Evaluation of liver fibrosis staging in patients with chronic hepatitis via the gadolinium washout rate: a comparative study with magnetic resonance elastography and pathology

Introduction

Liver fibrosis is a critical stage in the disease progression of patients with chronic hepatitis and is closely associated with increased morbidity, mortality, and healthcare costs (1). This process involves complex pathophysiological pathways, including cellular damage, release of inflammatory mediators, and accumulation of extracellular matrix components (2). Early diagnosis of liver fibrosis is crucial for determining treatment plans and assessing patient prognosis, especially when fibrosis reaches a stage F3 or beyond, which typically indicates an irreversible phase of the disease (3,4).

Although liver biopsy remains the gold standard for the diagnosis of liver fibrosis, its invasiveness, interobserver variability, and sampling error limit its widespread application (5). Noninvasive diagnostic tools, such as serological models, including the aspartate aminotransferase-to-platelet ratio (APRI), Fibrosis-4 index (FIB-4), and Forns index, while convenient and easy to perform, lack sufficient sensitivity in detecting early liver fibrosis (6). Additionally, although methods such as normalized iodine concentration (NIC) and extracellular volume (ECV) in quantitative computed tomography imaging have gained attention in liver fibrosis assessment, the potential risk of ionizing radiation restricts their clinical application (7,8). Ultrasound elastography is widely used due to its lack of ionizing radiation, noninvasiveness, and relatively low cost, but its diagnostic accuracy is limited by operator dependency and patient-related factors (9-11).

In recent years, magnetic resonance elastography (MRE) has been widely recognized as a reliable biomarker for fibrosis, but its high equipment costs and specific hardware requirements limit its widespread use (11). Multiphase T1 quantitative imaging with gadolinium-ethoxybenzyl diethylenetriamine pentaacetic acid (Gd-EOB-DTPA) has also gained recognition (12-18). However, its high cost has restricted its broader application in China.

This paper introduces a new metric, the gadolinium washout rate (GWR), based on multiphase T1 mapping and gadobenate dimeglumine (Gd-BOPTA), a liver-specific contrast agent. GWR quantifies the differences between washout and wash-in in delayed-phase liver gastric (GA) emptying (impeded by portal hypertension or increased ECV) and hepatobiliary-phase GA uptake (hampered by functional hepatocyte death). By comparing GWR with MRE and pathological biopsy results, we aimed to evaluate its potential in liver fibrosis staging and potentially develop a novel strategy for the diagnosis and treatment of chronic hepatitis. To our knowledge, this represents the first attempt to develop a noninvasive assessment of liver fibrosis. We present this article in accordance with the TRIPOD+AI reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2024-2621/rc).

Methods

Participants

Between March 2022 and June 2023, a total of 2,180 patients who underwent multiphasic abdominal enhanced magnetic resonance imaging (MRI) were prospectively included.

The inclusion criteria were as follows: (I) normal renal function (male creatinine level: <115 µmol/L; female creatinine level: <97 µmol/L); (II) age over 18 years; (III) uniform equipment and parameters (UIH 790; United Imaging Healthcare, Shanghai, China); and (IV) MRE and multiphasic T1 mapping sequence.

The exclusion criteria (Figure 1) were as follows: (I) lack of pathological meta-analysis of histological data in viral hepatitis (METAVIR) fibrosis results (n=1,798); (II) failure of MRE check (n=11); (III) liver transplantation (n=5); (IV) portal vein tumor thrombus (n=16); (V) acute hepatitis (n=6); (VI) massive lesion leading to obstruction of liver parenchyma evaluation (n=27); (VII) delayed phase (DLP) time error exceeding 1 minute (n=31); (VIII) hepatobiliary phase (HBP) time error exceeding 10 minutes (n=80); and (IX) lack of serological indicators (n=32).

Figure 1 Flowchart of the study procedure. F0, no fibrosis; F1, early fibrosis; F2, significant fibrosis; F3, advanced fibrosis; F4, cirrhosis. DLP, delayed phase; HBP, hepatobiliary phase; METAVIR, meta-analysis of histological data in viral hepatitis; MRE, magnetic resonance elastography.

This retrospective MRI study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments and was approved by the Ethics Committee of The Third Affiliated Hospital of Sun Yat-sen University (approval No. II2024-228-01). All patients provided written informed consent.

Experimental procedure and theoretical basis path for GWR

The GWR is based on two concepts. First, it is based on the hepatic metabolic features of GD-BOPTA. As fibrosis advances, hepatocyte apoptosis occurs, reducing the liver’s wash-in of gadolinium during the HBP. This results in decreased liver signal intensity (SI) and increased T1-MAP values. Second, it is based on the renal metabolic features of GD-BOPTA. Fibrosis activates hepatic stellate cells, causing extracellular matrix deposition, hepatocyte atrophy, and intrahepatic blood flow obstruction, resulting in portal hypertension. During the DLP, the specific contrast agent causes impaired gadolinium washout from the liver, thus increasing liver SI and decreasing T1-MAP values.

The normalization principle in GWR involves using unenhanced T1-MAP (Unenh-T1MAP) as the background for normalization. The differences between HBP and Unenh and between DLP and Unenh are calculated, and the ratio of the difference between the two groups is used to quantify the proportion of renal excretion and hepatic uptake, respectively. This enhances the generalizability of the data across different populations and parameters (Figure 2).

Figure 2 Diagram illustrating the principle of GWR. The vertical axis represents the longitudinal relaxation time, while the horizontal axis depicts the temporal characteristics of multiphase T1 mapping. The red line illustrates the foundational theory of the GWR metric, including the characteristics of gadolinium wash-out during the delayed phase and gadolinium uptake during the hepatobiliary phase (wash-in). GRE, gradient echo; GWR, gadobenate washout rate; MRE, magnetic resonance elastography; ROI, regions of interest; T1_DLP, longitudinal relaxation time of the 5-minute delay phase; T1_HBP, longitudinal relaxation time of the 120-minute hepatobiliary phase; T1_Unenh, unenhanced longitudinal relaxation time.

