Conceived and designed the experiments: WCL CPL YFC. Performed the experiments: WCL TWH CWW HLC CHL. Analyzed the data: TWH CPL. Contributed reagents/materials/analysis tools: KHC. Wrote the paper: WCL TWH.
The authors have declared that no competing interests exist.
Hepatic encephalopathy (HE) is a complex neuropsychiatric syndrome and a major complication of liver cirrhosis. Dysmetabolism of the brain, related to elevated ammonia levels, interferes with intercortical connectivity and cognitive function. For evaluation of network efficiency, a ‘small-world’ network model can quantify the effectiveness of information transfer within brain networks. This study aimed to use small-world topology to investigate abnormalities of neuronal connectivity among widely distributed brain regions in patients with liver cirrhosis using resting-state functional magnetic resonance imaging (rs-fMRI). Seventeen cirrhotic patients without HE, 9 with minimal HE, 9 with overt HE, and 35 healthy controls were compared. The interregional correlation matrix was obtained by averaging the rs-fMRI time series over all voxels in each of the 90 regions using the automated anatomical labeling model. Cost and correlation threshold values were then applied to construct the functional brain network. The absolute and relative network efficiencies were calculated; quantifying distinct aspects of the local and global topological network organization. Correlations between network topology parameters, ammonia levels, and the severity of HE were determined using linear regression and ANOVA. The local and global topological efficiencies of the functional connectivity network were significantly disrupted in HE patients; showing abnormal small-world properties. Alterations in regional characteristics, including nodal efficiency and nodal strength, occurred predominantly in the association, primary, and limbic/paralimbic regions. The degree of network organization disruption depended on the severity of HE. Ammonia levels were also significantly associated with the alterations in local network properties. Results indicated that alterations in the rs-fMRI network topology of the brain were associated with HE grade; and that focal or diffuse lesions disturbed the functional network to further alter the global topology and efficiency of the whole brain network. These findings provide insights into the functional changes in the human brain in HE.
Liver cirrhosis is frequently associated with a wide range of neuropsychiatric abnormalities including personality disorders and inappropriate affective, behavioral, and sleep disturbances. Patients with acute liver failure can succumb to neurologic death, with brain edema and intracranial hypertension
The results from some studies have suggested that patients with HE might have disturbed brain energy metabolism and intracranial hemodynamics
Neuroimaging studies in cirrhotic patients have described early impairment of the neural connectivity mechanism and abnormal coupling between visual judgment areas in HE
Several studies have demonstrated that the human brain is organized intrinsically as highly modular small-world architectures capable of efficiently transferring information at a low wiring cost, and forming highly connected hub regions
To investigate this hypothesis, patients with cirrhosis with 3 grades of encephalopathy (no HE, minimal HE, and overt HE) were enrolled for resting-state fMRI (rs-fMRI) analysis. The advantage of rs-fMRI is that it avoids the need for training and differential performances on cognitive tasks, and also allows the quantification of functional connectivity (FC) by measuring the correlations or interdependencies between intrinsic blood oxygenation level dependent (BOLD) signal fluctuations of distributed brain areas. Resting-state fMRI data were obtained to investigate the differences in the organizational patterns of functional networks between patients with differing grades of HE. Results suggested dissimilar modifications to the brain organization patterns, depending on focal insults or global alterations caused by the severity of HE. The affected global and local economic performances of functional brain networks in HE might be associated with the accumulation of ammonia.
From August 2009 to December 2010, patients with cirrhosis and awaiting liver transplantation were enrolled from the Surgery Outpatient Clinic of Chang Gung Memorial Hospital, Kaohsiung Medical Center. Thirty-five patients (26 men and 9 women; mean age, 54.2 years; range, 34 to 68 years) were recruited and received both neuroimaging and neuropsychological (NP) examinations. The hospital’s Institutional Review Committee on Human Research approved the study.
Patients with liver cirrhosis were excluded if they had any history of drug abuse, psychiatric or neurological illness, or head injury. Cirrhosis was diagnosed according to clinical and imaging features
For comparison, 35 healthy volunteers (22 men and 13 women; median age, 51 years; range, 26 to 67 years), without any medical history of neurological disease, were recruited through advertising within the hospital and served as the control group. All volunteers received detailed clinical and neurological examinations on the same day of the MRI scans.
