This study was approved by the Medical Ethics Committee at Tongji Medical College of Huazhong University of Science and Technology, and all participants gave written informed consent in agreement with the Declaration of Helsinki. From December 2015 to January 2017, 30 HBV-RC patients with MHE (24 males; mean age: 48.8 ± 12.2 years), 32 HBV-RC patients without MHE (NMHE, 28 males; mean age: 46.3 ± 9.2 years) and 64 healthy controls who were age-, gender- and education-matched (46 males; mean age: 46.8 ± 9.7 years) were recruited; all participants underwent clinical examinations and neuropsychological tests including the PHES tests and Mini-Mental State Examination (MMSE) before the MR scanning. The HBV-RC diagnosis was based on the medical histories, established clinical examinations and biochemical results as well as imaging findings (such as abdominal ultrasonography and MRI) with standard clinical practice guidelines (Ferenci et al., 2002). According to the Child–Pugh classification of liver function, 8, 8 and 14 of the 30 MHE patients as well as 11, 10 and 11 of the 32 NMHE patients had Child–Pugh grade A, B and C, respectively.

The inclusion criteria for recruitment of all patients were as follows: patients with Hepatitis B cirrhosis; the completion of the MMSE, PHES tests and the MRI scanning without any MRI contraindication; right-handedness. There were no control subjects with any types of liver disease. Exclusion criteria for all subjects included other types of virus-related cirrhosis, any carcinoma, severe metabolic diseases (e.g., thyroid dysfunction), drug intoxications, psychiatric or neurological diseases, head injury, positive human immunodeficiency virus status, bad vision, current overt HE or history of clinical manifestation of overt HE, taking psychotropic medications, probable dementia (MMSE scores ≤24; Cui et al., 2016), left-handedness, any MRI contraindication, and a translation of more than 2.0 mm or a rotation of more than 2.0° during the MRI scan (Chen et al., 2015).

The PHES, recommended by a commission of experts on the hepatic encephalopathy as test batteries suitable for the diagnosis of MHE (Randolph et al., 2009), is a normative set of five psychometric tests for assessing neurocognitive functions (such as motor, visual perception, memory and attention) in cirrhotic patients (Duarte-Rojo et al., 2011; Seo et al., 2012; Li et al., 2013). Five neuropsychological tests were completed by all participants: (1) the digit-symbol test (DST); (2) serial dotting test (SDT); (3) number connection test A (NCT-A); (4) number connection test B (NCT-B); and (5) line tracing test (LTT); PHES has been validated for identifying patients with MHE as reported in many previous studies in different countries (Duarte-Rojo et al., 2011; Seo et al., 2012). The SDT and LTT were available on the network of Hepatic Encephalopathy¹. The DST was a subtest of the Wechsler Adult Intelligence Scale-Revised for China (WAIS-RC). Because several participants did not understand the English alphabet, the English characters from A to L in the NCT-B were replaced with Chinese characters from 1 to 12 in the same order (Li et al., 2013). The NCT-A, the China Revised NCT-B, and the DST tests have also been validated and widely applied for diagnosing MHE in patients with cirrhosis in China (Zhong et al., 2001; Chen et al., 2012b; Lv et al., 2013a). A trained and experienced doctor performed the PHES tests on all enrolled subjects. The healthy controls were included in the five tests to gain a PHES “normative or reference value.” Regression models and formulas were obtained with Pearson’s correlations between the psychological tests scores and age as well as education level by applying multivariate linear regression analyses (Duarte-Rojo et al., 2011; Lv et al., 2013a). These regression formulas were then applied to predict values for patients. The difference between the observed and predicted test scores was divided by SDs for the reference group. Differences for each test were summarized in the following: the observed result ≥1 SD predicted values was scored as +1 points, results −1 SD and −2 SDs < the predicted were scored as −1 and −2 respectively as well as the observed ≤−3 SDs was scored as −3 points (Duarte-Rojo et al., 2011; Lv et al., 2013a). In all cirrhotic patients with no evidence of overt HE and no previous history of overt HE, those having a PHES score of ≤−4 points were considered to have MHE and others having a PHES score of >−4 points were considered to be without MHE (Duarte-Rojo et al., 2011; Lv et al., 2013a).

