Current role of ultrasound in the diagnosis of hepatocellular carcinoma.

Ultrasonography (US) is a major, sustainable hepatocellular carcinoma (HCC) surveillance method as it provides inexpensive, real-time, and noninvasive detection. Since US findings are based on pathological features, knowledge of pathological features is essential for delivering a correct US diagnosis. Recent advances in US equipment have made it possible to provide more information, such as malignancy potential and accurate localization diagnosis of HCC. Evaluation of malignancy potential is important to determine the treatment strategy, especially for small HCC. Diagnosis of blood flow dynamics using color Doppler and contrast-enhanced US is one of the most definitive approaches for evaluating HCC malignancy potential. Recently, a new Doppler microvascular imaging technique, superb microvascular imaging, which can detect Doppler signals generated by low-velocity blood flow, was developed. A fusion imaging system, another innovative US technology, has already become an indispensable technology over the last few years not only for US-guided radiofrequency ablation but also for the detection of small, invisible HCC. This article reviews the evidence on the use of ultrasound and contrast-enhanced ultrasound with Sonazoid for the practical management of HCC.


Introduction
Ultrasonography (US) is a simple and noninvasive real-time imaging method available worldwide. Thus, it is the most frequently used imaging tool for diagnosing liver diseases. US is important not only for surveillance but also characterization of hepatocellular carcinoma (HCC). Most patients with HCC show liver cirrhosis (LC) with poor liver function and a high recurrence rate. Thus, early detection of HCC, especially in patients with LC, is important for timely treatment, which minimizes damage and preserves liver function.
While both computed tomography (CT) and magnetic resonance imaging (MRI) require contrast agents to detect small HCCs in most cases, US can detect most HCCs even without contrast agents. The use of contrast agents for CT and MRI is restricted in elderly patients with declining renal function. In addition, US is not associated with radiation exposure concerns. As US involves repeated examinations over a long period, it is suitable for the surveillance of HCC. Of course, using contrast agents improves the sensitivity and specificity of US; therefore, this method is useful for evaluating the malignancy grade of HCC, which is important when determining a treatment plan. Moreover, many new technologies such as fusion have improved our ability to diagnose HCC.
This article reviews the evidence on use of US and contrast-enhanced US with Sonazoid (Sonazoid CEUS) for the diagnosis and practical management of HCC.

B-mode findings
In abdominal US, both detection and characterization play a role in diagnosing focal liver lesions. B-mode findings of HCC are the basis of diagnostic ultrasound, and the differential diagnosis is based on tumor shape, border and contour, tumor margin, and intratumoral and posterior echo [1]. The Terminology and Diagnostic Criteria Committee (TDCC) of Japan Society of Ultrasonics in Medicine (JSUM) summarizes HCC features of B-mode findings in detail (as presented in Table 1). These findings are useful for the differentiation of focal liver lesions.
The internal echoes of HCCs smaller than 10 mm are almost hypoechoic (low level) or isoechoic, and the number of such low-level echoes increases with cell density. When tumor growth occurs as multistep hepatocarcinogenesis, fatty change is most frequently observed (36.4%) at a tumor diameter of 10-15 mm [7], and internal echoes of these HCCs are hyperechoic. When the diameter of an HCC reaches 20 mm or more, typical US patterns such as the "mosaic pattern," "peripheral sonolucency (halo)," "lateral shadow," and "posterior echo enhancement" can be recognized [8][9][10][11]. Findings of "mosaic pattern," "posterior echo enhancement," and "lateral shadow" show a higher accuracy (≥ 70%) and specificity (≥ 90%) in diagnosing HCC than metastatic liver cancer. With an increase in the size of the tumor, the frequency of observation of these US findings increases. However, these typical US findings are less frequently observed in smaller HCCs. The "halo sign" corresponds to the thin fibrous capsule of the HCC [12][13][14][15] (Figs. 1a, b). Correspondence between the sonographic halo sign and a histological capsule has been reported to be 90.1% [16]. The "lateral shadow," which is a linear US feature observed at the edge of a tumor, represents the refraction that occurs when ultrasound passes through spherical tissue and the surrounding tissue at different speeds (Figs. 1c, d). Posterior echo enhancement arises posterior to any lesion that attenuates sound less than the surrounding tissue; the intensity of the transmitted ultrasound beam is relatively preserved distal to the lesion [3]. However, posterior echo enhancement is not specific to HCC; this finding is also associated with hemangiomas and cystic lesions. In addition, the macroscopic configuration is important for the prediction of recurrence and prognosis in patients with HCC. In the classification proposed by a research group from Japan, the macroscopic configuration of HCC is divided into five types: small nodular type with indistinct margins, simple nodular type, simple nodular type with extranodular growth, confluent multinodular type, and infiltrative type [17,18]. The potential for malignancy tends to change in accordance with the progression of the macroscopic configuration [11,18].

