Comparison of monoexponential,intravoxel incoherent motion diffusion-weighted imaging and diffusion kurtosis imaging for assessment of hepatic fibrosis

Abstract

Background

Hepatic fibrosis is a dynamic, reversible process which can result in liver failure. Diagnosis and monitoring of hepatic fibrosis are clinically important.

Purpose

To compare the diagnostic performance of diffusion kurtosis imaging (DKI), intravoxel incoherent motion (IVIM), and monoexponential diffusion-weighted imaging (DWI) to detect clinically significant fibrosis (≥ F2).

Material and Methods

This retrospective study was approved by Institutional Review Board and the requirement of informed consent was waived. One hundred and six patients were included who underwent liver multiple b-value DWI (10 b-values at 0–1000 s/mm²) at 1.5 T and were histologically diagnosed with hepatic fibrosis. Apparent diffusion coefficient (ADC), DKI-derived apparent kurtosis (Kapp) and diffusivity (Dapp), and IVIM-derived true diffusion (Dt), pseudodiffusion (D*), and perfusion fraction (f) were compared between no or early fibrosis (F0–1, n = 19) and clinically significant fibrosis (≥ F2, n = 87). Diagnostic performance was evaluated with receiver operating characteristic (ROC) analysis.

Results

F2–4 had a significantly lower D* (59.9±16.3 vs. 86.2±21.0 [×10⁻³ mm²/s]) and Dapp (3.46±0.79 vs. 4.07 ± 0.76 [×10⁻³ mm²/s]) but higher Kapp (1.10±0.18 vs. 0.98±0.12) than F0–1 (P < 0.01). ADC, Dt, and f did not show significant difference between two groups (P > 0.05). The area under the ROC curve for diagnosis of clinically significant fibrosis (≥ F2) was significantly larger in D* (0.89; 95% CI = 0.81–0.94) than Dapp (0.73; 95% CI = 0.63–0.81) and Kapp (0.75; 95% CI = 0.65–0.83) (P = 0.017 and 0.012, respectively).

Conclusion

IVIM-DWI might be more suitable for detecting hepatic fibrosis than the monoexponential and kurtosis model, and D* showed a better diagnostic performance to detect clinically significant fibrosis than other parameters.

Keywords

Liver cirrhosis hepatitis diffusion-weighted imaging kurtosis intravoxel incoherent motion

Introduction

Hepatic fibrosis is the final result of chronic liver injury, which ultimately causes liver cirrhosis and liver failure as well as hepatocellular carcinoma (HCC) (1,2). The current standard of reference to diagnose or grade hepatic fibrosis is a pathological assessment of a liver biopsy specimen. However, there is a growing clinical demand for a non-invasive diagnostic method due to the invasiveness, sampling bias, and relatively high intra- and inter-observer variation of fibrosis staging of liver biopsies (3 –5). Furthermore, recent studies have reported hepatic fibrosis is a dynamic, reversible disease although hepatic fibrosis has been regarded as a progressive, static phenomenon for a long time (6). Indeed, recent studies have demonstrated the increasing success of antiviral treatments in blocking or reversing the fibrogenic progression of chronic liver diseases as well as similar success with bariatric surgery in patients with non-alcoholic fatty liver disease (7 –9). Therefore, a non-invasive technique receives more attention as a tool for treatment monitoring of hepatic fibrosis, and as a prognostic tool for the clinical outcome of chronic liver disease and the identification of cirrhotic patients who are at the highest risk of future hepatic decompensation (10,11).

There have been attempts to diagnose hepatic fibrosis non-invasively using serum-based markers or imaging-based techniques including ultrasound and magnetic resonance imaging (MRI)-based elastography (12,13), and diffusion-weighted imaging (DWI) (14 –16). Because DWI is an entirely non-invasive technique which is routinely performed in clinical liver MR studies, and easy to repeat without the use of contrast media, it is worth examining DWI for staging liver ﬁbrosis (16,17). In hepatic fibrosis, the apparent diffusion coefficient (ADC) has been widely reported to decrease according to prior studies (14,18), which may be explained by the restricted movement of water molecules due to collagen deposition or diminished hepatic perfusion. The ADC value is calculated based on several assumptions, including that all movement in tissues is pure molecular diffusion, and has a Gaussian distribution of the displacement probabilities of the water molecules. However, tissues contain vessels, and the bulk motion of intravascular water protons may affect the ADC value (19,20). Furthermore, natural barriers in tissues including cell membranes and intracellular organelles may also contribute to changing the distribution probability of water molecules from the Gaussian distribution (21). Thus, more complicated models for DWI have been suggested to overcome the limitations of ADC for hepatic fibrosis staging, including the intravoxel incoherent motion (IVIM) model (19) and kurtosis model (21,22). The IVIM-DWI provides the pure molecular diffusion and perfusion component separately, using multiple b-values, and diffusional kurtosis imaging (DKI) provides information on the skewness of the distribution, which may be related to the complexity of tissue microstructure (23). Until now, there have been no studies that compared the diagnostic performance of IVIM-DWI and DKI for hepatic fibrosis staging.

