Whole brain atlas-based diffusion kurtosis imaging parameters for evaluation of minimal hepatic encephalopathy

Abstract

Background and purposes

Minimal hepatic encephalopathy (MHE) has no recognizable clinical symptoms, but patients have cognitive and psychomotor deficits. Hyperammonemia along with neuroinflammation lead to microstructural changes in cerebral parenchyma. Changes at conventional imaging are detected usually at the overt clinical stage, but microstructural alterations by advanced magnetic resonance imaging techniques can be detected at an early stage.

Materials and methods

Whole brain diffusion kurtosis imaging (DKI) data acquired at 3T was analyzed to investigate microstructural parenchymal changes in 15 patients with MHE and compared with 15 age- and sex-matched controls. DKI parametric maps, namely kurtosis fractional anisotropy (kFA), mean kurtosis (MK), axial kurtosis (AK) and radial kurtosis (RK), were evaluated at 64 white matter (WM) and gray matter (GM) regions of interest (ROIs) in the whole brain and correlated with the psychometric hepatic encephalopathy score (PHES).

Results

The MHE group showed a decrease in kFA and AK across the whole brain, whereas MK and RK decreased in WM ROIs but increased in several cortical and deep GM ROIs. These alterations were consistent with brain regions involved in cognitive function. Significant moderate to strong correlations (–0.52 to –0.66; 0.56) between RK, MK and kFA kurtosis metrics and PHES were observed.

Conclusion

DKI parameters show extensive microstructural brain abnormalities in MHE with minor correlation between the severity of tissue damage and psychometric scores.

Introduction

Minimal hepatic encephalopathy (MHE), a potentially reversible “subclinical” form, is seen in up to 80% of patients with cirrhosis.^1–3 MHE is the earliest form of hepatic encephalopathy, defined by a psychometric hepatic encephalopathy score (PHES) of less than –5. MHE has a severe detrimental impact on quality of life of patients by causing cognitive dysfunctions in early stages. ⁴ Patients with MHE have lack of attention, visuomotor incoordination, frequent falls, psychometric slowing and loss of mental skills, and show early changes in memory and learning. MHE is difficult to diagnose as there is often a lack of insight amongst patients and caregivers about cerebral changes from liver failure. Patients with late and undiagnosed MHE may convert to clinically manifest hepatic encephalopathy over time, which leads to an irreversible stage resulting in increased morbidity and mortality. For early diagnosis, many tests are employed (Table 1); however, consistency and reproducibility of a test is essential before it can be used in a clinical scenario. Psychometric tests are often used in a clinical setting but are time consuming and influenced by age, gender and education level of patients. Moreover, the application of these clinical tests depends on local expertise and resources. A learning effect may occur on repeated usage of these tests and therefore more objective and repeatable tests need to be formulated. Brain imaging methods such as morphometry, spectroscopy, perfusion and functional magnetic resonance imaging (MRI) have shown promising surrogate markers in diagnosing MHE and good correlations with clinical psychometric scores. Diffusion tensor imaging (DTI) has emerged as a powerful and widely used technique to measure microstructural changes in the brain for a variety of diseases. Diffusion kurtosis imaging (DKI) is a more improved method that addresses shortcomings of DTI and arrives at more detailed measures of tissue microstructure. ⁵ In this study, we acquired DKI data on the whole brain to investigate parenchymal microstructural abnormalities throughout the cerebrum and cerebellum and evaluated correlations of DKI measures with psychometric scores in MHE patients.

Table 1.

Tests for hepatic encephalopathy.

Clinical tests for grading of hepatic encephalopathy
1. West Haven criteria (WHC) 2. Hepatic encephalopathy scoring algorithm (HESA) 3. Modified orientation log (MO-Log) 4. Clinical hepatic encephalopathy staging scale (CHESS)5. Glasgow coma scale (GCS)
Neuropsychometric tests
A. Paper and pencil	1. Psychometric hepatic encephalopathy score (PHES)2. Number connection test (NCT) A/B3. Digit symbol test (DST)4. Repeatable battery for assessment of neuropsychological status (RBANS)
B. Computerized	1. Inhibitory control test2. EncephalApp Stroop test3. Scan test
C. Neurophysiological	1. Electroencephalography2. Evoked potentials3. Critical flicker frequency
Imaging techniques
1. Conventional MRI2. Volumetric MRI3. MR spectroscopy4. Functional MRI including perfusion5. Magnetization transfer imaging6. Diffusion-weighted imaging 7. Diffusion tensor imaging8. Diffusion kurtosis imaging
Nuclear medicine imaging techniques
1. Positron emission tomography2. Single photon emission computed tomography

MR: magnetic resonance; MRI magnetic resonance imaging.

