As a common process of chronic liver diseases, hepatic fibrosis is characterized by excess deposition of collagenous extracellular matrix (ECM) components which often eventually leads to liver dysfunction and hepatocellular carcinoma [1]. Recently studies show that early-stage liver fibrosis can be reversed by effective treatment, fibrosis and cirrhosis advanced by which are usually irreversible [2, 3]. Percutaneous liver biopsy, as a gold standard for staging hepatic fibrosis, is limited by its invasiveness, morbidity and mortality of the procedure [4-5]. Therefore, alternative non-invasive techniques of the early diagnosis and monitor of liver fibrosis progression are vital for clinical applications. Recently, diffusion time dependency of measured diffusion coefficients, being the evidence of restricted diffusion, has been observed in rat liver models. Moreover, the diffusion parameters with long diffusion times served as a more sensitive biomarker of the pathological alterations during firbogenesis [6]. Based on our previous study on the validation of diffusion time dependency of measured diffusion coefficients on human livers [7], here we tried to examine whether different diffusion times detect the pathological alterations during liver fibrogenesis in human livers at 3 T.

MRI: Nineteen patients with clinically histological-proven liver fibrosis disease and ten healthy volunteers were recruited for the experiment (range of age 35~65yrs, mean age 51.95 yrs., 12F/7M) from Jan, 2015 to July 2015. Local ethics committee approved this study and written informed consent was obtained from each participant. The distribution of patients in different fibrosis stages was: 1(n=4), 2(n=11), 3(n=4). All MRI experiments were performed on a Philips 3T clinical scanner (Philips, Best, The Netherlands). The liver images were acquired with a respiratory triggered single-shot simulated echo acquisition mode (STEAM) EPI DWI and a spin echo (SE) DWI sequence [7]. All liver DWI data with 5 b-values (12,200,300,400,600 s/mm²) and 3 different diffusion times (Δ=80,106,186ms corresponding to TM=50, 80,160ms) were acquired covering 3 slices with slice gap=9mm. Other parameters are: resolution =3×3 mm2, FOV=340×304mm², slice thickness=5mm, δ =8ms, TR/TE=1600/53ms, SENSE =2, trigger delay=500ms, repeated number=9 (STEAM)/3(SE), scan time≈130s.

Data analysis: ROIs excluding large blood vessels were drawn on the whole liver parenchyma including left lobe. Additionally, before data analysis, the motion corrupted data with severe signal voids on the liver parenchyma were rejected from the repeated scans semi-automatically. Then ADC_p maps of all ROIs were obtained by fitting the equation: $$$\frac{SI_{b}}{SI_{0}}=exp(-b\times ADC_{p})$$$ with the first two b values in Matlab (The Mathworks Inc., Natick, MA). From the histogram of the ADC_p map, a new perfusion-free ROI was then determined for further ADC and true diffusion coefficient (D_true) analyses by excluding pixels with extraordinarily high ADC_p values (with frequency lower than 8%). Finally, ADC and D_true were estimated by fitting the signal decay in ROIs with all and three larger b values (300,400,600 s/mm²) using equations $$$\frac{SI_{b}}{SI_{0}}=exp(-b\times ADC)$$$ and $$$\frac{SI_{b}}{SI_{0}}=(1-f)exp(-b\times D_{true})$$$, respectively. One-way ANOVA with Turkey’s multiple comparison tests was employed to compare ADC, and D_true measurements between the volunteers and patients with different Δs, p<0.05 was considered as statistical significant.

Fig. 1 shows that the liver ADC and D_true values decrease with Δ=80,106 and 186ms in both volunteers and fibrotic liver patients, respectively, which confirms their diffusion time dependency and spatially restricted diffusion in liver parenchyma. When the diffusion time increases, the signal decay becomes slower. Fig.2 compares the ADC and D_true values measured on the volunteers (the light grey) and the patients (marked with 3 kinds of darker grey colors) at 3 different Δs. It demonstrates that ADC and D_true with higher diffusion times exhibit larger significant differences compared with other the lower Δs.First, the mean ADC and D_true reductions on both the volunteers and patients with 3 Δs in the liver confirmed the diffusion time dependency of DWI quantification and the restricted diffusion behavior. Second, consistent with the previous study [6], the lower diffusion coefficients in the fibrotic liver may be caused by increased intracellular and decreased extracellular water fraction. Moreover, comparing with lower diffusion time, the significance of diffusion measurements between the volunteers and patients with Δ=186ms become larger. Especially for stage-2 and stag-3 groups, the mean ADC and D_true reductions between them were 20.5±9.7% and 18.6±7.5%; while those with Δ=80ms were 3±1.6% and 9.4±6.2%, respectively (Fig. 2). It confirmed the hypothesis in the previous study [6] there were more barriers such as increased cell membranes and accumulated collagen fibers in fibrotic livers. To conclude, diffusion measurements with higher diffusion times will be much more sensitive as a biomarker to detect the microstructure pathological alterations during human liver fibrogenesis.