Comparative study of 3D MR elastography and intravoxel incoherent motion for the evaluation of hepatocellular carcinoma grade

Abstract

Background and aims

Noninvasive preoperative radiologic prediction of histologic grade—a key prognostic factor—is invaluable. We aim to compare the diagnostic values of 3D magnetic resonance elastography (MRE), intravoxel incoherent motion (IVIM), and conventional contrast-enhanced magnetic resonance imaging (cMRI) in predicting the histologic grade of hepatocellular carcinoma (HCC).

Methods

This institutional review board-approved retrospective study included patients who underwent MRI between December 2014 and October 2021. Sixty-eight patients with pathologically confirmed HCCs who underwent MRE, IVIM, and cMRI imaging were included in the analysis. Two radiologists measured HCC stiffness volumetrically and over a single slice, and also measured apparent diffusion coefficient (ADC), IVIM-derived parameters, and enhancement ratio (ER) on arterial phase images via cMRI. Student’s t-test or the Mann–Whitney U test was used for group comparisons. Receiver operating characteristic (ROC) curve analyses were performed to evaluate the diagnostic performance.

Results

Histologically, fifty-three (78%) patients had well-differentiated or moderately differentiated HCCs, and fifteen (22%) patients had poorly differentiated HCCs. Both the volumetric stiffness and single-ROI tumor stiffness were significantly elevated in the poorly differentiated HCC group (P < 0.001, P = 0.001), and the volumetric stiffness was a better measurement of stiffness because it had a higher ROC curve value (0.816). However, the ADC, the true diffusion coefficient (D), the pseudodiffusion coefficient (D∗), the pseudodiffusion fraction (f), and ER during the arterial phases on cMRI were not significantly different between the two groups (P = 0.309, 0.187, 0.440, 0.350, and 0.714, respectively).

Conclusions

Stiffness measured with 3D MRE may be useful for noninvasively predicting HCC histologic grade, and the volumetric measuring method achieved the highest ROC curve value, outperforming single-ROI HCC stiffness, IVIM parameters, and arterial-phase ER on cMRI.

Highlights

• •MRE effectively differentiates the histologic grade of HCC noninvasively and outperforms single ROI-based method.
• •The difference of the IVIM parameters and the CE-MRI ER between different histologic grades had no statistical significance.
• •Volumetric MRE is superior to IVIM and contrast-enhanced MRI for predicting the histologic grade of HCC.

1. Introduction

Liver cancer is one of the most common malignancies worldwide.^1,2 Hepatocellular carcinoma (HCC) accounts for approximately 80% of liver cancer cases and is among the top five causes of cancer-related deaths in 90 countries, especially in Asia.³ Histologic grade is one of the most important factors related to patient prognosis.⁴ The main risk factor for HCC is cirrhosis.³ Multiple international liver disease organizations agree that the diagnosis of HCC in patients with cirrhosis can be made through imaging.⁵ Therefore, the noninvasive preoperative radiologic prediction of the histologic grade would be valuable. Some studies have already focused on conventional contrast-enhanced MRI (cMRI) imaging characteristics, such as tumor size or enhancement pattern, restricted diffusion on diffusion-weighted imaging (DWI) images, and uptake on hepatobiliary phase images to evaluate the histological grade of HCC.6, 7, 8

Magnetic resonance elastography (MRE) is a noninvasive MRI-based quantitative technique that has been used clinically most extensively in examining the liver.⁹ Several studies have suggested that tumor stiffness measured via MRE may be a candidate imaging biomarker for differentiating HCC tumor grade10, 11, 12; however, there is no consensus on this prospect. Thompson et al.¹⁰ used a 2D MRE technique and reported that well- or moderately-differentiated HCCs were stiffer than their poorly differentiated counterparts, whereas Wang J et al.¹² reported that tumor stiffness based on 3D MRE was greater in poorly differentiated HCCs than in moderately to well-differentiated HCCs. This discrepancy may be attributable to the fact that 3D MRE allows for a more extensive liver evaluation, capturing the varied spatial patterns of parenchymal alterations better than 2D MRE does.¹¹