Methods

Study design, enrollment and patient selection

This was an institutional ethics committee approved, single center prospective study. Between July 2018 and November 2019, 15 cirrhotic patients with MHE (PHES ≤ –5) and 15 age- and sex-matched healthy controls underwent MRI. Complete demographic information is given in Table 2. Patients with overt hepatic encephalopathy, history of upper gastrointestinal bleed, spontaneous bacterial peritonitis, known neuropsychiatric disease or history of consumption of psychoactive drugs/antibiotic usage were excluded. Patients with any malignancy or history of previous transjugular intrahepatic portosystemic shunt or shunt surgery were also excluded from the study.Consecutive MHE patients who fulfilled the above inclusion and exclusion criteria were enrolled into the study after informed consent was obtained.

Table 2.

Subject demographics.

	Control	MHE patients	p-value
Age (mean ± SD)	51.73 ± 4.17	51.33 ± 9.11	0.88
Sex	11 M / 4 F	11 M / 4 F
Etiology
Alcohol	NA	7 (46.7%)
HBV	NA	5 (33.3%)
HCV	NA	3 (20.0%)

F: female; HBV: hepatitis B virus; HCV: hepatitis C virus; M: male; MHE: minimal hepatic encephalopathy; NA: not applicable; SD: standard deviation.

Neuropsychometric testing was performed using a PHES battery for both patients and controls and the battery comprised five tests: number connection test (NCT) A, digit symbol test (DST), finger connection test (FCT) A (Indian version replacing NCT B ⁶ ), serial dotting test and line drawing test. Test values for controls was obtained in a randomized way for time (t) and error (e) and PHES for patients with cirrhosis were expressed as z scores (i.e. standard deviation between observed and expected scores for given age and education).

Imaging protocol

All subjects were scanned using a 3.0 Tesla MRI scanner (Siemens Healthcare, Erlangen, Germany) with a 32-channel head coil. DKI data were acquired using a single-shot twice-refocussed spin-echo imaging sequence with TR/TE 11500/98 ms, 3 b-values (0, 1000, 2000 s/mm²) along 30 diffusion encoding directions, field-of-view 222 mm × 222 mm, acquisition matrix 74 × 100, slice thickness of 3 mm with no gaps, 54 slices, in-plane resolution of 3 × 3 mm, and an acquisition time of 7 min. Data from conventional T1-weighted, T2-weighted and fluid-attenuated inversion recovery sequences were also obtained to rule out other structural abnormalities.

Data processing and image analysis

Diffusion-weighted imaging (DWI) data was processed on a 64-bit workstation using Diffusional Kurtosis Estimator (DKE) software. ⁷ Pre-processing steps embedded within DKE correct the images for eddy current distortions using a 12-parameter affine transformation. Automated skull stripping was also performed on the b0 image using BET skull-strip tool from the MRIcro software. DKE then calculates the kurtosis metrics comprising mean kurtosis (MK), axial kurtosis (AK), radial kurtosis (RK) and kurtosis fractional anisotropy (kFA), as shown in Figure 1. These metrics were evaluated at specific regions of interest (ROIs) in the brain using the JHU-MNI-SS type 2 atlas which contains 189 extensively segmented and labeled ROIs drawn on the “Eve” brain template in MNI space. The atlas was registered from the template to each subject space using a non-linear large deformation diffeomorphic metric mapping transformation with dual contrast (fractional anisotropy (FA) and b0). ⁸ This atlas-based approach has several advantages over voxel-based or white matter skeleton-based (e.g. tract-based spatial statistics ⁹ ) analytical methods by capturing the mean value from all the voxels in an ROI, improving measurement accuracy and increasing the effect size to detect changes. ¹⁰ Finally, the number of ROIs is reduced by removing irrelevant regions (e.g. cerebrospinal fluid), merging smaller subdivisions into single macro ROIs, and combining left and right brain ROIs. This process reduced the number of ROIs to 64 (Table S1 in the supplementary file). The Indian version of the PHES was used in which the NCT B was replaced with the FCT A. ⁶

Figure 1.

Diffusion kurtosis imaging parametric maps showing kurtosis fractional anisotropy (kFA), mean kurtosis (MK), axial kurtosis (AK) and radial kurtosis (RK) obtained from an axial slice of the brain of a minimal hepatic encephalopathy subject.

