Use of CEUS LI-RADS for the Accurate Diagnosis of Nodules in Patients at Risk for Hepatocellular Carcinoma: A Validation Study.

Purpose
To validate the contrast agent-enhanced US Liver Imaging Reporting and Data System (CEUS LI-RADS) algorithm for accurate diagnosis of hepatocellular carcinoma (HCC) and categorization of all nodules encountered in patients at risk for HCC.


Materials and Methods
A single-center retrospective review of 196 nodules in 184 patients at risk for HCC (consisting of 139 HCCs, 18 non-HCC malignancies, and 39 benign nodules) was performed in a three-reader blinded read format, with the use of the CEUS LI-RADS algorithm. Pathologic confirmation was available for 143 nodules (122 HCCs, 18 non-HCC malignancies, and three benign nodules). Nodule sizes ranged between 1.0 and 16.2 cm. Nodules assessed with contrast-enhanced US were assigned various CEUS LI-RADS categories by three blinded readers. CEUS LI-RADS categorization was then compared against histopathologic findings, concurrent CT, and/or MR images or follow-up imaging to assess diagnostic accuracy of CEUS LI-RADS. In addition, the proportion of HCC in all LI-RADS (LR) categories, univariable and multivariable feature analysis, and interrater agreement using Light κ were determined.


Results
The LR-5 category, determined through radiologist categorization of nodules using the CEUS LI-RADS criteria, showed sensitivity, specificity, positive predictive value, and negative predictive value of 86% (119 of 139), 96% (55 of 57), 98% (119 of 121), and 73% (55 of 75), respectively, for the diagnosis of HCC. Two false-positive cases of LR-5 included a cholangiocarcinoma and a combined hepatocellular and cholangiocarcinoma. The remainder of the cholangiocarcinomas in the sample (n = 8) were appropriately categorized as LR-M. Multivariable logistic regression analysis showed that washout of greater than 60 seconds was the contrast-enhanced US feature most predictive of HCC diagnosis, whereas washout of less than 60 seconds was the feature most predictive of nonhepatocellular malignancy. The proportion of HCC nodules categorized in the LR-M and LR-4 categories was 35% and 20%, respectively. Light κ agreement between readers for LI-RADS categorization was 90%.


Conclusion
This study showed excellent specificity for the CEUS LI-RADS LR-5 category, allowing for confident imaging diagnosis of HCC, without necessity for pathologic confirmation. In addition, there was accurate differentiation of HCC from non-HCC malignancies and benign nodules. Only a single cholangiocarcinoma was misdiagnosed as category LR-5, with the remainder of the cholangiocarcinomas in the sample appropriately characterized as category LR-M.Keywords: Abdomen/GI, Evidence Based Medicine, Liver, Neoplasms-Primary, Ultrasound-Contrast© RSNA, 2020.

H istoric inconsistencies in the interpretation of liver imaging and subsequent management of patients at risk for hepatocellular carcinoma (HCC) prompted the development of the Liver Imaging Reporting and Data System (LI-RADS) (1), which was initially designed for CT and MRI. Recently, LI-RADS has been expanded for use with contrast-enhanced US (2). LI-RADS categorization uses nodule size, enhancement characteristics in the arterial phase, and contrast agent washout characteristics for determining LR categories. Contrast agent-enhanced US LI-RADS (CEUS LI-RADS) is endorsed by the American College of Radiology and is now recognized by the American Association for the Study of Liver Diseases (3). LI-RADS currently includes a categorization algorithm, a lexicon of terminology, and an illustration guide.
Most current international guidelines do not include contrast-enhanced US as a valid tool for liver nodule management in at-risk patients, primarily because of concern over false-positive diagnosis of HCC in patients with cholangiocarcinoma (4). Motivation for our study included the recent U.S. Food and Drug Administration approval of a microbubble contrast agent for liver contrast-enhanced US in the United States and the endorsement of CEUS LI-RADS by the American College of Radiology. Successful integration of liver contrast-enhanced US mandates validation of the CEUS LI-RADS algorithm. The CEUS LI-RADS algorithm (https://www. (2) provides categories for all focal liver observations that may be encountered on a contrast-enhanced US scan of an at-risk patient, ranging from benign tumors (eg, hepatic hemangioma) to malignant tumors (eg, metastases and cholangiocarcinoma). Most importantly, it includes a diagnostic table that incorporates the entire spectrum of nodules that develop by hepatocarcinogenesis, culminating in the CEUS LR-5 category, which has the desired outcome of providing a noninvasive and confident diagnosis of HCC. Diagnosis of Hepatocellular Carcinoma Using CEUS LI-RADS US because contrast-enhanced US is a nodule-based imaging technique.