Regions name | Abbreviation | Classification | Regions name | Abbreviation | Classification |
Superior frontal gyrus, dorsalateral | SFGdor | Association | Superior parietal gyrus | SPG | Association |
Superior frontal gyrus, orbital | SFGorb | Paralimbic | Paracentral lobule | PCL | Association |
Superior frontal gyrus, medial | SFGmed | Association | Postcentral gyrus | PoCG | Primary |
Superior frontal gyrus, medial orbital | SFGmorb | Paralimbic | Inferior parietal gyrus | IPG | Association |
Middle frontal gyrus | MFG | Association | Supramarginal gyrus | SMG | Association |
Middle frontal gyrus, orbital | MFGorb | Paralimbic | Angular gyrus | ANG | Association |
Inferior frontal gyrus, opercular | IFGoper | Association | Precuneus | PCUN | Association |
Inferior frontal gyrus, triangular | IFGtri | Association | Posterior cingulate gyrus | PCC | Association |
Inferior frontal gyrus, orbital | IFGorb | Paralimbic | |||
Gyrus rectus | REG | Association | Insula | INS | Paralimbic |
Anterior cingulate gyrus | ACC | Paralimbic | Thalamus | THA | Subcortical |
Olfactory cortex | OLF | Paralimbic | |||
Superior temporal gyrus | STG | Association | |||
Precentral gyrus | PreCG | Primary | Superior temporal gyrus, temporal pole | STGp | Paralimbic |
Supplementary motor area | SMA | Association | Middle temporal gyrus | MTG | Association |
Rolandic operculum | ROL | Association | Middle temporal gyrus,temporal pole | MTGp | Paralimbic |
Median- and para- Cingulate gyrus | MCC | Paralimbic | Inferior temporal gyrus | ITG | Paralimbic |
Heschl gyrus | HES | Primary | |||
Calcarine fissure and surrounding cortex | CAL | Primary | Hippocampus | HIP | Paralimbic |
Cuneus | CUN | Association | Parahippocampal gyrus | PHIP | Paralimbic |
Lingual gyrus | LING | Association | Amygdala | AMG | Paralimbic |
Superior occiptal gyrus | SOG | Association | |||
Middle occiptal gyrus | MOG | Association | Caudate nucleus | CAU | Subcortical |
Inferior occiptal gyrus | IOG | Association | Lenticular nucleus, putamen | PUT | Subcortical |
Fusiform gyrus | FG | Association | Lenticular nucleus, pallidum | PAL | Subcortical |
Control | no HE | MHE | OHE | F or X2 | p-value | |
# of subjects | 35 | 17 | 9 | 9 | ||
Age (years) | 50.06 ± 10.23 | 49.95 ± 7.85 | 57.00 ± 5.70 | 60.56 ± 7.09 | 4.42 | 0.007 |
Gender | 23F/22M | 5F/12M | 1F/8M | 4F/5M | 1.83 | 0.149 |
Creatinine (mg/dL) | – | 0.71 ± 0.35 | 0.74 ± 0.19 | 1.06 ± 0.49 | 0.93 | 0.402 |
GOT (IU/L) | – | 66.65 ± 52.58 | 94.25 ± 102.07 | 92.00 ± 72.42 | 0.67 | 0.519 |
Bilirubin (mg/dL) | – | 1.80 ± 0.40 | 3.66 ± 3.04 | 6.98 ± 12.43 | 1.91 | 0.162 |
Venous ammonia (mg/dL) | – | 113.64 ± 47.75 | 139.10 ± 67.63 | 194.22 ± 81.93 | 3.22 | 0.053 |
Albumin (mg/dL) | – | 3.35 ± 0.64 | 2.73 ± 0.57 | 2.90 ± 0.31 | 4.10 | 0.026 |
International Normalized Ratio (INR) | – | 1.19 ± 0.18 | 1.29 ± 0.28 | 1.60 ± 0.63 | 4.31 | 0.021 |
Neuropsychiatric tests | – | |||||
Digit-symbol | 65.76 ± 22.49 | 59.12 ±17.46 | 27.22 ± 10.77 | – | 13.41 | 0.000 |
Block design | 35.60 ± 13.14 | 32.38 ± 10.65 | 21.80 ± 6.61 | – | 5.39 | 0.008 |
Results are given as mean ± standard deviation. MHE: minimal hepatic encephalopathy; OHE, overt hepatic encephalopathy. RSPM, Raven’s Standard Progressive Matrices; CASI, Cognitive Ability Screening Instrument.
