Abstract
Objective:
To evaluate the usefulness of quantitative diffusion-weighted (DW) imaging acquired by multivendor magnetic resonance units for predicting grade of hepatocellular carcinoma (HCC).
Methods:
83 patients with 100 histologically diagnosed HCCs who underwent pre-operative liver DW imaging with b = 0 and1000 s mm–2 or b = 0 and800 s mm–2 at any of six institutions were included. Two radiologists independently measured the apparent diffusion coefficient (ADC) of the lesion as well as non-ADC parameters, such as the relative contrast ratio and the contrast-to-noise ratio (CNR) between the lesion and the liver parenchyma on high b-value DW images. The diagnostic performance of the DW parameters in discriminating poorly-differentiated HCCs was compared using receiver operating characteristic (ROC) analysis.
Results:
The areas under the receiver operating characteristic curves for the CNR (86.4% [95% confidence interval (CI) (77.2–95.6] and 83.9% [95% CI 71.2–96.6] for b = 1000 and 800 s mm–2, respectively] and the relative contrast ratio (85.3% [95% CI 75.5–94.8] and 83.5% [95% CI 70.5–96.4]) tended to be superior to the ADC [71.1% [95% CI (56.9–85.2)] and 75.7% [95% CI (55.1–96.2)]; p < 0.05 for CNR vs ADC for b = 1000 s mm–2, but not significant for other parameters) for discrimination of poorly-differentiated HCCs.
Conclusion:
All DW parameters could discriminate HCC grade. Non-ADC parameters might be more useful than the ADC for predicting poorly-differentiated HCCs.
Advances in knowledge:
The utility of quantitative DW parameters for predicting HCC grade was demonstrated by using multivendor MR units.
Full title: Short title: Type of Manuscript: Author names and affiliations: Corresponding author: Yoshio Kitazume Department of Diagnostic Radiology, Medical Hospital, Tokyo Medical and Dental University, 1-5-45 Yushima, Bunkyo-ku, Tokyo 113-8510, Japan E-mail address: ktzmmrad@tmd.ac.jp Prediction of histological grade of hepatocellular carcinoma using quantitative diffusion-weighted magnetic resonance imaging: a retrospective multi-vendor study
Indroduction
Histological grade is one of the prognostic factors in patients with hepatocellular carcinoma (HCC).1–4 Metastases and recurrences after liver resection and transplantation are seen more frequently in poorly-differentiated HCC (pHCC).5–11 The grade of HCC has a relationship with the effectiveness of radiofrequency ablation and the survival rate thereafter.1, 12 Accurate radiological stratification of HCC grade is helpful for predicting the prognosis and selecting the therapeutic strategy.
The ability of MRI to predict histological grade of HCC has been evaluated using various methods. Gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid-enhanced MRI or diffusion-weighted imaging (DWI) has been used to discriminate the grade of HCC in most of the recent studies. The grade is qualitatively predictable to some extent by the enhancement pattern of the lesion, its size, and the degree of necrosis.13 However, in terms of quantitative analysis, the ability of gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid-enhanced MRI to discriminate pHCCs from other HCCs is inadequate.14–17 The apparent diffusion coefficient (ADC), which is the most commonly used index in DWI, has been used for quantitative assessment of malignant tumours. Several studies have evaluated the correlation between the ADC and histological grade of HCC.17–27 but yielded conflicting results. Recently, some studies showed that non-ADC parameters, i.e. the relative contrast ratio (RCR) or contrast-to-noise ratio (CNR) of HCC to the surrounding liver parenchyma on DW images, were superior to the ADC in predicting the grade of HCC.18, 19,21 However, they were single-centre studies using single MR units, and the interobserver reproducibility of these measurements has not been reported.
The aim of this study was to evaluate the usefulness and the reproducibility of DW parameters, including RCR, CNR and ADC using multivendor MR units for predicting grade of HCC.
Methods and materials
Patients
This multicentre, retrospective, multivendor study included patient data from six institutions (A, Medical Hospital of Tokyo Medical and Dental University; B, Ome Municipal General Hospital; C, Ochanomizu Surugadai Clinic; D, Advanced Imaging Center Yaesu Clinic; E, Japanese Red Cross Musashino Hospital; F, Medical Scanning Ochanomizu). The study protocol was approved by the institutional review boards at institutions A, B and E. Institutions C, D and F did not have ethical review committees, so delegated approvals to Institution A. The requirement for informed consent was waived.
We identified 131 consecutive patients with HCC that was histologically diagnosed after liver resection in Institution A from January 2010 to December 2016 who underwent pre-operative liver DW MRI acquired in the axial plane at b = 0 s mm–2 and a high b-value (1000 s mm–2 or 800 s mm–2). Patients were excluded if the quality of DWI was poor (n = 8), if the diameter of their HCC was less than 1.5 cm because of decreasing reliability of signal intensity (SI) measurement (n = 14), and if their HCC had already been treated with radiofrequency ablation, transcatheter arterial chemoembolization, or chemotherapy before MRI examination or surgery (n = 26). We finally enrolled 83 patients with 100 HCCs.
Table 1 shows the patients’ demographic and tumour characteristics. All patients were classified as having Child–Pugh class A disease. The HCCs were classified as well-, moderately-, or poorlydifferentiated according to the criteria of the Liver Cancer Study Group of Japan.28 When a HCC had multiple histological features, it was classified as the higher grade if the area of that grade accounted for over 10% of the entire tumour.
