Non-invasive assessment of liver fibrosis in autoimmune hepatitis: Diagnostic value of liver magnetic resonance parametric mapping including extracellular volume fraction

All imaging was performed on a clinical whole-body 1.5 T MRI system (Ingenia, Philips Healthcare) equipped with 32-channel abdominal coil with digital interface for signal reception. Besides morphological sequences, patients underwent MRE and parametric mapping of the liver.

Liver MRE was implemented by 2D gradient-recalled echo to acquire liver elasticity maps with motion-encoding gradients (MEGs). Sequence parameters were as follows: time of repetition (TR) 50 ms, time of echo (TE) 20 ms, flip angle (FA) 20°, parallel imaging factor 2.3, active driver frequency 60 Hz, active driver power 100%, acquired voxel size 1.5 × 4.74 × 10 mm, reconstructed voxel size 1.17 × 1.17 x 10 mm, scan duration/breath hold 15.3 s, and 3 slices. The system configuration was based on a pneumatically powered active wave driver and a tube-connected and strap-secured passive driver placed over the right liver lobe. Generated shear waves at a fixed vibration frequency were coursing through the liver and created tissue displacements, which could be detected to generate magnitude and phase images. Phase shift of magnetic resonance signal was measured at four different phase offsets over one cycle of motion. Further analysis by integrated software (MR elastography View, Philips Healthcare) allowed the creation of a quantitative elastogram (liver stiffness map). For hepatic T1 mapping a heart rate-independent 10-(2)-7-(2)-5-(2)-3-(2), modified Look-locker inversion recovery (MOLLI) acquisition scheme [29] with internal triggering was implemented. The following technical parameters were applied: TR 1.92 ms, TE 0.84 ms, FA 20°, parallel imaging factor 2, acquired voxel size 1.98 × 2.45 × 10 mm, reconstructed voxel size 1.13 × 1.13 × 10 mm, and scan duration/breath hold 14.0 s. Post-contrast T1 maps using the same imaging technique were performed 10 min after contrast injection in the same positions as pre-contrast examinations. For contrast enhancement, the extracellular contrast agent Gadobutrol (1.0 mmol/ml solution with 0.1 mmol per kilogram of body weight, Gadovist, Bayer Healthcare Pharmaceuticals) was injected as a bolus at a rate of 1.5 ml/s and followed by a 10 ml saline flush. For hepatic T2 mapping, a six-echo gradient spin echo sequence (GraSE) was used [30], and scan parameters: TR 450 ms, inter-echo spacing 16 ms, FA 90°, parallel imaging factor 2.5, acquired voxel size 1.98 × 2.01 × 10 mm, reconstructed voxel size 0.88 × 0.88 × 10 mm, scan duration/breath h hold 15/3 × 5 s. T1 and T2 mapping were performed in transversal views covering liver parenchyma at the level of the portal bifurcation. T1 and T2 relaxation maps were reconstructed at the scanner console. Maps of proton density fat fraction (PDFF) and T2* were achieved with a six-echo 3D gradient-echo sequence (mDixon Quant, Philips Healthcare).The following parameters were applied: TR 7.8 ms, TE 1.1 ms, FA 5°, parallel imaging factor 2, acquired voxel size 1.99 × 1.99 × 6 mm, reconstructed voxel size 0.99 × 0.99 × 3 mm, scan duration/breath hold 15.0 s.

Image analyses were performed by an experienced board-certified radiologist (J.A.L, 8 years of experience in abdominal MRI), blinded for the clinical information. For the assessment of T1 and T2 relaxation times and PDFF, three representative round regions of interest (ROIs) (minimum of one cm²) were drawn centrally in three hepatic segments (segments 2, 4a, and 7), and mean relaxation times were calculated [31]. T1 values of the blood pool were obtained from the abdominal aorta on the transversal maps. ECV values were normalized for hematocrit and calculated with regions of interest from pre- and post-contrast T1 values using the following equation [32]: ECV = (1 − hematocrit)*(1/T1 parenchyma post-contrast − 1/T1 parenchyma pre-contrast)/(1/T1 aortic post-contrast − 1/T1 aortic pre-contrast). Blood hematocrit levels were determined on the day of examination. Liver tissue stiffness values were derived from stiffness confidence map by drawing the largest possible freehand ROIs (minimum of one cm²) in three different representative regions of the liver. All patients were divided into two groups, without (< fibrosis stage (F) 2) and with significant fibrosis (≥ F2) according to the MRE-based liver stiffness. According to the literature, a cut-off of 3.66 kPa was chosen to differentiate between patients without and with significant liver fibrosis. The cut-off values for F2, F3, and F4 were 3.66, 4.11, and 4.71 kPa, respectively[4, 33].