Color/power Doppler
Color Doppler can help visualize blood flow signals inside and at the margin of a tumor, and the direction of blood flow can be determined on the basis of color. Color Doppler helps identify not only nutrient vessels but also volume of blood flow. When it is necessary to detect the blood flow orthogonal to the ultrasonic beam or to evaluate low flow velocity, power Doppler is effective. Previously, power Doppler could increase the detection sensitivity, but it could not detect the blood flow velocity or direction. Now, with advances in technology, power Doppler can also detect the direction of blood flow.
JSUM's TDCC also summarizes the HCC features of Doppler findings [1] (as presented in Table 2). In most HCCs smaller than 2 cm, blood flow is low and appears as lines or dots inside or around the tumor. Typical color Doppler features of a small HCC include afferent continuous waveform signals, which reflect a feeding portal flow [19]. When the diameter of a tumor is 2 cm or more, the blood flow volume increases. Especially in a moderately differentiated HCC that has a capsule and shows expansive growth, basket-pattern blood flow is observed. This pattern represents a fine network of arterial vessels that surrounds the tumor nodule [11,20,21]. Typical color Doppler features of an advanced HCC (massive type) include afferent pulsatile waveform signals associated with intratumoral continuous waveform signals and efferent continuous waveform signals [20].
Recently, a new Doppler microvascular imaging technique known as superb microvascular imaging (SMI; Canon Medical Systems, Otawara, Japan), which can distinguish Doppler signals generated by low-velocity blood flow from those generated by tissue movement, was developed. This technique can reduce motion artifacts and simultaneously provide a high level of sensitivity and imaging resolution at a high frame rate. As a result, ultrasound examiners using this technique can clearly observe low-velocity capillary blood flow without the need for a contrast agent. This increases confidence in the assessment of the nature of the tumors (Fig. 2) [22].
Similar technology, such as high-definition color (HDC; GE Healthcare, Chalfont St. Giles, UK), has also become available and can display very small tumor vessels (Fig. 3). Although there are still few papers on the evaluation of HCC using these new Doppler microvascular imaging techniques, these are already being used in clinical practice in Japan. Thus, these blood flow display technologies may be used more in the future for HCC diagnosis using ultrasonography.

Contrast agent and time phase
The first-generation agent Levovist (SH U 508A; Schering AG, Berlin, Germany) has been introduced not only in Japan but also in other countries [23]. However, the parenchymaspecific contrast yielded by Levovist is effective only when imaging is performed at high acoustic power using a high mechanical index (MI), and the effect is transient [24]. Therefore, it cannot be used for real-time examinations in  [27][28][29][30][31][32][33][34]. SonoVue is minimally phagocytosed by reticuloendothelial cells (Kupffer cells) [35,36]. Parenchyma-specific contrast images can be seen only 3-5 min after injection of SonoVue. Thus, its use is not approved in Japan [27].
In Japan, we can use only the second-generation ultrasound contrast agent Sonazoid (GE Healthcare, Amersham, UK); it is a lipid-stabilized suspension of perfluorobutane gas microbubbles, available since January 2007 [27]. Sonazoid is associated with a low incidence of side effects [37]. Because Sonazoid is metabolized in the lungs, it has no contraindication for patients with renal dysfunction or iodine allergy. Sonazoid CEUS is usually performed at a low MI. Because a low MI facilitates the prevention of microbubble destruction, obtaining perfusion images of hepatic lesions in the vascular phase using Sonazoid CEUS is easier. Sonazoid CEUS at a low MI also enables the user to scan the whole liver in the late phase, facilitating the detection of perfusion defect images of hepatic malignant lesions. With the employment of both the vascular and late phases, Sonazoid CEUS facilitates the characterization of liver tumors, histological grading of HCC lesions, and guided ablation of unresectable HCCs [27].
In humans, the reason for the enhancement of liver parenchyma and the non-enhancement of malignant tumors in the late phase using Sonazoid CEUS is not completely understood; however, it is thought to be a result of phagocytosis by reticuloendothelial (Kupffer) cells due to the adherence of microbubbles to the hepatic sinusoids and tumor vascular spaces or due to recirculation of Sonazoid microbubbles. Because malignant tumors contain few or no reticuloendothelial cells, they appear as perfusion defects in the late phase. The recommended dose of Sonazoid is 0.015 mL/ kg. However, with an increase in the quantity of contrast agent, the ultrasound signal of the background liver may be enhanced simultaneously with that of the tumor at the time of imaging, which may make diagnosis difficult. This dose was determined on the basis of clinical research conducted at the initial introduction of this agent. Due to advances in US techniques, the image quality of Sonazoid CEUS is sufficient at doses lower than the recommended dose. Therefore, most authors injected a decreased dose of Sonazoid to evaluate the vascularity of liver lesions, especially HCCs [27,[38][39][40][41]. With regard to Sonazoid CEUS, time phases may be categorized into three subphases: arterial, portal, and post-vascular (Kupffer) phases [42]. The arterial phase is the period from the point of arrival of perflubutane microbubbles into the hepatic artery to the point when the portal vein and the hepatic artery intersect in the time-intensity curve (TIC). The portal phase is the period from the point when the hepatic artery and the portal vein intersect in the TIC to the point when the portal vein and the parenchyma intersect in the TIC. The post-vascular (Kupffer) phase (from 10 min after injection) starts when the parenchyma of the liver is enhanced after the disappearance of the contrast effect from the vasculature. JSUM's TDCC also summarizes the HCC features of CEUS findings [1] (as presented in Table 3).