Table 1.
Patient demographic and tumour characteristics
Parameter | b = 1000 s mm–2(Institutions A–D) | b = 800 s mm–2(Institutions E, F) | p-value |
Mean age (y)a | 68.4 ± 11.0 | 69.6 ± 11.0 | 0.59 |
Sex (male/female) | 41/8 | 25/9 | 0.28 |
Background liver | |||
Hepatitis B virus | 12 | 8 | 1.00 |
Hepatitis C virus | 13 | 13 | 0.34 |
Alcoholism | 13 | 5 | 0.28 |
Resected tumours (single/multiple) | 45/4 | 29/5 | 0.48 |
Tumour locations (right/left) | 36/22 | 33/9 | 0.08 |
Subdiaphragmatic locations (right/left) | 13/6 | 11/2 | 1.00/1.00 |
Histological tumour grade (well-/moderately-/poorly-differentiated) | 8/31/19 | 10/20/12 | 0.50 |
Mean tumour diameter (mm)a | 44.0 ± 30.4 | 43.3 ± 29.2 | 0.99 |
Median AFP level (ng ml–1)b | 5.2 (1.5–484000) | 17.9 (1.3–15682) | 0.028 |
Median PIVKA-II level (mAU ml–1)b | 51.0 (10–77800) | 221.5 (17–239630) | 0.0055 |
AFP, alpha-fetoprotein; PIVKA, protein induced by Vitamin K absence or antagonists.
aThe data are shown as the mean ± standard deviation.
bThe data are shown as the median with the range in parentheses. All AFP and PIVKA-II levels were measured in Institution A before surgery.
MRI
The average time interval between pre-operative liver MRI and surgery was 41.0 days on b = 0 and 1000 s mm–2 (range, 2–134 days) and 42.7 days on b = 0 and 800 s mm–2 (range, 12–126 days) (p = 0.40). MRI examinations were performed on 1.5 T or 3.0 T MR units (1.5 T, Signa HDxt [GE Medical Systems, Milwaukee, WI] and Intera Achieva NovaDual [Philips Medical Systems, Best, Netherlands]; 3.0 T, Ingenia [Philips] and SPECTRA [Siemens Medical Solutions, Erlangen, Germany]). Table 2 shows the MR units at the institutions and their DWI protocols.
Table 2.
MR units in the six institutions and their DWI protocols
Institution | A | B | C | D | E | F |
MR unit | GE Signa HDxt | Philips Intera Achieva NovaDual | Philips Intera Achieva NovaDual | Philips Ingenia | GE Signa HDxt | Siemens Spectra |
Coil | 8-ch body upper | 32 cm SENSE/torso cardiac coil | 32 cm SENSE/torso Cardiac coil | 16-ch dStream Torso coil | 8-ch body Upper | 6-ch body coil + pine coil |
Magnetic field strength (T) | 1.5 | 1.5 | 1.5 | 3 | 1.5 | 3 |
b factor (s mm–2) | 1000 | 1000 | 1000 | 1000 | 800 | 800 |
Repetition time (ms) | 4500 | 1198.3–1336.7 | 2319–5000 | 6250 | 5500 | 1,3800 |
Echo time (ms) | 76.3 | 73 | 69 | 65 | 72.5 | 68 |
Acquisition matrix | 88 × 110 | 112 × 88 | 144 × 114 | 128 × 135 | 128 × 192 | 128 × 63 |
Image matrix | 256 × 256 | 256 × 256 | 256 × 256 | 384 × 384 | 256 × 256 | 256 × 168 |
Field of view (cm) | 40 | 35 | 35–40 | 36 | 40 | 38 |
Slice thickness (mm) | 7 | 8 | 7–11 | 6–7.5 | 6 | 6 |
Slice gap (mm) | 2 | 2 | 0.7–1.1 | 0–1 | 2 | 0 |
Respiratory motion compensation technique | Respiratory-triggered | Respiratory-triggered | Respiratory-triggered | Free-breathing | Respiratory-triggered | Free-breathing |
Fat suppression | CHESS | CHESS | CHESS | CHESS + SSGR | CHESS | CHESS + SSGR |
DWI, diffusion-weighted imaging; CHESS, chemical shift selective; SSGR, section-select gradient reversal.
Imaging analysis
The MRI data were evaluated using a Digital Imaging and Communications in Medicine viewer (OsiriX imaging software v. 8.0; OsiriX Foundation, Geneva, Switzerland) by two board-certified radiologists (YO and YI, with 7 and 10 years of experience in liver MRI, respectively) who worked independently and were blinded to the histological grade of HCC. Three oval regions of interest (ROIs) were placed on the “solid” areas where the SI values were considered to be the three highest in DWI on b = 1000 s mm–2 or 800 s mm–2 (Figure 1). Necrotic, haemorrhagic and cystic lesions were avoided by referring to other sequences. The ROIs were copied and pasted on the same locations in DWI on b = 0 s mm–2. Three round ROIs with a diameter of 10 mm were then set on the normal hepatic parenchyma near the three lesions in DWI on a high b-value, excluding visible large vessels by reviewing the other MR sequences.
Figure 1.
