Hepatocellular Carcinoma: In Vivo Evaluation of Water Percentage as a Prognostic Biomarker Using Magnetic Resonance Imaging 3D-VIBE Multiecho Dixon

Abstract

Introduction:

It is urgent to find an effective method to diagnose and prognose early hepatocellular carcinoma (HCC). The purpose of this study was to investigate the correlation between HCC histological degree and water percentage (WP) obtained from magnetic resonance imaging 3D-VIBE multiecho Dixon, and to evaluate the feasibility of WP in the postoperative prediction of early HCC recurrence.

Methods and Materials:

From June 2016 to July 2017, 76 patients with diagnostic HCC all underwent 3D-VIBE Multiecho Dixon and ultrahigh b value diffusion-weighted imaging (DWI) examination. Freehand regions of interests were placed to measure the WP and apparent diffusion coefficient (ADC) value. The Edmondson–Steiner (E-S) grades proved by histopathological results were acquired from all patients. Comparisons between mean WP and ADC with E-S grades I–IV were performed using Kruskal–Wallis test and one-way ANOVA. Least Significant Difference t-test (LSD-t test) was applied to compare particular pairs of mean ADC value between every two E-S groups. Correlations between WP, ADC, and E-S grades were assessed by Spearman's rank correlation test. The Mann–Whitney U test was utilized to compare the difference of mean WP between recurrence and nonrecurrence group. The receiver operating characteristic (ROC) curves were calculated to estimate the diagnostic effect of 3D-VIBE Multiecho Dixon and ultrahigh b value DWI to HCC. Kaplan–Meier method was used to evaluate the recurrence free survival (RFS) after surgical resection.

Results:

Mean WP values among groups E-S I to IV were 91.8%, 95.2%, 96.4%, and 97.7%, respectively. A positive correlation was exhibited between the WP and histopathological E-S grades (r = 0.480, p < 0.01). The ADC values based on E-S grades were 0.93, 0.82, 0.74, and 0.62 × 10⁻³ mm²/s, respectively. Significant differences were found between every two E-S groups (p < 0.01), and a negative correlation between ADC and E-S grades (r = −0.784, p = 0.000) was observed. Mean value of WP was 97.2% in recurrence group and 94.6% in nonrecurrence group (p < 0.01). The optimal cutoff value was 95.7%. The maximal area under the ROC curve was 0.747 ± 0.06 for WP, 0.631 ± 0.07 for ADC, and 0.753 ± 0.06 for the combination of WP and ADC. Mean RFS was 18.1 months in the lower WP and 10.7 months in higher WP group (p < 0.01).

Conclusions:

WP can be used as a potential biomarker for the diagnosis and prognosis of HCC. A lower value of WP may imply a better outcome in patients after surgical resection.

Introduction

Hepatocellular carcinoma (HCC) is the sixth most common cancer worldwide, with a rising incidence.¹ However, the overall prognosis of HCC is poor because the initial diagnosis is often at late stage or after the postoperative intrahepatic recurrence.² Accurate preoperative prediction of intrahepatic recurrence is of great importance, because the treatment and prognosis of these entities can be quite different.

Water is essential for tumor growth, which can convert the energy produced by contractile proteins into the chemical energy and keep the balance of ion accumulation and electron transport.³ Furthermore, intracellular water is useful in maintaining conformation and function of all biomolecules.^4,5 Recently, more and more studies have investigated the correlation between water content and tumor differentiation grade. Ali et al. demonstrated that water content was heavily associated with the differentiated grade of prostate cancer.³ Cerussi et al. reported that water content increased in breast cancer tissue compared with normal tissue, and it was also closely correlated with histological grade.^6,7 Thus, water content may be a valuable biomarker for the assessment and the prognosis of malignant tumors. As a quantitative metrics for water content, water percentage (WP) can be calculated by magnetic resonance imaging (MRI) 3D-VIBE multiecho Dixon that can quantify the water content in HCC noninvasively compared with cell refractometry and Terahertz spectroscopy.^8,9 Diffusion-weighted imaging (DWI) is useful in detecting the molecular diffusion and reflecting the aquaporin (AQP) 9 expression by applying ultrahigh b value. DWI can also determine the diffusion properties of water molecules. However, the relationship among WP, MRI 3D-VIBE multiecho Dixon, ultrahigh b DWI, and the histopathological characteristic is still inconclusive. To the authors' knowledge, there have been no reports on the applications of 3D-VIBE multiecho Dixon and DWI combined to investigate the relationship between WP and histopathological grade, and to predict the prognosis in HCC patients.