Magnetic resonance (MR) acquisition

MRI scans were performed on a 3-T system (UIH 790; United Imaging Healthcare) with standard abdominal and spinal matrix coils. All patients fasted for over 4 hours, adopted a supine position with feet advanced, and practiced respiratory coordination. The conventional sequences acquired included a fat-suppressed T2 sequence, diffusion-weighted imaging (DWI), multiphase T1 mapping acquisition, and multiphase quick three-dimensional (3D).

Multiphase T1 mapping acquisition

T1 mapping is a quantitative MR technique that measures the T1 values of tissues (voxels). The sequence belongs to the gradient echo (GRE)-MAPS protocol and was acquired in 3D mode under the following parameters: repetition time (TR) =3.33 ms, echo time (TE) =1.23 ms, bandwidth =650 Hz/pixel, field of view (FOV) =400 mm² × 300 mm², matrix =272×204, slice thickness =8 mm, echo chain length [ETL =20, and phase encoding direction = anterior − posterior (AP)]. The framework for acceleration strategy (FAST) was applied in two-dimensional (2D) mode 2, with flip angles of 3° and 15°, and an acquisition time of 18 seconds. Saturation bands were applied superiorly and inferiorly, and B1 correction was implemented. Patients were instructed to hold their breath during the procedure, and quantitative T1 mapping images and qualitative magnitude maps were generated. The acquisition included three phases: Unenh (T1_Unenh), 5-minute DLP (T1_DLP), and 120-minute HPB (T1_HBP).

Fat analysis and calculation technique sequence

The fat analysis and calculation technique 3D (FACT-3D) sequence is a type of GRE-FACT protocol and a multiecho water-fat separation sequence. In this study, it was acquired in 3D mode under the the following parameters: FOV =400 mm² × 300 mm²; TR =11 ms; TE1−TE6 =1.46 ms; 2.95 ms, 4.44 ms, 5.93 ms, 7.42 ms, and 8.91 ms, respectively; bandwidth =930 Hz/pixel; Dixon angle =240°; matrix =192×144; slice thickness =8 mm; interpolation = none; and ETL =16. The parallel acquisition technique (FAST) was applied in 2D mode with 2.4× acceleration, a flip angle of 3°, 1 number of excitation (NEX), an AP phase encoding direction, and a breath-hold time of 15 seconds. The images generated included fat fraction (FF) maps, R2* maps, and water-fat separation images.

Contrast agent injection

Gd-BOPTA (Bracco, Milan, Italy) was used as the paramagnetic contrast agent for MRI. For adult patients, a contrast agent dose of 0.75 mmol/kg (equivalent to 0.15 mL/kg) was administered at a rate of 1.5 mL/s. Dynamic enhanced multiphase scanning commenced when contrast agent was observed entering the right atrium.

MRE acquisition

A 19-cm-diameter passive pneumatic actuator was positioned at the level of the right rib cage and connected to the sound generator. The 2D spine echo sequence was used under the following parameters: TR =1,000.2 ms, TE =44.6 ms, bandwidth =1,500 Hz/pixel, FOV =420 mm² × 420 mm², driver frequency =60, driver strength =40, Motion ENCoding sensitivity (MENC)=20 µm/pi, and phase number =4. A 60-Hz shear wave was applied to the patient’s body surface. MRI sequences detected amplitude and phase information, generating a waveform image and outputting an elasticity image via an inverse fitting algorithm (19). An algorithm provided by Mayo Clinic was used to calculate the MRE hardness image, with tissue shear stiffness being determined and displayed on the elasticity image in kilopascals (kPa). A fully automated algorithm described by Dzyubak et al. (20) was used to analyze MRE waves, stiffness, and anatomical images.

MRI analysis

Three-phase T1 mapping images [unenhanced longitudinal relaxation time (T1_Unenh), longitudinal relaxation time of the 5-minute delay phase (T1_DLP), and longitudinal relaxation time of the 120-minute hepatobiliary phase (T1_HBP)] and MRE images were transferred to a dedicated workstation (uPMR 790, United Imaging Healthcare) and analyzed blindly by two radiologists (Y.G. and S.X., with 3 and 17 years of abdominal radiology experience, respectively) who were unaware of the pathological results. A simultaneous measurement and cross-checking method was used to minimize errors. In cases of disagreement, a third radiologist (J.Q., with 25 years of abdominal radiology experience) analyzed the case and made the final decision. The triphasic T1-MAP images were synchronized with the MRE images. Two circular 400 mm² regions of interests (ROIs) were placed in the right hepatic lobe, matching the corresponding confidence mask regions on the MRE images. The ROIs were positioned away from the liver edge, blood vessels, and abnormal tissues such as lesions. The mean and standard deviation of the T1_Unenh, T1_DLP, and T1_HBP phases displayed in the ROI were recorded in turn. The final reading was obtained by averaging the four layers combined (Figure 3).

Figure 3 Quantitative image summary of METAVIR fibrosis staging. (A-E) The T1-weighted unenhanced images, T1-weighted dynamic liver protocol images, T1-weighted hepatobiliary phase images, MRE stiffness maps, and pathological images, respectively, of a 29-year-old female patient with no hepatic fibrosis (F0 stage). (F-J) The same sequences and images, respectively, of a 69-year-old female patient with advanced hepatic fibrosis (F3 stage). According to the formula of GWR, the GWR for patient 1 was –0.63 and that for patient 2 was 0.32. According to the cutoff values of GWR [>–0.014 (≥ F1), >–0.1404 (≥ F2), > 0.00013 (≥ F3), and >0.1139 (≥ F4)], both patients were accurately diagnosed with their respective fibrosis stages, which was consistent with MRE and pathological findings. Sirius Red staining of liver fibrosis pathology at 40× magnification. GWR, gadobenate washout rate; METAVIR, meta-analysis of histological data in viral hepatitis; MRE, magnetic resonance elastography.