All MRI data were acquired on a 3.0T GE Discovery MRI 750 w (Grandview Blvd, Waukesha, WI, USA) using an eight-channel head coil. Tight foam padding was fixed to keep their head still and earphones were used to reduce scanner noise during the scan. Each subject was instructed to relax, lie quietly, keep their eyes closed but not fall asleep, not think of anything in particular, and remain still during the MRI scan. The rs-fMRI images were obtained by using an echo-planar imaging sequence. The rs-fMRI acquisition parameters were as follows: repetition time (TR), 3000 ms; echo time (TE), 30 ms; flip angle, 90°; matrix, 64 × 64; field of view (FOV), 256 mm × 256 mm; section thickness, 4.0 mm; number of sections, 40; and NEX, 1.0. The total scan time was 10 min. Axial T2-weighted fluid-attenuated inversion recovery (FLAIR) images(inversion time (TI), 2200 ms; TE/TR, 120/12,000 ms; and matrix, 320 × 224) and three-dimensional brain volume imaging (3D BRAVO) images (TI, 400 ms; TR, 9.1 ms; TE, 3.5 ms; FOV, 25.6 × 25.6 cm²; matrix, 256 × 256; 180 slices; slice thickness, 1.1 mm) were applied at the same orientation for acquiring the anatomical images. All of the images were evaluated by two experienced radiologists blinded to the group status.

Functional images preprocessing was performed by using the Data Processing and Analysis of Brain Imaging (DPABI) toolbox (Yan et al., 2016). The first 10 time points were discarded for adaptation of all participants to the scanning environment and scanner calibration. The remaining volumes were used for further data preprocessing, which included slice-timing correction and realignment correction for head movements. Any subject who had a maximum displacement in any of the three cardinal directions (x, y, z) >2.0 mm or a maximum spin (x, y, z) >2.0° was excluded from the study (Chen et al., 2015). Next, the resulting functional data were spatially normalized to the Montreal neurological institute (MNI) space and resampled to 3 × 3 × 3 mm³. Then, linear-trend discarding and temporal band-pass filtering (0.01 Hz < f < 0.08 Hz) were performed to reduce the effects of low- frequency drifts and high-frequency physiological noise. Several nuisance variables including head motion parameters, white matter and cerebrospinal fluid signals were regressed out were regressed out. We took the Friston 24-parameter model to remove artifacts caused by head motion (Yan et al., 2013). Regarding the global signal, previous studies have demonstrated that removing it may lead to enhances in negative correlations (Murphy et al., 2009; Weissenbacher et al., 2009) or affect differences in functional connection (Saad et al., 2013). Given its controversial biological interpretations, the global signal was not regressed out, which has been applied in prior studies (Murphy et al., 2009; Wang J. et al., 2013; Luo et al., 2015).

Following previous calculation procedures (Ni et al., 2012; Yan et al., 2016), we computed ReHo maps of all subjects by computing the Kendall coefficient of concordance (KCC; Zang et al., 2004). The KCC was used to measure the ReHo of the time series within one voxel and its most adjacent 26 neighboring voxels in a voxel-wise manner. A whole brain map of ReHo values for each subject was calculated. We normalized ReHo values by dividing the ReHo value of each voxel in one ReHo map by the mean whole-brain ReHo value, and then obtained the mean ReHo (mReHo) maps for each subject (Zhang et al., 2012). Then, all mReHo maps were smoothed with a Gaussian kernel of 4-mm full width at half maximum (FWHM; Cui et al., 2016). A one-way analysis of variance (ANOVA) was used on the mReHo maps to identify regions with significant differences among the three groups (p < 0.05, false discovery rate (FDR) correction with a minimum cluster size of at least 10 contiguous voxels; Mutschler et al., 2010; Plaza et al., 2012). These areas were then extracted as a mask. The post hoc t-test of the ReHo maps within this mask was used between each pair of the three groups (MHE vs. NMHE, MHE vs. healthy controls (HC), NMHE vs. HC), with nuisance covariates, i.e., age, education, gender and head motion. The threshold of the tests was set at p < 0.05 with FDR correction for multiple comparisons using the DPABI software.