Thus, the purpose of this study was to compare the diagnostic performance of monoexponential DWI, DKI, and IVIM-DWI in the detection of clinically significant hepatic fibrosis (≥F2).

Material and Methods

This retrospective study was approved by our institutional review board and the requirement for informed consent was waived. From January 2011 to July 2012, 118 patients with chronic hepatitis or liver cirrhosis underwent liver MRI at 1.5 T which included DWI with multiple b-values followed by hepatectomy or liver biopsy. After 12 patients with hepatic iron deposition (n = 7) or massive or infiltrative HCC (n = 5) that replaced the right lobe of the liver were excluded, 106 patients (81 men, 25 women; mean age = 55.4 ± 11.6 years [mean age in men = 56.1 ± 10.7 years, mean age in women = 53.2 ± 14.1 years]; age range =18–79 years) were included in this study. The median interval between MRI and a hepatectomy or liver biopsy was six days (range = 4–38 days). Ninety-six out of the 106 patients underwent total hepatectomy for liver transplantation (n = 30), hemihepatectomy (n = 27), trisectionectomy (n = 3), sectionectomy (n = 13), or segmentectomy (n = 27) for HCC (n = 82), cholangiocarcinoma (n = 3), colorectal metastases (n = 1), giant hemangioma (n = 1), intraductal papillary mucinous neoplasm of the bile duct (n = 1), lipoma (n = 1), and decompensated cirrhosis (n = 7). The remaining ten patients underwent percutaneous liver biopsy for hepatic fibrosis evaluation. Underlying diseases were hepatitis B (n = 82), hepatitis C (n = 9), non-B non-C cirrhosis (n = 3), alcohol liver disease (n = 10), and primary biliary cirrhosis (n = 2).

The serum markers including albumin, total bilirubin, and prothrombin time (PT) were recorded according to electronic chart review and classified into chronic liver disease or compensated cirrhosis and decompensated cirrhosis, based on histology and laboratory findings. Of the 106 patients, 13 (12.3%) had decompensated cirrhosis (11 men, 2 women; mean age = 52.1 ± 3.3 years; age range = 48–59 years): Child-Pugh class B (n = 9) and C (n = 4).

MRI acquisition

All patients underwent liver MRI at 1.5 T (Signa HDx, GE Healthcare, Waukesha, WI, USA) with an eight-channel torso phased-array coil. Routine liver MRI consisted heavily T2-weighted (T2W) image, T2W imaging, dual-echo T1-weighted (T1W) imaging, DWI, precontrast and dynamic T1W imaging using a hepatocyte-specific contrast agent. DWI was obtained in the axial plane with a free-breathing fat-suppressed single-shot echo-planar sequence. Ten b-values (0, 15, 25.4, 42.9, 72.5, 122.5, 207, 350, 592, and 1000 s/mm²) were applied in three orthogonal directions, except b-value of 0 s/mm², using a monopolar diffusion-encoding scheme. Other scan parameters were as follows: TR/TE = 6000/59.5 ms; bandwidth = 250 kHz; slice thickness = 7 mm; matrix = 128 × 128; field of view = 350–380 mm; number of excitations = 5; and acceleration factor = 2. The acquisition time was approximately 4 min 30 s.

Image analysis

The acquired data were post-processed to generate maps of IVIM-DWI (Eq. 1) and DKI (Eq. 2) using the following equations and all b-values, respectively (24).