Statistical analysis was performed using SPSS Statistics (Version 25, IBM) and MedCalc (Version 19.1.3, MedCalc Software). Patient characteristics are presented as mean ± standard deviation or as absolute frequency. Continuous variables between two groups were compared using Student t test. Dichotomous variables were compared using the χ2 test (with the cell count greater than five) and Fisher test (with a cell count less than or equal to five). The bivariate Pearson correlation coefficient (r) was used for a correlation analyses. Receiver operating characteristic analysis (ROC) was performed to calculate areas under the curve (AUC). AUCs were compared using the method proposed by DeLong et al. [34]. Sensitivity, specificity, and accuracy were calculated. MRE-based liver stiffness was the reference standard against which the diagnostic performance of MRI-derived mapping parameters of liver was tested. A P value < 0.05 was considered statistically significant.

A total of 27 patients with AIH diagnosis were included in this study. 9/27 (33.3%) patients had an overlap syndrome. At the time of MRI examination, 13/27 (48.1%) patients received immunosuppressive therapy with budesonide alone; 8/27 (29.6%) patients received a combination of budesonide with azathioprine; and 3/27 (11.1%) patients received immunosuppressive therapy with azathioprine alone. There were also 3/27 (11.1%) patients, who received no therapy at the time of MRI examination. 13/27 (48.2%) patients had no or not significant (< F2), and 14/27 (51.8%) had significant (≥ F2) fibrosis. 1/14 (7.1%), 4/14 (28.6%), and 9/14 (64.3%) patients had fibrosis stages F2, F3, and F4, respectively. The mean age of patients with no significant fibrosis according to the MRE was 46.6 ± 18.6 years (range: 20–74 years), with significant fibrosis 42.6 ± 18.6 years (range: 19–77 years). The mean body mass index (BMI) in patients with no significant fibrosis was 26.9 ± 4.3 kg/m², in patients with significant fibrosis 23.5 ± 2.9 kg/m² (P = 0.047). Age and sex did not differ in both groups (P > 0.05). No significant differences were found between clinical fibrosis scores APRI and AST/ALT ratio in both groups (P > 0.05). FIB-4 was higher in the group with significant fibrosis (≥ F2) compared to the group with no significant fibrosis (< F2) (P = 0.006). Clinical characteristics are summarized in Table 1.

Compared to patients with AIH and no significant fibrosis (< F2), patients with significant fibrosis (≥ F2) had markedly increased hepatic native T1 relaxations times (548.8 ± 40.7 ms vs. 620.3 ± 66.3 ms; P = 0.003) and hepatic ECV values (27.1 ± 3.2% vs. 38.7 ± 18.9%; P = 0.039). There were no significant differences in hepatic T2 relaxation times between both groups (50.7 ± 4.2 ms vs. 50.5 ± 6.5 ms; P = 0.920). Also, no significant difference in fat fraction was present in both groups (5.3 ± 4.6% vs. 3.2 ± 1.5%, P = 0.135). MRE-based liver stiffness and hepatic parametric MRI results are given in Table 2. Furthermore, we found a strong correlation between MRE-based liver stiffness and hepatic native T1 (r = 0.69, P < 0.001) as well as hepatic ECV (r = 0.80, P < 0.001, see also Fig. 1). There were also significant correlations between clinical fibrosis scores such as FIB-4 and APRI and hepatic native T1 (for both scores, r = 0.49, P < 0.05). Also, hepatic ECV showed a significant correlation with FIB-4 score (r = 0.39, P = 0.04). We found no correlations between hepatic T2 and MRE-based liver stiffness as well as clinical fibrosis scores (FIB-4 and AST/ALT Ratio). A correlation matrix is given in Table 3. Representative images from patients with and without significant fibrosis are given in Fig. 2.

Several parametric mapping parameters were evaluated regarding the diagnostic performance to diagnose significant fibrosis (≥ F2). Regarding the overall diagnostic performance, hepatic ECV revealed the highest diagnostic performance with an AUC of 0.885, a sensitivity of 85.7%, and a specificity of 84.6% (cut-off value: > 29.5%). There were no significant differences in diagnostic performance of hepatic ECV and clinical fibrosis scores for diagnosing significant fibrosis: APRI (P = 0.702), FIB-4 (P = 0.138), and AST/ALR ratio (de-Ritis) (P = 0.058). Hepatic native T1 showed also a high diagnostic performance with an AUC of 0.846, a sensitivity of 85.7%, and a specificity of 76.9% to diagnose significant fibrosis (cut-off value: > 565 ms). There were no significant differences in diagnostic performance of hepatic ECV and native T1 (P = 0.550). Diagnostic performance of hepatic native T1 also differs not significant compared to the clinical fibrosis scores: APRI (P = 0.956) and FIB-4 (P = 0.346), AAR (P = 0.123). Diagnostic performance of hepatic T2 (AUC: 0.566) was significantly lower than that of hepatic native T1 (P = 0.006) and ECV (P = 0.004). Parameters of diagnostic performance for all other evaluated parameters with sensitivities, specificities, accuracies, positive and negative predictive values are given in Table 4. A ROC curves graph for diagnosis of significant fibrosis is given in Fig. 3.