Intravoxel incoherent motion (IVIM) modeling is a widely used bi-exponential model for analyzing DWI signals, compared with conventional DWI, which uses a mono-exponential model.¹³ It can reflect tissue structural features and pathological alterations more precisely. IVIM has potential for classifying HCC histologically; however, findings from various studies on the associations between IVIM parameters and HCC grade are inconsistent. Some studies have revealed that the true diffusion coefficient (D) and the apparent diffusion coefficient (ADC) are negatively associated with the histologic grade of HCC.¹⁴ In contrast, values of the pseudodiffusion coefficient (D∗) and pseudodiffusion fraction (f) have not shown a significant association with histological grade.¹⁵

As of the date of writing, no published study has simultaneously compared 3D MRE, IVIM, and cMRI for the preoperative assessment of HCC histologic grade. Therefore, this study was performed to investigate the diagnostic performance of MRE using different ROI methods and to compare it with the performance of IVIM parameters and the cMRI-based ER for predicting the histologic grade of HCC.

2. Materials and methods

2.1. Ethical approval

This single-center retrospective study was conducted in accordance with the Declaration of Istanbul and the Declaration of Helsinki. The study protocol was approved by the Institutional Review Board of The Third Affiliated Hospital of Sun Yat-sen University (No. II2024-151-01). The board waived the requirement for informed consent due to the retrospective and anonymous nature of the study.

2.2. Patients

From December 2014 to October 2021, through a computerized search of the medical data, a total of 68 patients with surgically confirmed HCCs were enrolled (Fig. 1). The inclusion criteria were the presence of chronic liver disease, with MRE and IVIM examinations prior to surgery within one month, and pathologically confirmed HCC. We excluded patients according to the following exclusion criteria: patients with HCC due to needle-biopsy, patients absence of IVIM examination, patients with mixed HCC/cholangiocarcinoma confirmed by liver resection and pathology, patients with previous treatments (e.g., radiofrequency ablation, transarterial chemoembolization), patients with lesions less than 2.5 cm in size, patients with slice misregistration, and patients with distinct motion artifacts.

Fig. 1.

Flowchart of the patient selection process. Abbreviations: HCC, hepatocellular carcinoma; IVIM, intravoxel incoherent motion; MRE, magnetic resonance elastography.

2.3. 3D MRE and IVIM scanning

All the subjects underwent cMRI, 3D MRE, and IVIM examinations using a 3.0T MR system (Discovery MR750, GE Healthcare, Milwaukee, WI, USA) with an 8-channel, phased-array, torso coil. Patients were instructed to fast for a minimum of 4 h before the MRI exams to avoid potential physiological effects. The liver protocols and the detailed acquisition parameters of the conventional enhanced MRI, MRE, and IVIM sequences used are summarized in Table 1.

Table 1.

cMRI, MRE, and IVIM acquisition parameters.

Sequence	Matrix	TR/TE (ms)	FA (°)	FOV (cm2)	Bandwidth (kHz)	Slice thickness (mm)	Slice gap (mm)	Acquisition time	Number of slices
cMRI
FIESTA	224 × 256	3.1/1.1	45	38 × 38	125	8	1	10 s	16
SSFSE	384 × 160	1222/80.9	90	44 × 44	83.3	8	0	21 s	18
T2WI	320 × 320	6000/72.8	90	36 × 36	83.3	5	1	3 min 36 s	32
LAVA-Flex pre-contrast	260 × 224	3.7/1.7	15	36 × 36	200	5	1	14 s	80
LAVA-Flex Dynamica	260 × 224	3.7/1.7	12	36 × 36	200	5	1	–	80
MRE
60Hz
SE-EPI	80 × 80	334/52	90	44 × 44	250	3.6	0	64 s	32
IVIMb
DW single-shot EPI	128 × 128	6000-10,000/56	90	38 × 30	250	5	1	4–7 min	27–32

Abbreviations: cMRI, conventional enhanced-MRI; DW, diffusion-weighted; EPI, echo-planar imaging; FA, flip angle; FIESTA, fast imaging employing steady-state acquisition; FOV, field of view; IVIM, intravoxel incoherent motion; LAVA, liver acquisition with volume acceleration; MRE, magnetic resonance elastography; SE-EPI, spin-echo echo-planar imaging; SSFSE, single-shot fast spin echo; TE, echo time; TR, repetition time.

^aA fat-suppressed 3D gradient-echo sequence with flexible echo times, used for dynamic contrast-enhanced MRI. After the contrast-medium injection, arterial phase, portal venous phase, and delayed-phase images were subsequently acquired at 15–20 s, 60 s, and 180 s, respectively.