Significant was set at p<0.05.
The A. column displays the interregional correlation matrix for each group, obtained by calculating Pearson’s correlations between the regional brain areas across subjects within the group. The color bar indicates the correlation coefficient between regions. The B column displays the binary connectivity matrices with fixed network density thresholds of 10%
The functional brain networks showed higher local efficiency than that of the matched random networks, (A) and higher global efficiency than that of the matched regular networks (B) at a wide range of cost thresholds. Thus, the functional brain networks for each group exhibited small-world properties regardless of disease severity. The brain networks were also found to be economical since both the local and global efficiency were much higher than the required cost. Note that the regular and random networks in the plots had the same number of nodes and edges as the real networks.
Significant decreases in absolute global and local efficiency with increasing severity of HE are observed, with non-significant increases of integrated relative global efficiency.
The regional correlation strength of the nodes were redistributed and higher in healthy control compared with the HE group in the association, primary, and limbic/paralimbic cortices. The nodes shown in blue lines were 1 SD over the mean nodal connectivity strength in each group. Each box plot shows the range for the individual estimates of the regional correlation strength values in each group. The boxes are color coded to differentiate the primary sensory or motor cortex (red), heteromodal association cortex (blue), limbic or paralimbic cortex (orange), and subcortical nuclei (green).
The “hubs” were redistributed and the nodes that were 1 SD over the mean regional efficiency tended to be in the same regions of the association and primary cortices in each group. The HE groups had lower percentages of regional efficiency in the primary cortex for the HE (B, C, D) patients compared with the healthy controls. Each box plot shows the range for the individual estimates of the regional correlation strength values in each group. The boxes are color coded to differentiate the primary sensory or motor cortex (red), heteromodal association cortex (blue), limbic or paralimbic cortex (orange), and subcortical nuclei (green).
According to Ferenci’s report
Functional imaging data were acquired using a 3.0 T GE Signa MRI scanner (Milwaukee, WI, USA). Resting-state images, which resulted from 300 contiguous echo planar imaging whole brain functional scans (TR = 2 s, TE = 30 ms, FOV = 240 mm, flip angle 80°, matrix size 64×64, thickness = 4 mm), were collected. During the resting experiment, the scanner room was darkened and the participants were instructed to relax, with their eyes closed, without falling asleep. A 3D high-resolution T1-weighted anatomical image was also acquired using an inversion recovery fast spoiled gradient-recalled echo pulse sequence (TR = 9.5 ms; TE = 3.9 ms; TI = 450 ms; flip angle = 20°; field of view = 256 mm; matrix size = 512×512).
Prior to preprocessing, the first 10 volumes were discarded to reach a steady-state magnetization, and to allow the participants to adapt to the scanning noise. Resting-state fMRI data preprocessing was then performed using the Statistical Parametric Mapping (SPM8, Wellcome Department of Cognitive Neurology, London, UK;
To ensure that each rs-fMRI data set was the best possible representative of spontaneous neural activities, the effects of physiological sources were minimized by regressing out estimated predictors. Nine predictors were generated including white matter (WM), cerebral spinal fluid (CSF), the global signal, and 6 motion parameters
Nodes and edges are two basic elements of a network. To determine the nodes and edges of the brain networks, methods applied were similar to those described previously
For each subject data set, 90 regional mean time series were estimated by averaging voxel time series within each of the 90 anatomically defined regions (
Brain functional networks have economical small-world properties which support the efficient transfer of parallel information at relatively low cost
Three measures for the nodal (regional) characteristics of functional networks in liver cirrhosis were quantified: the nodal correlation strength, its regional absolute, and the relative efficiency. The regional strength of connectivity
The regional nodal efficiency of node i was defined as the inverse of the mean harmonic shortest path length between this node and all other nodes in the network.