The purpose of this study was to prospectively assess the relationship between WP and histopathological grade using 3D-VIBE multiecho Dixon and ultrahigh b DWI as measuring modalities, and to evaluate the feasibility of WP as biomarker for postoperative prediction of early recurrence of HCC.

Materials and Methods

Patients

This study was approved by the institutional review board and ethics committee of authors' university (approve no. 2016-(297)), and written informed consent was obtained from all patients. This study was conducted in accordance with the declaration of Helsinki. From Jun 2016 to July 2017, 130 consecutive patients with suspected HCC underwent MRI 3D VIBE multiecho Dixon and DWI examination. Among them, 54 patients were excluded for the following reasons: (1) patients with late-stage HCC who could not receive operation (n = 25), (2) patients' pathology did not confirm HCC (intrahepatic cholangiocarcinoma, ect. = 18), and (3) image qualities were not satisfied for diagnosis (n = 13). Eventually, 76 patients were enrolled in this study. Time interval between MRI and surgery was 2–3 d. All of the 76 patients had type B hepatitis, and 12 patients had liver Cirrhosis.

Imaging technique

All examinations were carried out using a 3.0T MRI system (MAGNETOM skyra, Siemens Medical Solutions, Germany) with 18 channels body phased-array coils. The 3D-VIBE multiecho Dixon was obtained with gradient echo (GRE) sequence using the following parameters: TR 9.15 ms, TE 1.05, 2.46, 3.69, 4.92, 6.15, 7.38 ms, acquisition matrix = 160 × 95; flip angle (FA) 4° to reduce the T1-weighting with six echoes Dixon reconstruction; partitions 72, partition thickness 3.5 mm, breath-hold acquisition time 13 s; pixel size 2.62 × 3.5 mm, field of view (FOV) 420 × 315 mm; bandwidth 1040 Hz/Px. A Levenberg-Marquardt nonlinear fitting was then used to fit the magnitude of the complex signal of the multiecho data. In brief, Equations (1) and (2) are the signal models for the fitting, depending on the acquisition scheme and reconstruction configuration: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} \vert {S_n} \left\vert { \; = \;} \right\vert \; \left( {{M_w} + {c_n}{M_f}} \right) \cdot {e^{ - R_{2eff}^*T{E_n}}} \vert \;. \tag{1} \end{align*} \end{document} \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} \vert {S_n} \left\vert { \; = \;} \right\vert { \rm{ }} ( {M_w} \cdot {e^{ - R_{2w}^*T{E_n}}} + {c_n}{M_f} \cdot {e^{ - R_{2f}^*T{E_n}}} ) { \rm{ }} \vert. \tag{2} \end{align*} \end{document}

In the equations, \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document}$${{ \rm{c}}_n}$$ \end{document} is the complex coefficient defined as \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document}$${{ \rm c}_n} = \sum {W_i}{e^{j \left( {2{ \rm{ \pi }} \Delta {{ \rm{f}}_i}} \right) T{E_n}}}$$ \end{document} for a seven-peak fat model. W_i and \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document}$$\Delta {{ \rm{f}}_i}$$ \end{document} are the weighting factor and the resonance frequency offset of the \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document}$${i_{th}}$$ \end{document} fat peak, respectively.¹⁰ The resultant images are generated into different series using six echoes, including fat percentage maps, WP maps, fat maps, and water maps. The higher the percentage of water in HCC tumor, the higher the signal intensity in the WP image.