S (b) / S (0) = (1 - f) \times exp (- b \times Dt) + f \times exp [(- b \times D^{*})]

(1)

S (b) / S (0) = exp (- b Dapp + b^{2} D ap p^{2} K app / 6)

(2)

On IVIM-DWI, three parametric maps were generated using Eq. 1: true diffusion coefficient (Dt); pseudodiffusion coefficient (D*); and perfusion fraction (f). For fitting, simultaneous non-linear least square fitting with three unknowns was used. On DKI, two parameters were obtained: the apparent mean diffusivity (Dapp) and the apparent mean kurtosis (Kapp). In addition, ADC values were calculated monoexponentially, with two b-values of 0 and 1000 s/mm² using the following equation:

S (b) / S (0) = exp (- b \times ADC)

(3)

One fellowship-trained radiologist (JHY) who was blinded to the pathologic result drew 3–5 regions of interest (ROIs) of approximately 1 cm² in the liver right lobe at the portal hilum level with careful attention to avoid vascular structures and focal liver lesions on the ADC map, and then ROIs were copied to the maps of all the other parameters from the same patients. The average value was taken as a representative value.

Histologic analysis

All specimens were embedded in a paraffin block after fixation in a formalin solution and serial sections of 4-mm slices were stained with hematoxylin and eosin (H&E). Specimens were analyzed by one hepatopathologist (KBL) who was blinded to the MR findings. The hepatic fibrosis stage (F stage) was graded based on standardized guidelines (25,26): no fibrosis (F0); portal fibrosis (F1); periportal fibrosis (F2); septal fibrosis (F3); and cirrhosis (F4). Inflammatory activity was graded as no (A0), mild (A1), moderate (A2), and severe (A3). According to histologic examination, patients had F0 (n = 13; A0 [n = 6], A1 [n = 6], and A2 [n = 1]), F1 (n = 6; A0 [n = 1], A1 [n = 5]), F2 (n = 19; A0 [n = 3], A1 [n = 10], and A2 [n = 6]), F3 (n = 18; A0 [n = 4]; A1 [n = 9], A2 [n = 4], and A3 [n = 1]), and F4 (n = 50; A0 [n = 9], A1 [n = 26]; A2 [n = 13]; and A3 [n = 2]).

Statistical analysis

Correlations between each parameter of DWI and the stage of hepatic fibrosis and the grade of inflammatory activity were evaluated with Spearman’s correlation coefficient; the coefficients were interpreted as follows: weak, ≥0.2; moderate, ≥0.5; and strong, ≥0.8 (27). The partial correlation coefficient was obtained between the DWI parameters and the fibrosis stage after excluding the effect of inflammation. Kruskal–Wallis test was used to compare DWI parameters among the fibrosis grades (F0–1, F2–3, and F4) followed by a pairwise comparison using Bonferroni correction. Between no or early fibrosis (F0–1) and clinically significant fibrosis (F2–4), categorical variables were analyzed with a χ² test; continuous variables were compared with either Student’s t-test or Mann–Whitney test after a normality test. For evaluation of the diagnostic performance to differentiate clinically significant fibrosis from no or early fibrosis, non-parametric receiver operating characteristic (ROC) analysis was performed. The area under the ROC curve (AUC) was compared for each parameter using the DeLong test follows by Bonferroni corrected pairwise comparisons. Intraclass correlation coefficient (ICC) and coefficient of variance (CV) were obtained in all MR parameters in three repeated measurements. ICC was calculated using a two-way model and the assumption of absolute agreement (28). Results with a P value < 0.05 were used to indicate statistical significance. Post-hoc power analysis was done for ADC, Dt, and f using G*Power (29). All statistical analyses were performed with commercially available software (IBM SPSS, version 23, SPSS Inc., IBM Company, Armonk, NY, USA; or MedCalc, version 12, MedCalc Software, Mariakerke, Belgium).

Results

Correlation between each parameter and hepatic fibrosis stage

Among the IVIM-DWI parameters, D* only showed a statistically significant correlation (r = −0.5; 95% confidence interval [CI] = −0.6–−0.3) with the fibrosis stage whereas Dt and f did not show any significant correlation (r − –0.01; 95% CI = −0.2–0.2] and r = -0.1 [95% CI = −0.3–0.1], respectively). Dapp and Kapp of DKI also showed a significant correlation with the fibrosis stage (r = −0.4; 95% CI = −0.5–−0.2 and r = 0.4; 95% CI = 0.2–0.6, respectively). ADC did not show a significant correlation with the fibrosis stage (r = −0.2; 95% CI = −0.3–0.02). Partial correlation coefficients between the fibrosis stage and each DWI parameter showed statistical significance in D* (−0.5, P < 0.0001), Kapp (0.3, P = 0.001), and Dpp (−0.3, P = 0.007). However, there was no statistical significance in ADC (0.13, P = 0.16), Dt (−0.1, P = 0.3), and f (−0.1, P = 0.7). None of the parameters showed a significant correlation with inflammatory activity.

