Spatial heterogeneity of hepatic fibrosis in primary sclerosing cholangitis vs. viral hepatitis assessed by MR elastography

Spatial heterogeneity of hepatic fibrosis in primary sclerosing cholangitis (PSC) in comparison to viral hepatitis was assessed as a potential new biomarker using MR elastography (MRE). In this proof-of-concept study, we hypothesized a rather increased heterogeneity in PSC and a rather homogeneous distribution in viral hepatitis. Forty-six consecutive subjects (PSC: n = 20, viral hepatitis: n = 26) were prospectively enrolled between July 2014 and April 2017. Subjects underwent multifrequency MRE (1.5 T) using drive frequencies of 35–60 Hz and generating shear-wave speed (SWS in m/s) maps as a surrogate of stiffness. The coefficient of variation (CV in %) was determined to quantify fibrosis heterogeneity. Mean SWS and CV were 1.70 m/s and 21% for PSC, and 1.84 m/s and 18% for viral hepatitis. Fibrosis heterogeneity was significantly increased for PSC (P = 0.04) while no difference was found for SWS of PSC and viral hepatitis (P = 0.17). Global hepatic stiffness was similar in PSC and viral hepatitis groups, but spatial heterogeneity may reveal spatial patterns of stiffness changes towards enhanced biophysics-based diagnosis by MRI.

www.nature.com/scientificreports/ MRE towards regional assessment of mechanical tissue properties 21 . Parameters are shear-wave speed (SWS in m/s), which is determined as a surrogate marker of stiffness, and the phase angle of the complex modulus (φ in rad) which is related to the tissue's viscous properties and is a surrogate of fluidity, where φ = 0 rad indicates pure solids and φ = π/2 rad pure fluids 22,23 . Aim. Based on reported visual impressions of experienced abdominal radiologists and pathologists, we hypothesize that fibrosis is heterogeneous in PSC and homogeneous in viral hepatitis 1,2,9 . To the best of our knowledge, this hypothesis has never been confirmed quantitatively by measuring viscoelastic tissue properties. Moreover, PSC and viral hepatitis can easily be distinguished using routine clinical testing and are thus well suited for comparison in this explorative study. Therefore, we conducted a study using MRE to evaluate the heterogeneity of hepatic fibrosis as a potential new quantitative biomarker.