Images from a 64-year-old male with a HCC measuring 9.0 cm in diameter, classified as poorly-differentiated. This tumour was composed mainly of moderately-differentiated HCC coexisting with poorly-differentiated HCC that accounted for over 10% of the entire tumour area. A 1.5 T magnetic resonance unit was used and diffusion-weighted images were acquired on b = 0 and 800 s mm–2. (a) Diffusion-weighted image, (b) T2 weighted image using a fast spin echo sequence, (c) a three-dimensional T1 weighted image using a gradient-recalled echo sequence with fat suppression (3D-FST1WI) during the arterial phase, and (d) a 3D-FST1WI during the hepatobiliary phase. Oval regions of interest are placed on the “solid” area where the signal intensity is considered to be highest on DWI at a high b-value and on the surrounding normal hepatic parenchyma. We avoided placing regions of interest in a necrotic part of this lesion (arrow). The measurements obtained were as follows: apparent diffusion coefficient, 0.85 × 10−3 mm2 s–1; relative contrast ratio, 10.9; and contrast-to-noise ratio, 80.1. DWI, diffusion-weighted imaging; HCC, hepatocellular carcinoma.
The ADC, RCR and CNR on DWI were calculated as follows: ADC = ln [SI (lesion on b = 0 s mm–2)/SI (lesion on high b-value [1000 or 800 s mm–2])]/1000 or 800; RCR = SI (lesion on high b-value)/SI (liver on high b-value); and CNR = [SI (lesion on high b-value) – SI (liver on high b-value)]/noise (liver on high b-value), where ln is the natural logarithm, the SI used in the RCR and CNR calculations is the mean SI of the 3 ROIs on each high b-value, and noise is the mean standard deviation of the SI of the 3 ROIs. The ADC was calculated at each ROI, and the mean ADC of the 3 ROIs was used for further analyses.
Statistical analysis
The statistical analysis was performed using R v. 3.3.1 software (R Foundation for Statistical Computing, Vienna, Austria). The patients were divided into two groups according to the b-values (0 and 1000 s mm–2 or 0 and 800 s mm–2) used for DWI acquisition. Differences between two quantitative variables were tested for significance using the Wilcoxon rank sum test. Qualitative MR variables were tested using Fisher’s exact test.
The relationships between histological grade of HCC and DW parameters were assessed using the Spearman’s rank correlation test, after which a correlation coefficient (ρ) was obtained for each b-value group. The degree of ρ was defined as follows: none (0.00–0.20), weak (0.21–0.40), moderate (0.41–0.70), or strong (0.71–1.00). A positive ρ-value was interpreted as a positive correlation and a negative ρ-value as a negative correlation. The Kruskal–Wallis test was used to assess for statistically significant differences in values for the DW parameters among the HCC grades, and subsequent pairwise comparisons were performed using the Wilcoxon rank sum test with Bonferroni correction. The diagnostic performance of the parameters in discriminating the grades; (1) well-differentiated HCCs (wHCCs) vs moderately-differentiated HCCs (mHCCs) and pHCCs, and (2) pHCCs vs wHCCs and mHCCs, was evaluated for each group using receiver operating characteristic (ROC) analysis. The areas under the ROC curves (AUCs) of the parameters for discriminating pHCCs from non-pHCCs, i.e. wHCCs and mHCCs were compared for statistical significance using the pROC package in R. The optimal cut-off values were chosen to minimize the (1-sensitivity)2+ (1-specificity)2 formula or to be the point closest to the top left part of the plot.29 When no significant difference was found between the paired AUCs, the power of the test (1-β) was calculated as a post hoc analysis.
In addition, the DW parameters for each grade were compared between b = 1000 s mm–2 and b = 800 s mm–2 using the Mann–Whitney U test.
Differences in the DW parameters between the institutions were evaluated for each b-value group and were stratified by grade, sensitivity and specificity for discriminating pHCCs. The values for the DW parameters at four institutions on b = 0 and 1000 s mm–2 were analysed using the Kruskal–Wallis test and at two institutions on b = 0 and 800 s mm–2 using the Mann–Whitney U test; sensitivity and specificity values were assessed using Fisher’s exact test.
The interobserver agreement values for the ADC, RCR and CNR were assessed using the intraclass correlation coefficient (ICC) and the Bland-Altman method. The ICC was graded as follows: poor (0.00–0.20), fair (0.21–0.40), moderate (0.41–0.60), strong (0.61–0.80), or almost perfect (0.81–1.00). The mean absolute difference (bias) and the 95% confidence interval (CI) of the mean difference (limits of agreement) between the two radiologists’ measurements were obtained using the Bland-Altman method.
A p-value of less than 0.05 indicated a statistically significant difference.
Results
Relationships between histological grade and DW parameters
Figure 2 and Table 3 show the relationship between the histological grades of HCC and DW parameters.
Figure 2.
Boxplots of the apparent diffusion coefficient (a, d), relative contrast ratio (b, e) and contrast-to-noise ratio (c, f) subgrouped into well-differentiated (well), moderately-differentiated (moderate) and poorly-differentiated (poor) hepatocellular carcinoma on b = 1000 s mm–2 (a–c) and b = 800 s mm–2 (d–f). *p < 0.05, **p < 0.01, ***p < 0.001 indicate significant differences. ADC, apparent diffusion coefficient; CNR, contrast-to-noise ratio; RCR. relative contrast ratio.
Table 3.