The integrated nodal efficiency enabled the characterization of nodal properties in the brain networks without the selection of a specific network threshold.
Correlation Strength | Absolute efficiency | Relative efficiency | |||||
Anatomical regions |
|
|
|
|
|
|
|
Association | Right rolandic gyrus | −4.815 | (<0.001) | −5.574 | (<0.001) | −5.083 | (<0.001) |
Right superior temporal gyrus | −4.484 | (<0.001) | −4.410 | (<0.001) | −3.923 | (<0.001) | |
Left rolandic gyrus | −4.279 | (<0.001) | −4.076 | (<0.001) | −4.056 | (<0.001) | |
Left superior temporal gyrus | −4.202 | (<0.001) | −4.440 | (<0.001) | −4.247 | (<0.001) | |
Left paracentral lobule | −3.110 | (0.003) | −3.338 | (<0.001) | −2.193 | (0.032) | |
Right supramarginal gyrus | −3.062 | (0.003) | −2.745 | (0.008) | NS | ||
Right gyrus rectus | −2.876 | (0.005) | NS | NS | |||
Right supplementary motor area | −2.815 | (0.006) | −2.156 | (0.035) | NS | ||
Left superior parietal gyrus | −2.740 | (0.016) | −2.690 | (0.009) | NS | ||
Right paracentral lobule | −2.632 | (0.011) | −2.917 | (0.005) | −2.606 | (0.011) | |
Left gyrus rectus | −2.381 | (0.020) | NS | NS | |||
Right superior paeietal gyrus | −2.226 | (0.029) | −2.462 | (0.016) | NS | ||
Left middle temporal gyrus | −2.191 | (0.032) | NS | NS | |||
Right superior frontal gyrus, dorsolateral | NS | NS | 2.846 | (0.06) | |||
Left middle frontal gyrus | NS | NS | 3.179 | (0.002) | |||
Right middle frontal gyrus | NS | NS | 2.610 | (0.011) | |||
Left inferior occiptial gyrus | NS | −3.033 | (0.003) | −2.137 | (0.036) | ||
Left supplementary motor area | NS | −2.728 | (0.008) | −2.554 | (0.013) | ||
Primary | Right heschl gyrus | −5.683 | (<0.001) | −5.648 | (<0.001) | −4.410 | (<0.001) |
Left heschl gyrus | −5.464 | (<0.001) | −4.697 | (<0.001) | −3.680 | (<0.001) | |
Right precentral gyrus | −3.950 | (<0.001) | −3.111 | (0.003) | −2.591 | (0.012) | |
Right postcentral gyrus | −3.730 | (<0.001) | −4.102 | (<0.001) | −3.937 | (<0.001) | |
Left precentral gyrus | −3.235 | (0.002) | −2.622 | (0.011) | −2.367 | (0.021) | |
Left postcentral gyrus | −3.233 | (0.002) | −3.803 | (<0.001) | −4.014 | (<0.001) | |
Right calcarine fissure and surrounding cortex | NS | −2.269 | (0.026) | NS | |||
Paralimbic | Right insula | −4.270 | (<0.001) | −4.882 | (<0.001) | −3.615 | (<0.001) |
Left insula | −3.818 | (<0.001) | −4.060 | (<0.001) | −4.428 | (<0.001) | |
Right amygdala | −3.552 | (0.001) | −4.602 | (<0.001) | −3.669 | (<0.001) | |
Right superior temporal gyrus, temporal pole | −3.190 | (0.002) | −4.414 | (<0.001) | −3.956 | (<0.001) | |
Right hippocampus | −2.892 | (0.005) | −4.024 | (<0.001) | −2.871 | (0.005) | |
Left amygdala | −2.861 | (0.006) | −3.614 | (0.001) | −2.667 | (0.010) | |
Left superior temporal gyrus, temporal pole | −2.745 | (0.008) | −3.425 | (0.001) | −2.059 | (0.043) | |
Right parahippocampus gyrus | −2.675 | (0.009) | −4.622 | (<0.001) | −4.019 | (<0.001) | |
Left anterior cigulate gyrus | −2.598 | (0.011) | −3.233 | (0.002) | −2.577 | (0.012) | |
Left parahippocampus gyrus | −2.524 | (0.014) | −3.387 | (0.001) | −2.020 | (0.047) | |
Right anterior cigulate gyrus | −2.239 | (0.028) | NS | NS | |||
Left median- and para-cigulate gyrus | NS | −2.839 | (0.006) | −2.228 | (0.029) | ||
Right median- and para-cigulate gyrus | NS | −2.099 | (0.040) | NS | |||
Left superior frontal gyrus, medial orbital | NS | −2.655 | (0.010) | NS | |||
Left hippocampus | NS | −3.055 | (0.003) | NS | |||
Subcortical | Right putamen | −3.880 | (<0.001) | −4.602 | (<0.001) | −3.317 | (0.001) |
Left putamen | −3.441 | (0.001) | −3.684 | (<0.001) | −2.934 | (0.005) | |
Left pallidum | −3.235 | (0.002) | −4.178 | (<0.001) | −3.092 | (0.003) | |
Right pallidum | −2.749 | (0.008) | −3.424 | (0.001) | −2.077 | (0.042) |
The t-value indicates significant decreases in nodal characteristics with increasing grade hepatic encephalopathy, and vice versa. NS = non-significant. The cortical and subcortical regions were classified as primary, association, paralimbic and subcortical.
The figures show significant between group increases in the stages of HE for the left and right hemisphere.
