Pretransplant prediction of microvascular invasion in patients with hepatocellular carcinoma: Added value of diffusion-weighted magnetic resonance imaging

Abstract

The goals of the meticulous selection of patients with hepatocellular carcinoma (HCC) for liver transplantation (LT) are to minimize the incidence of tumor recurrence after transplantation and to improve patients' survival. Tumor invasion into vascular structures is a known risk factor for tumor recurrence after LT. Macrovascular invasion, which is defined as the presence of a tumor thrombus in a hepatic or portal vein, can be detected on preoperative cross-sectional imaging. It is considered an absolute contraindication to LT. Conversely, microvascular invasion (MVI) by definition can be definitively diagnosed only by histopathology. Several studies have shown that MVI is associated with HCC recurrence within 1 year of transplantation.1, 2 Although the presence of MVI in an explanted specimen does not guarantee HCC recurrence, the recurrence rate may be as high as 63% when MVI is present.1 Naturally, there has been a lot of interest in trying to predict the presence or absence of MVI before LT. Gouw et al.3 summarized much of the published literature on this topic. It is known that a larger lesion size (>3-4 cm), a higher number of lesions (>3), and moderate-to-poor tumor differentiation are associated with MVI. Increased [18F]fludeoxyglucose uptake on positron emission tomography has also been associated with the detection of MVI by histopathology. In this issue of Liver Transplantation, Suh et al.4 suggest that a reduced apparent diffusion coefficient (ADC) and a higher lesion signal intensity on diffusion-weighted imaging (DWI) performed as a part of liver magnetic resonance imaging (MRI) may be predictive of MVI. ADC, apparent diffusion coefficient; b, gradient strength; DWI, diffusion-weighted imaging; HCC, hepatocellular carcinoma; LT, liver transplantation; MRI, magnetic resonance imaging; MVI, microvascular invasion. DWI was developed in the 1980s and originally was used for imaging of the brain. Hepatic applications of DWI were first described in the 1990s, and the interest in DWI for liver lesion detection, differentiation between benign and malignant processes, and the prediction of responses to liver-directed therapy has been increasing. Principles and applications of the DWI technique for liver imaging are described in a recent comprehensive literature review by Taouli and Koh.5 Briefly, diffusion is a physical process that results from the thermally driven, random motion of water molecules. In a container of water, molecules undergo free, thermally agitated diffusion (with a 3-dimensional Gaussian distribution). The width of the Gaussian distribution expands with the elapsed time, and the average square of this width per unit of time provides the ADC units (mm2/second). In tissues, the movement of water molecules is affected by their interactions with cell membranes and macromolecules. The DWI technique derives its image contrast from differences in the mobility of protons (primarily associated with water). In tissues that are highly cellular (eg, tumors), the tortuosity of the extracellular space and the higher density of hydrophobic cellular membranes restrict the diffusion of water protons. In such an environment, water diffusion is said to be relatively restricted. Conversely, in cystic or necrotic tissues, the apparent diffusion of water protons is relatively free. In effect, DWI provides information on tissue cellularity and the integrity of cell membranes. The technique for diffusion-weighted image acquisition has been evolving over time. Currently, both breath-hold and free-breathing techniques are in use. The breath-hold technique is faster and allows imaging of the entire liver in 2 breath holds of 20 to 30 seconds. The free-breathing DWI technique requires 5 to 6 minutes to image the entire liver but results in better image quality. DWI is typically performed at 2 or more different gradient strength (b) values. Although imaging at lower b values (b < 150 seconds/mm2) allows lesions to stand out through the suppression of signals from blood vessels, at higher b values (b > 500 seconds/mm2), there is usually less signal attenuation from cellular tumors containing protons with shorter diffusion distances in comparison with the normal liver. Suh et al.4 used a free-breathing, single-shot, echo planar sequence for DWI with b values of 50, 400, and 800 seconds/mm2. Performing DWI at 2 or more b values allows tumor detection and characterization based on differences in water diffusivity as well as the calculation of the ADC of tissues. The ADC values, in turn, can be used quantitatively as a way of differentiating benign lesions from malignant ones [ADC values are typically higher for benign lesions (eg, hemangiomas and cysts) versus malignant lesions6] and tracking responses to liver-directed therapy (treatment-induced necrosis leads to less diffusion restriction and an increase in ADC7). However, DWI does not appear to improve the detection of tumor recurrence at the treatment site after liver-directed