Subjects. The study was approved by the institutional review board of the Charité-Universitätsmedizin
Berlin and was conducted in accordance with relevant guidelines and regulations after obtaining oral and written patient informed consent. In this prospective single center study, we enrolled a total of 46 consecutive subjects between July 2014 and April 2017: 20 patients with PSC (mean age, 41.4 ± 14.1 years; 5 women) and 26 patients with viral hepatitis (mean age, 50.3 ± 14.5 years; 8 women). Inclusion criteria were multifocal bile duct strictures and segmental dilatations on magnetic resonance cholangiopancreatography in patients with elevated serum markers of cholestasis (alkaline phosphatase, γ-glutamyltransferase) for PSC, confirmation by routine laboratory blood testing for chronic viral hepatitis (e.g., surface antigen of the hepatitis B virus, hepatitis B core antigen, antibodies to the hepatitis C virus, and hepatitis C RNA), and age ≥ 18 years ( Fig. 1) 13,24,25 . Exclusion criteria were other concurrent liver or biliary diseases, untreated dominant strictures (common bile duct ≤ 1.5 mm; left or right hepatic duct ≤ 1 mm) to avoid potential bias due to increased biliary pressure 19,26 , history of liver transplantation and general contraindications for magnetic resonance imaging. Additionally, as liver biopsy is not suited to stage PSC-related hepatic fibrosis, two biological scores were used to further characterize subjects: the aspartate aminotransferase-to-platelet ratio index (APRI) and the Mayo risk score (MRS). APRI is a surrogate marker to assess the severity of disease in relation to the stage of hepatic fibrosis and is based on routinely available laboratory test results 27  Magnetic resonance elastography. Patients underwent multifrequency MRE with tomoelastography postprocessing at 1.5 T (Magnetom Aera, Siemens Healthineers, Germany) with an 18-channel phased-array coil in combination with the spine-array coil. The multifrequency MRE setup and components used in this study are described in detail in Hudert et al. 22 and Shahryari et al. 23 . Briefly, subjects were placed in supine position and a custom-designed piezoelectric driver was positioned at the level of the xiphoid process. Mechanical vibrations at drive frequencies of 35, 40, 45, 50, 55 and 60 Hz were sequentially applied, and image acquisition was performed for a total of 4:30 min with patients breathing freely 29 . MRE imaging parameters were as follows: 9 axial www.nature.com/scientificreports/ slices, 8 offsets, full wave field, 300 × 234 mm 2 field of view (fixed), 100 × 78 matrix size, 3 × 3 × 5 mm 3 resolution, 50 Hz motion-encoding gradient frequency, 30 mT/m motion-encoding gradient amplitude, TR 1610 ms, TE 54 ms, GRAPPA factor 2, and 2 averages. Furthermore, the scan protocol included the following conventional sequences: axial T1-weighted dual gradient-echo sequence for liver fat quantification 30 , axial and coronal T2-weighted half-Fourier acquisition single-shot turbo spin-echo sequence, and, for PSC patients, magnetic resonance cholangiopancreatography using paracoronal T2-weighted single-shot turbo spin-echo sequences in thin-slice multisection (3 mm) and thick-slab (40-60 mm) acquisition.
Data processing. Full-field-of-view elastograms-stiffness and fluidity maps-were computed based on 3D compound multifrequency processing 21 . The postprocessing pipeline is publicly available at https:// bioqicapps. chari te. de. Each elastogram slice was generated by compounding 216 images of multifrequency MRE raw data using 12 spatiotemporal filter directions, 3 field components, and 6 drive frequencies. Using a systematic approach, volumes of interest were manually drawn by one radiologist who was blinded to clinical parameters (R.R., 9 years of experience in abdominal MRE) by contouring livers on SWS maps. Major blood vessels and regions of insufficient wave excitation were consistently excluded by using a lower SWS threshold of 1 m/s, as previously described 31 . The same volume of interest was used for both SWS and fluidity maps. Fibrosis was staged using cutoff values as previously published: F1 (any fibrosis) ≥ 1.52 m/s, F2 (moderate fibrosis) ≥ 1.55 m/s, F3 (severe fibrosis) ≥ 1.67 m/s, and F4 (cirrhosis) ≥ 1.72 m/s 31 . For quantifying fibrosis heterogeneity, we assessed intrahepatic standard deviations (SD) of SWS and φ. Results were normalized by dividing individual intrahepatic SDs with their corresponding means to account for differences in fibrosis severity between the two groups, yielding the coefficient of variation (CV in %).

Statistical data analysis. Group values were calculated as mean and interindividual SD. A two-sided t-test
and Pearson's correlation coefficient were used to assess MRE parameters and biological scores. AUC analysis with 95% confidence intervals was used to assess diagnostic performance of CV. The level of significance was P < 0.05. Statistical analysis was conducted using Matlab version 9.0 R2016a (The Mathworks, Inc., United States).