The lower integrated relative and absolute local efficiencies that were significant and borderline significant correlated with higher ammonia levels.
Shown are the nodes of the networks, defined in terms of regional connectivity strength (A) and relative nodal efficiency (B) that were 1 SD over the mean nodal connectivity strength and the relative nodal efficiency in the healthy controls. The distribution of nodal correlation strength and efficiency in the liver cirrhotic groups showed increased weighting in the association cortex and decreased weighting in the primary cortex in the order of no HE, MHE, and OHE and correlated with the mean ammonia level. The heteromodal or unimodal association cortices (blue), primary cortices (red), limbic/paralimbic (green) and mean ammonia level (orange line) are shown.
Statistical analysis was performed using SPSS 13 software (SPSS Inc, Chicago, IL). The data analyses were conducted in two steps: demographic and main analyses. The demographic analyses compared the demographic variables with clinical characteristics of all subgroups of liver cirrhosis and healthy controls to identify any potentially confounding relationships. The main analyses determined the network topology between subgroups of liver cirrhosis with various degrees of severity, using linear regression and univariate ANOVA. The participants’ age and sex were applied as covariates. Network parameters,
The efficiency curves in the healthy controls and patients with cirrhosis were intermediate compared with those of the matched regular and random networks over a wide range of network costs (
Univariate ANOVA of small-world efficiency in the FC networks also revealed significant differences in the integrated absolute (F(3,66) = 4.228,
In this study, the regional correlation strength of each node was a measurement of its connectivity strength to all other nodes of the cortical network; with results for all 90 brain regional nodes sorted in order of descending regional correlation strength. The primary (sensory-motor), association (heteromodal and unimodal), and limbic/paralimbic cortices had relatively higher correlation strength values in the healthy controls than in the patients with liver cirrhosis (
The regional efficiency is an approach for measuring regions with highly connected nodes or hubs. High regional efficiency indicates a pivotal role in organizing network dynamics and the ability of the region to exert a strong influence on the state of additional peripheral nodes. The results for regional efficiency were similar to those for regional correlation strength (
The spatial reorganization in all of the study participants was increasingly weighted in the heteromodal or unimodal association cortices, and decreasingly weighted in the primary cortex; in the order of healthy, no HE, MHE, and OHE groups. Further comparison of the regional node characteristics, using a linear regression method of analysis, revealed that increased severity of HE was associated with significantly reduced node strength and integrated relative and absolute regional efficiency in the frontal and temporal cortices, including the limbic and paralimbic regions. These consist of the insula, hippocampus, parahippocampus gyrus, amygdala, temporal pole of the superior temporal gyrus, anterior and median cingulate cortex, and subcortical regions including the putamen and pallidum (
Low integrated relative efficiency was the only factor to significantly correlate with high ammonia levels (
(1) Disruption to the local and global topological organization of the FC network in patients with HE; with abnormal small-world properties and topological efficiency, especially in the OHE group; (2) alterations in the regional characteristics, including nodal efficiency and nodal strength, in patients with HE; predominantly in the numbers of the heteromodal or unimodal association, primary, and limbic/paralimbic regions; (3) that ammonia levels in patients with liver cirrhosis were associated with the alterations in local network properties. These findings supported the hypothesis that HE is characterized by the loss of small-world network characteristics.
