypertension alone in older adults. NeuroImage Clinical. 2014;5:36–41. [PMC free article] [PubMed]
27. Espeland MA, Bryan RN, Goveas JS, et al. Influence of Type 2 Diabetes on Brain Volumes and Changes in Brain Volumes: Results from the Women’s Health Initiative Magnetic Resonance Imaging Studies. Diabetes Care. 2012;36:90–97. [PMC free article] [PubMed]
28. Cui Y, Jiao Y, Chen YC, et al. Altered spontaneous brain activity in type 2 diabetes: a resting-state functional MRI study. Diabetes. 2014;63:749–760. [PubMed]
29. Chen YC, Jiao Y, Cui Y, et al. Aberrant brain functional connectivity related to insulin resistance in type 2 diabetes: a resting-state fMRI study. Diabetes care. 2014;37:1689–1696. [PubMed]
30. Machielsen WC, Rombouts SA, Barkhof F, et al. FMRI of visual encoding: reproducibility of activation. Hum Brain Mapp. 2000;9:156–164. [PubMed]
31. Ham T, Leff A, de Boissezon X, et al. Cognitive control and the salience network: an investigation of error processing and effective connectivity. J Neurosci. 2013;33:7091–7098. [PubMed]
32. Mortimer JA. Brain reserve and the clinical expression of Alzheimer’s disease. Geriatrics. 1997;52(Suppl 2):S50–53. [PubMed]