Statistical analysis

All statistical tests were implemented with R software. Diffusion parameters were first checked for normality by the Shapiro–Wilk method. Data outliers were also removed using standard z score cut-off values (z < –3; z > 3). For each ROI, group-wise comparisons between controls and MHE were performed using Student’s t-test. Finally, a Pearson test was used to find correlations between the PHES and DKI parameters. Results were considered as significant for a p < 0.05, corrected for multiple comparisons using the false discovery rate method.

Results

Patient characteristics

A total of 15 cases and 15 controls were age and sex matched, with mean age of 51.5 ± 6.6 years and 11 male (73.3%) subjects in each group. Etiology of cirrhosis was alcohol related in 7 patients, hepatitis B in 5 patients and hepatitis C in 3 patients. A PHES of –6 was seen in 9 patients, –7 in 3 patients, –10 in 2 patients and –9 in 1 patient.

DKI parameters

Consistent decrease in kFA was recorded in the MHE group throughout the brain in both white matter (WM) and gray matter (GM) regions (Figure 2) with significant decreases in 15 ROIs (Table S2 in the supplementary file). AK showed a trend of decrease in multiple regions of the brain (Figure 3), with 14 ROIs showing significant change (Table S3 in the supplementary file). Changes in MK values (Figure 4) and RK values (Figure 5) were less impactful. MK showed an overall decreasing trend in the WM with significant changes in 4 ROIs but no significant changes in cortical and deep GM areas (Table S4 in the supplementary file). RK values were mostly increasing in cortical and deep GM but decreasing in WM with significant changes in 6 ROIs (Table S5 in the supplementary file). Those regions in the brain with significant between-group differences are shown in Figure 6 on coronal MRI reformatted slices. The PHES showed a positive correlation with kFA values in the superior longitudinal fasciculus (r = 0.559, p = 0.03) and negative correlations with MK in the lateral fronto-orbital gyrus (r = –0.562, p = 0.029) and with RK in the lateral fronto-orbital gyrus (r = –0.66, p = 0.007), caudate (r = –0.524, p = 0.045) and thalamus (r = –0.515, p = 0.049). Figure 7 shows the above correlations with PHES. The DKI parameters of the remaining ROIs did not show significant correlations with the PHES.

Figure 2.

Bar plot showing group differences in kurtosis fractional anisotropy (kFA) between the control and minimal hepatic encephalopathy (MHE) groups at 64 brain regions of interest (ROIs). The bottom plot shows deep white matter (WM) ROIs and the top plot shows ROIs in the cortical areas, deep gray matter (GM) and the cerebellum. Asterisks indicate ROIs with a significant between-group difference (p < 0.05, false discovery rate (FDR)). Expansions for the abbreviations used for indicating the brain anatomical ROIs are provided in Table S1 (in the supplementary file).

Figure 3.

Bar plot showing group differences in axial kurtosis (AK) between the control and minimal hepatic encephalopathy (MHE) groups at 64 brain regions of interest (ROIs). The bottom plot shows deep white matter (WM) ROIs and the top plot shows ROIs in the cortical areas, deep gray matter (GM) and the cerebellum. Asterisks indicate ROIs with a significant between-group difference (p < 0.05, false discovery rate (FDR)). Expansions for the abbreviations used for indicating the brain anatomical ROIs are provided in Table S1 (in the supplementary file).

Figure 4.

Bar plot showing group differences in mean kurtosis (MK) between the control and minimal hepatic encephalopathy (MHE) groups at 64 brain regions of interest (ROIs). The bottom plot shows deep white matter (WM) ROIs and the top plot shows ROIs in the cortical areas, deep gray matter (GM) and the cerebellum. Asterisks indicate ROIs with a significant between-group difference (p < 0.05, false discovery rate (FDR)). Expansions for the abbreviations used for indicating the brain anatomical ROIs are provided in Table S1 (in the supplementary file).

Figure 5.

Bar plot showing group differences in radial kurtosis (RK) between the control and minimal hepatic encephalopathy (MHE) groups at 64 brain regions of interest (ROIs). The bottom plot shows deep white matter (WM) ROIs and the top plot shows ROIs in the cortical areas, deep gray matter (GM) and the cerebellum. Asterisks indicate ROIs with a significant between-group difference (p < 0.05, false discovery rate (FDR)). Expansions for the abbreviations used for indicating the brain anatomical ROIs are provided in Table S1 (in the supplementary file).

Figure 6.