A suitable balance between local specialization and global integration of brain functional activity
Local efficiency (an index of functional segregation), is predominantly associated with short range connections between nearby regions which mediate modularized information processing, or fault tolerance of a network
Although the declined FC network in liver cirrhosis from no HE to OHE, group differences in local small-world parameters occurred only between healthy controls and the most severe HE patients. This further supports the previous clinical observations. Patients with cirrhosis and subclinical HE exhibited a “normal” appearance, unless the patient had reached the state of OHE. Local alterations in regional FC networks, therefore, precede the global changes in the no HE and MHE groups; and their remodeling may further maintain clinical performance. Regional anatomical investigation in subclinical HE might serve to increase understanding of the development of OHE, and facilitate its early diagnosis and prevention.
In efficient absolute and relative networks, the nodes function as hubs for information processes. According to rs-fMRI findings, they exhibit a relatively high level of metabolism and receive a high blood flow supply. This might provide an explanation for these regions being the first to demonstrate involvement in HE and in degenerative processes
The reductions in regional correlation strength or efficiency in the patients with cirrhosis were most prominent in the heteromodal and unimodal association, primary, paralimbic, and subcortical cortices (
Ammonia is extremely toxic to the brain and is absorbed and metabolized by astrocytes, leading to cell swelling
For regional efficiency, hyperammonemia caused increased nodal weighting in the heteromodal or unimodal association cortices, and decreased weighting in the primary cortices of the HE, MHE, and OHE groups (
First, the BOLD signal is an indirect measurement of neuroelectrical activity. The hemodynamic response of the BOLD signal is quite limited, due to temporal resolution, and delayed by real-time modulations of neural activity. Although rs-fMRI is an important approach for understanding the normal human brain and psychiatric diseases, some noise, such as cardiac and/or respiratory cycle-related pulsations and instrumental and thermal sources of noise, is inevitable. Head movement (rotation or translation) of the participant during scanning may also affect the stability of the rs-fMRI signals. Methods such as regression can reduce noise; although new and improved methods need development to further reduce noise from the previously mentioned sources. Second, it is not possible to eliminate the effects of heterogeneity in clinical symptoms, duration of illness, severity of symptoms, and medication on patient findings. Third, the study sample size was small (9 MHE and 9 OHE patients), due to the generally low incidence of HE. Fourth, the human brain is a complex network on multiple spatial and time scales; therefore it is difficult to select the nodes, edges, and time scale which most appropriately represent the network in the natural state of the brain. The inappropriate representation of nodes and edges in a network, and failure to consider the dynamics of the system of interest, could lead to misleading conclusions and generally poor results
In clinical application, the rs-fMRI has advantages over task activation fMRI in that it avoids the need for training and differential performance on cognitive tasks; and also enables noninvasive measurement of human brain function in cirrhotic patients with decreased consciousness. When combined with traditional structural imaging techniques, the rs-fMRI is a powerful tool for detecting neural activation and functional connectivity in local and global brain regions in vivo. The present study’s results might contribute to future methods which facilitate the early detection, or the monitoring of progression, of HE in clinical practice.
The present study’s results indicate that the resting-state network topology of the brain relates to the grade of HE. Focal or diffuse lesions might cause decreased network efficiency in patients with severe HE, as well as alterations in the whole brain functional neuronal organization. These findings further elucidate changes occurring in the functional architecture of the human brain in patients with liver cirrhosis and HE.
(DOC)