Defect reperfusion imaging
Differentiation of necrotic and viable areas is sometimes difficult in the post-vascular (Kupffer) phase as both appear as perfusion defects. To solve this problem, Kudo et al. reinjected Sonazoid into HCCs that had previously shown a perfusion defect in the post-vascular (Kupffer) phase [43]. They devised a method called defect reperfusion imaging, which confirms blood flow into the defects. Using this method, Hatanaka et al. reported that Sonazoid CEUS showed a higher sensitivity (95.4%) and accuracy (94.7%) in the diagnosis of hepatic malignancies than contrastenhanced CT (sensitivity 85.2% and accuracy 82.3%) (both P < 0.005) [44].

Contrast-enhanced low-MI tissue harmonic imaging
Contrast harmonic imaging (CHI) exploits the nonlinear oscillations of microbubbles in contrast agents that produce harmonic overtones of the original sound wave [45,46]. Several contrast harmonic software applications have been developed for CEUS examination, with the most promising techniques being phase inversion (PI) and amplitude modulation (AM) [47,48]. These techniques usually sacrifice spatial and time resolution to improve sensitivity and specificity of contrast signals. Tissue harmonic imaging (THI), which is the most commonly used conventional B-mode imaging modality, is a form of native harmonic imaging that provides a better signal-to-noise ratio and reduced side lobe artifacts than the other types [49][50][51][52]. As this THI technique is based on the  [53]. This new CEUS method, also called low-MI harmonic imaging, is a technique that has been enabled by the recent evolution of ultrasound devices. It allows us to observe vessels in the vascular phase at high spatial and time resolutions using conventional B-mode US techniques such as compounding and time smoothing (Fig. 4c). Contrast low-MI THI using Sonazoid is also useful in the post-vascular (Kupffer) phase due to its high spatial and time resolutions (Fig. 5); however, the sensitivity is slightly lower than that of amplitude modulation, the most sensitive CEUS mode. Thus, low MI contrast THI is effective at providing detailed resolution, image quality, focal abnormality margin sharpness, and penetration for hepatic imaging.

High-MI intermittent imaging
Numata et al. performed intermittent imaging at the rate of 2 frames/s in coded harmonic angio (CHA) mode at a high MI (0.7-1.2) to depict the features of tumor vascularity in the late phase (> 5 min). They called this method "high MI intermittent imaging." When they used high-MI intermittent imaging to scan a tumor lesion, the Sonazoid microbubbles within and around the tumor may have been destroyed immediately, and the tumor vessels and tumor enhancements were possibly seen because of back flow into the tumor vessels and vascular spaces. In contrast, if the tumor became necrotic as a result of transcatheter arterial embolization, ablation therapy, or spontaneous necrosis, the lesions would exhibit neither tumor vessels nor enhancement when scanned using high-MI intermittent imaging. The advantages of this method include the acquisition of enough information from patients with multiple lesions. This method is regarded as an alternative to additional injection of contrast agents when tumor vascularity requires further evaluation, and it can be used when there is a lack of medical personnel or to save time [27,38]. High-MI Doppler methods, such as advanced dynamic flow (ADF), may be used for evaluating HCC in the postvascular (Kupffer) phase with the highest sensitivity and specificity, because microbubbles accumulated in the liver are destroyed by high-MI ultrasound exposure [54,55]. The CEUS method can be performed only once as it destroys almost all the bubbles around the tumor on its first usage. Therefore, it is necessary to be careful not to destroy the bubbles accidentally. However, recent improvements have made it possible to perform these high-MI Doppler methods using only simple operations (Fig. 4f).