Comparison of each parameter and the hepatic fibrosis stage

Among the fibrosis stages (F0–1, F2–3, and F4), D*, Dapp, and Kapp showed a statistically significant difference whereas ADC, Dt, and f did not show any significant difference (Table 1). Observed power of ADC, Dt, and f were 0.36, 0.17, and 0.08, respectively. In pairwise comparisons, D* had significantly lower values in F2–3 and F4 compared with that of F1–2, but it did not show any significant difference between F2–3 and F4. Dapp showed a significant difference only between F0–1 and F4. Kapp showed a significantly increased value in F4 compared with F0–1 and F2–3, but the values were overlapped between F0–1 and F2–3. In a comparison of the DWI parameters between no or early fibrosis (F0–1) and clinically significant fibrosis (F2–4), only D*, Dapp, and Kapp showed a statistically significant difference as well (Table 2). Only D* showed a significant difference between chronic liver disease or compensated cirrhosis and decompensated cirrhosis (66.8 ± 19 vs. 49.2 ± 20.0, P = 0.002, Table 3).

Table 1.

Comparison of parameters derived from IVIM-DWI and DKI in each fibrosis stage.

	ADC*	Dt*	D*	f (%)	Dapp*	Kapp
F0–1 (n = 19)	1.43 ± 0.2 (1.34–1.53)	1.12 ± 0.21 (0.98–1.18)	86.2 ± 21.0 (74.3–93.6)	25.9 ± 6.4 (22.5–30.2)	4.07 ± 0.76 (3.73–4.28)	0.98 ± 0.12 (0.88–1.05)
F2–3 (n = 36)	1.35 ± 0.23 (1.28–1.43)	1.06 ± 0.16 (0.95–1.13)	64.0 ± 12.8 (57.6–74.3)	24.9 ± 5.6 (21.2–29.5)	3.62 ± 0.80 (3.25–4.18)	1.04 ± 0.16 (0.94–1.13)
F4 (n = 50)	1.33 ± 0.3 (1.24–1.41)	1.06 ± 0.28 (0.92–1.19)	56.8 ± 17.9 (49.8–69.4)	25.5 ± 8.1 (21.0–28.4)	3.34 ± 0.77 (2.92–3.61)	1.14 ± 0.19 (1.04–1.23)
P value	0.31	0.60	<0.001	0.89	0.003	0.001
F0–1 vs. F2–3			<0.001		0.14	0.60
F0–1 vs. F4			<0.001		0.002	0.002
F2–3 vs. F4			0.16		0.31	0.02

Values are mean ± standard deviation (interquartile range [IQR] 25–75).

A P value < 0.05 indicates statistical significance among three groups. Post-hoc analysis was done only in variables with a P value < 0.05.

*(×10⁻³ mm²/s).

Table 2.

Comparisons of parameters derived from IVIM-DWI and DKI between no or early fibrosis (F0–1) and clinically significant fibrosis (≥F2).

	F0–1 (n = 19)	F2–4 (n = 87)	P value
ADC*	1.43 ± 0.2 (1.29–1.58)	1.34 ± 0.26 (1.20–1.45)	0.14
Dt*	1.12 ± 0.21 (0.98–1.18)	1.06 ± 0.23 (0.93–1.17)	0.31
D*	86.2 ± 21.0 (74.3–93.6)	59.9 ± 16.3 (54.4–70.5)	<0.001
f (%)	25.9 ± 6.4 (22.5–30.2)	25.2 ± 7.1 (21.1–29.0)	0.73
Dapp*	4.07 ± 0.76 (3.73–4.28)	3.46 ± 0.79 (3.05–3.82)	0.003
Kapp	0.98 ± 0.12 (0.88–1.05)	1.10 ± 0.18 (1.03–1.18)	0.008

Values are mean ± standard deviation (IQR 25–75%).

A P value < 0.05 indicates statistical significance between the two groups.

*(×10⁻³ mm²/s).

Table 3.

Comparisons of parameters derived from IVIM-DWI and DKI between chronic liver disease or compensated cirrhosis and decompensated cirrhosis.