^bThe repetition time of IVIM is automatically calculated on the basis of the respiratory rate and the respiratory interval number.

2.3.1. 3D MRE scan

3D MRE was performed before any intravenous contrast, as previously described,¹¹ using a multislice, flow-compensated, spin-echo echo-planar imaging (SE-EPI) MRE sequence. A pneumatic, passive, drum driver was positioned over the tumor-containing lobe (left or right) of the liver, at the level of the xiphisternum, and secured with an elastic belt. Continuous 60-Hz mechanical vibrations were generated by an active acoustic driver located outside the scanning room and transmitted through polyvinyl chloride tubing to the passive driver to produce shear waves in the liver. The acquisition, with a total time of approximately 64 s, was performed during either three 21-s breath-holds or six about 11-s breath-holds. Other key sequence parameters included a 90-degree flip angle, an 80 × 80 acquisition matrix, a field of view 44 × 44 cm, slice thickness 3.6 mm, inter-slice gap 0 mm, and 32 slices. The standard liver 3D MRE reconstruction, which utilizes direct inversion of the Helmholtz wave equation with spatiotemporal directional filtering, was automatically generated by the scanner.

2.3.2. IVIM scanning and postprocessing

IVIM was performed using a respiratory-triggered single-shot spin-echo echo-planar imaging (EPI) sequence in the transverse plane before contrast-enhanced imaging. The protocol acquired 27 to 32 slices per patient. Eleven b values from 0 to 1500 s/mm² (0, 30, 50, 100, 150, 200, 300, 500, 800, 1000, and 1500) were applied, with the number of excitations (NEX) for each b value being 1, 1, 1, 1, 1, 1, 2, 4, 4, 6, and 6, respectively.

The apparent diffusion coefficient (ADC) values were estimated by fitting diffusion-weighted signals at all b values (0–1500 s/mm²) to the following mono-exponential equation:

S_b/S₀ = exp(−b·ADC)

where S_b is the signal intensity at a given b value and S₀ is the signal intensity without diffusion weighting.

The IVIM model employs a bi-exponential fitting of the diffusion decay curve to separate true diffusion from “pseudodiffusion”, which reflects tissue microperfusion. The signal intensity curves from a DWI acquisition with multiple b values are described by the following equation:¹⁶

S_b/S₀ = (1−f)·exp(−b·D) + f·exp (−b·D∗)

where S_b and S₀ are the signal intensities with and without diffusion weighting, respectively, D (true diffusion coefficient) represents pure molecular diffusion; D∗ (pseudodiffusion coefficient) represents incoherent microcirculation within the voxel, and f (pseudodiffusion fraction) is the proportion of the pseudodiffusion. To robustly separate the diffusion and perfusion effects in equation,¹⁷ a typical multistep approach is applied.¹⁸ Initially, D is estimated by mono-exponential fitting of the diffusion-weighted signals at high b-values (b > 200 s/mm²), assuming that perfusion contributions are negligible in the equation.¹⁹

S_b/S₀ = exp(−b·D)

The pseudodiffusion fraction f and pseudodiffusion coefficient D∗ are subsequently estimated by fitting the measured signal intensity at all b values to an equation with a fixed D.

2.4. Image analysis and measurement of ER, tumor stiffness, and IVIM metrics

All images were interpreted in consensus by two board-certified abdominal radiologists (blinded to all the clinical data and histological results), each with 18 years of experience, to assess image quality and the reliability of the tumor parameter measurements. A workstation with FuncTool software (version AW 4.6, GE Healthcare, Chicago, IL, USA) was used for postprocessing.

For the volumetric analysis, regions of interest (ROIs) were manually drawn on every slice demonstrating the tumor on the magnitude images for 3D MRE, using T2-weighted and contrast-enhanced images as anatomical references. For HCCs with nonenhancing/necrotic components, ROIs were carefully drawn to include only the solid portions. ROIs were drawn to be as large as possible while excluding tumor margins (where partial-volume effects may affect the calculated stiffness and IVIM parameters), areas of significant wave interference (e.g., liquefactive necrosis in the tumor), and any other artifacts seen on the magnitude and phase images.