Based on the ReHo results after the one-way ANOVA test among the three groups, we chose nine brain regions with significant group differences obtained in ReHo as the ROIs for the ROI-based FC analysis: the right fusiform gyrus (FFG.R), right superior frontal gyrus, orbital part (SFGorb.R), right thalamus (THA.R), right calcarine (CAL.R), left middle frontal gyrus (MFG.L), left precentral gyrus (PreCG.L), right postcentral gyrus (PoCG.R), right superior frontal gyrus (SFG.R) and left paracentral lobule (PCL.L; FDR-corrected, p < 0.05). To uncover the functional connections patterns among these regions, we constructed their pair-wise connectivity matrix individually using the FC analysis; FC analysis calculates the functional correlations between any of each of the two ROI mean time courses among different ROI regions using the DPABI software. For each ROI region of nine ROIs, mean time series of all ROIs were calculated for each subject and then correlated with the other eight ROIs. We obtained 9 × 8/2 ROI correlation coefficients for each subject; the Fisher’s Z transformation was applied in all correlation coefficient matrixes to enhance the normality for comparisons (Tang et al., 2011). Next, for comparing the FC, we performed the one-way ANOVA test and post hoc t-tests to calculate the group differences of z values of ROI correlation coefficients across the three groups. The threshold of all tests was set at p < 0.05 with FDR correction for multiple comparisons using the GRETNA software toolbox (Wang et al., 2015). Alternatively, a threshold of p < 0.001 (uncorrected for multiple comparisons) was applied in the condition group comparisons based on the stringency of the group contrasts used in this exploratory study.

The Kolmogorov-Smirnov test, one-way ANOVA test and post hoc t-tests, two independent samples T-test, χ2 test and the Mann-Whitney U test were performed to analyze the group differences in the demographic and clinical data using SPSS 18.0 software (SPSS Inc., Chicago, IL, USA). Multivariate linear regression was applied to calculate the PHES normative data (Seo et al., 2012). The significance level was set at p < 0.05.

The clusters showing significant group differences in ReHo and FC between each pair of the three groups (MHE vs. NMHE, MHE vs. HC, NMHE vs. HC) were visualized with the BrainNet Viewer (Xia et al., 2013) and DPABI software viewer (FDR-corrected, p < 0.05). The mean ReHo values and FC correlation coefficient values of these abnormal clusters were extracted using DPABI software. Then, we applied Pearson’s correlation analysis to examine the relationship between altered ReHo/FC and with clinical variables (e.g., PHES tests scores and disease duration) in the MHE/NMHE patients. Considered this is an exploratory study, correlation analyses were not corrected by Bonferroni correction.

The demographic and clinical characteristics for all patients and healthy subjects are shown in Table 1. There were no significant differences in age, gender or education among the three groups. All participants in our study performed normally on the MMSE test.

Compared with the PHES in healthy controls (mean = −0.19; SD = 1.42; range: −2 to +3), HBV-RC patients with MHE had significantly worse PHES scores (mean = −8.00; SD = 3.10; range: −13 to −4). The 95% range of the normative values between the mean −2 SDs and mean +2 SDs was −3.03 to 2.65 points. The MHE was diagnosed when the PHES score was ≤−4 points (Li et al., 2013).

A one-way ANOVA test showed significant group differences in ReHo among the MHE, NMHE and HC groups in the following nine cortical and subcortical regions: FFG.R, SFGorb.R, MFG.L, THA.R, CAL.R, PoCG.R, PreCG.L, SFG.R and PCL.L (FDR-corrected p < 0.05; Figure 1, Table 2).

Compared with HC, the MHE patients had significantly altered ReHo mainly in the visual network (VN, FFG.R, CAL.R), somatomotor network (SMN, PoCG.R, PCL.L), frontal parietal control (FPC) network (SFG.R) and THA.R (FDR-corrected p < 0.05; Figure 2, Table 3). For NMHE vs. HC, we found a significant ReHo decrease in the THA.R and CAL.R, as well as ReHo enhancement in the right frontal pole, dorsal attention network (DAN, MFG.L), SMN (PreCG.L), and SFG.R in NMHE patients (FDR-corrected p < 0.05; Figure 3, Table 3). MHE patients showed increased mean ReHo in the PoCG.R compared to the NMHE group (p < 0.001, uncorrected; Table 3).

A one-way ANOVA test exhibited significant group differences in FC among the three groups (FDR-corrected p < 0.05). As is shown in Figure 4A and Table 4, compared with HC, the MHE patients showed significantly lower FC between the CAL.R and FFG.R/THA.R, as well as greater FC between the MFG.L and FFG.R/CAL.R (FDR-corrected p < 0.05). We detected significantly greater FC between CAL.R and MFG.L in MHE patients compared to the NMHE group (FDR-corrected p < 0.05; Figure 4B; Table 4). When compared to HC, significantly altered FC in NMHE patients were not observed in the present study after FDR correction.

