Introduction

Liver fibrosis is the result of chronic liver diseases, and it can progress to cirrhosis and hepatocellular carcinoma¹. In recent years, several MRI techniques has been proposed to non-invasively characterize and stage liver fibrosis, which overcome some limitations of liver biopsy, the current golden standard. MRE offers a new sight to depict tissue mechanical properties, which could be an alternative to liver biopsy because it reduces the risk of complications and sampling errors. However, the use of external mechanical vibration equipments increases the time and money costs. DWI, as another non-invasive MRI approach, offers water diffusivity information of liver at a microscopic level. We hypothesize that water diffusivity was strongly correlated with mechanical properties in liver fibrosis² and investigated the potential of DWI as an alternative to MRE to quantitatively characterize liver mechanical properties without external devices in patients with different types of liver fibrosis - Hepatitis B and Nonalcoholic fatty liver disease (NAFLD).

Methods

This study was approved by the institutional review board of both participating institutions. Between May and October 2020, a total of 73 patients with liver fibrosis were enrolled in this retrospective research (42 Hepatitis B patients, 16 NAFLD patients and 15 other diseases). MRI examinations were performed at a 3T scanner (uMR 780, United Imaging Healthcare, Shanghai, China) with a 32-channel body array. The MRI protocols were as follows: 1) 2D MRE sequence was acquired with the following parameters: TR/TE = 50/20.63ms, flip angle = 30°, slice thickness = 10mm, FOV = 300mm × 300mm, all examinations were carried out with an electromagnetic actuation frequency of 60 Hz³; 2) DWI was acquired based on a single-shot echo-planar imaging (EPI) sequence: TR/TE=4000/78ms, slice thickness = 5mm, FOV = 300mm × 380mm, 11b values were performed but only b = 0, 200, 500, 800, and 1200 s/mm² were considered in this study. Tissue stiffness (μ_MRE) was processed from the original MRE data. Different combinations of b values (200&500, 200&800, 200&1200) were used to calculate the apparent diffusion coefficient (ADC) with the following equation: ADC=ln(S_Lb/S_Hb)/(Hb-Lb) (Equation 1), where Hb and Lb represent high b value and low b value respectively. Both μ_MRE and ADC were processed with corresponding in-house prototype software developed by MATLAB (Mathworks, Natick, Mass). Linear regression and Spearman’s rank correlation coefficient were performed to give an exact expression of ADC and μ_MRE. Comparison of ADC values between Hepatitis B and NAFLD at different stages was evaluated using two samples T-test. ROC analysis was performed to evaluate the diagnostic accuracy of ADC for staging liver fibrosis. Inter-method reliability for diffusion-based elastography and MRE were evaluated by using Bland-Altman plot.

Results

The processing flowchart was shown in Figure 1. Using MRE, the stages of fibrosis in all patients were distributed from F0 to F4. Figure 2(a)-(c) demonstrated that ADC and μ_MRE had a strong linear inverse relationship. ADC_200&500 had the best correlation with μ_MRE (R² = 0.869, p<0.001), the linear regression result was: ADC_200&500=-8.3176x10^-5μ_MRE+1.1x10^-3. By substituting Equation 1 into the above result, stiffness from diffusion-based elastography can be obtained: μ_diff=-40.0767ln(S₂₀₀/S₅₀₀)+13.225. A moderate correlation between ADC_200&800 & ADC_200&1200 and μ_MRE was found. (R² = 0.527 and 0.477, respectively). Between Hepatitis B and NAFLD, no significant difference was found at different stages (Table 1). Figure 2(d) showed the mean ADC values for all 73 patients at different stage of liver fibrosis stages. With the progression of fibrosis, ADC value decreased significantly (p<0.001). The ADC values from F0 to F4 were: 0.923, 0.864, 0.818, 0.796, 0.712 x10^-3 mm²/s, respectively. ROC analysis was displayed in Figure 3, the AUC for distinguishing F4 and F3 or lower was 0.91, and the AUC for distinguishing F2 or lower and F3 & F4 was 0.771. Bland-Altman plot was shown on Figure 4. Inter-method agreement between diffusion-based elastography and MRE was high (mean difference = 0.112±0.3 kPa, p<0.001) and showed no difference.

Discussion & Conclusion

In our study, an exact expression between μ_MRE and ADC was well established, which indicated that tissue stiffness can be obtained from ADC. This inverse relationship between μ_MRE and ADC can be explained that: with the progression of liver fibrosis, tissue stiffness increases due to the proliferation of collagen fibers and cell shrinkage, leading to the restricted diffusion of both extracellular and intracellular water molecules, thus the decreased ADC. In this study, no significant difference was found between Hepatitis B and NAFLD patients. The high inter-method agreement validated the feasibility of diffusion-based elastography for characterizing liver fibrosis. Thus, instead of MRE, tissue stiffness can be characterized from a DWI scan without the need of external devices, which helps reduce the complexity of the examination clinically. In conclusion, this study demonstrated that DWI has the potential to act as an alternative to MRE and can quantitatively characterize liver stiffness in patients with liver fibrosis.

Acknowledgements

Funding support from grant 31870941 from National Natural Science Foundation of China (NSFC) and grant 1944190700 from Shanghai Science and Technology Committee (STCSM) are acknowledged.