acr.org/Clinical-Resources/Reporting-and-Data-Systems/LI-RADS/CEUS-LI-RADS-v2017)
The purpose of this study was to validate the CEUS LI-RADS algorithm, particularly for the accurate diagnosis of HCC. We hypothesized that the CEUS LI-RADS tool provides the ability to confidently differentiate HCC from nonhepatocellular malignancies as well as benign entities. We showed that category assignment of nodules within CEUS LI-RADS correlated well with reference standard diagnoses, validating the ability of CEUS LI-RADS to accurately diagnose HCC by imaging, and accurately stratified all nodules seen in patients at risk for the development of HCC, including non-HCC malignancies and benign nodules.

Materials and Methods
Study Design This retrospective validation study was approved by the local Health Research Ethics Board. Informed consent was waived for this retrospective review which involved evaluation of images obtained in the context of routine imaging performed at our institution, without necessity for additional imaging or patient visits. Cases were enrolled from a database of patients who underwent liver contrast-enhanced US between 2008 and 2016 in a single tertiary care center prior to the implementation of CEUS LI-RADS. This time frame corresponded with initiation of contrast-enhanced US of the liver within our practice in 2008. A total of 987 consecutive contrast-enhanced US examinations of the liver for nodule evaluation were reviewed. Forty-two cases from the same database were previously reported in 2019 in a study that looked at the contribution of contrast-enhanced US to the characterization of nodules with indeterminate MRI characteristics in patients who were at risk for HCC (9). The purpose of this study was to highlight the features of contrast-enhanced US that could be a problemsolving tool in the event of an equivocal MRI result, essentially unrelated to the purpose of the current study.
All cases included in this study were from patients with known risk factors, as defined by the American Association for the Study of Liver Diseases, and a diagnostic-quality contrast-enhanced US scan with sufficient imaging data for the blinded read. An appropriate reference standard was additionally required, defined as confirmation of diagnosis with pathologic findings, concordant imaging at contrast-enhanced US and CT/MRI, or follow-up imaging of 24-60 months. Cases (nodules) were excluded based on the lack of risk criteria (n = 375), absence of appropriate reference standard (n = 150), and inadequate quality or insufficient imaging data for the blinded read (n = 266). A total of 196 individually evaluated nodules in 184 patients (mean age, 62 years [ age range, 27-87 years]; 138 men) were included in this study. A total of six patients had two nodules, of which four were concurrent and two were sequential. The study population is detailed in Figure 1.