The DWI was conducted with ultrahigh b value to validate the slow diffusion due to the correlation between apparent diffusion coefficient (ADC) and water content.¹¹ The applied parameters were as follows: TR 5600 ms, TE 63 ms, slices 25, thickness 5.0 mm, and acquisition time 154 s. Bandwidth 1924 Hz/Px, acquisition matrix = 100 × 76. b = 0, 100, 700, 1400, and 2100 (mm²/s). FOV 380 mm, voxel size 3.8 × 3.8 × 5 mm. The equation for DWI fitting was as follows: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\usepackage{upgreek}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*} \frac { S } { { S0 } } = \exp \left( { - { \rm { b } } \bullet { \rm { ADC } } } \right) , { \rm { b } } \; \ge \;2000 \; \; { \rm { s } } / { \rm { m } } { { \rm { m } } ^2 } . \tag { 3 } \end{align*} \end{document}

S is the diffusion-weighted signal intensity for the b value, and S0 is the signal intensity obtained with the b0 value.¹²

Imaging analysis

MR images were evaluated by two radiologists independently who were blinded to the clinical history and histopathological results. When the two radiologists could not fully agree on the findings, consensus was achieved by discussion. Simultaneously, parameters regarding tumor diameter, WP, and ADC were measured by the third radiologist utilizing a Siemens syngo through VA30A_HF07 workstation. In this study, the longest diameter of the tumor on the transverse plane was saved as tumor size. Resultant images were classified into three groups according to the longest diameter: tumor size ≤2 cm, tumor size 2–5 cm, and tumor size 3 > 5 cm.¹³

Regions of interests (ROIs) were placed manually slice by slice over the whole tumor on 3D six echoes Dixon WP maps. The tumor boundary was unclear when the WP of HCC tumor was close to the background liver. Similarly, in b = 2100 s/mm², the ADC image had a poor signal-to-noise ratio, and the HCC tumor display was not clear. So image (b = 0 s/mm²) was set as a reference. The freehand ROI was placed to cover as much of the solid part of the tumors as possible to avoid the hemorrhagic and necrotic area. The WP, which was defined as W/(F + W) (W represents water content, F represents fat content), was automatically calculated from WP maps in specific ROI. The gray value 1 on the WP map corresponds to 0.1%. On ADC map, three ROIs were carefully placed on three difference slices, and mean ADC value was calculated. The averages size of the ROI ranged from 2.4 to 28.5 cm² (12.8 ± 7.1 cm²).

Clinicopathological data and patients' follow-up

Seventy-six patients were divided into two groups (<200 and >200 ng/mL) based on the serum levels of alphafetoprotein (AFP) and the risk of early recurrence.¹⁴ All HCC patients were monitored after resection for recurrence according to their serum AFP levels. Ultrasound or contrast computed tomography (CT)/MRI was scanned in the first month, and after every 3 months during the first year. The endpoint was set at the time when recurrence happened, which was defined as new tumors with typical imaging features of HCC found within liver parenchyma.

Histopathological examination

All the specimens were appropriately fixed with formalin for 12–24 h, and the fixed specimens were sectioned transversely, which were corresponded with the MRI plane. The specimens were embedded with paraffin and cut in a 5 μm slice. Then, the slices were stained with hematoxylin-eosin.¹⁵ An experienced pathologist who was blinded to the MR findings identified the histopathological grade of HCC according to the Edmondson–Steiner (E-S) classification.¹⁶