For MRE delineation, ROIs with fewer than 1,000 pixels were excluded due to suboptimal image quality. ROIs were manually drawn on the four elastic images, with liver edges, blood vessels, bile ducts, and focal lesions being avoided (21). Average liver stiffness (m) was calculated with the following weighted arithmetic mean formula:

MRE=(m1w1+m2w2+m3w3+m4w4)(w1+w2+w3+w4)[1]

where m1–m4 are the average liver hardness values of single-layer ROIs, and w1–w4 are the corresponding ROI pixel values. ROI readings were obtained by averaging multiple measurements. The calculation formulas of the relative enhancement rate (RLE) for the HBP (RLE_HBP) (22), RLE for the DLP (RLE_DLP) (23), and GWR are as follows:

RLEHBP=R1HBP−R1UnenhR1Unenh=1T1HBP−1T1Unenh1T1Unenh[2]

RLEDLP=R1DLP−R1UnenhR1Unenh=1T1DLP−1T1Unenh1T1Unenh[3]

where R1=1/T1.

The principle of GWR is the difference in gadolinium clearance in the kidneys during the 5-minute DLP and the gadolinium uptake in the liver at 120 minutes. These two factors are expressed as changes relative to the unenhanced scan and referred to as “washout” and “wash-in”, respectively. Detailed explanations can be found in Figure 2. The GWR was calculated as follows (24):

GWR=1−R1HBP−R1UnenhR1DLP−R1Unenh=1−1/T1HBP−1/T1Unenh1/T1DLP−1/T1Unenh[4]

where R1=1/T.

The GWR formula normalizes HBP gadolinium uptake (T1_HBP) against DLP clearance (TI_DLP), with unenhanced T1 values (T1_Unenh) serving as a baseline. A lower GWR indicated impaired hepatocyte function and elevated portal pressure

Akaike information criterion (AIC)

The AIC is a method used for model selection in statistical modeling. Proposed by Japanese statistician Hirotugu Akaike in 1974, it aims to identify the model that best explains the data while incorporating the fewest parameters. The core idea of AIC is to seek a balance between model fit and model complexity to avoid overfitting. This study included 17 variables: gender, age, height, weight, body mass index (BMI), FF, R2*, hematocrit (HCT), APRI, FIB-4, Forns index, T1_Unenh, T1_DLP, T1_HBP, RLE_DLP, RLE_HBP, and GWR. An AIC model was established via Python (Python Software Foundation, Wilmington, DE, USA).

Pathology

Sample evaluation was performed by an experienced pathologist, Na Cheng, with 15 years of expertise in liver pathology, who was blinded to the clinical and radiological information. Biopsy samples were considered insufficient if the core size was less than 1.5 cm or displayed fewer than 10 complete portal tracts. Fibrosis staging was determined according to the METAVIR classification, which ranged from F0 (no fibrosis) to F4 (cirrhosis). Objective quantification of fibrosis was conducted via Sirius red staining and Image ProPlus version 6.1 image analysis software (MediaCybernetics, Rockville, MD, USA) (25).

Serological indices

Laboratory tests were recorded for each patient within 2 weeks after MRE. The formulae for FIB-4 and APRI are as follows:

FIB-4=age×aspartateaminotransferase(AST)(U/L)age×aspartateaminotransferase(AST)(U/L)×alanineaminotransferase(ALT)(U/L)[5]

APRI=(AST/ULN)×(100/PLT(109/L))[6]

Age—patient’s age in years; aspartate aminotransferase (AST)—liver enzyme, measured in IU/L; alanine aminotransferase (ALT)—another liver enzyme, measured in IU/L; platelet count (PLT)—number of platelets in blood, measured in 10⁹/L (26).

Statistical analysis

Statistical analysis was performed with SPSS version 26 (IBM Corp., Armonk, NY, USA). The 174 samples were divided into four groups (≥ F1, ≥ F2, ≥ F3, and ≥ F4). Descriptive statistics were used to describe the demographic characteristics of participants with chronic liver disease. Pearson correlation was used to analyze the correlation coefficient between multiple independent variables and the METAVIR fibrosis stage (r <0.3 for weak correlation, 0.3< r <0.7 for moderate correlation, and r >0.7 for strong correlation). The Kruskal-Wallis test, Chi-squared test, and independent-samples t-test were used to compare differences in various independent variables between fibrosis groups (P<0.05 was considered statistically significant). Univariate and multivariate regression analyses were applied to assess hazard ratios (HRs). Python was used for univariate screening, and the GWR was selected based on stepwise selection and the AIC (P<0.05 indicated statistical significance). The evaluation of diagnostic performance was divided into two phases: non-cross-validation and cross-validation. In the non-cross-validation phase, receiver operating characteristic (ROC) curves were generated to obtain the area under the curve (AUC) for each model. DeLong tests were then employed to compare the diagnostic accuracy among various models across different fibrosis stages (≥ F1, ≥ F2, ≥ F3, and ≥ F4). This initial comparison provided a preliminary assessment of model performance.

In the cross-validation phase, a fivefold cross-validation approach was implemented with stratified sampling (test set ratio =0.2 and random seed =42) to ensure balanced representation across fibrosis stages. Logistic regression models were evaluated via AUC, sensitivity, specificity, and F1 score. All statistical analyses were conducted with R version 4.2.3 (The R Foundation for Statistical Computing) and Python version 3.11.4.