Relationship between histological grade of hepatocellular carcinoma and diffusion-weighted parameters with results of statistical analysis
b = 1000 s mm–2 | b = 800 s mm–2 | |||||
ADC (×10−3 mm2 s–1) | RCR | CNR | ADC (×10−3 mm2 s–1) | RCR | CNR | |
wHCCa | 1.08 ± 0.15 | 1.76 ± 1.09 | 9.05 ± 8.86 | 1.33 ± 0.19 | 2.22 ± 0.74 | 13.03 ± 8.54 |
mHCCa | 1.19 ± 0.23 | 2.74 ± 1.45 | 15.27 ± 11.56 | 1.20 ± 0.21 | 2.85 ± 1.83 | 17.29 ± 12.72 |
pHCCa | 1.01 ± 0.19 | 5.04 ± 2.25 | 34.54 ± 18.56 | 1.06 ± 0.39 | 5.25 ± 3.25 | 38.22 ± 24.18 |
Spearman’s rank correlation coefficientb | −0.27b1 | 0.62b3 | 0.64b3 | −0.44b2 | 0.53b3 | 0.53b3 |
Kruskal–Wallis test p-value | 0.027 | <0.001 | <0.001 | 0.017 | 0.002 | 0.002 |
Wilcoxon rank sum test p-value | ||||||
wHCC vs mHCC | 1.000 | 0.137 | 0.706 | 0.654 | 0.843 | 1.000 |
mHCC vs pHCC | 0.031 | <0.001 | <0.001 | 0.154 | 0.015 | 0.005 |
wHCC vs pHCC | 0.478 | <0.001 | <0.001 | 0.019 | 0.005 | 0.008 |
ROC analysis | ||||||
wHCC vs mHCC + pHCC cut-off value | 1.13 | 1.83 | 13.70 | 1.27 | 2.18 | 14.87 |
AUCc | 0.504 (0.299, 0.708) | 0.820 (0.659, 0.981) | 0.760 (0.591, 0.929) | 0.720 (0.562, 0.878) | 0.728 (0.551, 0.905) | 0.703 (0.511, 0.895) |
wHCC + mHCC vs pHCC cut-off value | 1.13 | 3.22 | 17.66 | 1.10 | 2.87 | 16.99 |
AUCc | 0.711 (0.569, 0.852) | 0.853 (0.757, 0.947) | 0.864 (0.773, 0.955) | 0.757 (0.552, 0.962) | 0.835 (0.705, 0.964) | 0.839 (0.712, 0.966) |
AUC, area under the curve; ADC, apparent diffusion coefficient; CNR, contrast-to-noise ratio; mHCC, moderately-differentiated hepatocellular carcinoma; pHCC, poorly-differentiated hepatocellular carcinoma; RCR, relative contrast ratio; wHCC, well-differentiated hepatocellular carcinoma.
aThe data are shown as the mean ± standard deviation.
bSignificant differences are indicated as follows: b1p < 0.05, b2p < 0.01, b3p < 0.001.
cData in parentheses are the 95% confidence intervals.
The ADC on b = 0 and 1000 s mm–2 (ADC1000) showed a weak negative correlation with grade of HCC (ρ = −0.27, p = 0.041), and the ADC on b = 0 and 800 s mm–2 (ADC800) showed no correlation (ρ = −0.44, p < 0.01). The RCR and CNR on b = 1000 s mm–2 (RCR1000 and CNR1000) or 800 s mm–2 (RCR800 and CNR800) showed moderate positive correlations (ρ = [0.51–0.62], p < 0.001).
In multiple comparisons between the histological grades, there was a significant difference in the ADC1000 between mHCCs and pHCCs (p = 0.003) and a significant difference in the ADC800 between wHCCs and pHCCs (p = 0.019). Significant differences in RCR and CNR were found between pHCCs and the other grades of HCC (p < 0.001 for b = 1000 s mm–2 and p < 0.05 for b = 800 s mm–2).
Figure 3 shows the results of the ROC curve analysis for DW parameters on each high b-value for discriminating pHCC. The AUCs of RCR1000 (85.3%) and CNR1000 (86.4%) was greater than those of ADC1000 (71.1%; not statistically significant [NS] with a 1–β value calculated as a post hoc analysis of 0.20 for ADC1000 vs RCR1000 and p = 0.034 for ADC1000 vs CNR1000). The AUCs for RCR800 (83.5%) and CNR800 (83.9%) were greater than those for ADC800 (75.7%; NS with a 1–β value of 0.06 for both ADC800 vs RCR800 and ADC800 vs CNR800).
Figure 3.
Receiver operating characteristic curves discriminating poorly-differentiated HCCs from well- and moderately-differentiated HCCs for diffusion-weighted parameters on b = 1000 s mm–2 (a) and 800 s mm–2 (b). The areas under the curve are shown in the boxes. The numbers in parentheses indicate the 95% confidence intervals. ADC, apparent diffusion coefficient; CNR, contrast-to-noise ratio; HCCs, hepatocellular carcinomas; RCR. relative contrast ratio.
When the AUCs of the DW parameters between b = 1000 s mm–2 and 800 s mm–2 were compared for their ability to discriminate pHCC, the RCR1000 and CNR1000 were shown to be slightly higher than the RCR800 and CNR800, whereas the ADC1000 was lower than the ADC800 (NS for all).