Contrast-enhanced US Technique
All scans were performed with the contrast agent perflutren lipid microspheres (Definity; Lantheus Medical Imaging, Bill-Accurate assignment of the LR-5 category to a nodule evaluated with contrast-enhanced US allows for tumor management without the requirement for invasive tissue diagnosis. An important feature of the CEUS LI-RADS algorithm is that HCC may be found within any category according to a probabilistic scale ranging from 0% HCC within category LR-1 (definitely benign nodules) to 100% HCC for category LR-5 (definitely HCC), in which this categorization may allow treatment without biopsy (5). The important category of LR-M (definitely or probably malignant tumor) is designed to include all malignancies of nonhepatocellular origin and also many HCCs that do not meet the stringent criteria for category LR-5. Knowledge of this probabilistic scale for diagnosis of HCC is contributory to management decisions, including short-interval surveillance and consideration for biopsy, especially for nodules within the categories LR-4 and LR-M.
Several similarities exist between CT/MRI LI-RADS and CEUS LI-RADS. However, the differences between these tools highlight the compelling features of contrast-enhanced US, which are all reflected in the CEUS LI-RADS algorithm. One of the main advantages of contrast-enhanced US is the superior temporal resolution afforded by real-time dynamic imaging. This is particularly helpful in depicting the enhancement patterns of hepatic tumors, allowing for differentiation of benign entities (eg, flash-filling hemangiomas) from hypervascular malignant tumors (eg, HCC) (6,7). The second major advantage of contrast-enhanced US is the use of a purely intravascular contrast agent. In malignant tumors, such as massforming intrahepatic cholangiocarcinoma, CT and MRI fail to demonstrate washout because of pooling of contrast agent in the interstitium. However, on contrast-enhanced US, clear washout is visible (8). Finally, pseudolesions, such as arterioportal shunts, are typically not visualized on contrast-enhanced Abbreviations AP = arterial phase, APHE = arterial phase hyperenhancement, AUC = area under the receiver operating characteristic curve, AVI = audiovisual, CEUS LI-RADS = contrast-enhanced US Liver Imaging Reporting and Data System, cHCC-CCA = combined hepatocellular-cholangiocarcinoma, HCC = hepatocellular carcinoma, LP = late phase, NPV = negative predictive value, PPV = positive predictive value, PVP = portal venous phase Summary CEUS LI-RADS provided high specificity for noninvasive imagingbased diagnosis of hepatocellular carcinoma, removing the need for tissue diagnosis; in addition, it allowed for accurate categorization of all nodules encountered in patients at risk for hepatocellular carcinoma.

Key Points
n The contrast-enhanced US Liver Imaging Reporting and Data System allows for accurate diagnosis of hepatocellular carcinoma and its differentiation from intrahepatic cholangiocarcinoma.
n Logistic regression analysis shows that late washout of contrast agent (>60 seconds) is the contrast-enhanced US feature most indicative of hepatocellular carcinoma, and rapid washout of contrast agent (<60 seconds) is the contrast-enhanced US feature most indicative of nonhepatocellular malignancy. evaluation for its arterial enhancement was performed by injecting contrast agent "on top" so that the enhancement was superimposed on the existing PVP, with a search for a nodule in this location at gray-scale US.

Blinded Read
A computerized set of contrast-enhanced US images was numbered and fully anonymized. A resident (A. Makoyeva) and a staff radiologist (A. Medellin), who were not blinded to the results, selected the cases for the blinded read and developed the image and audiovisual (AVI) files. Each case included a single AVI movie demonstrating contrast material arrival to the point of peak enhancement and gray-scale, AP, PVP, and LP images. The number of still and PVP and LP images varied based on the timing of washout. For lesions present in association with venous thrombus, an additional set of images and an additional AVI file were attached. If an image category was unavailable or of inadequate quality, the case was not selected. Three expert readers from two tertiary care centers, all members of the CEUS LI-RADS working group, each with more than 10 years of individual contrast-enhanced US experience, independently performed the blinded read (S.R.W., T.K.K., H.J.J.). One of the readers was a contrast-enhanced US expert at the center in which the study was performed (S.R.W.) but had no participation or action in construction or analysis of the blinded read. The read was performed at least 6 months after the acquisition of the original images. Reader responses were documented on an electronic Microsoft Access response sheet, through drop-down menus for gray-scale nodule characteristics, observations on contrastenhanced US, and CEUS LI-RADS categorization. Readers assigned a CEUS LI-RADS category based on the overall evaluation of each nodule.
Interpretation of the image files was based on the CEUS LI-RADS algorithm version 2016, which was available at the time of the read. There were no major differences in the content of version 2016 and the more recent version 2017. The only substantive change was the renaming of the LR-5V category to LR-TIV for tumor in vein.