Coronal slices of T1 template image in MNI space showing the location of all the ROIs where significant structural changes in MHE patients were seen using each of the 4 DKI metrics: kurtosis fractional anisotropy (kFA) top left, axial kurtosis (AK) bottom left, radial kurtosis (RK) top right, and mean kurtosis (MK) bottom right. With kFA and AK all the regions of interest (ROIs) highlighted throughout the brain had a significant decrease. White matter ROIs highlighted for MK and RK also had a significant decrease, whereas gray matter ROIs (i.e. putamen (Put) and Cuneus gyrus (Cu)) had an increase in RK. Expansions for the abbreviations used for indicating the brain anatomical ROIs are provided in Table S1 (in the supplementary file).

Figure 7.

Scatter plots showing Pearson correlations between psychometric hepatic encephalopathy score (PHES) and diffusion kurtosis imaging metrics. Psychometric scores had positive correlation with kurtosis fractional anisotropy (kFA) in the superior longitudinal fasciculus (r = 0.559, p = 0.03), negative correlation with mean kurtosis (MK) in the lateral fronto-orbital gyrus (r = –0.562, p = 0.029), and negative correlation with radial kurtosis (RK) in the lateral fronto-orbital gyrus (r = –0.66, p = 0.007), caudate (r = –0.524, p = 0.045), and thalamus (r = –0.515, p = 0.049).

Discussion

Hepatic encephalopathy is a devastating disease with morbid consequences. PHES is the commonly used clinical test for identification of MHE that is performed in an outpatient setting within 20–30 min; however, it is not specific for MHE due to an increase in false positives in testing patients with a low pre-test likelihood of the disease. Objective quantification of tissue structural, metabolic and physiological alterations due to MHE in involved brain anatomical regions using advanced brain imaging methods may help us to solve the diagnostic conundrum. Various mechanisms described for cerebral changes in hepatic encephalopathy include an osmotic effect ¹¹ due to ammonia after crossing the blood–brain barrier directly along with hyponatremia, dysregulation of amino acid metabolism, changes in extracellular concentration of neurotransmitters ¹² and free radical damage from pro-inflammatory state. MRI techniques have been used to evaluate the structural and functional changes in hepatic encephalopathy. Besides conventional MRI, magnetization transfer imaging has shown a decrease in magnetization transfer ratio from increased cerebral water content in MHE patients. ¹³ Failure of autoregulation and its resultant changes in cerebral perfusion have been reproduced by magnetic resonance perfusion techniques ¹⁴ with studies showing an increase in putaminal blood flow. ¹⁵ T2*-based and susceptibility-weighted imaging have shown abnormal iron deposition in the frontal-basal ganglia-thalamocortical circuit which has been linked with neurocognitive decline in these patients. ¹⁶ Metabolic alterations from conversion of cerebral ammonia (glutathione synthase activity of astrocytes) produce a characteristic triad of increased glutamine, decreased myo-inositol (usually the earliest and most consistent finding) and decreased choline on magnetic resonance spectroscopy. ¹³ Advanced imaging techniques such as blood-oxygenation-level-dependent functional MRI have also been used to assess functional changes in spatial working memory. Involvement of bilateral prefrontal cortices, bilateral premotor areas, supplementary motor areas and bilateral parietal areas in MHE has been documented previously. ¹⁷ Resting-state functional MRI studies have also shown weaker functional connectivity of bilateral putamen, pallidum and thalami (subcortical areas). ¹⁸

Diffusion MRI techniques assess the increased cerebral water content which alters the diffusion properties of water molecules such as decreased FA and increased mean diffusivity and axial diffusivity on DTI studies in MHE patients. ¹³ A counter-regulatory volume effect occurs with a resultant increase in extracellular water to combat the astrocytic swelling. Because biological tissues are composed of both hindered and restricted components, a more advanced diffusion model than DWI and DTI (i.e. kurtosis imaging) is essential. DKI assesses the degree of non-Gaussian diffusion with higher order description of the water diffusion process in the cerebral microstructure. The water molecules in the brain microstructures show non-monoexponential b-value dependency at higher b-values (b > 1000 s/mm²), ¹⁹ which can be completely characterized by a 2nd order three-dimensional (3D) diffusivity tensor (as in conventional DTI) along with a 4th order 3D kurtosis tensor used in DKI. DKI has shown promising results in a variety of central nervous system pathologies (e.g. Parkinson’s disease ²⁰ and Alzheimer’s dementia ²¹ ). Therefore, we used DKI in this study to investigate microstructural alterations in the brain of patients with MHE.