Diagnosis of malignancy grade
Evaluation of malignancy potential is important for determining the treatment strategy, especially for small HCC. Although B-mode US can reflect the macroscopic morphology of HCC, only one study has investigated the relationship between B-mode ultrasonograms and the histological differentiation grade in patients with small HCC [56]. Moribana et al. classified small HCC into two groups ultrasonographically: type 1 (with halo) and type 2 (without halo). Type 2 was classified into three subgroups: type 2a, homogenous hyperechoic; type 2b, hypoechoic with a smooth margin; and type 2c, hypoechoic with an irregular or unclear margin. They showed that the malignancy potential of type 2a was the lowest and that of type 2c was the highest [56]. Their study is simple and important; however, many more studies have demonstrated the utility of CEUS in the assessment of malignancy.

Evaluation in arterial phase
Details regarding the vascularity of an HCC are important, because blood supply and the grade of HCC malignancy are closely related. Hayashi et al. found a strong correlation between the intranodular arterial and portal supply evaluated by CT during hepatic arteriography and CT during arterial portography, and the malignancy grade of the hepatocellular nodules. In other words, the intranodular portal supply relative to the surrounding liver parenchyma observed by CT during arterial portography was decreased, whereas the intranodular arterial supply revealed by CT during hepatic arteriography was first decreased during the early stage of hepatocarcinogenesis, and then increased according to the malignancy grade of the nodules [57]. Thus, detection of hypervascularity in the arterial phase is important for assessing the malignancy grade of HCC. The sensitivity of CEUS in detecting the tumor vascularity of HCC nodules is equal to [58] or higher than [59] that of contrast-enhanced CT. Numata et al. demonstrated that the detection rate for hypervascularity of early HCC using CEUS (32.7%, 17/52) was significantly higher than that obtained using contrastenhanced CT (21.2%, 11/52) (P < 0.005) [59]. Maruyama et al. also showed that CEUS detected hypervascularity in 26% (7/27) of the tumors characterized as non-hypervascular using contrast-enhanced CT [60].

Evaluation in Kupffer phase
Superparamagnetic iron oxide (SPIO) (ferucarbotran [Resovist]; Bayer, Osaka, Japan) is a tissue-specific MRI contrast agent similar to Sonazoid; it is phagocytized by the liver  [40]. In the Kupffer phase with Sonazoid CEUS, all moderately and poorly differentiated HCCs showed a hypoechoic pattern and were detected as perfusion defects; most (69.2%) of the well-differentiated HCCs showed an isoechoic pattern [40]. Sonazoid CEUS is also useful for estimating the histological grades of HCCs.
Gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid (Gd-EOB-DTPA) has the properties of an extracellular and hepatocyte-specific contrast agent, and has recently played a leading role in the diagnosis of HCCs [62,63]. Ohama et al. compared the histological enhancement patterns of the post-vascular (Kupffer) phase of Sonazoid CEUS and the hepatobiliary phase of Gd-EOB-DTPA-enhanced MRI, as well as uptake of Sonazoid and Gd-EOB-DTPA by HCC [64]. They evaluated the uptake of Sonazoid by TruAgent detection 30 min after administration of Sonazoid using Sonazoid enhancement indices [log (Doppler signals in the tumorous lesion) -log (mean of Doppler signals in the non-tumorous area]. TruAgent detection, the technique used to detect Doppler signals from the microbubbles of contrast agents, is based on the loss of correlation caused by the destruction of bubbles with high mechanical indices, ranging from 0.8 to 1.0. The median values of the Sonazoid enhancement indices were also decreased with a decrease in HCC histological differentiation: DNs 0 (range − 1.10 to 3.20); hypovascular well-differentiated HCCs 1.05 (range 5.65 to 3.60); hypervascular well-differentiated HCCs − 2.60 (range, − 0.82 to 0.40); moderately to poorly differentiated HCCs − 5.00 (range, − 8.60 to 2.20). In contrast, there was no difference in Gd-EOB-DTPA uptake across stages of HCC differentiation.
In addition, Kupffer images of Sonazoid CEUS can also be used to evaluate the gross HCC types that are closely related to malignancy potential. Hatanaka et al. classified HCC using Sonazoid CEUS into three types based on the macroscopic classification of the Liver Cancer Study Group of Japan [17,65]: single nodular (SN) type, single nodular with extranodular growth (SNEG) type, and confluent multinodular (CMN) type. The ability of Sonazoid CEUS to correctly depict the gross HCC types was assessed. The sensitivity, specificity, and accuracy of Sonazoid CEUS were 96%, 80%, and 90%, respectively [66]. The diagnostic accuracy was 86.9% (53/61) for Sonazoid CEUS and 65.6% (40/61) for CECT [67].