Reference Standard of Nodules by Histopathologic Analysis, Concurrent CT and/or MRI, and Follow-up Imaging
To generate reference standards, all contrast-enhanced US scans were compared against histopathologic diagnosis, concurrent CT and/or MRI, or follow-up imaging over a period of at least 24 months. erica, Mass), on multiple standard US units with contrast capability and low mechanical index technique, according to previously published protocols (10). The standard dose of perflutren lipid microspheres was 0.2 mL delivered as a bolus injection, followed by a 10-mL saline flush. Injections were repeated as needed to a maximum dose of 10 µL/kg. Details regarding contrast-enhanced US technique for nodule characterization are available in recently published guidelines (11). To note, perflutren lipid microspheres are different than the 2016 U.S. Food and Drug Administration-approved sulfur hexafluoride (Lumason; Bracco Diagnostics, Monroe Township, NJ). It was, however, of the same generation as perflutren microspheres and was used interchangeably in our facility.
Suspicious nodules were assessed with targeted contrast-enhanced US. Cine-loop storage began after visualization of the first bubble in the field of view and continued until just beyond peak arterial phase (AP) enhancement, which was assessed visually. Subsequent intermittent brief imaging of the nodule was performed, with acquisition of single frames to show the presence or absence of washout at 1 minute and at 30-to 60-second intervals, depending on the ability of the liver to retain contrast agent, for up to 5 minutes (AP 10-45 seconds, portal venous phase [PVP] 45-120 seconds, and late phase [LP] 120 seconds to 5 minutes) (12). After evaluation of target lesions, sweeps of the entire liver were done in the PVP and LP to look for other regions of washout that may have been occult on baseline assessment. If a washout lesion was identified in the PVP or LP,  Diagnosis of Hepatocellular Carcinoma Using CEUS LI-RADS Austria). To validate the diagnostic accuracy of the CEUS LI-RADS algorithm for the diagnosis of HCC, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were determined for the LR-5 category alone ( Fig  2); LR-5 and LR-5V categories (Fig 3); and categories LR-5, LR-5V, and LR-4 (Fig 4). The diagnostic performance of category LR-M ( Fig 5) for nonhepatocellular malignancy was similarly assessed. All assessments were performed for individual readers, as well as for the majority ( two of three readers). Exact methods were used to calculate 95% confidence intervals. The proportion of HCC, nonhepatocellular malignancy, and benign diagnosis was tested for all LR categories. We further scrutinized gray-scale and contrast-enhanced US characteristics for each nodule with correlation to the reference standard.
Univariable and multivariable logistic regression analysis for parameters predictive of HCC and nonhepatocellular malignancy at gray-scale and contrast-enhanced US were performed. For HCC, the parameters examined included (a) hypoechogenicity at gray-scale, (b) size of greater than 20 mm, (c) background liver cirrhosis, (d) non-rim arterial phase hyperenhancement (APHE), (e) washout of greater than 60 seconds, and (f ) mild washout. For Of the 196 nodules in 184 patients, 139 nodules in 129 patients were confirmed as HCC on the basis of histopathologic (n = 122) or imaging findings on multiphasic contrast-enhanced CT or MRI performed concurrently, according to American Association for the Study of Liver Diseases practice guidelines (n = 17).
Eighteen nodules were confirmed histopathologically as nonhepatocellular malignancies in 18 patients, and included nine cholangiocarcinomas, one combined hepatocellular and cholangiocarcinoma (cHCC-CCA), six metastases (colon, lung, stomach, and breast), one perivascular epithelioid cell neoplasm, and one lymphoma.
Three regenerative nodules in three patients were confirmed with histopathologic findings. The remaining 36 nodules in 34 patients were considered benign on the basis of stability at follow-up imaging of 24-60 months (median follow-up, 29.6 months). Stability at 24 months was used because it is an ancillary feature indicating benignancy in both the CT/MRI LI-RADS and the CEUS LI-RADS algorithms.