Statistical analysis

All statistical analyses were performed using SPSS 19.0 software package (Chicago, IL) and MedCalc (MedCalc Software, Ostend, Belgium). Data were presented as mean ± standard deviation. Normality and homogeneity tests of data distribution were conducted. Intraclass correlation coefficient (ICC) was used to test the intraobserver repeatability for the observers of the WP and ADC measurements. Comparisons between mean WP and ADC and E-S grades I–IV were performed using Kruskal–Wallis test and one-way ANOVA. Least Significant Difference t-test (LSD-t test) was applied to compare particular pairs of mean ADC value between every two E-S groups. The number and percentages of cases in the subgroups were presented as categorical variables. The Fisher Exact test was performed to assess the differences of categorical variables between subgroups. Correlations between WP, ADC, and E-S grades were evaluated by the Spearman rank correlation test. The mean WP between recurrence and nonrecurrence group was compared using the Mann–Whitney U test. The receiver operating characteristic (ROC) curves were calculated to estimate the diagnostic effect of 3D-VIBE Multiecho Dixon and ultrahigh b value DWI to HCC. The Kaplan–Meier method was used to estimate the recurrence free survival (RFS) 2 years after surgical resection. The differences between Kaplan–Meier survival curves were assessed using log-rank test. p < 0.05 was considered statistically significant.

Results

Clinical characteristics of HCC patients with recurrence and without recurrence

The clinical characteristics of the 76 patients (60 men, 16 females, mean age 50.6 ± 10.3 and 50.1 ± 12.1 years) are shown in Table 1. There were 29 patients with HCC recurrence 2 years after surgical resection and 47 patients without recurrence. In recurrence group, six patients were classified as E-S grade II, 13 as E-S grade III, and 10 as E-S grade IV. The numbers of patients grouped as E-S grades I–IV in the nonrecurrence group were 10, 22, 12, and 3, respectively. Seventy-one of 76 patients (93.4%) were graded as Child-Pugh class A in this study. The corresponding tumor sizes in recurrence group and nonrecurrence group were 7.0 ± 3.4 and 5.9 ± 2.7 cm, respectively. There were no significant differences in age, sex, Child-Pugh classification, mean tumor size, and serum AFP levels between the recurrence and nonrecurrence group (p = 0.145–0.773).

Table 1.

Baseline Characteristics of Patients with Hepatocellular Carcinoma with Recurrence and Without Recurrence

Characteristics	Recurrence group	Nonrecurrence group	Total	p
Mean age (years) (±SD)	49.7 ± 12.7	50.9 ± 9.3	50.5 ± 10.6	0.664
Sex
Male	22 (36.7)	38 (63.3)	60 (100)	0.773
Female	7 (43.8)	9 (56.2)	16 (100)
Histopathological grade				0.000
Edmondson–Steiner I	0 (0.0)^b(0.0)^a	10 (21.3)^b(100.0)^a	10 (100)^a
Edmondson–Steiner II	6 (20.7)^b(21.4)^a	22 (46.8)^b(78.6)^a	28 (100)^a
Edmondson–Steiner III	13 (44.8)^b(52)^a	12 (25.5)^b(48)^a	25 (100)^a
Edmondson–Steiner IV	10 (34.5)^b(76.9)^a	3 (6.4)^b(23.1)^a	13 (100)^a
Total	29 (100)^b	47 (100)^b
Child-Pugh classification				0.281
Class A	26 (89.7)^b(36.6)^a	45 (95.7)^b(63.4)^a	71 (100)^a
Class B	3 (10.3)^b(60)^a	2 (4.3)^b(40)^a	5 (100)^a
Class C	0	0	0
Total	29 (100)^b	47 (100)^b
Tumor size (cm)	7.0 ± 3.4	5.9 ± 2.7 cm	6.3 ± 2.6	0.145
MVI positive	7 (28)	18 (72)	25 (100)	0.222
MVI negative	22 (43.1)	29 (56.9)	51 (100)
AFP (ng/mL)
AFP ≤200	19 (44.2)	24 (55.8)	43 (100)	0.243
AFP >200	10 (30.3)	23 (69.7)	33 (100)

Data reported in parentheses are percentages; ^aRow percentage, ^bColumn percentage.

AFP, alpha fetoprotein; MVI, microvascular invasion.

Repeatability of WP and ADC measurements

The ICC for WP and ADC were 0.9484 (95% CI, 0.9200–0.9670) and 0.9606 (95% CI, 0.9388–0.9747), respectively. The results indicated excellent reliability for WP and ADC.