This two-step approach allowed for a comprehensive evaluation of model performance, ensuring both robustness and clinical relevance.

Results

A total of 174 participants were ultimately recruited in this study, all of whom underwent the MR two-phase T1-mapping sequence and MRE sequence and had complete pathological staging and serological data. The demographic and clinical characteristics of the participants are summarized in Table 1. Age, gender, APRI, and FIB-4 showed statistically significant differences between the fibrosis groups (P<0.05), indicating their potential for diagnosing liver fibrosis. In contrast, BMI and categorical indicators such as hepatitis B virus, hepatitis C virus, autoimmune hepatitis, and nonalcoholic steatohepatitis showed no statistically significant differences (P>0.05).

For the four independent fibrosis groups (≥ F1, ≥ F2, ≥ F3, and ≥ F4), the correlation coefficients of MRE (R=0.454, R=0.561, R=0.661, and R=0.683, respectively; P<0.001) and GWR (R=0.485, R=0.526, R=0.53, and R=0.391, respectively; P<0.001) were higher than those of the other indicators (Table 2 and Figure S1). MRE showed increasing correlation with fibrosis progression, peaking at cirrhosis. GWR also showed increasing correlation with fibrosis progression, peaking at advanced fibrosis (≥ F3). Other indicators, including relative enhancement rate of the hepatobiliary phase in T1 mode (RLE_HBP), exhibited similar patterns with advanced fibrosis, highlighting the clinical value of the model at this stage.

The intraclass correlation coefficient (ICC) statistical results ranged from 0.811 to 0.925, indicating high interrater consistency (Table 2).

Further analysis indicated that MRE, T1_HBP, RLE_HBP, GWR, APRI, and FIB-4 were significantly different between the groups, while T1_Unenh and T1_DLP were only significantly different between the ≥ F3 and ≥ F4 stages (Table S1; P<0.05).

To determine the predictive power of each index for fibrosis severity, univariate and multivariate regression analyses were performed. GWR was similar to MRE in the ≥ F2 stage (17.24 vs. 20.03) and exceeded MRE for the ≥ F3 (31.32 vs. 7.35) and ≥ F4 stages (25.84 vs. 3.86). This suggests that GWR’s diagnostic ability for fibrosis improves with disease progression, which is consistent with the correlation findings. In the ≥ F1 stage, the HR of GWRs was markedly lower than that of MREs (8.85 vs. 141.66), suggesting portal hypertension factors are insignificant in early fibrosis. Additionally, the HR was highest in the ≥ F3 stage, in line with the correlation analysis (Figure 4).

Figure 4 Forest plot of univariate and multivariate regression analyses. The HR for GWR in the univariate analysis progressively increased from ≥ F1 to ≥ F3 and then declined in ≥ F4. The HR for GWR was higher than that for MRE only for ≥ F3 according to both univariate and multivariate analyses. *, statistical significance with a P value <0.05 in the univariate regression analysis; **, statistical significance with a P value <0.05 in the multivariate regression analysis. APRI, aminotransferase-to-platelet ratio; BMI, body mass index; FF, fat fraction; FIB-4, fibrosis-4 index; GWR, gadobenate washout rate; MRE, magnetic resonance elastography; R2*, indicator of iron deposition; RLE_DLP, relative enhancement rate of the delayed phase in T1 mode; RLE_HBP, relative enhancement rate of the hepatobiliary phase in T1 mode; T1_DLP, longitudinal relaxation time of the 5-minute delay phase; T1_HBP, longitudinal relaxation time of the 120-minute hepatobiliary phase; T1_Unenh, unenhanced longitudinal relaxation time.

In the modeling with 17 variables via the AIC criterion, GWR was included in the AIC model for ≥ F2, ≥ F3, and ≥ F4 (all P values <0.05). Both AIC modeling and univariate and multivariate regression analyses indicated that with fibrosis progression, portal hypertension, and ECV deterioration directly lead to impaired delayed gadolinium washout (Table 3).

The evaluation of the diagnostic performance of various models was divided into two steps.

In the non-cross-validation phase, MRE had the highest AUC in the diagnosis of fibrosis stages F1–F4 (0.95, 0.91, 0.901, and 0.916, respectively), followed by GWR (0.82, 0.834, 0.864, and 0.844, respectively). DeLong tests confirmed that the diagnostic accuracy of MRE and GWR was comparable in the diagnosis of advanced fibrosis (P=0.277) (Table 4 and Figure 5).

Figure 5 ROC curves for MRE and GWR. (A-D) The ROC curves for fibrosis stages ≥ F1 to ≥ F4, respectively. (C) The closest AUC results to those of MRE. AUC, area under the curve; FIB-4, Fibrosis-4 index; GWR, gadobenate washout rate; MRE, magnetic resonance elastography; RLE_DLP, relative enhancement rate of the delayed phase in T1 mode; RLE_HBP, relative enhancement rate of the hepatobiliary phase in T1 mode; ROC, receiver operating characteristic; T1_DLP, longitudinal relaxation time of the 5-minute delay phase; T1_HBP, longitudinal relaxation time of the 120-minute hepatobiliary phase; T1_Unenh, unenhanced longitudinal relaxation time.

To further validate the robustness of our findings, we conducted a fivefold cross-validation analysis on the MRE and GWR, examining their performance on the test sets to corroborate the conclusions drawn in the non-cross-validation phase.

In the diagnosis of early fibrosis (≥ F1 and ≥ F2), MRE had higher AUC, accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and F1 scores compared to GWR. The PPV and NPV of the GWR test set showed a high misdiagnosis and missed diagnosis rate.

In advanced fibrosis (≥ F3), GWR had higher AUC, accuracy, sensitivity, specificity, PPV, NPV, and F1 scores compared to MRE (Figure S2).