Maximum intensity projection
Micro-flow imaging (MFI), which is one of the CEUS methods for maximum intensity projection (MIP), is an accumulative imaging technique that reveals blood vessels after a flash with high-transmission power ultrasound exposure. In this modality, after the flash generated by the flash-replenishment sequence (FRS), the microbubbles in the scanning area are completely destroyed by the high energy output from the ultrasound beam. In succession, the area is re-perfused with microbubbles from the adjacent blood vessels, and the microbubbles are visible, because the imaging mode quickly shifts into a low-MI mode. Through maximumholding image processing, the trace of the microbubbles in a temporal dimension can be clearly depicted, reflecting the vascularity in the scanning area [32]. HCC is known to undergo changes in vascular structure as it progresses. MIP can help visualize this fine vascular structure without an angiography of the liver, and the spatial resolution of MIP remains higher than that of SMI (Fig. 6).
Sugimoto et al. reported the feasibility of diagnosis of the histological differentiation of HCC using MIP with SonoVue CEUS. Briefly, the maximum-hold processing began immediately after the burst scan. The burst scan consisted of high-MI (1.3-1.6) scanning of five frames. Low-MI (0.16-0.30) scanning begun immediately after the MI burst scanning to visualize fresh microbubble contrast agent flowing into the scanning volume. The maximum intensity holding sequence started at the same time as the flash-replenishment low-MI imaging, which maintained the maximum brightness level on each pixel and displayed a persistent vision. They classified the image patterns as follows: normal, cotton, vascular, and dead-wood. In the normal pattern, the border between tumoral and non-tumoral regions is slightly indistinct, and the vascular architecture in the tumoral region is similar to that in the adjacent non-tumoral region. In the cotton pattern, the border between the tumoral and non-tumoral regions is distinct, but tumoral blood vessels are not clearly visualized, and the tumor appears pale as a whole as if it was stained. In the vascular pattern, tortuous and meandering tumoral blood vessels are clearly visualized, and the tumor is imaged as a whole. In the dead-wood pattern, tumoral blood vessels are clearly visualized, but they gradually taper off and are suddenly interrupted. HCCs showing normal and cotton patterns are classified as well differentiated; those showing vascular or dead-wood patterns are classified as moderately or poorly differentiated; the sensitivity, specificity, and accuracy of these assessments were found to be 85%, 92.7%, and 90%, respectively [69].
We also examined the clinical utility of the malignancy grading system to detect histologically advanced HCC using the MIP pattern. Our malignancy grading system was used to evaluate HCC histological grade with a combination of two key features of Sonazoid CEUS, i.e., Kupffer imaging and MIP pattern. Kupffer imaging was classified as an isoechoic or hypoechoic pattern. MIP patterns were evaluated using MFI and were classified as one of the following three patterns: fine (tumor vessels are not clearly visualized and only fine vessels are visualized), vascular (tumor vessels are clearly visualized), and irregular (tumor vessels are thick and irregular). An advantage of our MIP classification is the simplicity of its classification scheme. If some irregular vessels can be detected, the case is classified as an "irregular pattern". Similarly, if some tumor vessels thicker than the surrounding fifth or sixth branch can be detected, the case is classified as a "vascular pattern". Other cases were classified into a "fine pattern," including cases in which the difference between the tumor vessels and surrounding hepatic parenchyma could not be detected. These MIP patterns of "fine," "vascular," and "irregular" are nearly equivalent to, respectively, Sugimoto's MIP patterns of "normal or cotton," "vascular," and "dead-wood." Simplification to three classifications has also made classification easier, especially for "normal or cotton" and "cotton or vascular" patterns, which are sometimes difficult to distinguish. Based on the combination of MIP patterns and Kupffer imaging, we were able to classify HCC into four grades: Grade 1 (iso-fine/vascular), Grade 2 (hypo-fine), Grade 3 (hypo-vascular), and Grade 4 (hypo-irregular) ( Table 4). Reportedly, the distribution of moderately and poorly differentiated HCCs was as follows: Grade 1, 4% (1/24); Grade 2, 52% (15/29); Grade 3, 85% (44/52); and Grade 4, 100% (16/16). In this study, all patients with irregular patterns had either moderately or  Fig. 4 A case of newly developed HCC (maximum diameter 11 mm) in Segment 2. Hepatobiliary phase contrast-enhanced MRI with gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid (Gd-EOB-DTPA) in the arterial phase shows a high-intensity area (a). This tumor is clearly enhanced with contrast-enhanced ultrasound with Sonazoid in the arterial phase using the phase inversion method (b). However, it is more clearly depicted by low mechanical index (MI) contrast-enhanced tissue harmonic imaging (THI) because of its high spatial and time resolutions (c). In the post-vascular (Kupffer) phase, the tumor is in the hypoechoic area by amplitude modulation, which is the mode most sensitive to contrast-enhanced ultrasound (d). We could also delineate the tumor with the same sensitivity by low MI contrast-enhanced THI (e). In addition, high-MI intermittent Doppler imaging could most clearly depict the tumor in the post-vascular (Kupffer phase) (f). The arrows point to the location of the identified hypoechoic tumor on imaging 0% (0/4); Grade 2, 13% (1/8); Grade 3, 23% (11/48); and Grade 4, 67% (8/12) [70]. Takada et al. also reported the irregular MIP pattern, using Sonazoid CEUS, to be the most important independent risk factor for a poor outcome after successful radiofrequency ablation (RFA) in patients with early stage HCC (hazard ratio: 8.26, 95% confidence interval: 2.24-30.3, P = 0.002) [71].
As these study results demonstrate, MIP patterns are useful for estimating the malignancy potential of HCC. However, there remain some limitations to MIP analysis. First, the accuracy depends on the location, because it cannot be assessed without being clearly depicted. Second, there remains the possibility that key images cannot be detected, as only two-dimensional images of MIP patterns instead of the entire tumor are detectable. Therefore, these cases may be classified as being of a less malignant grade. Maruyama et al. summarized in detail the utility of Sonazoid CEUS for HCC treatment [41]. RFA is the most frequently used local treatment for HCC. Sonazoid CEUS could increase RFA applicability (from 21% [n = 95 cases] to 32% [n = 219 cases]) [72] and reduce the number of RFA sessions (from 1.49 ± 0.76 [historical control] to 1.33 ± 0.45) [73]. Using Sonazoid CEUS, Dohmen et al. demonstrated an increase in the non-local recurrence rate from 66.4 to 85.3% 2 years after RFA [74]. A recent study showed that Sonazoid CEUS performed 3 h after RFA could show the outline of the coagulated tumors in 78/87 patients (89.7%), and that the 5-year cumulative local recurrence rate was very low (2.3%) with a 5-year cumulative survival rate of 58.4% [75]. However, we should know two signs that indicate the risk of recurrence. The first sign is a linear-shaped positive enhancement, which was observed in the RFA-treated area in 33 lesions (18.4%) using Sonazoid CEUS; 17 of them were followed-up with no treatment, and three of these 17 (17.6%) showed local tumor progression that corresponded to linear enhancement [58]. The second sign is gradual intra-tumor enhancement in the pre-treatment early arterial phase, which indicates potential risk of distant recurrence after RFA [76].