With respect to differences in DW parameters for each grade between b = 1000 s mm–2 and 800 s mm–2, the ADC1000 for wHCC was found to be significantly lower than the ADC800 for wHCC (p = 0.029). There were no significant differences in the other parameters.
Differences between institutions
There was no significant difference in ADC, RCR, or CNR for each grade of HCC between the institutions (p = 0.12–1.00; Supplementary file 1, supplementary files available online). The sensitivity and specificity values of the DW parameters for discriminating pHCC were in the ranges of 62.5–100.0% and 33.3–100.0%, respectively, between the institutions; however, there were no significant differences (p = 0.19–1.00; Supplementary file 2).
Interobserver agreement
There was strong agreement for ADC1000 [ICC 0.650, 95% CI (0.283–0.820), p < 0.001] and almost perfect agreement for RCR1000 [ICC 0.926, 95% CI (0.879–0.956), p < ]0.001 and CNR1000 [ICC 0.811, 95% CI (0.700–0.884), p < 0.001]. There was strong agreement for ADC800 [ICC 0.754, 95% CI (0.355–0.892), p < 0.001] and CNR800 [ICC 0.628, 95% CI (0.405–0.780), p < 0.001], and almost perfect agreement for RCR800 [ICC 0.871, 95% CI (0.774–0.929), p < 0.001].
Figure 4 shows the results of the Bland–Altman plot for each DW parameter. The mean bias ± limits of agreement values were −0.13 × 10−3 mm2 s–1 ±0.353 for ADC1000, 0.04 ± 1.514 for RCR1000, and 0.48 ± 19.951 for CNR1000, and were −0.12 × 10−3 mm2 s–1 ±0.316 for ADC800, 0.16 ± 2.139 for RCR800, and 2.25 ± 27.165 for CNR800.
Figure 4.
Bland–Altman plots of the diffusion weighted parameters between two readers’ measurements for the apparent diffusion coefficient (a, d), relative contrast ratio (b, e) and contrast-to-noise ratio (c, f) on b = 1000 s mm–2 (a–c) and 800 s mm–2 (d–f). The central dashed lines indicate the mean absolute difference and the upper and lower lines indicate the 95% limits of agreement. ADC, apparent diffusion coefficient; CNR, contrast-to-noise ratio; RCR. relative contrast ratio.
Discussion
This multivendor study demonstrated that all DW parameters were useful for evaluating the grade of HCC, and that the RCR and CNR tended to be superior to the ADC in predicting the histological grade of HCC, particularly for discriminating pHCCs from other HCCs.
HCC is known to increase T2 elongation and to restrict diffusion with increasing histological grade.18, 30,31 The SIs of DWI, which are presented as the RCR and the CNR, are influenced by T2 elongation as well as restriction of diffusion. Therefore, the RCR and CNR might be more appropriate for predicting the histological grade of HCC than measurement of the ADC alone. Previous studies have shown similar results.18, 19,21
ROC analyses revealed that the RCR and CNR on b = 1000 s mm–2 tended to be superior to the RCR and CNR on b = 800 s mm–2 for discrimination of pHCCs. Restriction of diffusion is more strongly related to the contrast in DWI on b = 1000 s mm–2 than on b = 800 s mm–2 and is considered to be a more important reflection of histological grade. A previous study that used methods similar to those in the present study showed that the AUCs of ADC on b = 0 and 1500 s mm–2 were greater than those on b = 800–1000 s mm–2 in the present study.18
However, the AUC of the ADC on b = 0 and 1000 s mm–2 tended to be inferior to that of the ADC on b = 0 and 800 s mm–2 in discriminating pHCCs. We speculate that the reason for this was that the ADC1000 does not correctly measure restriction of diffusion on wHCC when compared with the ADC800; the ADC1000 on wHCC was shown to be slightly lower than that for mHCC and closer to that for pHCC, whereas the ADC800 on wHCC was highest in all grades. The ADC1000 on wHCC must have the highest values because the RCR and CNR values on wHCC were shown to be lowest for DWI on both b = 800 s mm–2 and b = 1000 s mm–2. This discrepancy can be explained by Rician noise, which increases at a higher b-value and causes a more marked decrease in the ADC on b = 0 and 1000 s mm–2 than on b = 0 and 800 s mm–2 for wHCC. However, it is difficult to determine the precise reason for this discrepancy because the ADC could be affected by various factors, e.g. the microcapillary perfusion effect.
The previous studies used two main methods to set the ROIs for measuring the ADC values of hepatic tumours. One method was to set the ROIs on the “solid” lesion where the ADC was considered to be the lowest in the entire tumour while avoiding necrotic, haemorrhagic and cystic lesions (known as the “minimum” ADC). The other method was to set the ROIs so that they included the entire lesion (known as the “mean” ADC). To our knowledge, the minimum ADC has consistently been reported to have high diagnostic accuracy for prediction of the histological grade of HCC,22, 23 whereas studies that have used the mean ADC have yielded inconsistent results.17, 19,21,26 In the present study, the ROI was set on the solid part of lesion where the SIs seemed to be highest. This method is similar to that used to obtain the minimum ADC and is considered to be appropriate for prediction of HCC grade.