For cirrhosis (≥ F4), MRE had higher AUC, accuracy, specificity, PPV, and F1 scores as compared to GWR. The PPV and NPV of the GWR test set indicated a high misdiagnosis and missed diagnosis rate.

In the diagnostic assessments of stages ≥ F1, ≥ F2, ≥ F3, and ≥ F4, GWR’s AUC, accuracy, specificity, PPV, and F1 scores were consistent with the correlation findings: diagnostic efficacy increased with fibrosis progression and peaked at advanced fibrosis. All the results show that the liver wash-in and washout features captured by GWR are most significant in late fibrosis (≥ F3). This is likely because portal hypertension is not prominent in the earlier stages and is overshadowed by higher liver stiffness in the later stages (Table 5).

In summary, the first step of our study confirmed the feasibility of GWR in diagnosing fibrosis and importantly, demonstrated its optimal diagnostic efficacy in advanced fibrosis.

Discussion

This study confirmed the feasibility of GWR in diagnosing liver fibrosis (≥ F1 to ≥ F4) through dual screening with the AIC and logistic regression model, as well as dual validation with the DeLong test and 10-fold cross-validation. The results show that GWR is superior to MRE in diagnosing advanced fibrosis (≥ F3).

The AIC-based model and univariate logistic regression analysis demonstrated that GWR performs well in staging liver fibrosis from ≥ F2 to ≥ F4, indicating its high value in precise staging of hepatic fibrosis.

GWR effectively leverages the quantitative differences in hepatic and renal metabolism of liver-specific contrast agents, which are closely related to hepatic fibrosis. Hepatic metabolism is reflected in hepatic gadolinium wash-in during the HBP, indicating residual functional hepatocytes. Renal metabolism is reflected in delayed gadolinium washout, indicating the progression of hepatocyte atrophy, extracellular matrix deposition, and portal hypertension. The combination of these features is crucial for assessing the reversibility of hepatic fibrosis.

Gd-BOPTA vs. Gd-EOB-DTPA

Both Gd-BOPTA and Gd-EOB-DTPA are liver-specific GA contrast agents. The differences between these agents are as follows: the hepatic uptake proportion of Gd-BOPTA is 5%, while that of Gd-EOB-DTPA is 50%. Furthermore, the price of Gd-BOPTA is CNY ¥260, while that of Gd-EOB-DTPA is CNY ¥1,600, with regional and temporal fluctuations. In terms of advantages, Gd-BOPTA excels in vascular display, while Gd-EOB-DTPA is better for nodular liver diseases. Finally, the HBP of Gd-BOPTA is 120 minutes, while that of Gd-EOB-DTPA is 20 minutes.

These drugs are actively transported into functional hepatocytes and then excreted into bile. The hepatic uptake difference affects liver lesions and the diagnosis of fibrosis (27). Numerous studies (16-18) have confirmed that Gd-EOB-DTPA is effective in diagnosing liver fibrosis staging in RLE-HBP with satisfactory results. However, due to its high cost and limited vascular feature display ability, more cost-effective contrast agents and improved diagnostic strategies are needed in clinical practice.

Our study used Gd-BOPTA, and we found that compared with the Gd-EOB-DTPA literature, we observed that RLE_HBP derived from Gd-BOPTA showed a modestly weaker correlation with METAVIR fibrosis stage. The lower hepatocyte uptake rate led us to anticipate and accept this result, which is also one of the reasons for incorporating the DLP. Indeed, the diagnostic efficacy of our GWR in fibrosis stages (AUC 0.84–0.959) outperformed that reported in several studies using Gd-EOB-DTPA, including that by Breit et al. (AUC 0.74–0.76) (28), Yang et al. (AUC 0.809–0.849) (29), and Li et al. (AUC 0.811–0.894) (30).

Our findings expand upon the existing research on GD-BOPTA in the field of liver fibrosis and provide a means to effectively reduce the cost of patient examinations.

The positive correlation between hepatic fibrosis progression and portal venous pressure aligns with the other results in our study. GWR’s correlation and diagnostic AUC increased with fibrosis severity, peaking in the advanced stages. This indicates that the diagnostic weight of hepatic and renal metabolic features grows with the progression of fibrosis, being most significant in the advanced stages. This likely reveals that the portal hypertension feature reflected by GWR and the liver stiffness feature reflected by MRE have optimal stages in hepatic fibrosis.

Initiation of the nonenhanced phase

The multiphase data we obtained were from different populations, and there were thus natural differences in height, weight, gender, and age. Although we eliminated the differences in equipment and parameters, a normalization process was still necessary. The inclusion of T1_Unenh and the calculation of the differences between the HBP and DLP resulted in GWR being an excellent model with a normalized background, ensuring generalizability. Similar models, such as ECV, which use the aorta as a normalization background, have been widely used in liver fibrosis diagnosis for a period of time. Our model outperformed those in several studies that used ECV in the diagnosis of fibrosis stage, including that by Obmann et al. (AUC 0.69–0.72) (31) and Wang et al. (AUC 0.801–0.907) (32).

It is worth noting that the GWR is a single-modality model and does not require the HCT data necessary for the dual-modality model ECV.

Comparison with MRE

Costa-Silva et al. reported an AUC of 0.928 for MRE in diagnosing advanced liver fibrosis, with a sensitivity of 0.909 and specificity of 0.973 (33). Meanwhile, Xiao et al. reported an AUC of 0.96 in a study evaluating multiple indicators for diagnosing advanced liver fibrosis, with a sensitivity of 0.84 and a specificity of 0.9 (34). Finally, Loomba et al. reported an AUC of 0.981 for 3D MRE in diagnosing advanced fibrosis, with a sensitivity of 1 and a specificity of 0.94 (35).