ROIs were placed on the focal lesions on the ADC map at the same or similar level wherever possible to the stiffness map, and then the ROIs were copied to D, D∗, and f maps. We imported the images in DICOM format into the ITK-SNAP software (version 3.8.0, Cognitica, Philadelphia, PA, USA; https://www.itksnap.org) to manually draw the tumor boundary slice by slice. The software then provided the 3D reconstruction and its volume value (Fig. 2, Fig. 3). Moreover, the single ROI-based stiffness and the ER in the arterial phase at the maximum tumor cross-section were recorded. The signal intensity (SI) of the tumor was measured for the ER analysis of enhancement characteristics in the arterial phase on cMRI.

Fig. 2.

A surgically confirmed, well-differentiated HCC in a 35-year-old man. The circles in Figure A–I represent the ROIs. (A) Axial, fat-suppressed, T2WI images. (B) Axial, diffusion-weighted images with b = 800 s/mm². (C) Axial, fat-suppressed, precontrast T1WI image shows a tumor signal intensity of 383.0. (D) Axial, fat-suppressed, enhanced T1WI in the arterial phase image shows a tumor signal intensity of 509.9. (E) The 3D MRE reveals a tumor stiffness of 3.72 ± 0.18 kPa. (F–I) The ADC, D, D∗, and f maps show that the mean values of ADC, D, D∗, and f in the tumor were 0.91 × 10⁻³mm²/s, 0.72 × 10⁻³mm²/s, 3.36 × 10⁻³mm²/s, and 25.12%, respectively. (J) Coronal, fat-suppressed, T1W contrast-enhanced images. (K) The VOI was created by merging all the ROIs covering the entire tumor, and the VOI was 318546.97 mm³. Abbreviations: HCC, hepatocellular carcinoma; VOI, volume of interest.

Fig. 3.

A surgically confirmed, poorly differentiated HCC in a 42-year-old woman. The circles in Figure A–I represent the ROIs. (A) Axial, fat-suppressed, T2WI image. (B) Axial, diffusion-weighted image with b = 800 s/mm². (C) Axial, fat-suppressed, precontrast T1WI image shows a tumor signal intensity of 385.8. (D) Axial, fat-suppressed, enhanced T1WI in the arterial phase image shows a tumor signal intensity of 504.9. (E)The 3D MRE reveals a tumor stiffness of 6.46 ± 0.32 kPa. (F–I) The ADC, D, D∗, and f maps show that the mean values of ADC, D, D∗, and f in the tumor were 0.98 × 10⁻³mm²/s, 0.86 × 10⁻³mm²/s, 4.39 × 10⁻³mm²/s, and 26.45%, respectively. (J) Coronal, fat-suppressed, T1W contrast-enhanced images. (K) The VOI was created by merging all the ROIs covering the entire tumor, the VOI was 28230 mm³. Abbreviations: HCC, hepatocellular carcinoma; VOI, volume of interest.

The ER in the arterial phase based on cMRI was calculated according to the established formula:⁸

ER=(tumor signal _postcontrast−tumor signal _precontrast)/tumor signal _precontrast.

2.5. Histological analysis

In patients with multiple tumors, the lesion with the largest cross-sectional area was selected for imaging analysis and histological grading. All specimens were obtained by surgical resection.

Histologic evaluation was performed independently by two experienced hepatopathologists who were blinded to clinical and MRI data. On hematoxylin-and-eosin-stained sections, tumor grade was classified as either poorly differentiated HCC or well-/moderately differentiated HCC according to the 2020 World Health Organization classification.²⁰ When heterogeneous differentiation coexisted, the predominant pattern (>50% of the tumor area) determined the final grade;¹¹ discrepancies between the two observers were resolved by consensus.

2.6. Statistical analyses

Categorical variables were summarized as counts and percentages; group comparisons were performed with χ² or Fisher’s exact test. For continuous variables, the data were tested for normality via the Shapiro-Wilk test. Data with normal distribution are expressed as mean ± standard deviation (SD) and compared using independent-sample t-tests. Data without normal distribution are expressed as median (interquartile range) and compared using Mann-Whitney U tests. The intraclass correlation coefficient (ICC) was calculated to assess the inter- and intra-observer agreement.

ROC curve analyses were performed to evaluate the diagnostic performance of each parameter in distinguishing the histologic grade of HCC. All the statistical analyses were performed using SPSS v.22.0 (IBM, Armonk, NY, USA). P < 0.05 for two-sided tests was considered statistically significant.