Statistical Analysis
Statistical analysis was performed using R-3.3.2 (Statistical Computing and Graphics Software, R Foundation, Vienna,

Test Characteristics of LI-RADS Categories Accurately Reflect Diagnosis of HCC and Non-HCC Malignancy
Histopathologic results, concurrent CT and/or MR images, and stability at follow-up imaging of 24-60 months for each nodule were compared against the assigned LI-RADS categories to determine the sensitivity, specificity, PPV, and NPV for various CEUS LR categories. The accuracy metrics for contrast-enhanced US LR-5, LR-5V, LR-4, and LR-M are shown nonhepatocellular malignancy, parameters examined included (a) hypoechogenicity on gray-scale, (b) size of greater than 20 mm, (c) background liver cirrhosis, (d) rim APHE, (e) washout of less than 60 seconds, and (f ) marked washout of less than 120 seconds. Variables that were significant (P < .05) at the univariable analysis were included in the multivariable model. Area under the receiver operating characteristic curve (AUC) was calculated for each of the models to give an estimate of model fit.
Overall agreement for LI-RADS categorization, as well as the proportion agreement between readers and majority opinion, were evaluated. Agreement was calculated using Light k. Confidence intervals were determined using a bootstrap approach. Agreement between individual readers and majority opinion was calculated as a proportion.
To avoid complexity of data presentation, we reported results calculated from majority opinion only. In instances of disagreement between the three readers, the case was not considered for analysis. This was encountered in feature analysis only. in Table 1. The proportions of HCC, non-HCC malignancy, and benign nodules in each LR category are shown in Table 2.
For nodules categorized as CEUS LR-M (n = 26), the sensitivity, specificity, PPV, and NPV for the diagnosis of nonhepatocellular malignancy were 89% (16 of 18), 95% (169 of 178), 64% (16 of 25), and 99% (169 of 171), respectively. Among the 26 cases categorized as LR-M, there were 16 non-HCC malignancies (62%), including six metastases, eight cholangiocarcinomas, one lymphoma, and one perivascular epithelioid cell neoplasm. Nine cases of HCC were categorized as LR-M (35%): 33% (three of nine) on the basis of early washout of less than 60 seconds, 56% (five of nine) by demonstrating both early washout and washout characterized as marked, and the remaining 11% (one of nine) by showing rim enhancement and early washout. A single benign regenerative nodule was categorized histologically as LR-M. This nodule measured 2.1 cm and demonstrated APHE and rapid washout by 30 seconds after contrast agent administration. This nodule was biopsied on two separate occasions, showing no malignant cells in a regenerative nodule, and was reassessed again 7 months after the initial US, demonstrating stability. The patient harboring this nodule was subsequently lost to follow-up.
Twenty-three nodules were categorized as CEUS LR-3. The majority (22 of 23, 96%) of the nodules were designated as benign with either histopathologic findings (two of 22, 9%) or interval stability of at least 24 months (20 of 22, 91%). One of the CEUS LR-3 observations (one of 23, 4%) was diagnosed as HCC with pathologic findings. This 1.9-cm HCC nodule was hypoenhancing in the arterial phase, becoming isoenhanced in  ). B, At 14 seconds following contrast agent administration, lesion showed arterial phase hyperenhancement. C, At 1 minute, lesion became isoenhanced relative to the background liver parenchyma. D, At 5 minutes, lesion remained isoenhanced and was essentially nonperceptible, never demonstrating evidence of washout. Imaging findings were consistent with LR-4 (suspicious for hepatocellular carcinoma, but not filling criteria for LR-5). Pathologic findings confirmed diagnosis of hepatocellular carcinoma.
the PVP, and showing mild late washout. All 12 LR-1 (nine of 12) and LR-2 (three of 12) observations were considered benign by exhibiting classic benign characteristics and/or interval stability of at least 24 months. No HCC nodules or other malignant nodules were characterized as LR-1 or LR-2.