Our results (AUC 0.959; sensitivity 0.8929; specificity 1) were only slightly inferior to the diagnostic efficacy of 3D MRE and comparable to the other two studies. In our study, GWR surpassed MRE in diagnosing advanced fibrosis (≥ F3) (AUC: 0.934 vs. 0.921; sensitivity: 0.80 vs. 0.76; specificity: 1 vs. 0.964). The higher sensitivity and specificity reflected lower rates of false negatives and false positives.

Notably, certain factors can impact MRE results, such as biliary obstruction, ascites, hepatitis, and obesity. Our aim is not to discredit MRE as the gold standard for liver fibrosis diagnosis. Yet, chronic liver disease is highly complex, with various comorbidities that may produce unpredictable risks. The nonphysical index GWR can avoid the above-mentioned interference and can supplement or even replace MRE assessment in certain patients.

Comparison with transient elastography (TE) and serological models

A study (36) involving 19,199 patients from 63 studies using TE reported AUCs of 0.83, 0.8, 0.87, and 0.94 for diagnosing liver fibrosis (≥ F1–F4). Meanwhile, a study with 1,484 patients from 14 studies using MRE reported AUCs of 0.89, 0.92, 0.89, and 0.94 for diagnosing liver fibrosis (≥ F1–F4).

In our study, In the cross-validated test cohort, GWR achieved AUCs of 0.848, 0.882, 0.959, and 0.840 for discriminating METAVIR stages ≥ F1, ≥ F2, ≥ F3, and F4, demonstrating better results for advanced fibrosis (≥ F3) as compared to MRE. Moreover, GWR had significantly higher diagnostic AUCs across all fibrosis stages (≥ F1–F4) as compared to APRI and FIB-4 (0.613–0.715). Our findings regarding serological models are consistent with the work of Kim et al. (37) and Huttman et al. (6).

Comparison with frontier research

Our findings align with recent efforts to combine multiparametric MRI biomarkers for fibrosis staging. For instance, Wang et al. (20,23) achieved similar AUC improvements (0.89–0.93) by integrating T1 mapping and DWI, yet their model required additional sequences. In contrast, in another study, GWR achieved comparable accuracy with standard Gd-BOPTA protocols, enhancing clinical feasibility (38).

Future prospects for GWR

GWR, after rigorous screening, exhibited advantages in diagnosing late-stage hepatic fibrosis in our study and others. In the future, our team plans to use numerous multiphase T1 mapping sequences to build a deep learning model. This will capture GWR’s features from a higher-dimensional perspective and facilitate the development of real-time measurement-guided clinical diagnostic software.

We also plan to conduct precise independent subgroup analyses for patients with limitations (e.g., ascites, biliary obstruction, and iron overload affecting MRE and TE). This will clarify and explain GWR’s generalization and overfitting and promote the popularization of blood flow dynamics-guided liver fibrosis diagnosis and treatment.

This study involved several limitations that should be acknowledged. First, the relatively small sample size increased the risk of overfitting; larger samples are needed to validate GWR’s diagnostic accuracy across liver fibrosis stages. Second, selection bias arising due to the single-center, retrospective design and exclusion of MRE measurements with ROIs <1,000 pixels in size might have skewed the data, necessitating future validation in larger, multicenter studies with rigorous protocols. Third, due to the sample size, subgroup analyses were sparse, and despite the completion of supplementary multivariate linear stratified regression of GWR across subgroups, further validation of GWR’s generalizability needs to be conducted with larger subgroup samples.

Conclusions

Based on the pathophysiology of hepatic fibrosis (portal hypertension and functional hepatocyte death), GWR—grounded in hemodynamic theory (washout + wash-in)—is capable of diagnosing liver fibrosis of all stages. In our study, it surpassed MRE in diagnosing advanced fibrosis (≥ F3), offering higher predictive value and clinical net benefit, and may thus serve as a cost-effective alternative.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD+AI reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2024-2621/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2024-2621/dss