3. Results

3.1. Patient characteristics

The characteristics of the patients evaluated in this study are summarized in Table 2. Among the 68 patients ultimately enrolled, 51 patients underwent local tumor resection, 13 patients underwent hemihepatectomy, and 4 underwent orthotopic liver transplantation. Patients were divided into two groups on the basis of tumor grade. Well- or moderately differentiated HCCs were detected in 53 patients (78%) and poorly differentiated HCCs were detected in 15 patients (22%). There were 23 (33.8%) lesions with obvious necrosis and hemorrhage. The majority of patients (48/68, 70.6%) had a single HCC lesion. The diameter and volume of the tumors ranged from 26 to 90 mm (median: 53 mm) and (27.40–220.50) × 10³ mm³ (median:74.16 × 10³ mm³), respectively.

Table 2.

Demographic and clinical characteristics of the 68 patients with HCC.

Characteristics	All patients (n = 68)	Poorly differentiated HCC (n = 15)	Well-or moderately differentiated HCC (n = 53)	P- value
Demographics
Age, year	49.24 ± 10.78	48.46 ± 8.24	49.43 ± 11.40	0.771
Sex, male, n (%)	62 (91.1)	14 (93.3)	48 (90.6)	1.000
BMI (kg/m2)	22 (20–25)	21 (19–23)	22 (22–24)	0.110
Chronic liver disease
HBV	65	15	50
HCV	2	0	2
Alcoholic	1	0	1
Radiological characteristics
Number of tumors				0.370
Single, n (%)	48 (70.6)	12 (80.0)	36 (67.9)
Multiple, n (%)	20 (29.4)	3 (20.0)	17 (32.1)
Tumor size, mm	53 (26–90)	48 (32–90)	58 (26–85)	0.580
Tumor size group				0.270
< 50 mm	32 (47.1%)	9 (60.0%)	23 (43.4%)
≥ 50 mm	36 (52.9%)	6 (40.0%)	30 (56.6%)
Tumor volume, × 103mm3	74.16（27.40–220.50）	64.35 (28.23–298.39)	57.3（23.52-136.26）	0.640
Child-Pugh
A	65 (95.6%)	14 (93.3%)	51 (96.2%)	1.000
B	3 (4.4%)	1 (6.7%)	2 (3.8%)

Data are expressed as n (%), mean ± standard deviation (SD), or median (interquartile range).

Abbreviations: BMI, body mass index; HBV, hepatitis B virus; HCV, hepatitis C virus; HCC, hepatocellular carcinoma.

3.2. Interobserver reproducibility

For all parameters, ICCs ranged from 0.85 to 0.95, indicating very good to excellent reliability. The ICC values were 0.983 (95% CI: 0.973–0.990) for the volumetric method of 3D MRE stiffness, 0.900 (95% CI: 0.837–0.938) for the ADC, 0.896 (95% CI:0.885–0.917) for D, 0.887 for D∗ (95% CI: 0.876–0.912), 0.856 for f (95% CI: 0.847–0.870), 0.940 (95% CI: 0.921–0.949) for the single-ROI method of MRE stiffness and 0.885 for ER (95% CI: 0.860–0.901).

3.3. Comparison of MRI parameters with histopathological grade

The mean values of stiffness based on the single-ROI method and the volumetric method, the mean values of ADC, D, D∗, f, and the ER in the arterial phase for the two HCC groups are summarized in Table 3. The mean stiffness values based on the volumetric and single-ROI methods were significantly elevated in the poorly differentiated HCC group (P < 0.001 and P = 0.001, respectively; Fig. 4). The ADC, D, D∗, f, and the ER values in the arterial phase based on the single-ROI method were not significantly different between the two groups (P = 0.309, 0.187, 0.440, 0.350, and 0.714, respectively; Table 3).

Table 3.

Comparison of MRI parameters between poorly differentiated and well-or moderately differentiated HCC groups.