In terms of interobserver variability, the ICCs for the RCR were in almost perfect agreement and were the highest of the DW parameters on each high b-value. The RCR is regarded as clinically useful because of its high diagnostic accuracy and high degree of interobserver reproducibility. The thresholds of RCR for predicting pHCC were approximately “3” at a b-value of 800 and 1000 s mm–2. Moreover, this threshold could be used to predict pHCC at up to 1500 s mm–2s/mm2, because the threshold was very similar to that in a previous study using a b-value of 1500 s mm–2.18
Almost all of the previous studies of the relationship between DWI and HCC grade were performed using a single MR unit.17–27 Our multivendor study has confirmed and validated the clinical value and relevance of DWI in prediction of HCC grade.
There are several limitations to this study, in particular its retrospective design and small patient population. Further prospective studies in large cohorts are needed. Another limitation is that the reproducibility of RCR and CNR among the different magnetic field strengths and respiratory motion compensation techniques available is unknown. The minimum ADC for hepatic tumours is not significantly different between 1.5 T and 3.0 T,32 and no significant differences in ADC have been identified between the different respiratory motion compensation techniques at 1.5 or 3.0 T.33–35 Further assessment of the reproducibility of RCR and CNR using other measurement methods under various conditions of MRI acquisition might be needed. Finally, some of the HCCs were located in the subdiaphragmatic area, where image quality is known to deteriorate because of cardiac motion artefacts.
In conclusion, this multivendor study demonstrated that RCR, CNR and ADC with a b-value of 800 and 1000 s mm–2 could discriminate HCC grade, and that DW parameters were able to distinguish well- and moderately-differentiated HCC from poorly-differentiated HCC with non-ADC parameters being more useful than ADC itself. A RCR threshold of more than 3 might predict pHCC with higher interobserver reproducibility.
Figure Legends
Footnotes
This work was supported in part by grants from Scientific Research Expenses for Health and Welfare Programs, the Grant-in-Aid for Cancer Research from the Ministry of Health, Labor and Welfare, No. 15K09885, the Scientific Research Expenses for Health and Welfare Programs, No. 29-A-3 (Takashi Terauchi and Ukihide Tateishi: squad leaders), Practical Research for Innovative Cancer Control and Project Promoting Clinical Trials for Development of New Drugs by Japan Agency for Medical Research and Development (AMED).
Contributor Information
Yusuke Ogihara, Email: yuogi47@gmail.com.
Yoshio Kitazume, Email: ktzmmrad@tmd.ac.jp.
Yoshihiro Iwasa, Email: iwsmrad@tmd.ac.jp.
Shinichi Taura, Email: taura-s@mghp.ome.tokyo.jp.
Yoshiro Himeno, Email: himeno@musashino.jrc.or.jp.
Tomo Kimura, Email: sc7711@gazo.or.jp.
Seishi Sawano, Email: sawano@m-satellite.ac.jp.
Shigehiko Terada, Email: terashige3725@yahoo.co.jp.
Minoru Tanabe, Email: tana.msrg@tmd.ac.jp.
Yukihisa Saida, Email: siddrnm@tmd.ac.jp.
Ukihide Tateishi, Email: ttisdrnm@tmd.ac.jp.
REFERENCES
- 1.Schwarz RE, Smith DD. Trends in local therapy for hepatocellular carcinoma and survival outcomes in the US population. Am J Surg 2008; 195: 829–36. doi: 10.1016/j.amjsurg.2007.10.010 [DOI] [PubMed] [Google Scholar]
- 2.Zhou L, Rui JA, Wang SB, Chen SG, Qu Q, Chi TY, et al. Outcomes and prognostic factors of cirrhotic patients with hepatocellular carcinoma after radical major hepatectomy. World J Surg 2007; 31: 1782–7. doi: 10.1007/s00268-007-9029-z [DOI] [PubMed] [Google Scholar]
- 3.Zhou L, Rui JA, Wang SB, Chen SG, Qu Q, Chi TY, et al. Factors predictive for long-term survival of male patients with hepatocellular carcinoma after curative resection. J Surg Oncol 2007; 95: 298–303. doi: 10.1002/jso.20678 [DOI] [PubMed] [Google Scholar]
- 4.Haratake J, Takeda S, Kasai T, Nakano S, Tokui N. Predictable factors for estimating prognosis of patients after resection of hepatocellular carcinoma. Cancer 1993; 72: 1178–83. doi: [DOI] [PubMed] [Google Scholar]
- 5.Oishi K, Itamoto T, Amano H, Fukuda S, Ohdan H, Tashiro H, et al. Clinicopathologic features of poorly differentiated hepatocellular carcinoma. J Surg Oncol 2007; 95: 311–6. doi: 10.1002/jso.20661 [DOI] [PubMed] [Google Scholar]