	Chronic liver disease or compensated cirrhosis(n=93)	Decompensated cirrhosis(n=13)	P-value
ADC*	1.34±0.2 (1.22, 1.46)	1.44±0.5 (1.20, 1.54)	0.37
Dt*	1.06±0.17 (0.85, 1.16)	1.13±0.50 (0.91, 1.23)	0.38
D*	66.8±19.0 (57.3, 76.3)	49.2±20.0 (35.1, 59.0)	0.002
f (%)	25.3±7.2 (21.3, 29.2)	25.4±5.3 (20.7, 29.8)	0.97
Dapp*	3.57±0.74 (3.13, 4.04)	3.54±1.27 (2.88, 3.63)	0.88
Kapp	1.07±0.17 (0.99, 1.15)	1.13±0.32 (1.02, 1.26)	0.49

Values are mean ± standard deviation (IQR = 25–75%).

A P value < 0.05 indicates statistical significance between the two groups.

*(×10⁻³ mm²/s).

Diagnostic performance for diagnosing clinically significant fibrosis (≥F2)

The AUCs were 0.89 in D*, 0.73 in Dapp, and 0.75 in Kapp (Table 4, Fig. 3). D* had greater AUC values than that of Dapp and Kapp (P = 0.017 and 0.012, respectively), but there was no significant difference in the AUCs between Dapp and Kapp (P = 0.65). Sensitivities, specificities, and positive predictive values at the cut-off values for the three parameters are summarized in Table 4.

Table 4.

Diagnostic performance of DWI parameters in diagnosing clinically significant fibrosis (≥F2).

	AUC	Cut-off	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)
D*	0.89 (0.81–0.94)	≤70.7	77 (67/87) [66.8–85.4]	89.5 (17/19) [66.9–98.7]	97.1 (67/69) [89.9–99.7]	46.0 (17/37) [29.5–63.1]
Dapp*	0.73 (0.63–0.81)	≤3.68	71.3 (62/87) [60.6–80.5]	73.7 (14/19) [48.8–90.9]	92.5 (62/67) [83.4–97.5]	35.9 (14/39) [21.2–52.8]
Kapp	0.75 (0.65–0.83)	>1.03	67.8 (59/87) [56.9–77.4]	79.0 (15/19) [54.4–94.0]	93.7 (59/63) [84.5–98.2]	34.9 (15/43) [21.0–50.9]

Numbers in brackets are 95% CI.

*(×10⁻³ mm²/s).

AUC, area under the curve; PPV, positive predictive value; NPV, negative predictive value.

Repeatability of parameters in three DWI models

ICCs were slightly low in D* (0.74; 95% CI = 0.61–0.84) compared to the others parameters including ADC, Dt, f, Dapp, and Kapp (0.84–0.89, Table 5). CVs were in the range of 6.5% in Kapp to 20.4% in D*. Values were summarized in Table 5.

Table 5.

Intraclass coefficient (ICC) and coefficient of variation (CV) in each parameters of three diffusion models.

	ICC	CV (%)
ADC*	0.86 (0.78–0.91)	6.97 ± 3.7 (4.10–9.20)
Dt*	0.86 (0.79–0.92)	9.39 ± 5.25 (5.85–11.93)
D*	0.74 (0.61–0.84)	20.40 ± 10.34 (13.40–25.78)
f (%)	0.84 (0.76–0.90)	11.79 ± 7.57 (6.63–15.73)
Dapp*	0.89 (0.84–0.93)	8.82 ± 5.98 (4.58–11.0)
Kapp	0.87 (0.81–0.92)	6.50 ± 3.83 (3.30–8.70)

ICC is shown with its 95% CI. CVs are mean ± standard deviation (IQR = 25–75%).

*(×10⁻³ mm²/s).

Discussion

Information on the presence and degree of liver fibrosis is useful before making therapeutic decisions or predicting disease outcomes in patients with chronic liver diseases (30). The crucial point is that clinicians have to identify the F2 stage of fibrosis early, because antifibrotic treatment is generally required for patients with moderate to advanced stage of fibrosis (F2–F4) (31). In our study, we compared three models of monoexponential, IVIM, and kurtosis of DWI to investigate their diagnostic performance to detect clinically significant fibrosis (≥F2). We found that D* and Dapp showed significantly lower values and Kapp had significantly higher values in clinically significant fibrosis compared to the no or early fibrosis, whereas ADC tended to decrease in clinically significant fibrosis, but did not show any statistically significant results. Indeed, the DWI has been reported to be useful to detect hepatic fibrosis (14 –16), but ADC showed limited diagnostic ability to detect clinically significant fibrosis compared with ultrasound or MR elastography in the literature (30). The non-Gaussian distribution in hepatic fibrosis and the significant contribution of microscopic vascular perfusion on ADC calculation at low b-values have been indicated as important causes of the limited capability of conventional DWI and, as a result, variable diffusional models have been tried for the liver (15,32–34). Our study results showed a presence of non-Gaussian behavior of diffusion in hepatic fibrosis; we believe that the non-Gaussian diffusivity might be captured with IVIM-DWI and DKI to improve the diagnostic performance of the DWI for hepatic fibrosis.