Overview of Gray-Scale Nodule Characteristics
Gray-scale characteristics of echogenicity and size were documented, and their frequency was determined for HCC, non-HCC malignancy, and benign nodules (Table 3). Only nodules in which there was majority agreement were considered for the analysis (three nodules of 196 excluded for evaluation of echogenicity because of disagreement between all readers). A total of 109 of 193 nodules (56%) were hypoechoic at gray-scale US. Of these, the majority were HCC (75 of 109, 69%). The majority of the hyperechoic and heterogeneous nodules were also HCC (29 of 46, 63% and 29 of 31, 94%, respectively).
There was a single nodule measuring less than 10 mm, as 10 mm is the accepted threshold for the performance of contrast-enhanced US. In our study sample, larger nodule sizes were more frequently associated with the diagnosis of HCC; 58% (45 of 77) for nodules measuring 10-20 mm and 80% (94 of 118) for nodules measuring greater than 20 mm.

Overview of Arterial Phase Enhancement at Contrastenhanced US
Enhancement characteristics in the arterial phase were documented for all 196 nodules (139 HCC, 18 non-HCC malignancy, and 39 benign nodules). Twenty nodules demonstrated no APHE of any morphology, rather showing isoenhancement or hypoenhancement.
Non-rim APHE, referred to as APHE at contrast-enhanced US, was observed in 160 nodules. A total of 134 of 160 nodules (84%) were HCC, 12 of 160 (8%) were non-HCC malignancy (six cholangiocarcinomas, one lymphoma, one perivascular epithelioid cell neoplasm, one mixed cHCC-CCA and cholangiocarcinoma, three metastases), and 14 of 160 (9%) were benign.  Diagnosis of Hepatocellular Carcinoma Using CEUS LI-RADS Rim APHE was observed in a total of seven nodules. Six of the seven nodules (86%) were confirmed as non-HCC malignancy (three cholangiocarcinomas and three liver metastases). A single HCC (14%) demonstrated rim-APHE. No benign nodules demonstrated rim enhancement.
Peripheral discontinuous globular APHE was seen in nine of 39 benign nodules (23%). Of these, all were diagnosed as hemangioma. Peripheral discontinuous globular APHE was not identified in the HCC and the non-HCC malignancy categories.
Overview of Timing and Intensity of Washout at Contrast-enhanced US Contrast agent washout characteristics were evaluated for the 196 nodules. Forty nodules demonstrated no washout. Two of these nodules were HCC.
The remaining 27 nodules did not fill criteria for marked washout occurring by 120 seconds or mild washout.

Univariable and Multivariable Logistic Regression Analysis
Results of univariable and multivariable regression analysis for gray-scale and contrast-enhanced US features are detailed in Table 4 for HCC and Table 5 for non-HCC malignancy.
For univariable logistic regression analysis of gray-scale US features predictive of HCC, background liver cirrhosis and size of greater than 20 mm were statistically significant (P = .02 and .001, respectively). AUC was 0.58 and 0.63, respectively. On contrast-enhanced US, non-rim APHE, washout of greater than 60 seconds, and mild washout were statistically significant for Note.-Data are incidences in percentages, with numbers in parentheses and 95% confidence intervals in brackets. HCC = hepatocellular carcinoma, ICC = intrahepatic cholangiocarcinoma. Note.-Data are percentages with 95% confidence intervals in parentheses. CEUS LI-RADS = contrast-enhanced US Liver Imaging Reporting and Data System, NPV = negative predictive value, PPV = positive predictive value.   predicting HCC (all with P values of < .0001) with AUC values of 0.75, 0.94, and 0.84, respectively. For gray-scale features predictive of non-HCC malignancy, only background liver cirrhosis was statistically significant with a P value of .01 and AUC of 0.66. All three contrast-enhanced US features, including rim APHE, washout of less than 60 seconds, and marked washout (<120 seconds) were statistically significant (P < .001, < .001, and .02, respectively) with AUC values of 0.66, 0.97, and 0.60, respectively.
Multivariable logistic regression analysis showed that washout of greater than 60 seconds had the highest association with HCC (P < .001). With consideration of all statistically significant parameters in the univariable analysis, AUC increased to 0.96 (relative to AUC of 0.94 for washout of greater than 60 seconds alone). Washout of less than 60 seconds had the highest association with non-HCC malignancy (P < .001). AUC remained 0.97 following multivariable regression analysis.