Mean WP and ADC value in E-S I–IV groups, and their correlation with histopathological grade

Water content is directly associated with signal intensity on WP map. Hypointensity on the WP map indicates a lower WP value; on the contrary, hyperintensity indicates a higher WP (Fig. 1).

FIG. 1.

WP in HCC tumors corresponding to groups E-S I–IV: (A) a homogeneous hyposignal intensity round tumor (arrowhead) located in right lobe of liver, WP 80.6%, histopathological grade E-S I; (B, C) heterogeneous signal intensity tumor located in right lobe of liver, WP 86.7% and 95.5%, respectively, corresponding histopathological grades E-S II and III, hypersignal intensity represents higher WP in the tumor (arrowhead); (D) a homogeneous signal intensity tumor located in left lobe of liver (arrowhead), WP 97.8%, pathological grade E-S IV. E-S, Edmondson–Steiner; HCC, hepatocellular carcinoma; WP, water percentage.

WP of E-S I, II, III, and IV in HCC tumors were 91.8% ± 6.7%, 95.2% ± 2.4%, 96.4% ± 3.3%, and 97.7% ± 1.0%, respectively. Significant differences were observed in three pairs of mean WP comparisons (Fig. 2A), including E-S I versus III (p = 0.019), E-S I versus IV (p = 0.004), and E-S II versus IV (p = 0.02). However, no statistical differences were found in the comparisons between E-S I and II (p = 1.00), E-S II and III (p = 0.112), E-S III and IV (p = 1.00). The expression of WP revealed a positive association with pathological E-S grades (r = 0.480, p < 0.01). The mean ± standard deviations of ADC values for HCC tumor in each E-S group are shown in Figure 2B. In E-S grade I, ADC value was 0.93 ± 0.07 × 10^–3 mm²/s. As E-S grades increased, the ADC value decreased accordingly. Mean ADC values in E-S grades II–IV were 0.82 ± 0.08 × 10^–3, 0.74 ± 0.08 × 10^–3, and 0.62 ± 0.07 × 10^–3 mm²/s, respectively. There was a significant difference between every two E-S groups (p < 0.01). The findings of this study demonstrated a negative correlation between ADC and E-S grades (r = −0.784, p = 0.000).

FIG. 2.

The correlation between WP, ADC, and E-S grades: (A) Mean WP of entire HCC tumor from E-S I to E-S IV. (B) Mean ADC of HCC tumor from E-S I to IV. Positive and negative correlations are observed between WP, ADC, and E-S grades, respectively (p < 0.01). Statistical differences of mean WP and ADC are observed between every two groups (p < 0.01). ADC, apparent diffusion coefficient; E-S, Edmondson–Steiner; HCC, hepatocellular carcinoma; WP, water percentage.

Prognostic value of WP and ADC for HCC patients after resection

Mean WP values based on the entire HCC tumor were 97.2% ± 2.1% in the recurrence and 94.6% ± 4.2% in nonrecurrence group (Fig. 3A). Statistical difference was found between the two groups (p < 0.01).

FIG. 3.

The prognostic values of WP for HCC patients after surgical resection in 2 years: (A) A statistical difference of mean WP observed between recurrence and nonrecurrence group (p = 0.01). (B) A moderate diagnostic efficacy by ROC curve of WP to predict the recurrence for HCC patients, AUC = 0.747 (p < 0.001). (C) Lower WP group has a longer recurrence free survival than higher WP group (p < 0.01). AUC, area under the curve; HCC, hepatocellular carcinoma; ROC, receiver operating characteristic; WP, water percentage.

ROC analyses were conducted to evaluate the diagnostic performance of MRI 3D VIBE multiecho Dixon, ultrahigh b value DWI, and MRI 3D VIBE multiecho Dixon and DWI in combination. The area under the curve (AUC) of the receiver operating characteristic (ROC) was 0.747 ± 0.06 (95% CI, 0.63–0.84, p < 0.01) for WP, 0.631 ± 0.07 for ADC, and 0.753 ± 0.06 for WP and ADC combined. Significant difference of AUC was only found in the comparison between ADC and the combined biomarkers (Fig. 4). The optimal cutoff value was 95.7% for WP. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were 86.2%, 61.7%, 58.1%, and 87.8%, respectively (Fig. 3B). For ADC, the cutoff value was 0.77, and the successive sensitivity, specificity, PPV, and NPV were 69.0%, 59.6%, 51.3%, and 75.7%. The sensitivity, specificity, PPV, and NPV for MRI 3D VIBE multiecho Dixon and DWI in combination were 82.8%, 61.7%, 57.2%, and 85.3%, respectively, with a cutoff value of 0.61 (Fig. 3B).