Evaluation of treatment
Sonazoid CEUS is also useful for assessing therapeutic effects in transcatheter arterial chemoembolization (TACE), another important procedure for treating HCC. Xia et al. demonstrated that detection rates of residual tumors using Sonazoid CEUS 1 week after TACE were better than those using CT (58.1% vs. 39.5%, P < 0.001) [77]. A subsequent prospective study confirmed this result; it showed that the detection rate of residual HCC nodules using Sonazoid CEUS 1 day after TACE was significantly higher than that  Fig. 6 A case of newly developed HCC (maximum diameter 21 mm) in Segment 5. An isoechoic tumor was found at the same site as the contrast-enhanced CT image in the arterial phase using fusion imaging (a). Maximum intensity projection (MIP) can help visualize the fine vascular structure with the superior vessel continuity (b). The arrow points to the location of the identified isoechoic tumor on imaging using contrast-enhanced CT 1 month after TACE (95.7% vs. 78.7%, P < 0.05) [78].
The first oral multi-targeted tyrosine kinase inhibitor, Sorafenib (Nexavar; Bayer Healthcare, Leverkusen, Germany), is recommended for unresectable advanced HCC. However, it is an expensive drug associated with certain adverse events. Therefore, evaluation of the early response is required. Sugimoto et al. demonstrated that tumor perfusion parameters were statistically significant on the basis of the area under the time-intensity curve (AUC) during wash-in on day 14, the most relevant factor indicating tumor response (P = 0.0016) [79]. Shiozawa et al. also examined the mean arrival time of Sonazoid using parametric imaging before and 2 weeks after treatment and suggested that an extended mean arrival time improves the median survival time (arrival time extend group 792 days, arrival time not extend group 403 days) [80].