Discussion
Validation of the CEUS LI-RADS algorithm is critical for effective and accurate differentiation of various nodules encountered in patients who are at risk for HCC. Although several recent publications have addressed the use of CEUS LI-RADS, they have mainly focused on interobserver agreement for categorization (13,14). Our validation study confirmed the high specificity required for LR-5 (definitely HCC), permitting diagnosis and treatment without biopsy. Further, it demonstrated prevalence of HCC in other high categories of the algorithm: LR-4 (probably HCC) and LR-M (probably or definitely malignant but not HCC-specific). Contrary to previous publications, our study emphasized individual contrast-enhanced US features and their importance for diagnosis of HCC and nonhepatocellular malignancy.
The LR-5 category requires three essential components in a nodule of greater than 10 mm, which include non-rim APHE, late washout (>60 seconds), and mild washout, in which some bubble presence remains within the nodule (2). Our univariable and multivariable logistic regression analysis demonstrated that late washout (>60 seconds) was the contrast-enhanced US feature most predictive of HCC. As previously stated, in the presence of a nodule, contrast-enhanced US has a higher specificity for APHE than CT and MRI because it does not include pseudolesions from arterioportal shunts (6,15,16).
There were two false-positive cases of LR-5 (two of 121): a biopsy-proven cholangiocarcinoma and a cHCC-CCA, both showing classic LR-5 characteristics at contrast-enhanced US. In this study, we showed excellent specificity of LR-5 for the diagnosis of HCC and accurate diagnosis of all but a single cholangiocarcinoma, categorized as LR-5 on both contrast-enhanced US and MRI. This contradicts the prior belief that contrastenhanced US does not differentiate HCC from cholangiocarcinoma (4), leading to its exclusion from major international guidelines for diagnosis of HCC. A recent publication validating diagnosis of HCC with MRI LI-RADS showed a specificity of Diagnosis of Hepatocellular Carcinoma Using CEUS LI-RADS 87%, with a higher rate of false-positive findings than in this study, mainly attributed to the presence of mixed cHCC-CCA tumors (17). In our experience, supported by previous publications, mixed tumors often exhibit imaging features suggestive of nonhepatocellular tumor origin, while acknowledging that overlap may occur (18). LR-5V, recently renamed LR-TIV (version 2017), indicates tumor thrombus, which is a characteristic that is highly associated with HCC. Recognition of arterial vascularity within tumor thrombi is a well-recognized strength of contrast-enhanced US and is not infrequently a challenge at CT and MRI (19). Stringent criteria for LR-5 are not met by all HCC. However, the algorithm addresses all nodules of hepatocellular origin in a diagnostic table which categorizes them as LR-3, with "intermediate malignancy probability"; LR-4, as probable HCC; and LR-5, as definite HCC. Categorization is based on lesion size, arterial phase enhancement, and presence or absence of washout. Management of LR-3 and LR-4 has great variation.
Although hypoechogenicity at gray-scale US is the most common appearance of HCC, as was shown in our gray-scale feature analysis, virtually any appearance on gray-scale US may ultimately represent an HCC. Last, larger nodule size was associated with a diagnosis of HCC in our study sample.
LR-M requires any of the following features: rim-APHE, rapid washout (< 60 seconds), or marked washout, temporally close to the first observation of washout (< 120 seconds). Univariable and multivariable logistic regression analysis demonstrated that rapid washout (< 60 seconds) was the contrast-enhanced US feature most predictive of nonhepatocellular malignancy and that the presence of additional features did not increase the likelihood of diagnosis of a non-HCC malignancy.
All LR-M lesions undergo biopsy for two main reasons: imaging features do not differentiate the type of non-HCC malignancy, and a fair number of HCCs can be classified as LR-M when they do not meet the stringent criteria for LR-5. More than one-third of the lesions categorized as LR-M in our sample were HCC, showing similar incidence to a recent Italian publication by Terzi et al on CEUS LI-RADS (20).