FIG. 4.

ROC analysis for WP, ADC, and WP combined with ADC in diagnostic performance of early recurrence of HCC. AUC of ROCs for WP was 0.747, AUC of ROCs for ADC was 0.631, and AUC of ROCs for combined factors was 0.753, the difference of AUC among them (WP vs. ADC, ADC vs. combined, WP vs. combined) was 0.117 (p = 0.1056), 0.122 (p = 0.0377), and 0.00550 (p = 0.7682). AUC, area under the curve; HCC, hepatocellular carcinoma; ROC, receiver operating characteristic; WP, water percentage.

A total of 76 patients were divided into two subgroups on the basis of the optimal cutoff value 95.7%. Patients with WP levels ≤95.7% were classified into the lower WP group, whereas patients with WP >95.7% were defined as higher WP. The mean follow-up times of all patients were 20 months. The mean RFS was 18.1 ± 0.9 (95% CI, 16.4–19.9) months in the lower WP group and 10.7 ± 1.0 (95% CI, 8.7–12.8) months in higher WP group (Fig. 3C). The difference between the two RFS groups was statistically significant (p < 0.01).

Discussion

Water content is essential for tumor growth, which makes it a potential biomarker for the diagnosis and prognosis of different kinds of tumors. Currently, the most common way for water content measurement is near-infrared (NIR) spectral imaging. NIR is an intrinsic property of water. Based on the different water absorption in cancer and normal tissue (prostate and breast cancer), and the transmission and scattering of visible, NIR is sensitive to water content. However, NIR is only generally used to identify the cancer tissue and normal tissue. For the qualitative diagnosis of malignant tumors, CT and MRI are the preferred examination. The reason why NIR spectral imaging is insufficient is that NIR is difficult to reflect the detailed characterization of water in thick tissues due to the subtle spectral shifts.⁷ In addition, NIR findings cannot reveal the information of the entire tumor. Furthermore, NIR scan must rely on the tumor location from standard X-ray mammography or ultrasound.⁷ Importantly, in previous studies, the water content was represented by water concentration, which indicated the cellularity instead of real water content. Overall, NIR is not the primary choice for water content evaluation in clinical practice. At present, 3D multiecho Dixon is another novel modality to quantitatively estimate the water content in tumor. The reproducibility of the WP is robust compared with NIR since WP is independent of tissue thickness and it can reflect the water proportion of the entire tumor. Moreover, WP map generated from the 3D-VIBE multiecho Dixon can clearly display the tumor size and location, making it easier to be accepted by clinicians. Also, WP can reveal both intracellular and extracellular water content in the tumor instead of water concentration.

The findings of this study exhibited a positive correlation between WP and histopathological classification for HCC tumors. Data showed that the mean value of WP was lowest in E-S grade I group and highest in E-S grade IV patients. The mechanism of this correlation still remains unclear. However, it was hypothesized that it might be related to the expression of AQP, which is a kind of mediator facilitating water transportation across the cell membrane. Among all the existing AQP, AQP9 is a unique AQP in hepatocyte. The study of Tan et al. reported a positive correlation between AQP9 and histopathological grade in human astrocytic tumors.¹⁷ However, to date, no study has explored the association between AQP9 and histopathological grade of HCC. In addition, another possible cause of increased WP in recurrence group may be hyaluronan. As hyaluronan always overexpresses in HCC tumors,¹⁷ more free water is retained in extracellular space. Meanwhile, the overexpression of hyaluronan is involved with the progression of HCC.¹⁷ This study only focused on disclosing the relationship between WP and E-S classification. The mechanism behind the increased WP was not assessed. However, the correlation between AQP9, hyaluronan, and WP would be another interesting point to explore in future studies.