Parameters	Poorly differentiated HCC (n = 15)	Well-or moderately differentiated HCC (n = 53)	P-value
MRE
Stiffness (kPa)
Single-ROI	6.31 ± 1.49	4.85 ± 1.39	0.001
Volumetric	6.85 ± 1.53	5.03 ± 1.29	<0.001
IVIM
ADC ( × 10−3mm2/s)	0.66 ± 0.15	0.62 ± 0.11	0.309
D ( × 10−3mm2/s)	0.90 ± 0.18	0.84 ± 0.15	0.187
D∗( × 10−3mm2/s)	18.76 ± 7.53	17.28 ± 6.22	0.440
f (%)	30.38 ± 10.35	28.05 ± 7.88	0.350
cMRI
ER	0.95 ± 0.48	1.11 ± 0.62	0.714

Data are expressed as mean ± standard deviation (SD).

Abbreviations:cMRI, conventional MRI; ER, enhancement ratio; HCC, hepatocellular carcinoma; IVIM, intravoxel incoherent motion; MRE, magnetic resonance elastography; ROI, region of interest.

Fig. 4.

Box and whisker plot of different parametersvs.histologic differentiation.(A) Tumor stiffness measured by the volumetric method; (B) Tumor stiffness measured by the single-ROI method;(C) ADC values; (D) D values; (E) D∗ values;(F) f values.

Differences in the measured mean stiffness of HCC between the volumetric method and the single-ROI method were observed within each histological subgroup. In the poorly differentiated HCC group, the mean stiffness values were 6.85 ± 1.53 kPa and 6.31 ± 1.49 kPa for the volumetric and single-ROI methods, respectively, with a mean difference of 0.54 kPa (P = 0.028). In the well- or moderately differentiated HCC group, the corresponding values were 5.03 ± 1.29 kPa and 4.85 ± 1.39 kPa, with a mean difference of 0.18 kPa (P = 0.54).

3.4. Diagnostic performance of 3D MRE and IVIM parameters

The areas under the receiver operating characteristic (ROC) curves for distinguishing the poorly differentiated HCCs from well-or moderately differentiated HCCs are shown in Fig. 5. The volumetric stiffness had the highest area under the curve (AUC) with a value of 0.816 (P < 0.001). The AUC values of single-ROI stiffness, and the ADC, D, D∗, and f values derived from IVIM were 0.770 (P = 0.002), 0.565 (P = 0.446), 0.621 (P = 0.156), 0.579 (P = 0.355), and 0.544 (P = 0.605), respectively.

Fig. 5.

ROC curves of each parameters. The AUC values of volumetric stiffness, single-ROI stiffness, ADC, D, D∗, and f values were 0.816, 0.770, 0.565, 0.621, 0.579, and 0.544, respectively.

4. Discussion

Our study revealed that the mean stiffness value for the whole tumor of the poorly differentiated HCC group was significantly greater than that of the well-differentiated or moderately differentiated HCC groups on the basis of both the volumetric method and single-ROI method. Volumetric stiffness also had the highest AUC in distinguishing the two groups. Statistical analysis revealed no significant differences in the IVIM-derived parameters (ADC, D, D∗, f) or the ER between the two groups. In summary, mean stiffness is markedly higher in the poor-grade group, and the volumetric technique outperforms all other tested parameters. Whole-tumor stiffness provides a quantitative, reproducible, and biologically plausible marker of tumor aggressiveness. This finding offers clinicians a non-invasive indicator of histologic grade that can guide patient counseling, surgical planning, and the intensity of post-treatment surveillance.

Several studies have attempted to evaluate the histologic grade of HCC with MRE imaging markers. Our study revealed that tumor stiffness was elevated in poorly differentiated HCCs, a finding consistent with Wang et al.¹² and that volumetric analysis demonstrated a higher AUC and may therefore provid higher diagnostic value than a single-ROI method. The elevated stiffness in poorly differentiated HCCs can be explained by their distinct tumor composition according to the WHO classification.²⁰ Well-differentiated lesions frequently exhibit fatty changes, and moderately differentiated HCCs often contain bile or proteinaceous fluid. In contrast, poorly differentiated HCCs usually have a solid pattern without distinct sinusoid-like blood spaces. This shift towards a compact, cellular-solid architecture logically corresponds to the increased stiffness we observed. In contrast to our findings, Thompson et al.¹⁰ reported that well or moderately differentiated HCCs were stiffer than poorly differentiated HCCs. The discrepancy between our findings and those of Thompson et al. may be attributed to several methodological differences. First, their study had a relatively small sample size (n = 21) and relied on biopsy for pathological grading, which is susceptible to sampling error given tumor heterogeneity; in contrast, our study used surgical resection specimens, ensuring precise grading. Second, their cohort had diverse etiologies, while ours was predominantly HBV-related. Third, they employed 2D MRE, whereas we used 3D MRE. Finally, and perhaps most importantly, 3D MRE provides more accurate and comprehensive stiffness measurements by assessing a larger tissue volume and minimizing partial volume effects.10, 11, 12^,21,22