- 6.Zimmerman MA, Trotter JF, Wachs M, Bak T, Campsen J, Wright F, et al. Predictors of long-term outcome following liver transplantation for hepatocellular carcinoma: a single-center experience. Transpl Int 2007; 20: 747–53. doi: 10.1111/j.1432-2277.2007.00505.x [DOI] [PubMed] [Google Scholar]
- 7.Pérez-Saborido B, de los Galanes SJ, Menéu-Díaz JC, Romero CJ, Elola-Olaso AM, Suárez YF, et al. Tumor recurrence after liver transplantation for hepatocellular carcinoma: recurrence pathway and prognostic factors. Transplant Proc 2007; 397: 2304–2307. doi: 10.1016/j.transproceed.2007.06.059 [DOI] [PubMed] [Google Scholar]
- 8.Shah SA, Cleary SP, Wei AC, Yang I, Taylor BR, Hemming AW, et al. Recurrence after liver resection for hepatocellular carcinoma: risk factors, treatment, and outcomes. Surgery 2007; 141: 330–9. doi: 10.1016/j.surg.2006.06.028 [DOI] [PubMed] [Google Scholar]
- 9.Ariizumi S, Takasaki K, Yamamoto M, Ohtsubo T, Katsuragawa H, Katagiri S. Histopathologic differentiation of the main nodule determines outcome after hepatic resection for synchronous multicentric hepatocellular carcinomas. Hepatogastroenterology 2004; 51: 500–4. [PubMed] [Google Scholar]
- 10.Lauwers GY, Terris B, Balis UJ, Batts KP, Regimbeau JM, Chang Y, et al. International Cooperative Study Group on Hepatocellular Carcinoma Prognostic histologic indicators of curatively resected hepatocellular carcinomas: a multi-institutional analysis of 425 patients with definition of a histologic prognostic index. Am J Surg Pathol 2002; 26: 25–34. [DOI] [PubMed] [Google Scholar]
- 11.Jonas S, Bechstein WO, Steinmüller T, Herrmann M, Radke C, Berg T, et al. Vascular invasion and histopathologic grading determine outcome after liver transplantation for hepatocellular carcinoma in cirrhosis. Hepatology 2001; 33: 1080–6. doi: 10.1053/jhep.2001.23561 [DOI] [PubMed] [Google Scholar]
- 12.Kim SH, Lim HK, Choi D, Lee WJ, Kim SH, Kim MJ, et al. Percutaneous radiofrequency ablation of hepatocellular carcinoma: effect of histologic grade on therapeutic results. AJR Am J Roentgenol 2006; 186(5 Suppl): S327–S333. doi: 10.2214/AJR.05.0350 [DOI] [PubMed] [Google Scholar]
- 13.Jeong YY, Yim NY, Kang HK. Hepatocellular carcinoma in the cirrhotic liver with helical CT and MRI: imaging spectrum and pitfalls of cirrhosis-related nodules. AJR Am J Roentgenol 2005; 185: 1024–32. doi: 10.2214/AJR.04.1096 [DOI] [PubMed] [Google Scholar]
- 14.Kim HY, Choi JY, Kim CW, Bae SH, Yoon SK, Lee YJ, et al. Gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid-enhanced magnetic resonance imaging predicts the histological grade of hepatocellular carcinoma only in patients with child-pugh class a cirrhosis. Liver Transplantation 2012; 18: 850–7. doi: 10.1002/lt.23426 [DOI] [PubMed] [Google Scholar]
- 15.Choi JY, Kim MJ, Park YN, Lee JM, Yoo SK, Rha SY, et al. Gadoxetate disodium-enhanced hepatobiliary phase MRI of hepatocellular carcinoma: correlation with histological characteristics. AJR Am J Roentgenol 2011; 197: 399–405. doi: 10.2214/AJR.10.5439 [DOI] [PubMed] [Google Scholar]
- 16.Kogita S, Imai Y, Okada M, Kim T, Onishi H, Takamura M, et al. Gd-EOB-DTPA-enhanced magnetic resonance images of hepatocellular carcinoma: correlation with histological grading and portal blood flow. Eur Radiol 2010; 20: 2405–13. doi: 10.1007/s00330-010-1812-9 [DOI] [PubMed] [Google Scholar]
- 17.Chang WC, Chen RC, Chou CT, Lin CY, Yu CY, Liu CH, et al. Histological grade of hepatocellular carcinoma correlates with arterial enhancement on gadoxetic acid-enhanced and diffusion-weighted MR images. Abdom Imaging 2014; 39: 1202–12. doi: 10.1007/s00261-014-0168-z [DOI] [PubMed] [Google Scholar]
- 18.Iwasa Y, Kitazume Y, Tateishi U, Saida Y, Ban D, Tanabe M, et al. Hepatocellular Carcinoma Histological Grade Prediction: A Quantitative Comparison of Diffusion-Weighted, T2-Weighted, and Hepatobiliary-Phase Magnetic Resonance Imaging. J Comput Assist Tomogr 2016; 40: 463–70. doi: 10.1097/RCT.0000000000000367 [DOI] [PubMed] [Google Scholar]
- 19.Le Moigne F, Boussel L, Haquin A, Bancel B, Ducerf C, Berthezène Y, et al. Grading of small hepatocellular carcinomas (≤2 cm): correlation between histology, T2 and diffusion-weighted imaging. Br J Radiol 2014; 87: 20130763: 20130763. doi: 10.1259/bjr.20130763 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.An C, Park M-S, Jeon H-M, Kim Y-E, Chung W-S, Chung YE, et al. Prediction of the histopathological grade of hepatocellular carcinoma using qualitative diffusion-weighted, dynamic, and hepatobiliary phase MRI. Eur Radiol 2012; 22: 1701–8. doi: 10.1007/s00330-012-2421-6 [DOI] [PubMed] [Google Scholar]