There was a positive correlation between the increased mean ReHo values in FFG.R and disease duration in patients with MHE (r = 0.441, p = 0.015; Figure 5A). For NMHE patients, the mean ReHo values in the right frontal pole were not only positively correlated with disease duration (r = 0.425, p = 0.015; Figure 5B) but also negatively correlated with the PHES scores (r = −0.485, p = 0.005; Figure 5C). Significantly reduced z values of FC between the CAL.R and THA.R in MHE were also positively correlated with poor PHES performance (r = 0.390, p = 0.033; Figure 6A). Within the VN and DAN, the increased z values of FC between the CAL.R and MFG.L in MHE patients were not only negatively correlated with the PHES test scores, in which a lower score suggested worse neuropsychological performance (r = 0.474, p = 0.008; Figure 6B) but also negatively correlated with the disease duration (r = 0.481, p = 0.007; Figure 6C). The abovementioned correlations were not corrected for multiple comparisons, due to the relatively small-size patient group and the exploratory feature of this correlation analysis.

Altered resting-state neural activity is considered to relate to functional disturbances in cirrhosis with and without MHE. Several previous rs-fMRI studies have revealed that some brain regions with disturbed local coherence are involved in the disease progress in MHE or NMHE, including cortical regions of the frontal cortex, PoCG and subcortical regions of the thalamus, which were consistent with our results (Qi et al., 2012, 2013; Lv et al., 2013a; Chen Q.-F. et al., 2016). Compared with HC, the cirrhotic patients with or without MHE exhibited increased mean ReHo in the FFG.R and PoCG.R, and decreased mean ReHo in the CAL.R, which indicate aberrant brain activity synchronization in these brain areas. In the VN, the CAL and FFG play essential roles in visional association functions, such as visual attention, processing and discrimination (Lee et al., 2000; Lv et al., 2013b). The PreCG and PoCG (components of the SMN) are primarily involved in motor control. Therefore, the disturbed regional synchronization of the VN and SMN may account for the visual and motor dysfunction in the MHE/NMHE patients, which were also reported in prior studies (Lockwood et al., 2002; Schiff et al., 2005; Weissenborn et al., 2007; Ni et al., 2012; Qi et al., 2012). Some positron emission tomography studies also displayed abnormal cerebral glucose utilization in the visual association regions (Lockwood et al., 2002), and motor-related regions (Weissenborn et al., 2007) in cirrhosis. Moreover, in line with previous neuroimaging studies in cirrhosis (Chen et al., 2012b; Lv et al., 2013a; Ni et al., 2014), MHE/NMHE patients showed enhanced mean ReHo in the DAN (MFG.L) and SFG.R compared with healthy controls. Attention and cognitive control deficits are also regarded as MHE-related characteristics (Bajaj et al., 2009; Chen et al., 2014), and can affect working memory and learning ability (Amodio et al., 2005; Felipo et al., 2012). The DAN is associated with the orienting of attention (Corbetta and Shulman, 2002; Chen et al., 2014), and the SFG within the FPC network is involved in cognitive control (du Boisgueheneuc et al., 2006; Kompus et al., 2009; Gao and Lin, 2012). Therefore, we speculate that altered local synchronization in the DAN and FPC network regions may account for the attention and cognition deficits in MHE/NMHE. In addition, MHE patients exhibited increased mean ReHo in the SMN (PoCG.R; voxel p < 0.001, uncorrected) compared with the NMHE group, indicating further motor dysfunction with the disease progression from NMHE to MHE; this finding was similar to one previous study (Ni et al., 2012). The result of group differences in ReHo between the two patient groups is exploratory because they did not pass multiple comparisons correction. Considering that the sample size for patient groups is relatively small and the variations in disease duration, we do not exclude the possibility that these factors might influence the validity of the statistical analysis and findings of the present study.

A striking finding was the identification of a positive correlation between the PHES tests scores and the enhanced mean ReHo values in the VN (FFG.R) in MHE patients, which was in line with a previous study (Chen et al., 2012a). For NMHE patients, the increased mean ReHo values in the prefrontal cortex (right frontal pole) were correlated with both the PHES scores and disease duration. PHES, a set of the neuropsychological tests, was widely used to assess the neurocognitive function in cirrhotic patients (Giménez-Garzó et al., 2017). The prefrontal cortex plays an essential role in cognitive control (Miller and Cohen, 2001). These correlations suggest that the ReHo indicator in these brain regions may be a potential biomarker to reflect neuropsychological alterations in cirrhosis and the ReHo in the right frontal pole may reflect the clinical progression of NMHE.