The ADC, D, D∗, and f values calculated from IVIM images of the whole tumor were not as useful as volumetric stiffness for predicting the histological grade of HCC in our study. The current results of different studies on the associations between IVIM-derived parameters and the histologic grade of HCC are inconsistent.¹³ Some studies have revealed that D and ADC values are negatively associated with the histologic grade of HCC.^14,23 During the formation of tumors, the increased cellular density and nuclear/cytoplasmic ratio may restrict the diffusion process; these changes caused the ADC and D values to decrease as the histologic grade increases.²³ Some studies have also used the conventional single exponential model DWI. One study revealed that the ADC and D values of the poorly differentiated HCCs were lower than those of well- and moderately differentiated HCCs,²⁴ whereas Xu et al.²⁵ reported that poorly differentiated HCCs had significantly higher ADC values than well-differentiated HCCs. However, Nasu et al.²⁶ reported that the histologic grade of HCCs was not correlated with ADC values, which is similar to our findings. The inconsistencies between different studies may have several possible causes. First, heterogeneity of HCC tumors could be a contributing factor.²⁷ Most of the patients (52.9%) in our study had tumors with diameters ≥50 mm. Tumors with large volumes are prone to have necrosis and hemorrhage, and ADC variations are accompanied by an increase in the true molecular component of diffusion D in itself, which is correlated with tumor necrosis and hemorrhage.²⁸ Furthermore, in our study, 78% of patients had well-differentiated or moderately differentiated HCCs. A higher tumor histological grade results in greater variability and a more complicated microstructure, leading to greater differences in the IVIM parameters. Second, we used the volumetric ROI analysis to cover the whole tumor without the necrosis and hemorrhage, which is different from the methods used in most other studies, in which a single ROI based on the largest tumor cross-section was used.^15,23 Although Wei et al.²⁹ reported that the mean ADC and D values derived from IVIM by the whole-tumor volume method were able to assess tumor grade, future studies with larger sample sizes via volumetric analysis are warranted to explore the relationships between tumor histological grade and IVIM metrics. Third, the previous studies have different case compositions. Our unique patient population (high number of HBV cases) is different from those in European and Western countries where cases due to hepatitis C and metabolic dysfunction-associated fatty liver disease (MAFLD) predominate.^14,15,23 Finally, we examined 11 b values, including only 1 low b value (0 = 30 s/mm²) in our study. One study recommend at least two low b (0 < b < 50 s/mm²) values for liver IVIM-DWI, as pseudodiffusion metrics tend to be underestimated in the liver.³⁰

There were several limitations in our study. First, the retrospective, single-center design and small sample size of our study limit the generalizability of the findings. The cohort included only six female patients, indicating a significant gender imbalance that may introduce selection bias and reduce the representativeness of the patient population. Second, our unique patient population (high number of HBV cases) is different than that in European and Western countries, where cases due to HCV and MAFLD predominate. We also excluded lesions less than 2.5 cm in size; therefore, the study population does not represent the entire clinical spectrum of HCC and imaging results. Third, grouping into the two histological grades (poorly differentiated and moderately differentiated or well-differentiated) reflects the heterogeneity spectrum of HCCs. Finally, this study selected the largest tumor to evaluate its imaging marker, which may not always reflect the differentiations in other smaller HCCs.

5. Conclusions

The mean stiffness measured by 3D MRE via both the volumetric method and the single-ROI method is useful for predicting the histological grade of HCC preoperatively, and the volumetric measuring method achieved the highest AUC value, outperforming single-ROI HCC stiffness, IVIM parameters, and arterial-phase ER on cMRI.

Authors' contributions

Weimin Liu: Writing – original draft, Methodology, Formal analysis, Data curation. Sidong Xie: Methodology, Data curation. Wenjie Tang: Writing – review & editing. Ka Zhang: Writing – review & editing, Visualization, Supervision.

Data availability statement

The data that support the findings of this study are available on request from the corresponding author.

Declaration of competing interest

The authors declare that there are no conflicts of interest.