Funding: This study was supported by the Guangdong Provincial Natural Science Foundation (No. 2017A030313841); Hospital National Natural Science Foundation Cultivation Project (No. 2021GZRPYM06); Five-Five Project of The Third Affiliated Hospital of Sun Yat-sen University (No. 2023WW605).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2024-2621/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This retrospective study complied with ethical committee standards and was approved by the Ethics Committee of The Third Affiliated Hospital of Sun Yat-sen University (approval No. II2024-228-01) and all patients provided written informed consent.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Sharma S, Khalili K, Nguyen GC. Non-invasive diagnosis of advanced fibrosis and cirrhosis. World J Gastroenterol 2014;20:16820-30. [Crossref] [PubMed]
2. Rockey DC, Bell PD, Hill JA. Fibrosis--a common pathway to organ injury and failure. N Engl J Med 2015;372:1138-49. [Crossref] [PubMed]
3. Abubakr S, Hazem NM, Sherif RN, Elhawary AA, Botros KG. Correlation between SDF-1α, CD34 positive hematopoietic stem cells and CXCR4 expression with liver fibrosis in CCl4 rat model. BMC Gastroenterol 2023;23:323. [Crossref] [PubMed]
4. Strader DB, Wright T, Thomas DL, Seeff LBAmerican Association for the Study of Liver Diseases. Diagnosis, management, and treatment of hepatitis C. Hepatology 2004;39:1147-71. Erratum in: Hepatology 2004;40:269. [Crossref] [PubMed]
5. Chowdhury AB, Mehta KJ. Liver biopsy for assessment of chronic liver diseases: a synopsis. Clin Exp Med 2023;23:273-85. [Crossref] [PubMed]
6. Huttman M, Parigi TL, Zoncapè M, Liguori A, Kalafateli M, Noel-Storr AH, Casazza G, Tsochatzis E. Liver fibrosis stage based on the four factors (FIB-4) score or Forns index in adults with chronic hepatitis C. Cochrane Database Syst Rev 2024;8:CD011929. [Crossref] [PubMed]
7. Marri UK, Das P. Shalimar, Kalaivani M, Srivastava DN, Madhusudhan KS. Noninvasive Staging of Liver Fibrosis Using 5-Minute Delayed Dual-Energy CT: Comparison with US Elastography and Correlation with Histologic Findings. Radiology 2021;298:600-8. [Crossref] [PubMed]
8. Shinagawa Y, Sakamoto K, Sato K, Ito E, Urakawa H, Yoshimitsu K. Usefulness of new subtraction algorithm in estimating degree of liver fibrosis by calculating extracellular volume fraction obtained from routine liver CT protocol equilibrium phase data: Preliminary experience. Eur J Radiol 2018;103:99-104. [Crossref] [PubMed]
9. Srinivasa Babu A, Wells ML, Teytelboym OM, Mackey JE, Miller FH, Yeh BM, Ehman RL, Venkatesh SK. Elastography in Chronic Liver Disease: Modalities, Techniques, Limitations, and Future Directions. Radiographics 2016;36:1987-2006. [Crossref] [PubMed]
10. Dietrich CF, Bamber J, Berzigotti A, Bota S, Cantisani V, Castera L, Cosgrove D, Ferraioli G, Friedrich-Rust M, Gilja OH, Goertz RS, Karlas T, de Knegt R, de Ledinghen V, Piscaglia F, Procopet B, Saftoiu A, Sidhu PS, Sporea I, Thiele M. EFSUMB Guidelines and Recommendations on the Clinical Use of Liver Ultrasound Elastography, Update 2017 (Short Version). Ultraschall Med 2017;38:377-94. [Crossref] [PubMed]
11. Smith AD, Porter KK, Elkassem AA, Sanyal R, Lockhart ME. Current Imaging Techniques for Noninvasive Staging of Hepatic Fibrosis. AJR Am J Roentgenol 2019;213:77-89. [Crossref] [PubMed]
12. Qu Y, Middleton MS, Loomba R, Glaser KJ, Chen J, Hooker JC, et al. Magnetic resonance elastography biomarkers for detection of histologic alterations in nonalcoholic fatty liver disease in the absence of fibrosis. Eur Radiol 2021;31:8408-19. [Crossref] [PubMed]
13. Charatcharoenwitthaya P, Sukonrut K, Korpraphong P, Pongpaibul A, Saiviroonporn P. Diffusion-weighted magnetic resonance imaging for the assessment of liver fibrosis in chronic viral hepatitis. PLoS One 2021;16:e0248024. [Crossref] [PubMed]
14. Breit HC, Block KT, Winkel DJ, Gehweiler JE, Henkel MJ, Weikert T, Stieltjes B, Boll DT, Heye TJ. Evaluation of liver fibrosis and cirrhosis on the basis of quantitative T1 mapping: Are acute inflammation, age and liver volume confounding factors? Eur J Radiol 2021;141:109789. [Crossref] [PubMed]
15. Shaikh S. Editorial for "T1 Mapping on Gd-EOB-DTPA-Enhanced MRI for the Prediction of Oxaliplatin-Induced Liver Injury in a Mouse Model". J Magn Reson Imaging 2021;53:903-4. [Crossref] [PubMed]
16. Kirchin MA, Pirovano GP, Spinazzi A. Gadobenate dimeglumine (Gd-BOPTA). An overview. Invest Radiol 1998;33:798-809. [Crossref] [PubMed]
17. Poetter-Lang S, Bastati N, Messner A, Kristic A, Herold A, Hodge JC, Ba-Ssalamah A. Quantification of liver function using gadoxetic acid-enhanced MRI. Abdom Radiol (NY) 2020;45:3532-44. [Crossref] [PubMed]
18. Brismar TB, Dahlstrom N, Edsborg N, Persson A, Smedby O, Albiin N. Liver vessel enhancement by Gd-BOPTA and Gd-EOB-DTPA: a comparison in healthy volunteers. Acta Radiol 2009;50:709-15. [Crossref] [PubMed]
19. Yoshimitsu K, Mitsufuji T, Shinagawa Y, Fujimitsu R, Morita A, Urakawa H, Hayashi H, Takano K. MR elastography of the liver at 3.0 T in diagnosing liver fibrosis grades; preliminary clinical experience. Eur Radiol 2016;26:656-63. [Crossref] [PubMed]
20. Dzyubak B, Glaser K, Yin M, Talwalkar J, Chen J, Manduca A, Ehman RL. Automated liver stiffness measurements with magnetic resonance elastography. J Magn Reson Imaging 2013;38:371-9. [Crossref] [PubMed]