Cost-effectiveness
Surveillance of HCC using US improves the prognosis of patients [81][82][83], and it is a cost-effective method for evaluating patients with HCV-related cirrhosis [84,85], even after a sustained virologic response to therapy [86]. The costeffectiveness of CEUS for HCC surveillance in patients with LC was also reported [87]. Using the Markov model, we demonstrated that CEUS surveillance could cost effectively extend the expected survival time. Sensitivity analysis showed that the annual incidence of HCC and sensitivity of CEUS were two critical parameters. When the annual incidence of HCC was more than 2% and/or the CEUS sensitivity was more than 80%, the incremental cost-effectiveness ratio was less than the commonly accepted threshold of $US 50,000/QALY, indicating cost-effectiveness. However, economic conditions could vary over time, and the substantial effect is yet to be evaluated in each country.

Fusion images
Fusion imaging is a novel technology that accurately combines real-time US images with real-time CT or MRI volume data and displays them on the same monitor, side by side. This means that a clinician can visualize both registered multiplanar reconstruction (MPR) images on the same monitor to make diagnostic or procedural decisions in real time. The fusion imaging system is composed of a positionsensing unit mounted on an ultrasound unit and an electromagnetic field transmitter. Some of the latest probe models are equipped with a position sensor.
US is the first choice for percutaneous interventional procedures as it provides real-time, noninvasive, repeatable, and non-radiation-based imaging. However, this process can be difficult if the ideal US scanning plane is different from that of the CT or MR image. Moreover, breathing and displacement and deformation of the abdominal structures due to pressure from the US probe can affect the process of mental registration. Lesions located deep in the most distal regions can be blurred and difficult to identify. In addition, US is affected by the presence of bone, gas, and fat. Real-time fusion imaging capitalizes on the strengths of all imaging modalities simultaneously. In addition, it allows for images acquired by means that are not affected by such issues, such as CT scanning and MR imaging, to be placed alongside or overlaid on the acquired images for better detection of lesions. US fusion imaging can also be associated with advanced US imaging techniques such as color/power Doppler, SMI, and CEUS for better localization and characterization of the lesions to be treated [88,89].
Some hepatic tumors that can be visualized by CT or MR cannot be seen on US due to their small size, their location, or their echogenicity. In these cases, fusion imaging has been shown to make HCC nodules more conspicuous and to increase the feasibility of percutaneous RFA of HCCs not visible on conventional US [88,90,91] (Fig. 6). If HCCs remain non-visible after fusion imaging, anatomical landmarks surrounding the lesions can be used to guide correct needle placement [88]. Thus, with the use of fusion imaging, a larger population can benefit from US-guided ablation procedures instead of undergoing a CT-guided ablation or a major surgical operation, which are more invasive and expensive techniques. The average procedure duration for a US and CT/MRI fusion-guided liver biopsy was approximately half of that of a CT-guided liver biopsy (31.63 min vs 61.67 min, P = 0.003) [92]. Fusion imaging can also minimize false-positive lesion detection during US-guided RFA and consistently improve the detection of HCCs, especially when these are smaller than 20 mm [93]. The ability of fusion imaging to reduce false positives also applies to the evaluation of local tumor progression after RFA and TACE [94].
In recent years, next-generation microwaves systems, such as the Emprint Ablation System with Thermosphere Technology (Covidien, Boulder, CO), have been favored for their ability to consistently provide high intratumoral temperatures, fast ablation times, and large ablation volumes. [95]. Microwave ablation (MWA) is less susceptible to the heat sink effect because of its higher temperatures and shorter ablation times. Therefore, MWA needs to be used more carefully than RFA for its complications such as bile duct injuries, vascular damage, and visceral damage (colon, stomach, intestine, gallbladder, kidney, abdominal wall). To prevent such complications, a more accurate assessment of the predicted ablation area is necessary for MWA.
Recently, we have also been able to use some new 3D display methods, such as 3D-GPS marker, which is one function of volume navigation (V Nav; GE Healthcare, Chalfont St. Giles, UK). This 3D-GPS marker may easily change the size and shape of an ellipse. Using this function, operators may make more accurate simulations of how they ablate the tumor with sufficient margins in three dimensions. By visually demonstrating the predicted ablation area in three dimensions, the operator can grasp the positional relationship between the ablation area and major blood vessels or other organs prior to ablation (Fig. 7b).
In addition, by combining a virtual needle-tracking system (VirtuTRAX; CIVCO Medical Solutions, Kalona, IA, USA) in which a magnetic sensor is attached to the distal tip of an ablation needle, a 3D-GPS marker may provide a more accurate prediction of the ablation area immediately prior to ablation [96]. This needle-tracking system is able to virtually visualize the tip of the needle and the needle path on US. Thus, the combined use of this needle-tracking system and 3D-GPS marker at the tip of the RFA or MWA needle may successfully visually demonstrate the predicted ablation region (Fig. 7c). A significantly bowed needle, however, may cause the virtual tract not to accurately match the actual needle. This problem may be solved using a more rigid needle such as a MWA needle.
It should be noted that the positional information derived by these 3D-GPS markers is not entirely consistent due to respiratory fluctuations, among other factors.