Furthermore, a previous study demonstrated an inverse linear relationship between ADC and the increasing brain water content in cytotoxic brain edema,¹⁶ and they also considered that the diffusion of water accumulating in cells was slowed. Other literature assumed that the mechanism of visualizing body malignancies on DWI was basically the same as that of visualizing a fresh brain infarction.¹⁷ From those investigations, it was hypothesized that ADC might correlate with WP in HCC. In this study, ultrahigh b value (b = 2100 s/mm²) was used in DWI evaluation. Xueying et al.¹² reported that ADC map was calculated from a monoexponential model when b ≥ 2000 s/mm², which minimized the effect of perfusion and was closer to the true diffusion. In the cases of b < 2000 s/mm², the diffusion-weighted signal S remains fit to the biexponential equation, and perfusion effect still exists. Therefore, b = 2100 s/mm² was finally chosen for DWI acquisition. This study demonstrated that ADC value decreased significantly as the E-S grade rised. Mean ADC values of poorly differentiated HCC were significantly lower than those of well and moderately differentiated HCC. This finding was inconsistent with the earlier reports.^17,18 Accordingly, ADC decreased in low E-S grade HCC, corresponding to the low WP in this study.

The results of this study revealed a positive association between WP and E-S grades. Although the exact mechanism was inconclusive, it was speculated that it might be related to the following factors: First, previous studies^18,19 suggested that a lower differentiated tumor resulted in high cellularity and a lower ADC value, which was in accordance with the finding of this study. Meanwhile, Sevick et al.¹¹ reported that ADC value was negatively correlated with water content, which means the lower the ADC levels, the higher the water content in tumor. Therefore, it was assumed that the poorly differentiated tumor with a higher E-S grade would have higher water content, which would be expressed as a higher WP. This might be a cause of the positive correlation between the WP and the E-S classification. In addition, E-S classification is an independent risk factor affecting the early recurrence of HCC. Patients with a high E-S grade might have a poor prognosis after surgery. Based on the correlation between the WP and E-S classification, it was presumed that WP can also predict the early recurrence of HCC postoperatively. The results of Kaplan–Meier analyses, which showed a significant shorter RFS in the high WP group, confirmed assumption of this study. The comparison between WP and WP combined with ADC on ROC demonstrated that there was no significant difference between the two groups. Although the results showed that WP might only have a moderate performance in the prognosis of postoperative HCC, it still proved the value of WP as a prognostic biomarker. Therefore, WP can be used as a predictor of early postoperative recurrence of HCC.

Although we found a moderately high sensitivity and NPV of WP in the prediction of recurrence for HCC patients after resection, the specificity and PPV were relatively low. The combination of many factors would affect the prognosis of HCC after surgery, including tumor size, serum AFP levels, E-S grades, and MVI.^14,20,21 Therefore, simply evaluating WP might result in a low sensitivity and PPV. In this study, no statistical differences were observed among the above clinicopathological parameters, which might decrease the confounding effect on WP in estimating prognosis for HCC patients.

There are still some limitations in this study: First, the sample size of recurrence patients in this study is relatively small, which may affect the reproducibility of the results. Second, AQP 9 and hyaluronan were not analyzed using immunohistochemistry (IHC) staining, because AQPs and hyaluronan IHC were not available in the hospital. Third, the follow-up time in this study is not long enough to observe the long-term survival of patients. Thus, a longer follow-up duration will be needed to determine the relationship between WP and survival.

In conclusion, WP obtained from 3D-VIBE Dixon can be used to assess HCC histopathological grade preoperatively. WP can also serve as biomarker for early recurrence of HCC.

Footnotes

Disclosure Statement

No competing financial interests exist.

References

Tang

, Hallouch

, Chernyak

, et al. Epidemiology of hepatocellular carcinoma: Target population for surveillance and diagnosis. Abdom Radiol, 2018; 43:13–25.