Discussion
We conducted a study using MRE to evaluate the spatial heterogeneity of hepatic fibrosis as a potential new biomarker. Our results quantitatively confirm known visual impressions of experienced abdominal radiologists and pathologists 1,2,9 by demonstrating increased stiffness-based heterogeneity in PSC as compared to viral hepatitis where a rather homogenous fibrosis distribution was found. The CV of MRE-based stiffness has the potential to quantify heterogeneity in PSC-related hepatic fibrosis as a new biomarker besides global stiffness, whereas the CV of fluidity was not found to be sensitive. Despite a tendency towards higher mean SWS in the viral hepatitis group, a significantly increased fibrosis heterogeneity was found in the PSC group. However, only a moderate diagnostic performance was found in this explorative study. For PSC and viral hepatitis, an increase in both SD and CV of SWS with the severity of fibrosis was shown. This relationship has been reported previously for SD 31 www.nature.com/scientificreports/ www.nature.com/scientificreports/ as the development of regenerating nodules, synthesis and connection of free collagen branches, and architectural distortions in the focal area of cell damage [34][35][36] . Our results also reproduce the correlation between mean stiffness measured by MRE or ultrasound elastography and APRI reported by earlier investigations 32,35 . It is a stimulating result of our study that, compared to mean SWS values, an even stronger correlation has emerged between APRI and the CV of SWS in the PSC group but not in the viral hepatitis group. This finding might indicate an impact of heterogeneity on the stiffness-based staging of PSC-related fibrosis. However, larger studies are necessary to confirm this preliminary finding and to determine the potential diagnostic benefit for the assessment of disease activity or prognosis of PSC. The lack of correlation with MRS, except for CV of φ, might be attributable to the fact that the MRS is a statistical model for predicting long-term disease outcome rather than characterizing the current status of hepatobiliary disease activity. Another contributing factor might be the low risk profile of our study population based on MRS.
Habibabadi et al. reported initial experience in detecting heterogeneity of hepatic fibrosis 37 . They investigated 128 patients with a wide range of underlying chronic liver diseases and suspected fibrosis using MRE. Heterogeneity of fibrosis was defined as present when the first and the second most predominant fibrosis stages were more than one category apart. Comparing region-of-interest-based and volumetric measurements, they found that global liver stiffness may not represent the entire spectrum of hepatic fibrosis 37 . Jhaveri et al. validated MRE in a group of 67 PSC patients using transient elastography as reference standard. For the differentiation of early and moderate fibrosis from cirrhosis, they found a good diagnostic performance with a sensitivity, specificity and accuracy of 87.5%, 96.1% and 94.0%, respectively 18 . Bookwalter et al. showed a low significant correlation between the stiffness of liver segments and corresponding segmental bile duct strictures in a group of 55 PSC Table 1. Baseline characteristics of subjects. If not otherwise specified, data are means and numbers in parentheses are standard deviations. APRI aspartate aminotransferase-to-platelet ratio index, ALP alkaline phosphatase, γGT gamma-glutamyltransferase, ALT alanine transaminase, AST aspartate transaminase, ULN upper limit of normal.  www.nature.com/scientificreports/ patients (R = 0.18, P < 0.001) 19 . At the same time, they did not find a correlation between stiffness and conventional magnetic resonance cholangiopancreatography findings such as segmental bile duct dilatation, thickening or enhancement. In alignment with our study, no correlation between stiffness and MRS was found. However, published results on the relationship of stiffness and MRS are mixed [18][19][20]38,39 . For instance, Idilman et al. observed a positive correlation between hepatic stiffness and MRS (R = 0.646, P < 0.0001) in a large cohort of 266 patients 39 . Moreover, Tafur et al. found that stiffness had a superior discriminatory ability than morphological intrahepatic stricture severity to distinguish MRS-based risk groups (AUC of 0.779 and 0.718; respectively) 38 . In contrast to our study, they found a moderately positive correlation between stiffness and MRS (R = 0.6; P ≤ 0.001), which may be related to an increased median MRS group value of − 0.07 in their study versus − 0.50 in our study. However, a moderately negative correlation between CV of φ and MRS was found in our study (R = − 0.45; P = 0.05). Corpechot et al. performed a study using transient elastography in 73 PSC patients with biopsy-proven fibrosis 40 . They found a good diagnostic performance of transient elastography with accuracy values for identification of severe fibrosis and cirrhosis of 83% and 88%, respectively 40 . Moreover, they reported the diagnostic performance of transient elastography to be superior to APRI and MRS in patients with significant or severe fibrosis. Of potential interest for the further development of heterogeneity as a promising new biomarker in MRE might be the so-called 1-Norm technique, which is based on a mathematical characterization of shear wave front geometry 41 . It provides a quantitative measure of the magnitude of wave scattering, which can potentially be used as a (in-)homogeneity index of biological tissue. This method was not available for in vivo application at the time of our study but will be implemented for future work. Moreover, also the assessment of heterogeneity of mapping and diffusion parameters should be investigated in the future.
Although encouraging, our study has limitations. First, there is no accurate reference standard available for the evaluation of disease activity in PSC, which in turn motivated this study. Fibrosis staging was based on MRE thresholds that have been established in a cohort of patients with mixed underlying chronic liver diseases 31 . Second, we investigated a small number of patients, especially for fibrosis stages F1-F3, as PSC is a rare disease. However, the preliminary results of this explorative study motivate further investigation of heterogeneity in hepatic fibrosis by elastography, and larger studies to confirm our findings should be performed. Third, some subjects were included in a previous study to investigate the diagnostic performance of MRE for staging hepatic fibrosis (nine patients with PSC and eleven patients with viral hepatitis) 31 . Finally, measuring CV of stiffness might not necessarily mean measuring heterogeneity of fibrosis given the fact that stiffness is an indirect measure of fibrosis which can be confounded by many factors including portal hypertension 42 and inflammation 43 . Nonetheless, this study is a first step towards quantitatively relating fibrosis heterogeneity to PSC.
In conclusion, our results show an increased stiffness-based heterogeneity in PSC as compared to viral hepatitis. Global hepatic stiffness was similar in PSC and viral hepatitis groups, but spatial heterogeneity may reveal spatial patterns of stiffness changes towards enhanced biophysics-based diagnosis by MRI.