Computer-aided diagnostic (CAD) systems
CEUS is useful for HCC diagnosis, but differential diagnosis is sometimes difficult. When microbubble destruction and variations in tumor behavior and liver enhancement occur, experienced sonographers (or radiologists) are required to characterize tumors reliably and accurately; however, the a b c Fig. 7 A case of recurrent HCC (maximum diameter 8 mm) in Segment 6. This tumor is detected by contrast-enhanced MRI with Gd-EOBDTPA only in the hepatobiliary phase. Because this tumor was very small, detection was also difficult with only B-mode ultrasound. However, using fusion technology, a high echo tumor could be visualized (a). Since the reconstructed magnetic resonance image can be enlarged and reduced freely, it is possible to search for lesions with ultrasound, while grasping the whole image with MRI (a). The 3D-GPS marker may easily change the size and shape of an ellipse.
Using this function, operators may make more accurate simulations of how they ablate the tumor with sufficient margins in three dimensions (b). In addition, by combining a virtual needle-tracking system, a 3D-GPS marker may provide a more accurate prediction of the ablation area immediately prior to ablation. This needle-tracking system is able to virtually visualize the tip of the needle and the needle path on US. Thus, the combined use of this needle-tracking system and 3D-GPS marker at the tip of the RFA or MWA needle may successfully visually demonstrate the predicted ablation region (c) number of sonographers (or radiologists) experienced in performing and interpreting CEUS studies is limited, and interobserver agreement remains an issue. Computer-aided diagnostic (CAD) systems are a potential solution to these problems. Generally, CAD systems extract features from the B-mode and/or CEUS videos and train machine-learning algorithms to associate these features with the known diagnoses to predict the diagnosis of unknown lesions. Sugimoto et al. proposed a method for classifying focal liver lesions into five classes (well-differentiated HCC, moderately differentiated HCC, poorly differentiated HCC, liver metastasis, and hepatic hemangioma) using four artificial neural networks (ANNs) for CAD analysis of focal liver lesions with Sonazoid CEUS. The accuracies of classifying hepatic hemangioma, HCC, and liver metastasis were 93.3%, 98.6%, and 84.8%, respectively. They also classified historical malignancy by CAD analysis using ANNs. The accuracies of classifying well-differentiated, moderately differentiated, and poorly differentiated HCC were 82.9%, 88%, and 100%, respectively [97].
Kondo et al. also proposed an automatic classification method based on machine learning for the interpretation of focal liver lesions on Sonazoid CEUS. This method yields spatial and temporal features in the arterial phase, portal phase, and Kupffer phase, as well as max-hold images. The lesions are classified as benign or malignant and again as benign, HCC, or metastatic liver tumor using support vector machines (SVM) with a combination of selected optimal features. Experimental results from 98 subjects indicated that benign and malignant classification showed 94.0% sensitivity, 87.1% specificity, and 91.8% accuracy, and that the accuracies of the benign, HCC, and metastatic liver tumor classifications were 84.4%, 87.7%, and 85.7%, respectively [98].
Reports on CAD are already available, and the practice is expected to be applied in clinical practice in the near future. There is no doubt that the introduction of deep learning will improve its accuracy. It will be necessary to develop a new diagnostic system to generate CAD-based results that enable experts to make reliable judgements.

Conclusion
This comprehensive review clearly demonstrates the magnitude of the importance of US in the management of HCC. US examination of HCC has evolved significantly due to advances in equipment. First, B-mode has resulted in a marked improvement in image quality. In addition, using hepatobiliary phase contrast-enhanced MRI with Gd-EOB-DTPA and fusion technology, an innovative USassisted system, even small lesions that were otherwise difficult to detect can be reliably analyzed. The visualized HCC can be evaluated for the very delicate hemodynamics of the tumor without the use of contrast agents by utilizing color Doppler, power Doppler, or SMI, which continue to evolve. Furthermore, when a contrast agent is used, a contrast pattern for each characteristic time phase can be evaluated, so that a more accurate diagnosis can be made. Although there remain certain parts that are difficult to visualize on ultrasonic examination, the high temporal and spatial resolution of US cannot be achieved using CT or MRI. Therefore, the usefulness, cost-effectiveness, and safety of US are indispensable. In the future, CAD will be incorporated into ultrasonic examinations. However, since tumor shape and blood flow evaluation remain unchanged, US will continue to play a central role in the diagnosis of HCC.