In addition, we found that D* showed a significantly better diagnostic performance to detect clinically significant hepatic fibrosis than that of Dapp and Kapp. Indeed, D* had a moderate negative correlation with fibrosis degree whereas Dapp and Kapp showed a weak correlation with hepatic fibrosis. Although there were several studies which reported the technical or clinical feasibility of DKI in the liver (22,33,35,36), our study showed that the DKI model had a lower diagnostic performance to detect clinically significant fibrosis and worse model fit compared with IVIM-DWI. We speculate that the worse diagnostic performance of DKI compared with IVIM-DWI could be related to several reasons. First, the limitation of applying the highest b-value in the liver could be one reason. The reliability of Kapp depends on the number and range of b-values (23), and ultra-high b-values up to 2000–3500 s/mm² have been applied to organs in prior studies (33 –38) whereas we used 1000s/mm². Thus, the highest b-value in our study might not be sufficient to demonstrate the non-Gaussian distribution of water molecules in the liver. However, applying ultra-high b-values is not feasible in the liver because of signal-to-noise deterioration at ultra-high b-values (>1000 s/mm²). Indeed, noise contamination at high b-value would result in a parabolic signal intensity decay curve and cause errors in parameters (38). Second, an inherent limitation of the DKI model for hepatic fibrosis could be another reason. The physiologic basis of Kapp has not been elucidated yet, but it is believed to reflect the tissue complexity (39). Although hepatic fibrosis causes structural abnormalities due to collagen deposition and sinusoidal obliteration, the current concept of liver cirrhosis adopts hemodynamic changes in the liver. Indeed, alterations of perfusion-related parameters (D*and f) in hepatic fibrosis have been consistently reported in the literature (15,16). Thus, if the hemodynamic alteration is an early main change during development of the hepatic fibrosis, the DKI model may not sensitively address the changes in the microcirculation. According to our study results, Kapp was significantly higher in F4 when compared to the others, but there was no significant difference between F0–1 and F2–3. Given that cirrhosis, especially decompensated cirrhosis, may be accompanied with prominent structural changes, we cautiously postulate that tissue complexity due to structural changes may appear in the late stage of fibrosis, and DKI might be more suitable for structurally more complex tissues such as cirrhosis or tumors. Further studies are warranted to confirm our postulation.

Our study results suggest that alterations in perfusion, which are represented as variations in the D* values, may have a key role in detecting the physiological changes of hepatic fibrosis rather than pure diffusion restriction (Dt). The reason for the superior performance of the perfusion-related parameters of IVIM-DWI in the detection of clinically significant hepatic fibrosis is not completely understood yet. However, consistent reports of notable changes in the perfusion-related parameters in hepatic fibrosis for IVIM-DWI including our results could be explained in relation to the recent concept of cirrhosis (7,15,16,20,32,40). According to recent studies, not only architectural disruption due to the deposition of matrix proteins in sinusoids but also increased hepatic vascular tone due to a deficit of vasodilatory substances and/or excessive vasoconstrictors contribute to increasing hepatic resistance in cirrhosis (2,6,7). Of note, in our study, only D* showed a significant difference among the study population whereas f did not show a major change in clinically significant fibrosis. In fact, although D* has been reported to decrease in clinically significant fibrosis, the results of f have been controversial in several previous studies (15,16,32,40). According to the literature, f and D* may represent the blood volume and blood flow, respectively (41). Together with the arterial buffer response and increased hepatic vascular tone which affects the sinusoids predominantly, a decline of D* could be a more prominent change than that of f in hepatic fibrosis. Further studies including a large number of patients with a variable degree of fibrosis are warranted, especially D* shows a relatively large variation compared to the other parameters.