Limitations
The biggest limitation of this study was related to a disproportionate number of HCCs in our sample. On reflection, we believe that this problem occurred for several reasons. HCC surveillance and nodule evaluation long precede the implementation of CEUS LI-RADS. Retrospective review for recruitment to this study required identification of a nodule at gray-scale, and its enhancement in the AP, PVP and LP, to comply with study design. Because HCC and nonhepatocellular malignancies show clear and well-defined enhancement patterns, even in retrospect, patients with these nodule characteristics are easily recruited, whereas precancerous hepatocellular nodules (LR-3, LR-4) tend to show variable enhancement and washout characteristics, which makes their documentation and retrospective inclusion difficult. Presently, our contrast-enhanced US summary file includes the essential components of a baseline grayscale image, an image obtained at the peak of AP enhancement, and serial images showing the presence or absence of washout regardless of suspected diagnosis. A 60-second image stratifies rapid and late washout. We believe that if we were to collect data today, the systematic acquisition of data would yield a much larger sample of nodules, particularly the precancerous LR-3 and LR-4 nodules, which were notoriously difficult to recognize and document in the past. Integration of the CEUS LI-RADS diagnostic table, integrated with knowledge of hepatocarcinogenesis, allows for recognition and documentation of virtually all nodules in patients who are at risk for HCC.
We included essentially no LR-2 (probably benign) lesions in our study because the current criteria in CEUS LI-RADS for category LR-2 is a size of less than 10 mm. Because 10 mm is the standard threshold above which contrast-enhanced US is performed for nodule categorization, there was an obvious conflict. We believe this will be resolved in future revision of the algorithm. LR-1 (definitely benign), composed mainly of hemangiomas, was not frequent in this high-risk population. In the few cases in which venous thrombus was present, we added an extra set of images and an additional AVI file, inevitably introducing bias. Finally, all blinded readers were current members of the LI-RADS contrast-enhanced US working group and were highly familiar with this topic and imaging technique.

Conclusion
Retrospective validation of the CEUS LI-RADS algorithm showed excellent diagnostic performance of key malignant and premalignant categories of the algorithm. Univariable and multivariable regression analysis, ranking contrast-enhanced US features for diagnosis of HCC and nonhepatocellular malignancy, contributed to the performance and implementation of the CEUS LI-RADS algorithm by increasing awareness of its key diagnostic features. In addition, the recognition of HCC occurrence outside of the LR-5 category is critical for management strategies.
Author contributions: Guarantor of integrity of entire study, S.R.W.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; approval of final version of submitted manuscript, all authors; agrees to ensure any questions related to the work are appropriately resolved, all authors; literature research, A. Makoyeva, T.K.K., A. Medellin; clinical studies, all authors; statistical analysis, A. Makoyeva; and manuscript editing, all authors.

Disclosures of Conflicts of Interest:
A. Makoyeva disclosed no relevant relationships. T.K.K. Activities related to the present article: disclosed no relevant relation-ships. Activities not related to the present article: employed by Medical Imaging, Toronto General Hospital. Other relationships: disclosed no relevant relationships. H.J.J. disclosed no relevant relationships. A. Medellin disclosed no relevant relationships. S.R.W. Activities related to the present article: partial drug support for performance of initial contrast-enhanced US liver examinations (This retrospective study is performed at a single institution, the University of Calgary. Review of many cases would have included some patients with CEUS performed many years ago who had pharmaceutical supply of the contrast agent Definity, Lantheus Medical Imaging. Activities not related to the present article: institution received partial support research grant for an ongoing study relating to ablative therapy monitoring with contrast-enhanced US from Samsung; author paid for lectures and for development of educational presentations by Philips; institution receives small research grant to support student research on contrast-enhanced US in broad terms from Lantheus Medical Imaging; equipment support from Samsung, Siemens, and Philips. Other relationships: disclosed no relevant relationships.