Ayuso

, Rimola

, García-Criado

. Imaging of HCC. Abdom Imag, 2012; 37:215.

Ali

, Wang

, Zevallos

, et al. Near infrared spectroscopy and imaging to probe differences in water content in normal and cancer human prostate tissues. Technol Cancer Res Treat, 2004; 3:491.

Privalov

, Crane-Robinson

. Role of water in the formation of macromolecular structures. Eur Biophys J, 2017; 46:203.

Marques

, de Carvalho

, Sakai

, et al. Intracellular water–an overlooked drug target? Cisplatin impact in cancer cells probed by neutrons. Phys Chem Chem Phys, 2017; 19:2702–2713.

Cerussi

, Shah

, Hsiang

, et al. In vivo absorption, scattering, and physiologic properties of 58 malignant breast tumors determined by broadband diffuse optical spectroscopy. J Biomed Opt, 2006; 11:44005.

Chung

, Cerussi

, Klifa

, et al. In vivo water state measurements in breast cancer using broadband diffuse optical spectroscopy. Phys Med Biol, 2008; 53:6713.

Ross

, Gordon

. Water in malignant tissue, measured by cell refractometry and nuclear magnetic resonance. J Microsc, 1982; 128:7.

Stanley

SASH

. Terahertz spectroscopy of liver cirrhosis: Investigating the origin of contrast. Phys Med Biol, 2010; 55:7587.

10.

Ren

, Dimitrov

, Sherry

, et al. Composition of adipose tissue and marrow fat in humans by 1H NMR at 7 Tesla. J Lipid Res, 2008; 49:2055.

11.

Sevick

, Kanda

, Mintorovitch

, et al. Cytotoxic brain edema: Assessment with diffusion-weighted MR imaging. Radiology, 1992; 185:687.

12.

Xueying

, Zhongping

, Zhoushe

, et al. Investigation of apparent diffusion coefficient from ultra-high b-values in Parkinson's Disease. Eur Radiol, 2015; 25:2593.

13.

Clark

, Carson

, Kavanagh

, et al. Staging and current treatment of hepatocellular carcinoma. Radiographics, 2005; 25 Suppl 1:S3.

14.

Peng

, Chen

, Lai

, et al. High α‐fetoprotein level correlates with high stage, early recurrence and poor prognosis of hepatocellular carcinoma: Significance of hepatitis virus infection, age, p53 and β-catenin mutations. Int J Cancer, 2004; 112:44.

15.

Cong

W-M

, Bu

, Chen

, et al. Practice guidelines for the pathological diagnosis of primary liver cancer: 2015 update. World J Gastroenterol, 2016; 22:9279.

16.

Edmondson

, Steiner

. Primary carcinoma of the liver: A study of 100 cases among 48,900 necropsies. Cancer, 1954; 7:462.

17.

, Wang

, Qin

, et al. Over expression of hyaluronan promotes progression of HCC via CD44-mediated pyruvate kinase M2 nuclear translocation. Am J Cancer Res, 2016; 6:509.

18.

Ludwig

, Camacho

, Kokabi

, et al. The role of diffusion-weighted imaging (DWI) in locoregional therapy outcome prediction and response assessment for hepatocellular carcinoma (HCC): The new era of functional imaging biomarkers. Diagnostics (Basel), 2015; 5:546.

19.

Tang

, Wang

, Ma

, et al. Diffusion-weighted imaging of hepatocellular carcinomas: A retrospective analysis of correlation between apparent diffusion coefficients and histological grade. Abdom Radiol, 2016; 41:1539.

20.

Earl

, Chapman

. Tumor size remains key for prediction of hepatocellular carcinoma recurrence after liver transplantation. Ann Surg Oncol, 2011; 18:1217.

21.

Masaki Kaibori

, Morihiko Ishizaki

, Kosuke Matsui

, et al. Predictors of microvascular invasion before hepatectomy for hepatocellular carcinoma. J Surg Oncol, 2010; 102:462.