Our study has several limitations. First, the retrospective nature of this study has an inevitable bias. Second, a small number of no or early fibrosis, especially F1, was included which might underestimate statistical power. Third, the diagnostic performance of the parameters might be overestimated because the cut-off values were derived from the same study population. Fourth, we did not include high b-values (≥1500 s/mm²). Although it was based on the concern for poor signal-to-noise ratio at high b-values, our selection of b-value may affect the diagnostic performance of kurtosis imaging. Lastly, the study results may not be valid for parameters obtained from the different IVIM-DWI fitting algorithms. Further studies that include a larger number of F0–1 and different IVIM-DWI fitting algorithms are warranted.

In conclusion, our study showed that IVIM-DWI might be more suitable for detecting hepatic fibrosis than the monoexponential and kurtosis model, and D* had a better diagnostic performance for detecting clinically significant fibrosis compared with other parameters.

Footnotes

Declaration of conflicting interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: DK and HK are employees of GE Healthcare and provided technical support for software implementation. The authors not associated with GE Healthcare (JHY, JML, KBL, and JKH) maintained full control of the data at all times.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Friedman

SL.

Liver fibrosis – from bench to bedside.

J Hepatol 2003; 38 (Suppl. 1):S38–53.

Friedman

SL.

Hepatic fibrosis—overview.

Toxicology 2008; 254:120–129.

Abdi

Millan

Mezey

Sampling variability on percutaneous liver biopsy.

Arch Intern Med 1979; 139:667–669.

Bedossa

Dargere

Paradis

Sampling variability of liver fibrosis in chronic hepatitis C.

Hepatology 2003; 38:1449–1457.

McGill

Rakela

Zinsmeister

, et al. A 21-year experience with major hemorrhage after percutaneous liver biopsy. Gastroenterology 1990; 99:1396–1400.

Bosch

Groszmann

Shah

VH.

Evolution in the understanding of the pathophysiological basis of portal hypertension: How changes in paradigm are leading to successful new treatments.

J Hepatol 2015; 62:S121–130.

Trautwein

Friedman

Schuppan

, et al. Hepatic fibrosis: Concept to treatment. J Hepatol 2015; 62:S15–24.

Marcellin

Gane

Buti

, et al. Regression of cirrhosis during treatment with tenofovir disoproxil fumarate for chronic hepatitis B: a 5-year open-label follow-up study. Lancet 2013; 381:468–475.

Lassailly

Caiazzo

Buob

, et al. Bariatric surgery reduces features of nonalcoholic steatohepatitis in morbidly obese patients. Gastroenterology 2015; 149:379–388.

10.

Asrani

SK.

Incorporation of noninvasive measures of liver fibrosis into clinical practice: diagnosis and prognosis. Clin

Gastroenterol Hepatol 2015; 13:2190–2204.

11.

Schmeltzer

Talwalkar

JA.

Noninvasive tools to assess hepatic fibrosis: ready for prime time?

Gastroenterol Clin North Am 2011; 40:507–521.

12.

Ferraioli

Tinelli

Dal Bello

, et al. Accuracy of real-time shear wave elastography for assessing liver fibrosis in chronic hepatitis C: A pilot study. Hepatology 2012; 56:2125–2133.

13.

Yoon

Lee

Joo

, et al. Hepatic fibrosis: prospective comparison of MR elastography and US shear-wave elastography for evaluation. Radiology 2014; 273:772–782.

14.

Annet

Peeters

Abarca-Quinones

, et al. Assessment of diffusion-weighted MR imaging in liver fibrosis. J Magn Reson Imaging 2007; 25:122–128.

15.

Chung

Lee

Kim

, et al. Intravoxel incoherent motion MRI for liver fibrosis assessment: a pilot study. Acta Radiol 2015; 56:1428–1436.

16.

Yoon

Lee

Baek

, et al. Evaluation of hepatic fibrosis using intravoxel incoherent motion in diffusion-weighted liver MRI. J Comput Assist Tomogr 2014; 38:110–116.

17.

Taouli

Koh

DM.

Diffusion-weighted MR imaging of the liver.

Radiology 2010; 254:47–66.

18.

Bonekamp

, et al. Assessing liver fibrosis: comparison of arterial enhancement fraction and diffusion-weighted imaging. J Magn Reson Imaging 2014; 40:1137–1146.

19.

Le Bihan

Breton

Lallemand

Seperation of diffusion and perfusion in intravoxel incoherent motion MR imaging.

Radiology 1988; 168:497–505.

20.

Koh

D-M

Collins

Orton

MR.

Intravoxel incoherent motion in body diffusion-weighted MRI: reality and challenges.

Am J Roentgenol 2011; 196:1351–1361.