- 21.Saito K, Moriyasu F, Sugimoto K, Nishio R, Saguchi T, Akata S, et al. Histological grade of differentiation of hepatocellular carcinoma: comparison of the efficacy of diffusion-weighted MRI with T2-weighted imaging and angiography-assisted CT. J Med Imaging Radiat Oncol 2012; 56: 261–9. doi: 10.1111/j.1754-9485.2012.02374.x [DOI] [PubMed] [Google Scholar]
- 22.Nakanishi M, Chuma M, Hige S, Omatsu T, Yokoo H, Nakanishi K, et al. Relationship between diffusion-weighted magnetic resonance imaging and histological tumor grading of hepatocellular carcinoma. Ann Surg Oncol 2012; 19: 1302–9. doi: 10.1245/s10434-011-2066-8 [DOI] [PubMed] [Google Scholar]
- 23.Nishie A, Tajima T, Asayama Y, Ishigami K, Kakihara D, Nakayama T, et al. Diagnostic performance of apparent diffusion coefficient for predicting histological grade of hepatocellular carcinoma. Eur J Radiol 2011; 80: e29–e33. doi: 10.1016/j.ejrad.2010.06.019 [DOI] [PubMed] [Google Scholar]
- 24.Heo SH, Jeong YY, Shin SS, Kim JW, Lim HS, Lee JH, et al. Apparent diffusion coefficient value of diffusion-weighted imaging for hepatocellular carcinoma: correlation with the histologic differentiation and the expression of vascular endothelial growth factor. Korean J Radiol 2010; 11: 295–303. doi: 10.3348/kjr.2010.11.3.295 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Nasu K, Kuroki Y, Tsukamoto T, Nakajima H, Mori K, Minami M. Diffusion-weighted imaging of surgically resected hepatocellular carcinoma: imaging characteristics and relationship among signal intensity, apparent diffusion coefficient, and histopathologic grade. AJR Am J Roentgenol 2009; 193: 438–44. doi: 10.2214/AJR.08.1424 [DOI] [PubMed] [Google Scholar]
- 26.Muhi A, Ichikawa T, Motosugi U, Sano K, Matsuda M, Kitamura T, et al. High-b-value diffusion-weighted MR imaging of hepatocellular lesions: estimation of grade of malignancy of hepatocellular carcinoma. J Magn Reson Imaging 2009; 30: 1005–11. doi: 10.1002/jmri.21931 [DOI] [PubMed] [Google Scholar]
- 27.Guo W, Zhao S, Yang Y, Shao G. Histological grade of hepatocellular carcinoma predicted by quantitative diffusion-weighted imaging. Int J Clin Exp Med 2015; 8: 4164–9. [PMC free article] [PubMed] [Google Scholar]
- 28. Liver Cancer Study Group of Japan. General Rules for the Clinical and Pathological Study of Primary Liver Cancer. third english ed Tokyo: The British Institute of Radiology.; 2010. [Google Scholar]
- 29.Perkins NJ, Schisterman EF. The inconsistency of “optimal” cutpoints obtained using two criteria based on the receiver operating characteristic curve. Am J Epidemiol 2006; 163: 670–5. doi: 10.1093/aje/kwj063 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Ebara M, Fukuda H, Kojima Y, Morimoto N, Yoshikawa M, Sugiura N, et al. Small hepatocellular carcinoma: relationship of signal intensity to histopathologic findings and metal content of the tumor and surrounding hepatic parenchyma. Radiology 1999; 210: 81–8. doi: 10.1148/radiology.210.1.r99ja4181 [DOI] [PubMed] [Google Scholar]
- 31.Kadoya M, Matsui O, Takashima T, Nonomura A. Hepatocellular carcinoma: correlation of MR imaging and histopathologic findings. Radiology 1992; 183: 819–25. doi: 10.1148/radiology.183.3.1316622 [DOI] [PubMed] [Google Scholar]
- 32.Tang Y, Wang H, Wang Y, Li J, Jia R, Ma L, et al. Quantitative comparison of MR diffusion-weighted imaging for liver focal lesions between 3.0T and 1.5T: Regions of interest of the minimum-spot ADC, the largest possible solid part, and the maximum diameter in lesions. J Magn Reson Imaging 2016; 44: 1320–9. doi: 10.1002/jmri.25277 [DOI] [PubMed] [Google Scholar]
- 33.Choi JS, Kim MJ, Chung YE, Kim KA, Choi JY, Lim JS, et al. Comparison of breathhold, navigator-triggered, and free-breathing diffusion-weighted MRI for focal hepatic lesions. J Magn Reson Imaging 2013; 38: 109–18. doi: 10.1002/jmri.23949 [DOI] [PubMed] [Google Scholar]
- 34.Kim SY, Lee SS, Byun JH, Park SH, Kim JK, Park B, et al. Malignant hepatic tumors: short-term reproducibility of apparent diffusion coefficients with breath-hold and respiratory-triggered diffusion-weighted MR imaging. Radiology 2010; 255: 815: 815–23. doi: 10.1148/radiol.10091706 [DOI] [PubMed] [Google Scholar]
- 35.Kandpal H, Sharma R, Madhusudhan KS, Kapoor KS. Respiratory-triggered versus breath-hold diffusion-weighted MRI of liver lesions: comparison of image quality and apparent diffusion coefficient values. AJR Am J Roentgenol 2009; 192: 915–22. doi: 10.2214/AJR.08.1260 [DOI] [PubMed] [Google Scholar]