The most commonly used kurtosis metrics are AK, MK, RK and kFA. AK is the directional kurtosis parameter parallel to the principal diffusion direction and reflects axonal integrity. MK is the overall average apparent diffusion kurtosis coefficient and reflects the microstructural complexity of brain tissues. RK is the average apparent kurtosis coefficient measured in equatorial plane. kFA is a measure of anisotropy (similar to conventional FA) but uses the kurtosis tensor for calculation, making it less prone to errors from complex WM fiber arrangements such as crossing fibers. In our study, there was a mean decrease in kFA and AK in the whole brain, an increase in GM MK and RK, and a decrease in WM MK and RK in MHE group versus control group in most ROIs. The most affected WM areas were the corpus callosum, cingulum, cerebral peduncle, corticospinal tract, corona radiata, internal capsule and fronto-occipital fasciculus. The decrease in kFA in those regions shows clear signs of WM axonal damage, suggesting definite involvement of these areas in MHE. In cortical regions and in deep GM we see most often the significant alterations in the temporal gyrus, amygdala, hippocampus, globus pallidus and putamen, all showing a consistent decrease in kFA and AK. The corpus callosum is one of the major interhemispheric connections and a crucial component of the cognitive nervous pathway. Impairment in corpus callosum affects the interhemispheric functional interactions. Similarly, involvement of superior longitudinal and inferior fronto-occipital fasciculi is commonly associated with cognitive dysfunction and executive, visuospatial-visuoconstructive impairments. ²² , ²³ These findings elucidate the role of these important structures in neurocognitive decline in MHE.

The reduction in AK indicates impairment of the axon function in WM. Li et al. ²⁴ have shown reduced AK in the anterior thalamic radiation, decreased MK in left thalamus and body of corpus callosum and reduced RK in the body and splenium of corpus callosum in MHE patients. Sun et al. ²⁵ showed reduced AK in the left cingulum, right caudate nucleus, hippocampus, inferior temporal gyrus, fusiform gyrus and right thalamus. Chen et al. ²⁶ reported reductions of all kurtosis metrics connoting decrease in microstructural complexity in GM and WM in MHE patients. However, in our study, some structures showed an opposite pattern of change in MK and RK with predominant increase in mean values from most GM ROIs. DKI-derived metrics change in any disease with time. Whether these changes are attributable to alterations in cortical synaptic plasticity in early stages of hepatic encephalopathy ²⁷ or from difference in neurotransmitter effects in different brain areas or from higher reactive astrogliosis ²⁸ is debatable and a matter of further research. Longitudinal follow up to document further changes with pattern of normalization followed by decline in MK and RK values will confirm the DKI metric, one of the major limitations of our study.

Correlation of DKI metrics with clinical PHES showed positive correlation with kFA in the superior longitudinal fasciculus, suggesting that severe PHES (lower scores) correlate with more severe axonal damage (lower kFA). We also note negative correlations with MK in the lateral fronto-orbital gyrus, and negative correlation with RK in the lateral fronto-orbital gyrus, caudate and thalamus. In the study by Chen et al., ²⁶ the average AK, MK and RK values in GM and WM were correlated with the PHES rather than individual ROIs.

An interesting finding in our study which incites further research is the possibility of DKI to assess derangements in cerebral microarchitecture to the level of aquaporin-4 channels, ²⁹ which regulate blood brain barrier permeability. We found significant DKI parametric changes in the ventricles in our study on MHE patients which was not present in controls. This has also been reported previously in patients with Alzheimer’s dementia ³⁰ and sheds light on the potential role of DKI in assessing, quantifying and understanding pathophysiology of cerebral edema in MHE patients.

Our study has a few limitations. First, due to the small sample size, there is a limit to the statistical power of our study for group comparisons, and to our correlation analysis as few MHE patients have severely low psychometric scores (PHES < –8). Second, we did not evaluate the effects of different etiologies of cirrhosis on DKI metrics and cerebral microstructure, as the final pathway for hepatic encephalopathy is nearly similar in all etiologies. As previously discussed, considering the disease progression of MHE, a longitudinal study is necessary to further explore the pathogenic mechanisms and neuroimaging features. Further, cirrhotic patients without MHE also present with brain structural changes. However, we ruled out the involvement of these microstructural changes by elaborating the new areas which are involved in MHE by a careful review of literature. The areas of involvement in our study did not corroborate with those due to cirrhosis alone and thus represent true involved areas in MHE.

In conclusion, the practice to perform neuropsychological functions for evaluation of MHE gives little information about the involvement of brain structures and its severity. This study elucidates the utility of DKI, a newer technique to show the microstructural involvement in MHE. The correlation of DKI and neurocognitive abnormalities may provide early and more accurate localization of the brain involvement and may help in predicting prognosis and neuropsychological sequelae.