21.

Jensen

Helpern

Ramani

, et al. Diffusional kurtosis imaging: the quantification of non-gaussian water diffusion by means of magnetic resonance imaging. Magn Reson Med 2005; 53:1432–1440.

22.

Yang

Rao

Wang

, et al. Staging liver fibrosis with DWI: is there an added value for diffusion kurtosis imaging? Eur Radiol 2018; 28:3041–3049.

23.

Rosenkrantz

Padhani

Chenevert

, et al. Body diffusion kurtosis imaging: Basic principles, applications, and considerations for clinical practice. J Magn Reson Imaging 2015; 42:1190–1202.

24.

Le Bihan

Turner

MacFall

JR.

Effects of intravoxel incoherent motions (IVIM) in steady-state free precession (SSFP) imaging: application to molecular diffusion imaging.

Magn Reson Med 1989; 10:324–337.

25.

Bedossa

Poynard

An algorithm for the grading of activity in chronic hepatitis C. The METAVIR Cooperative Study Group. Hepatology 1996; 24:289–293.

26.

Park

Chon

Park

, et al. Histological grading and staging of chronic hepatitis standardized guideline proposed by the Korean Study Group for the Pathology of Digestive Diseases. Korean Journal of Pathology 1999; 33:337–346.

27.

Zou

Tuncali

Silverman

SG.

Correlation and simple linear regression.

Radiology 2003; 227:617–628.

28.

Park

Han

Sung

, et al. Selection and reporting of statistical methods to assess reliability of a diagnostic test: conformity to recommended methods in a peer-reviewed journal. Korean J Radiol 2017; 18:888–897.

29.

Faul

Erdfelder

Lang

, et al. G*Power 3: a flexible statistical power analysis program for the social, behavioral, and biomedical sciences. Behav Res Methods 2007; 39:175–191.

30.

Wang

Zhu

Liu

, et al. Performance of magnetic resonance elastography and diffusion-weighted imaging for the staging of hepatic fibrosis: A meta-analysis. Hepatology 2012; 56:239–247.

31.

Palmucci

Cappello

Attina

, et al. Diffusion-weighted MRI for the assessment of liver fibrosis: principles and applications. Biomed Res Int 2015;2015:874201.

32.

Jeng

, et al. Assessing hepatic fibrosis: comparing the intravoxel incoherent motion in MRI with acoustic radiation force impulse imaging in US. Eur Radiol 2015; 25:3552–3559.

33.

Anderson

Barry

Soto

, et al. Characterizing non-gaussian, high b-value diffusion in liver fibrosis: Stretched exponential and diffusional kurtosis modeling. J Magn Reson Imaging 2014; 39:827–834.

34.

Anderson

Barry

Soto

, et al. Quantifying hepatic fibrosis using a biexponential model of diffusion weighted imaging in ex vivo liver specimens. Magn Reson Imaging 2012; 30:1475–1482.

35.

Rosenkrantz

Sigmund

Winnick

, et al. Assessment of hepatocellular carcinoma using apparent diffusion coefficient and diffusion kurtosis indices: preliminary experience in fresh liver explants. Magn Reson Imaging 2012; 30:1534–1540.

36.

Goshima

Kanematsu

Noda

, et al. Diffusion kurtosis imaging to assess response to treatment in hypervascular hepatocellular carcinoma. Am J Roentgenol 2015; 204:W543–549.

37.

Noda

Kanematsu

Goshima

, et al. Diffusion kurtosis imaging of the pancreas for the assessment of HbA1c levels. J Magn Reson Imaging 2016; 43:159–165.

38.

Roethke

Kuder

Kuru

, et al. Evaluation of diffusion kurtosis imaging versus standard diffusion imaging for detection and grading of peripheral zone prostate cancer. Invest Radiol 2015; 50:483–489.

39.

Rosenkrantz

Sigmund

Johnson

, et al. Prostate cancer: feasibility and preliminary experience of a diffusional kurtosis model for detection and assessment of aggressiveness of peripheral zone cancer. Radiology 2012; 264:126–135.

40.

Luciani

Vignaud

Cavet

, et al. Liver cirrhosis: intravoxel incoherent motion MR imaging—pilot study. Radiology 2008; 249:891–899.

41.

Sigmund

Vivier

Sui

, et al. Intravoxel incoherent motion and diffusion-tensor imaging in renal tissue under hydration and furosemide flow challenges. Radiology 2012; 263:758–769.