There are distant and local interconnections between different regions, shaping brain networks. To our knowledge, regional synchronization and functional connection pattern alterations in MHE/NMHE have not been fully studied in prior rs-fMRI studies at the same time. Thus, the current FC results may extend and confirm recent neuroimaging findings in patients with and without MHE.

When compared with healthy controls, MHE patients exhibited functional connections abnormalities mainly between the VN (CAL.R) and some networks, including the VN, DAN and THA.R, which are in line with previous studies that disruption of the DAN, VN and thalamic functional connections have been reported in MHE (Qi et al., 2012, 2013, 2015). These results suggest that aberrant cooperation of their corresponding function, such as attention control and visual cognition, were disturbed in MHE. Furthermore, the MHE group showed increased FC between the DAN (MFG.L) and VN (CAL.R) compared to NMHE patients. This finding indicates that long-range FC between the two networks had progressively increased functional connections from NMHE to MHE, and the FC values in those regions might be a potential marker for assessing the disease progression but be unspecific for the presence of MHE. Additionally, it has been detected cirrhotic patients’ disturbances in attention control, visual processing, memory, and other cognitive functions (Weissenborn et al., 2001; Weissenborn, 2008). The thalamus is an important subcortical gray matter area, and the thalamic FC network plays an essential role in integrating information across the brain circuits (Qi et al., 2013). Therefore, alterations in the FC within the thalamus, VN and DAN regions may be associated with the impaired attention and visual-spatial capabilities in patients with or without MHE. In addition, it is worth noting that there was no significant difference in FC between NMHE patients and the HC group. The possible reason is that functional connections in NMHE patients might be intact and robust due to functional compensatory or reorganization of the neural mechanism in the brain. However, this result was not in accordance with one prior neuroimaging study by Chen et al. (2014) in which NMHE patients showed an abnormality of FC in the attention-related network compare with controls. One reason for the difference of FC results in NMHE may be the inconsistency of admission criteria. The etiologies of cirrhosis in their study include alcoholic and HBV-related cirrhosis. Some studies suggested that the brain was very susceptible to alcohol action, and consequently low exposure to moderate alcohol could affect neural function (Baglietto et al., 2011). Therefore, we believe that alcohol is a potential confounder in the analysis of FC, which might lead to the results of our two studies being different. Furthermore, we found the correlations between clinical variables and aberrant FC in the VN, DAN and thalamus, which were consistent with prior studies (Qi et al., 2012, 2013). The PHES test scores in MHE patients were negatively correlated with the enhanced FC between the DAN (MFG.L) and VN (CAL.R) and positively correlated with FC between the VN (CAL.R) and THA.R; these correlations indicate that aberrant FC within those networks in MHE patients may be associated with dysfunctions in visual processing, attentive activity and information integration during cognitive tasks. For NMHE patients, FC values between the CAL.R and MFG.L were positively correlated with the disease duration, suggesting that functional connection impairments between the VN and DAN may represent the progression of the disease.

The current study has several limitations. First, the group differences in the ReHo were found in the PoCG.R between the MHE and NMHE groups (voxel p < 0.001 uncorrected), while they were no longer significant after FDR correction. We speculate that the current study is a cross-sectional study with a relatively small sample size, which may affect the statistical power and the interpretation of the final results. Therefore, further longitudinal studies with larger groups of participants are needed to further confirm our findings and provide additional information about the abnormalities of the functional network altered patterns with the progression of MHE and NMHE. Second, although we instructed all participants to stay still during the rs-fMRI scan, we cannot completely eliminate the possibility of some uncontrolled factors, such as involuntary head movement, affecting our results during the scanning. Third, the results of abnormal ReHo/FC correlation to clinical variables are exploratory because they did not pass stringent Bonferroni correction, and they thus need to be validated in large-size subjects. Fourth, in order to ensure the homogeneity of the case, all of the patients recruited in our study were HBV-RC subjects with MHE/NMHE. Therefore, our findings may not be generalized to other type of cirrhosis, such as the autoimmune cirrhosis.