21. Hoffman DH, Ayoola A, Nickel D, Han F, Chandarana H, Shanbhogue KP. T1 mapping, T2 mapping and MR elastography of the liver for detection and staging of liver fibrosis. Abdom Radiol (NY) 2020;45:692-700. [Crossref] [PubMed]
22. Besa C, Bane O, Jajamovich G, Marchione J, Taouli B 3D. T1 relaxometry pre and post gadoxetic acid injection for the assessment of liver cirrhosis and liver function. Magn Reson Imaging 2015;33:1075-82. [Crossref] [PubMed]
23. Hojreh A, Lischka J, Tamandl D, Ramazanova D, Mulabdic A, Greber-Platzer S, Ba-Ssalamah A. Relative Enhancement in Gadoxetate Disodium-Enhanced Liver MRI as an Imaging Biomarker in the Diagnosis of Non-Alcoholic Fatty Liver Disease in Pediatric Obesity. Nutrients 2023;15:558. [Crossref] [PubMed]
24. Nagayama Y, Kato Y, Inoue T, Nakaura T, Oda S, Kidoh M, Ikeda O, Hirai T. Liver fibrosis assessment with multiphasic dual-energy CT: diagnostic performance of iodine uptake parameters. Eur Radiol 2021;31:5779-90. Erratum in: Eur Radiol 2021;31:8823-4. [Crossref] [PubMed]
25. Segnani C, Ippolito C, Antonioli L, Pellegrini C, Blandizzi C, Dolfi A, Bernardini N. Histochemical Detection of Collagen Fibers by Sirius Red/Fast Green Is More Sensitive than van Gieson or Sirius Red Alone in Normal and Inflamed Rat Colon. PLoS One 2015;10:e0144630. [Crossref] [PubMed]
26. Ucar F, Sezer S, Ginis Z, Ozturk G, Albayrak A, Basar O, Ekiz F, Coban S, Yuksel O, Armutcu F, Akbal E. APRI, the FIB-4 score, and Forn's index have noninvasive diagnostic value for liver fibrosis in patients with chronic hepatitis B. Eur J Gastroenterol Hepatol 2013;25:1076-81. [Crossref] [PubMed]
27. Wan Q, Peng H, Liu F, Liu X, Cheng C, Tie C, Deng J, Lyu J, Jia Y, Wang Y, Zheng H, Liang D, Liu X, Zou C. Ability of dynamic gadoxetic acid-enhanced magnetic resonance imaging combined with water-specific T1 mapping to reflect inflammation in a rat model of early-stage nonalcoholic steatohepatitis. Quant Imaging Med Surg 2024;14:1591-601. [Crossref] [PubMed]
28. Breit HC, Vosshenrich J, Heye T, Gehweiler J, Winkel DJ, Potthast S, Merkle EM, Boll DT. Assessment of hepatic function employing hepatocyte specific contrast agent concentrations to multifactorially evaluate fibrotic remodeling. Quant Imaging Med Surg 2023;13:4284-94. [Crossref] [PubMed]
29. Yang L, Ding Y, Rao S, Chen C, Wu L, Sheng R, Fu C, Zeng M. Staging liver fibrosis in chronic hepatitis B with T(1) relaxation time index on gadoxetic acid-enhanced MRI: Comparison with aspartate aminotransferase-to-platelet ratio index and FIB-4. J Magn Reson Imaging 2017;45:1186-94. [Crossref] [PubMed]
30. Li Q, Chen T, Shi N, Ye W, Yuan M, Shi Y. Quantitative evaluation of hepatic fibrosis by fibro Scan and Gd-EOB-DTPA-enhanced T1 mapping magnetic resonance imaging in chronic hepatitis B. Abdom Radiol (NY) 2022;47:684-92. [Crossref] [PubMed]
31. Obmann VC, Ardoino M, Klaus J, Catucci D, Berzigotti A, Montani M, Peters A, Todorski I, Wagner B, Zbinden L, Gräni C, Ebner L, Heverhagen JT, Christe A, Huber AT. MRI Extracellular Volume Fraction in Liver Fibrosis-A Comparison of Different Time Points and Blood Pool Measurements. J Magn Reson Imaging 2024;60:1678-88. [Crossref] [PubMed]
32. Wang HQ, Jin KP, Zeng MS, Chen CZ, Rao SX, Ji Y, Fu CX, Sheng RF. Assessing liver fibrosis in chronic hepatitis B using MR extracellular volume measurements: Comparison with serum fibrosis indices. Magn Reson Imaging 2019;59:39-45. [Crossref] [PubMed]
33. Costa-Silva L, Ferolla SM, Lima AS, Vidigal PVT, Ferrari TCA. MR elastography is effective for the non-invasive evaluation of fibrosis and necroinflammatory activity in patients with nonalcoholic fatty liver disease. Eur J Radiol 2018;98:82-9. [Crossref] [PubMed]
34. Xiao G, Zhu S, Xiao X, Yan L, Yang J, Wu G. Comparison of laboratory tests, ultrasound, or magnetic resonance elastography to detect fibrosis in patients with nonalcoholic fatty liver disease: A meta-analysis. Hepatology 2017;66:1486-501. [Crossref] [PubMed]
35. Loomba R, Cui J, Wolfson T, Haufe W, Hooker J, Szeverenyi N, Ang B, Bhatt A, Wang K, Aryafar H, Behling C, Valasek MA, Lin GY, Gamst A, Brenner DA, Yin M, Glaser KJ, Ehman RL, Sirlin CB. Novel 3D Magnetic Resonance Elastography for the Noninvasive Diagnosis of Advanced Fibrosis in NAFLD: A Prospective Study. Am J Gastroenterol 2016;111:986-94. [Crossref] [PubMed]
36. Chon YE, Jin YJ, An J, Kim HY, Choi M, Jun DW, Kim MN, Han JW, Lee HA, Yu JH, Kim SU. Optimal cut-offs of vibration-controlled transient elastography and magnetic resonance elastography in diagnosing advanced liver fibrosis in patients with nonalcoholic fatty liver disease: A systematic review and meta-analysis. Clin Mol Hepatol 2024;30:S117-33. [Crossref] [PubMed]
37. Kim MN, Han JW, An J, Kim BK, Jin YJ, Kim SS, et al. KASL clinical practice guidelines for noninvasive tests to assess liver fibrosis in chronic liver disease. Clin Mol Hepatol 2024;30:S5-S105. [Crossref] [PubMed]
38. Kupczyk PA, Mesropyan N, Isaak A, Endler C, Faron A, Kuetting D, Sprinkart AM, Mädler B, Thomas D, Attenberger UI, Luetkens JA. Quantitative MRI of the liver: Evaluation of extracellular volume fraction and other quantitative parameters in comparison to MR elastography for the assessment of hepatopathy. Magn Reson Imaging 2021;77:7-13. [Crossref] [PubMed]