therapy in comparison with the conventional dynamic contrast-enhanced technique.8 The addition of DWI to standard MRI sequences such as T2-weighted imaging and dynamic contrast-enhanced T1-weighted imaging has been found to increase the ability to detect small (<2-cm) HCC lesions,9 and it may improve differentiation between well-differentiated HCC and dysplastic nodules against the background of a cirrhotic liver.10 Furthermore, lower ADC values have been associated with a higher HCC lesion grade11 and the presence of tumor thrombi.12 The study by Suh et al.,4 which included a cohort of 65 patients with 67 HCCs in all, is the first retrospective study that demonstrated a correlation between lower ADC values and MVI. The authors suggest that with a cutoff of 1.11 × 10−3 mm2/second, the ADC provided a sensitivity of 93.5% and a specificity of 72.2% for the prediction of MVI with an odds ratio of 24.5 (95% confidence interval = 4.1-144.8). Furthermore, the authors report that all lesions with ADC values of 0.97 × 10−3 mm2/second or less had MVI according to histopathology, whereas none of the lesions with ADC values greater than 1.15 × 10−3 mm2/second had MVI. The authors conclude that there may be a threshold below which all or most lesions are likely to have MVI and above which few or no lesions will demonstrate this histopathological finding. This study addresses a clinically important issue in the management of patients with HCC. The ability to use diffusion imaging–derived threshold values would help in determining the likelihood of MVI at the time of transplantation and, therefore, the risk of HCC recurrence. Despite the promising preliminary results reported in this study, several technical considerations and limitations should be addressed in the future. The quantification of ADC values remains a challenge, although improved signal-to-noise ratios with greater field strengths and the ability to apply multiple b values would be expected to improve data reproducibility. Absolute ADC values are dependent on technical parameters and have been shown to vary between vendors; this makes it difficult to derive meaningful quantitative data. Interestingly, the authors have used a ratio of the signal intensity and ADC with the background liver as the denominator. This could serve as an internal ratio; however, advanced liver fibrosis and cirrhosis would be expected to reduce the liver ADC to a variable amount and might, therefore, affect the results. DWI has an inherently low signal-to-noise ratio that is accentuated at high b values. A relative loss of signal intensity from less cellular tissues permits greater visualization of small lesions in comparison with conventional T2-weighted imaging. Diffusion imaging should be interpreted in conjunction with the ADC map because the signal intensity may be high in lesions with T2 shine, but unlike malignant tumors, such lesions also appear to have high signal intensity rather than low signal intensity on ADC maps. The authors' example of HCC with a high signal intensity on the ADC map is difficult to explain. A retrospective study without direct anatomical correlations of imaging and pathology results for matching each identified lesion could result in the potential misidentification of nodules. The question that arises is not whether a tumor was present but whether the correct nodule was interpreted as a tumor on imaging. The authors used a 0.8-cm slice thickness for the diffusion protocol. The lower end of the size range of lesions described in the results narrative was 0.7 cm. The potential of volume averaging and the ability to accurately quantify ADC values for subcentimeter lesions are concerns. An assessment of interobserver variability would be helpful for future studies of this topic. ADC measurements within the liver can be challenging. The authors present a narrow range of ADC values for the detection of MVI. A broader range of ADC values has been described for the background liver and liver tumors,13 and overlapping with the ADC values of solid tumors has been described.6 Considerable variations in ADC values may occur even within regions of the liver, and they may be related to the distance from the surface coil and susceptibility and motion artifacts.5 In particular, diffusion imaging and ADC measurements are unreliable in proximity to the heart in the left lobe of the liver. Regional variations in ADC measurements can be corrected to some extent by the use of ratios; however, this will not be able to correct for motion artifacts. An assessment of data reproducibility would be helpful in understanding the impact of regional variations in ADC values and the effects of lesion sizes on ADC measurements. The uncertainty about the true ADC measurement suggests the need for greater standardization of imaging parameters to minimize the measurement variability across platforms in order to allow more meaningful comparisons of results, facilitate multicenter studies, and ultimately influence clinical practice. Until ADC measurements can be standardized, the clinical utility of DWI in predicting pathological findings will remain limited.